Novel approaches in prognosis and personalized treatment of cancer

Expression of nuclear FIH independently predicts overall survival of clear cellrenal cell carcinoma patients Chapter 3 1

Novel approaches in prognosis and personalized treatment of cancerJoost Vermaat 2012

Part I Molecular prognosticators2


Novel approaches in prognosis and personalized treatment of cancer Joost Vermaat 2012


Novel approaches in prognosis and personalized treatment of cancer

Thesis with summary in Dutch, Utrecht University, The Netherlands

Proefschrift met Nederlandse samenvatting, Universiteit Utrecht

Joost Vermaat

ISBN/EAN: 978-94-6169-236-8

Cover design and lay-out by:

W. de Jong

Printed by:

Optima Grafische Communicatie, Rotterdam

The printing of this thesis was financially supported by:

Stichting Het Scholten-Cordes Fonds, Tridelta Development Ltd, Von Hippel Lindau Belangenvereniging,

J. E. Jurriaanse Stichting, Maatschap Interne Geneeskunde van het St. Antoniusziekenhuis Nieuwegein,

Janssen Biologics BV, Amgen BV, Novartis Oncology, Bayer BV, Pfizer Oncology, Boehringer Ingelheim BV,

Roche Nederland BV, GlaxoSmithKline BV, Sanofi Aventis, ChipSoft en Vermaat Groep BV

Copyright © 2012 J.S.P. Vermaat, the Netherlands.

All rights reserved. No part of this thesis may be reproduced or transmitted,

in any form or by any means without prior permission of the author.


Novel approaches in prognosis and personalized treatment of cancer

Nieuwe benaderingen van

prognose en gepersonaliseerde behandeling van kanker

(met een samenvatting in het Nederlands)

Proefschriftter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof. dr. G.J. van der Zwaan

ingevolge het besluit van het college van promoties in het openbaar te verdedigen op donderdag 10 mei 2012 des middags te 4.15 uur

doorJonathan Stephanus Philippus Vermaatgeboren op 11 maart 1980 te Veenendaal


Promotor

Prof. dr. E.E. Voest


Promotiecommissie

Prof. dr. J.H. Beijnen Prof. dr. E.E. Cuppen Prof. dr. P.H. van Diest Prof. dr. J.B.A.G. Haanen Prof. dr. G.J.P. Kops

Paranimfen C.A. van Antwerpen and G. Cornelisse


Table of Contents

Chapter 1 General introduction

Chapter 2 Outline of the thesis

PART I MoleCu lAR P Ro g nosTICAToR s

Chapter 3 Expression of nuclear FIH independently predicts overall survival of clear cell renal cell carcinoma patients

Chapter 4 Regulation of E2F1 by the von Hippel-Lindau tumor suppressor protein predicts survival in renal cell cancer patients

PART I I P RoTeoM ICs Chapter 5 Two-protein signature of novel serological markers apolipoprotein-A2 and serum amyloid alpha predicts prognosis in patients with metastatic renal cell cancer and improves the currently used prognostic survival models

Chapter 6 Validation of serum amyloid alpha as an independent biomarker for progression free- and overall survival in metastatic renal cell cancer patients

11

21

27

45

69

91


PART I I I M IToX

Chapter 7 Circulating cell-free mitochondrial DNA as prognostic but not predictive marker for various types of cancer

PART IV g e noM ICs

Chapter 8 Primary colorectal cancers and their subsequent hepatic metastases are genetically different: implications for selection of patients for targeted treatment

Chapter 9 General discussion and future perspectives

Dutch summary / nederlandse samenvatting

Contributing Authors

Acknowledgements / Dankwoord

list of Publications

Curriculum Vitae

111

133

155

169

177

181

185

189


Chapter 1 general introduction

12

g loBAl CAnCe R sTATI sTICs

Today, cancer is the leading cause of death in economically developed nations and the second cause of death in developing countries (1). Cancer-associated lifestyles, such as smoking, ‘westernized’ diets and physical inactivity, as well as population expansion and aging are attributable to this phenomenon. Approximately 12.7 million people were universally diagnosed with cancer in 2008 (1). Accordingly, about 7.6 million estimated deaths were registered as cancer-associated in the same year. Overall, leading world health organizations anticipate that these incidences and mortality rates will enhance over the next decade. To challenge these increasing cancer statistics, improvements in cancer prevention, diagnostic processes, prognostications and novel therapeutic approaches are indisputably necessary. This thesis primarily deals with kidney (Chapters 3-6) and colorectal cancer (Chapter 8) and therefore statistics of these malignancies are briefly reviewed here. Accordingly, the role of biomarkers in prognostication and prediction of therapeutic responsiveness is glanced. Next, the upcoming targeted therapies indicating the direction in personalized cancer medicine is discussed. Furthermore, as a genomics advance the novel technology ‘Next Generation Sequencing’ is mentioned, which contributes to a better understanding of mutational variations that drive tumor evolution and the current upcoming targeted therapies in cancer treatment.

R e nAl Ce ll CAnCe R

Kidney cancer accounts for approximately 3% of all adult malignancies. The worldwide incidence of renal cell carcinoma (RCC) is more than 200,000 newly diagnosed cases and is gradually rising per annum (2;3). About 30% of RCC patients have metastasized disease at time of diagnosis and another 30% will develop local or distant recurrences within 5 years after resection of the malignant kidney with initial curative intention (2-4) . Observational follow-up remains the postoperative standard of care for patients who were diagnosed without any metastasis. Unfortunately, metastatic RCC (mRCC) is almost irresponsive to cytotoxic chemotherapy and cytoreductive surgery seems to have controversy efficacy (5;6), making treatment options for mRCC limited. The past decades Interferon-alpha (IFN) and interleukin-2 (IL-2) have been adopted as first-line treatment for clinical management of mRCC patients, but the outcomes were poor with objective response rate (ORR) of 10-15% (3;6;7). Recently, these immune-based approaches were surpassed by new treatment modalities including tyrosine kinase inhibitors (TKIs), targeting the Vascular Endothelial Growth Factor (VEGF)-receptor, as well as the mammalian target of rapamycin (mTOR)-inhibitors (8-13). Accordingly, the addition of bevacizumab (a humanized monoclonal VEGF-antibody) to IFN has improved patient outcomes. The ORR for these targeted therapies have improved significantly, for some novel drugs even up to 30-40% (14). These patient outcome advances have unfortunately been accompanied by the expense of substantial toxicities for considerable patient numbers. Furthermore, despite these therapeutic improvements, prognosis of mRCC patients remains poor with a corresponding 5-year survival rate of 9% in metastatic disease (1;4;15). Therefore RCC is currently the most lethal of all genitourinary malignancies (1;16).

13General introduction Chapter 1

C oloR eCTAl CAnCe R

The global estimated incidence of colorectal cancer (CRC) in 2008 was about 1.2 million new cases (1). Therefore CRC is the second most commonly diagnosed cancer for females and third for males. More than 600,000 deaths were attributable to CRC in 2008 making this cancer type the second most common/predominant cause of cancer-related death (1;17;18). Resection of the primary tumor is the first therapy of choice to cure CRC. Depending on disease stage primary tumor resection is followed by (neo)adjuvant chemotherapy and/or radiation. CRC predominantly metastasizes to the liver (18). More than 25% of CRC patients are diagnosed with liver metastasis and approximately 50% of all CRC patients will develop liver metastasis in time (19;20). If metastasized disease is only limited to the liver, hepatic metastasectomy is moreover considered with curative intention (21;22). Conventional chemotherapy is given to patients with metastatic colorectal cancer (mCRC). Recently, chemotherapeutic strategies for CRC were expanded with targeted therapies which improved patients’ outcome. For example, the antibodies cetuximab and panitumumab antagonizing the anti-epidermal growth factor receptor (EGFR), improved overall survival of mCRC patients, if combined with chemotherapy regimens (23;24). Accordingly, chemotherapy is combined with bevacizumab, a monoclonal antibody of the vascular endothelial growth factor (VEGF), improving patients’ survival (25). Unless this progress many mCRC patients do not respond to these modernized therapeutic strategies (26). Subsequently, there is an indisputable need for appropriate applicability of targeted therapies. Besides for improving patient selection there is a great need for novel biomarkers (27).

B IoMAR K e R s I M P RoVe ‘R I s K-DI R eCTe D’ TH e RAP I e s

Despite improvements in treatment strategies, for many cancer patients disease course continues to be variable and prognosis for metastasized cancer remains very poor. For decades many cancer patients have been exposed to toxic schemes before valuable clinical response evaluation. Obviously, these treatment modalities are not void of side effects and consequently it is important to select only patients that may benefit from these toxic systemic therapies. Novel risk factors, also called biomarkers, predicting prognosis and/or therapeutic responsiveness, might be of great importance to select individual patients for appropriate therapy. These predictors and prognosticators are increasingly guiding therapeutic strategies and contribute to enhancement of personalized treatment strategy. In general, biomarkers can be divided in tumor tissue-based factors and substances determined in patients’ blood, such as proteins or nucleic acids. Cancer biology studies have unraveled entire molecular pathways that trigger tumorigenesis and accordingly these investigations have assisted the search for novel tissue biomarkers. Molecular factors, for example HER2-neu in breast cancer, are currently used in clinical cancer staging systems consequently adjusting treatment strategy (28). However, optimizing and refining these cancer staging systems with novel tissue biomarkers would advantage the clinical management of patients (Chapter 3 and 4). Currently, in the pre-phase of treatment initiation cancer patients are categorized in risk groups, discriminating between favourable and poor overall survival. This categorization is used to

14

select patients in advance for ‘risk-directed’ targeted therapy, thereby improving clinical outcome and avoiding long-term schemes of toxic treatment for poor risk patients who will not benefit. The risk classification systems are generally based on disease stage, patients’ performance status and laboratory blood tests. For example, mRCC patients stratified in the favourable and intermediate risk groups are treated with TKIs, where mTOR-inhibitors are the therapeutic choice for poor risk categorized patients (7). However, it remains a challenge to optimize and develop risk stratification that improves the a-priori patient selection for ‘risk-directed’ targeted therapies (Chapter 5-7). It is important that biomarkers should meet two requirements. First, biomarkers are supposed to display the highest sensitive and specific robustness, thereby outperforming the predictive accuracy of the currently used clinical evaluation examinations. Second, biomarkers should be economically justified. The monitoring of biomarkers is an elegant and noninvasive method that can easily be implemented in clinical settings. Biomarkers will increasingly be used in clinical practice for cancer staging and risk categorization, thereby adjusting specific therapies and thus improving patient management, survival outcome, best supportive care and quality of life.

TARg eTe D TH e RAP I e s e nC ou RAg e ‘P e R sonAlI Z e D’ CAnCe R M e DICI n e

The main challenge for the future of cancer treatment is providing every individual patient with the most effective drug tailored to their unique cancer. However this approach is still in the early stages, novel targeted therapies have pushed the development in the personalization of cancer medicine. The first example of this personalized approach is trastuzumab which is given to patients with breast and gastric cancer that have HER2-amplified tumors (29). Other examples are receptor tyrosine kinases, which are amenable to pharmaceutical intervention by tyrosine kinase inhibitors, TKIs (30). A kinase commonly targeted in clinical oncology management is the epidermal growth factor receptor (EGFR). Oncogenic EGFR overexpression leads to uncontrolled cell division. Gefitinib is an effective EGFR TKI for 10-20% of non-small cell lung cancer (NSCLC) patients presenting with an activating EGFR mutation (31;32). The EGFR mutational status is incorporated into clinical management of the modern oncologist to adapt personalized treatment with Gefitinib correlated to patients’ particular genotype. Similarly, the monoclonal antibody drugs Cetuximab and Panitumumab also target EGFR and have been approved for patients with CRC tumors (26;33;34). However, researchers discovered that mCRC patients with a KRAS mutation will not benefit from Cetuximab and that only patients with wild-type (wt) KRAS respond to these EGFR-antibodies (26;35-38). For that reason the mutational KRAS status is routinely screened in clinical settings. In addition it is becoming clear that wild-type BRAF and PTEN related to the EGFR pathway are required for response to Cetuximab (33;34). This finding will most likely result in Cetuximab only being prescribed to mCRC patients with wild-type KRAS, BRAF and PTEN, saving patients with KRAS/BRAF/PTEN mutations from the side-effects of Cetuximab. Besides, this specific drug use also has advantages when it comes to cost-effectiveness. These examples demonstrate that individual genetic mutations drive therapeutic strategies more towards a ‘personalized targeted therapy’ approach nowadays. Nevertheless, this progress has its limitations as therapeutic responsiveness

15

is only partially explained by these individual mutated genes (39). Moreover, it is unlikely that one single mutated gene is responsible for entire therapeutic efficacy; in other words a whole panel of mutated genes will be involved in this response process (40). This emphasizes the need for an integrated assessment of more complete genetic profiling of relevant cancer-associated pathways within the tumor to optimize the personalized treatment strategy (Chapter 8).

‘n e XT g e n e RATIon s eQu e nCI ng’ : A noVe l An D soP H I sTICATe D g e noM ICs AP P lICATIon

Besides targeted therapies, genetic cancer research has provided useful genomic biomarkers for diagnosis and prediction of therapy response. Sanger sequencing was the most comprehensive technology available for decades, used also in the clinic for diagnostics, e.g. BRCA1/2 genotyping in breast cancer patients. When the Human Genome Project (HGP) started to sequence the human genome in 1990, the capacity of the available techniques was pushed to and over its limits. It is ironic that one of the greatest milestones of the 21st century, completing the human genome in 2003 (41), was accomplished with relatively ‘old’ technology, although it stimulated the development of novel sequencing techniques (42). Next Generation Sequencing (NGS) technology, also called massively parallel, 2nd generation or high throughput sequencing, has so greatly reduced the costs and increased the sequencing output that –only 5 years later- sequencing the human genome is possible in a single laboratory within a few weeks (43;44). The immense advance which is made in Next Generation Sequencing technology is driving progress to the point that it is now possible to generate complex and detailed analysis of human tumor material (45-48). With the advent of NGS technologies, genetic cancer research has moved beyond single gene analysis to the investigation of hundreds of genes and complete downstream signaling pathways simultaneously (49). NGS technologies make it now possible to read billions of DNA bases -a large part of the human genome- in one single analysis (50). Approaches in genomics research are changing accordingly, including mutation discovery, whole genome sequencing, small RNA discovery and detection, gene expression profiling and DNA-protein interactions. Currently there are several commercially available NGS platforms (43;51). Each system has developed its own sample preparation, technology, chemistry and data acquisition as has been reviewed in great depth (43;44;52). Every platform has its pros and cons making it more or less suitable for certain research questions. Nevertheless the continuing technological development is likely to result in a status quo at which the accuracy, capacity and cost of all platforms will be comparable. The human genome comprises about 3 billion basepairs. Roughly 1% of the genome however, approximately 30 megabases, represented by 180,000 exons, encodes for a total of about 23,000 proteins (41). Although technically possible, sequencing the complete genome of every patient at a statistically significant level of coverage requires multiple runs. From a clinical point of view the protein-coding regions, only a small percentage of the genome, are the most relevant DNA regions thereby greatly enhancing the capacity and reducing costs. Reducing the DNA content to be sequenced, and thus reducing the complexity of the sample and increasing the fold-coverage per gene, can be accomplished by pre-selecting relevant genes and/or regions, which is also called ´target-enrichment sequencing strategies´. For example, one could compose a tumor mini-

General introduction Chapter 1

16

genome containing all relevant cancer-associated genes for a specific tumor type. Enrichment strategies are already widely used in many NGS studies (53). Today, NGS enables the generation of large mutational profiles, which might provide better guidelines for individualized less toxic and more appropriate therapy. It is not known whether the primary tumor is the best reflection of actual disease status. It is therefore relevant to question whether screening a metastasis might be a better alternative. Mutational profiling of large numbers of genes will also facilitate the understanding of tumor evolution. NGS permits to study tumor evolution in time by analyzing genetic profiles of primary tumors compared to the mutational outline of the comparative metastasis (Chapter 8). NGS is an extremely appropriate method to anticipate the unquestionable need for the analysis of entire panels of mutations to investigate new biomarkers guiding personalized targeted therapy (40).

17

R e f e R e nCe lI sT

(1) Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011;

61(2):69-90.

(2) Hollingsworth JM, Miller DC, Daignault S, Hollenbeck BK. Rising incidence of small renal masses: a need to

reassess treatment effect. J Natl Cancer Inst 2006; 98(18):1331-1334.

(3) Cohen HT, McGovern FJ. Renal-Cell Carcinoma. N Engl J Med 2005; 353(23):2477-2490.

(4) Gupta K, Miller JD, Li JZ, Russell MW, Charbonneau C. Epidemiologic and socioeconomic burden of metastatic

renal cell carcinoma (mRCC): a literature review. Cancer Treat Rev 2008; 34(3):193-205.

(5) Motzer RJ, Bander NH, Nanus DM. Renal-cell carcinoma. N Engl J Med 1996; 335(12):865-875.

(6) Pantuck AJ, Belldegrun AS, Figlin RA. Cytoreductive nephrectomy for metastatic renal cell carcinoma: is it still

imperative in the era of targeted therapy? Clin Cancer Res 2007; 13(2 Pt 2):693s-696s.

(7) Motzer RJ, Bacik J, Murphy BA, Russo P, Mazumdar M. Interferon-alfa as a comparative treatment for clinical

trials of new therapies against advanced renal cell carcinoma. J Clin Oncol 2002; 20(1):289-296.

(8) Escudier B, Eisen T, Stadler WM et al. Sorafenib in advanced clear-cell renal-cell carcinoma. N Engl J Med

2007; 356(2):125-134.

(9) Motzer RJ, Hutson TE, Tomczak P et al. Sunitinib versus interferon alfa in metastatic renal-cell carcinoma.

N Engl J Med 2007; 356(2):115-124.

(10) Hudes G, Carducci M, Tomczak P et al. Temsirolimus, interferon alfa, or both for advanced renal-cell carcinoma.

N Engl J Med 2007; 356(22):2271-2281.

(11) Motzer RJ, Escudier B, Oudard S et al. Efficacy of everolimus in advanced renal cell carcinoma: a double-blind,

randomised, placebo-controlled phase III trial. Lancet 2008; 372(9637):449-456.

(12) Escudier B, Pluzanska A, Koralewski P et al. Bevacizumab plus interferon alfa-2a for treatment of metastatic

renal cell carcinoma: a randomised, double-blind phase III trial. Lancet 2007; 370(9605):2103-2111.

(13) Rini BI, Halabi S, Rosenberg JE et al. Bevacizumab Plus Interferon Alfa Compared With Interferon Alfa

Monotherapy in Patients With Metastatic Renal Cell Carcinoma: CALGB 90206. J Clin Oncol 2008;

26(33):5422-5428.

(14) Patel PH, Chaganti RS, Motzer RJ. Targeted therapy for metastatic renal cell carcinoma. Br J Cancer 2006;

94(5):614-619.

(15) Motzer RJ, Bacik J, Schwartz LH et al. Prognostic factors for survival in previously treated patients with

metastatic renal cell carcinoma. J Clin Oncol 2004; 22(3):454-463.

(16) D’Hondt V, Gil T, Lalami Y, Piccart M, Awada A. Will the dark sky over advanced renal cell carcinoma soon

become brighter? Eur J Cancer 2005; 41(9):1246-1253.

(17) Jemal A, Siegel R, Ward E, Hao Y, Xu J, Thun MJ. Cancer statistics, 2009. CA Cancer J Clin 2009;

59(4):225-249.

(18) Field K, Lipton L. Metastatic colorectal cancer-past, progress and future. World J Gastroenterol 2007;

13(28):3806-3815.

(19) Nadal C, Maurel J, Gascon P. Is there a genetic signature for liver metastasis in colorectal cancer? World J

Gastroenterol 2007; 13(44):5832-5844.

(20) Boyle P, Leon ME. Epidemiology of colorectal cancer. Br Med Bull 2002; 64:1-25.

(21) Komborozos VA, Skrekas GJ, Pissiotis CA. The contribution of follow-up programs in the reduction of mortality

of rectal cancer recurrences. Dig Surg 2001; 18(5):403-408.

(22) Choti MA, Sitzmann JV, Tiburi MF et al. Trends in long-term survival following liver resection for hepatic

colorectal metastases. Ann Surg 2002; 235(6):759-766.


18

(23) Van CE, Kohne CH, Hitre E et al. Cetuximab and chemotherapy as initial treatment for metastatic colorectal

cancer. N Engl J Med 2009; 360(14):1408-1417.

(24) Amado RG, Wolf M, Peeters M et al. Wild-type KRAS is required for panitumumab efficacy in patients with

metastatic colorectal cancer. J Clin Oncol 2008; 26(10):1626-1634.

(25) Hurwitz H, Fehrenbacher L, Novotny W et al. Bevacizumab plus irinotecan, fluorouracil, and leucovorin for

metastatic colorectal cancer. N Engl J Med 2004; 350(23):2335-2342.

(26) Tol J, Koopman M, Cats A et al. Chemotherapy, bevacizumab, and cetuximab in metastatic colorectal cancer.

N Engl J Med 2009; 360(6):563-572.

(27) Walther A, Johnstone E, Swanton C, Midgley R, Tomlinson I, Kerr D. Genetic prognostic and predictive markers

in colorectal cancer. Nat Rev Cancer 2009; 9(7):489-499.

(28) Shepard HM, Jin P, Slamon DJ, Pirot Z, Maneval DC. Herceptin. Handb Exp Pharmacol 2008;(181):183-219.

(29) Romond EH, Perez EA, Bryant J et al. Trastuzumab plus adjuvant chemotherapy for operable HER2-positive

breast cancer. N Engl J Med 2005; 353(16):1673-1684.

(30) McDermott U, Settleman J. Personalized cancer therapy with selective kinase inhibitors: an emerging paradigm

in medical oncology. J Clin Oncol 2009; 27(33):5650-5659.

(31) Lynch TJ, Bell DW, Sordella R et al. Activating mutations in the epidermal growth factor receptor underlying

responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med 2004; 350(21):2129-2139.

(32) Paez JG, Janne PA, Lee JC et al. EGFR mutations in lung cancer: correlation with clinical response to gefitinib

therapy. Science 2004; 304(5676):1497-1500.

(33) Loupakis F, Pollina L, Stasi I et al. PTEN expression and KRAS mutations on primary tumors and metastases in

the prediction of benefit from cetuximab plus irinotecan for patients with metastatic colorectal cancer. J Clin

Oncol 2009; 27(16):2622-2629.

(34) Di NF, Martini M, Molinari F et al. Wild-type BRAF is required for response to panitumumab or cetuximab in

metastatic colorectal cancer. J Clin Oncol 2008; 26(35):5705-5712.

(35) Lievre A, Bachet JB, Le CD et al. KRAS mutation status is predictive of response to cetuximab therapy in

colorectal cancer. Cancer Res 2006; 66(8):3992-3995.

(36) Garm Spindler KL, Pallisgaard N, Rasmussen AA et al. The importance of KRAS mutations and EGF61A>G

polymorphism to the effect of cetuximab and irinotecan in metastatic colorectal cancer. Ann Oncol 2009;

20(5):879-884.

(37) Bardelli A, Siena S. Molecular mechanisms of resistance to cetuximab and panitumumab in colorectal cancer.

J Clin Oncol 2010; 28(7):1254-1261.

(38) Hawkes E, Cunningham D. Relationship between colorectal cancer biomarkers and response to epidermal

growth factor receptor monoclonal antibodies. J Clin Oncol 2010; 28(28):e529-e531.

(39) Ginsburg GS, Willard HF. Genomic and personalized medicine: foundations and applications.

Transl Res 2009; 154(6):277-287.

(40) Chin L, Gray JW. Translating insights from the cancer genome into clinical practice. Nature 2008; 452(7187):

553-563.

(41) Szekely M. Sequencing DNA. Nature 1977; 267(5607):104.

(42) Sanger F, Coulson AR. A rapid method for determining sequences in DNA by primed synthesis with DNA

polymerase. J Mol Biol 1975; 94(3):441-448.

(43) Metzker ML. Sequencing technologies - the next generation. Nat Rev Genet 2010; 11(1):31-46.

(44) Venter JC, Adams MD, Myers EW et al. The sequence of the human genome. Science 2001; 291(5507):

1304-1351.

(45) Comprehensive genomic characterization defines human glioblastoma genes and core pathways.

Nature 2008; 455(7216):1061-1068.

19

(46) Jones S, Zhang X, Parsons DW et al. Core signaling pathways in human pancreatic cancers revealed by global

genomic analyses. Science 2008; 321(5897):1801-1806.

(47) Sjoblom T, Jones S, Wood LD et al. The consensus coding sequences of human breast and colorectal cancers.

Science 2006; 314(5797):268-274.

(48) Wood LD, Parsons DW, Jones S et al. The genomic landscapes of human breast and colorectal cancers.

Science 2007; 318(5853):1108-1113.

(49) Ocana A, Pandiella A. Personalized therapies in the cancer “omics” era. Mol Cancer 2010; 9:202.

(50) Rothberg JM, Leamon JH. The development and impact of 454 sequencing. Nat Biotechnol 2008; 26(10):

1117-1124.

(51) Meyerson M, Gabriel S, Getz G. Advances in understanding cancer genomes through second-generation

sequencing. Nat Rev Genet 2010; 11(10):685-696.

(52) Bennett S. Solexa Ltd. Pharmacogenomics 2004; 5(4):433-438.

(53) Gnirke A, Melnikov A, Maguire J et al. Solution hybrid selection with ultra-long oligonucleotides for massively

parallel targeted sequencing. Nat Biotechnol 2009; 27(2):182-189.



Chapter 2 outline of the thesis

22

One of the major ambitions of present clinical cancer investigations is to enlarge and optimize ‘personalized’ cancer therapy. Essentially, this can be achieved by two different investigational methodologies; a) improving risk-directed therapy by modifying prognostication in advance (Chapter 3-7) and b) expanding and refining targeted therapies (Chapter 8).

PART I Molecular prognosticators

Clinical and pathologic factors are presently used in various prognostic models for disease staging of Renal Cell Cancer (RCC), such as the University of California Integrated Staging System (UISS), Stage, Size, Grade and Necrosis (SSIGN) score and Tumor-Node-Metastasis (TNM) stage including the Fuhrman grade. Every individual risk stratification based on these clinical and pathological variables has its own predictive accuracy although important differences in robustness have been demonstrated. To improve these prognostic models we embarked on two independent investigations to evaluate novel tissue biomarkers in large cohorts of RCC patients. The molecular pathways behind tumorigenesis of RCC have been increasingly unraveled the past years and this facilitates the search for novel prognostic tissue biomarkers. In RCC the hypoxia inducible factor (HIF) pathway dominates tumorigenic evolution by pressing on cellular processes like cell proliferation, cell survival, angiogenesis and metastases formation. In Chapter 3, the upstream targets of the HIF pathway -prolyl hydroxylases domain proteins (PHD) 1, 2 and 3 and factor inhibiting HIF (FIH) have been assessed for their independent prognostic importance in RCC patients. Inactivating mutations of the von Hippel-Lindau (VHL) gene are frequently present in RCC patients. Since the evidence that the transcription factor E2F1 is regulated by VHL is growing, we investigate in Chapter 4 whether E2F1 is associated with patients’ survival and attempt to elucidate the regulation mechanism of E2F1 by VHL in RCC cell lines and a zebrafish disease model.

PART I I Proteomics

The Memorial Sloan-Kettering Cancer Center (MSKCC) model is used in clinical management to assign metastasized RCC patients to three prognostic risk groups (favourable, intermediate or poor). According to these risk categories mRCC patients are exposed to different systemic targeted therapies, such as Tyrosine Kinase Inhibitors (TKIs) or mTOR-inhibitors. Currently the MSKCC model is composed of five independent variables; patients’ performance status, elevated lactate dehydrogenase, decreased hemoglobin, raised ‘corrected’ calcium and time between diagnosis and start of systemic treatment. However, refinement of this risk model is indisputably essential to improve patient selection in advance for ‘risk-directed’ therapies. In the search for new independent prognostic indicators a proteomics approach was applied to

23Outline of the thesis Chapter 2

identify novel prognostic proteins in serum of interferon-based treated patients with mRCC as described in Chapter 5. Subsequently, three specific proteins (Apolipoprotein-A2, Serum Amyloid Alpha and Transthyretin) are analyzed for the question if they can improve the currently used MSKCC risk model. The objective of the subsequent study (Chapter 6) was to validate the independent prognostic value of SAA and ApoA2 prospectively in an independent cohort of mRCC patients treated with TKIs. Accordingly, in a subset of patients the predictive accuracy of changes in SAA-levels after 6-8 weeks of TKI-treatment for therapeutic responsiveness was investigated.

PART I I I MITOX

Genomic nucleic acids circulating in blood (including all genomic DNA, mRNA, microRNA, mitochondrial DNA, viruses etc.) have demonstrated diagnostic and prognostic importance in patients with several cancer types. Elevated genomic nucleic acids levels in blood of cancer patients have been found due to several processes like apoptosis, necrosis, cell lysis and spontaneous release by tumor cells. However the prognostic and predictive significance of mitochondrial-(mt)DNA in cancer patients is still unclear. In this study the prognostic and predictive significance of mtDNA as pan-tumor marker in different human cancers was investigated (Chapter 7). First, a straightforward quantitative Polymerase-Chain-Reaction (qPCR) was developed, accurately quantifying circulating mtDNA in plasma of cancer patients. The next aim was to show that mtDNA has prognostic significance as a pan-tumor indicator in patients with several cancer types in a large retrospective cohort. Subsequently, these results are validated in a prospective cohort of diverse cancer patients and evaluated whether mtDNA has importance in predicting chemotherapy response.

PART IVGenomics

Single gene analyses established the first strategies in individualized cancer treatment, such as the KRAS mutation status in colorectal cancer (CRC) patients. For example, if wild-type KRAS is detected in a primary CRC tumor, cetuximab or panitumumab treatment will be considered; however, if the lesion harbors mutated KRAS, the antibodies will not be given to these patients. Nevertheless the power of single-gene analysis is limited. Genomic instability is a hallmark of cancer; because of the constant selection pressure, tumors rapidly change their genetic make-up over time. Secondly, specific populations of tumor cells may be more prone to metastasize than others, which is likely to result in an enrichment of these cells and consequently their genetic aberrations in the metastases. Thirdly, systemic treatment may induce selection pressure toward a specific genetic phenotype or induce additional genetic changes. Overall, tumors are genetically dynamic, which suggests that selecting patients for targeted treatments based on the characteristics of the primary

24

tumor and not their metastases may not be optimal. Therefore, in the era of DNA-guided personalized cancer treatment, it is essential to perform predictive analysis on the tissue that matters. Next generation sequencing (NGS) is an extremely powerful technology for genetic analysis of complete signaling pathways in large patient cohorts. The objective of the investigation presented in Chapter 8 is to analyze the genetic differences of a composed ‘cancer genome’ between the primary colorectal cancer (CRCs) tumor and their respective liver metastasis to show which tissue best reflects the presence of a target for therapy. The designed ‘cancer genome’ consists of 1,264 genes known as most relevant cancer-associated pathways. Applying the novel technology targeted deep-sequencing (NGS) the genetic make-up of the ‘cancer genome’ of primary CRC tumors of 21 patients are elucidated and compared to their corresponding liver metastasis.

25Outline of the thesis Chapter 2


Chapter 3 expression of nuclear fIH independently predicts overall survival of clear cell renal cell carcinoma patients

Stephanie G Kroeze, Joost S Vermaat, Aram van Brussel, Harm H van Melick, Emile E Voest, Trudy G Jonges, Paul J van Diest, John Hinrichs, J Ruud Bosch, Judith J Jans

Eur J Cancer. 2010 Dec;46(18):3375-82.


sTATe M e nT of TRAn s lATIonAl R e leVAnCe

The hypoxia inducible factor (HIF) pathway strongly stimulates the evolution of renal cell carcinoma (RCC). Using a tissue microarray, including primary RCC tumor tissue with clear cell histology of 100 RCC patients who underwent nephrectomy, this investigation demonstrates that factor inhibiting HIF (FIH) -an upstream target of the HIF pathway- significantly predicted overall survival (OS). Additionally, low nuclear expression of FIH is a strong independent prognostic factor for a poor OS. However other upstream targets as the prolyl hydroxylases domain proteins (PHD) 1, 2 and 3 were not correlated with survival. In conclusion, these results highlight the importance to study the upstream regulatory proteins of the HIF pathway besides its downstream targets. Furthermore, when incorporated with other clinical and pathological factors FIH expression levels might improve currently used prognostic models for disease staging of patients with non-metastasized RCC with clear cell histology.

29Nuclear FIH independently predicts overall survival of ccRCC patients Chapter 3

AB sTRACT

Background and objectiveThe hypoxia inducible factor (HIF) pathway plays an important role in sporadic clear cell renal cell carcinoma (ccRCC) by stimulating processes of angiogenesis, cell proliferation, cell survival and metastases formation. Here, we evaluate the significance of upstream proteins directly regulating the HIF pathway; the prolyl hydroxylases domain proteins (PHD)1, 2 and 3 and factor inhibiting HIF (FIH), as prognostic markers for ccRCC.

Patients and methodsImmunohistochemical marker expression was examined on a tissue microarray containing tumor tissue derived from 100 patients who underwent nephrectomy for ccRCC. Expression levels of HIF, FIH and PHD1, 2 and 3 were correlated with overall survival (OS) and clinicopathological prognostic factors.

ResultsHIF-1α was positively correlated with HIF-2α (p<0.0001), PHD1 (p=0.024), PHD2 (p<0.0001), PHD3 (p=0.004), FIH (p<0.0001) and VHL (p=0.031). HIF-2α levels were significantly associated with FIH (p<0.0001) and PHD2 (p=0.0155). Mutations in the VHL gene, expression variations of HIF-1α, HIF-2α and PHD1, 2, 3 did not show a correlation to OS or clinicopathological prognostic factors. Tumor stage, grade, diameter, metastastic disease and intensity of nuclear FIH were significantly correlated to OS in univariable analysis (p=0.023). Low nuclear FIH levels remained a strong independent prognostic factor in multivariable analysis (p=0.009).

ConclusionThese results show that low nuclear expression of FIH is a strong independent prognostic factor for a poor overall survival in ccRCC.


AB B R eVIATIon s

(cc)RCC; (clear cell) Renal Cell Carcinoma, mRCC; metastatic Renal Cell Carcinoma, HIF; hypoxia inducible factor, PHD; prolyl hydroxylases domain protein, FIH; factor inhibiting HIF, VHL; von Hippel-Lindau, OS; Overall Survival, ECOG; Eastern Cooperative Oncology Group Performance Status, FFPE; formalin-fixed paraffin embedded, TMA; tissue micro array, TNM stage; tumor-node-metastasis stage I nTRoDuCTIon

With a worldwide incidence of approximately 200,000 new cases and a mortality of 102,000 patients each year renal cell carcinoma (RCC) is one of the most lethal genitourinary malignancies (1). Of all RCC patients approximately 30% will develop local or distant recurrences within 5 years after initial curative treatment (2). Observational follow-up, succeeded by targeted therapy once metastases are present, remains the postoperative standard of care for these patients. Even with first line therapy, 5 year survival of metastasized RCC (mRCC) is less than 20% (3). Improvements in identification of patients at risk of developing metastatic disease are therefore important. Currently available prognostic models for non-metastasized disease, such as the University of California Integrated Staging System (UISS) and Stage, Size, Grade and Necrosis (SSIGN) score, are mainly based on clinical and pathologic variables (4, 5). The most important conventional features are tumor-node-metastasis (TNM) stage, Fuhrman grade (6) and Eastern Cooperative Oncology Group performance status (ECOG PS). Risk stratification with these features is possible, although the described accuracies vary (7-9). In addition to clinical and pathologic prognostic parameters, tissue biomarkers can play a significant role in predicting prognosis. For example in breast cancer, tissue marker detection for gene expression profiles indicating prognosis and treatment response has become common practice (10). Since the understanding of molecular pathways underlying RCC has increased dramatically in recent years, the search for prognostic tissue biomarkers has been facilitated. In clear cell RCC (ccRCC), the molecular hypoxia response pathway, in which hypoxia-inducible factors-α (HIF-1α and HIF-2α) play a central role, is crucial (11) (figure 1). HIF-α stimulates tumor growth and survival, angiogenesis, metastatic spread and glucose metabolism, amongst others. HIF-α levels are dependent of cellular oxygen levels and presence of the von Hippel-Lindau tumor suppressor protein (pVHL). In normoxia, the prolyl hydroxylase domain proteins (PHD1, PHD2 and PHD3) stimulate HIF-α degradation by enabling its recognition by pVHL by hydroxylation. Factor inhibiting HIF (FIH) adds a further level of control by reducing the transcriptional activity of HIF-α (12). Approximately 50-70% of sporadic ccRCC cases have VHL gene mutations (13). The subsequent absence of functional pVHL causes an overexpression of HIF-α (both HIF-1α and HIF-2α) independent of oxygen concentrations. Expression of HIF-α and various downstream proteins of the HIF pathway in RCC have been studied for their prognostic value (14, 15). While of great importance for HIF-α regulation, proteins upstream in the HIF pathway have been less extensively investigated in RCC, despite recent reports on associations of levels of these proteins with aggressiveness of other tumor types (16-18). The aim of this study was to determine whether the upstream factors of the HIF pathway PHD1, 2 and 3 and FIH have independent prognostic value as tissue biomarker for ccRCC.

31

PATI e nTs An D M eTHoDs

Patients This retrospective study included patients who underwent nephrectomy for ccRCC between 1994 and 2006 at the University Medical Center Utrecht (UMCU). The study was carried out in accordance with the ethical guidelines of our institution concerning informed consent about the use of patient’s materials after surgical procedures. Patients with von Hippel-Lindau’s disease, tuberous sclerosis, Wilms’ tumor or RCC subtypes other than ccRCC were excluded. Hereafter, a random ccRCC population of 100 patients was included of which the clinicopathologic characteristics are summarized in Table 1. Patient clinical, pathologic and survival data were obtained by reviewing hospital records and by information from the general practitioner, and

figure 1 - The hypoxia pathway

Nuclear FIH independently predicts overall survival of ccRCC patients Chapter 3

figure 1 Under normoxic conditions the prolyl hydroxylase domain proteins (PHD1, PHD2 and PHD3) stimulate HIF-a degradation by enabling its recognition by the von Hippel–Lindau protein (pVHL) through hydroxylation. Factor-inhibiting HIF (FIH) adds a further level of control by reducing the transcriptional activity of HIF-a, until severe hypoxia occurs. In the pre-sence of hypoxia or VHLmutations, HIF-a accumulates and mediates the transcription of factors stimulating tumour growth and survival, angiogenesis and glucose metabolism, amongst others. Ub: Ubiquitin, HRE: hypoxia response element.

cytoplasm PHD-1 HIF-α

HIF-α

HIF-α

HIF-αHIF-ß

HIF-ß

P300

HIF-α

HIF-α

HIF-α

HIF-α

VHL

VHL

VHL

HIF-α

PHD-2

PHD-3

FIH

O2

O2

O2

OH

VHL mutation

proteolytic degradation

Angiogenesis (e.g.VEGF

Glucose metabolism (e.g. Glut-1/CA9)

HIF mediated transcriptional activity

accumulation

nucleus

Erythropoiesis (e.g. EPO)

OH

OH

OH

OH

OH

OH

OHUb

Ub

Ub

HRE

HIF-α

HIF-α

HYPoXIAnoRMoXIA


Table 1

Features Median n (%)(Inter-Quartile Range)

No. of patients 100

Age (years) 64 (54 - 73)

Gender Female 32 (32)Male 68 (68)

Overall Survival (Months) 55 (22 - 93)Tumor Classification

pT1a 18 (18)pT1b 14 (14)

pT2 12 (12)pT3a 17 (17)

pT3b/c 35 (35)pT4 3 (3)

Fuhrman nuclear gradeTotal 88 (88)

1 12 (12)2 45 (45)3 28 (28)4 3 (3)

Tumor size 7.5 (4.5 - 10)< 5 cm 26 (26)≥ 5 cm 74 (74)

Lymph node involvementpNx + pN0 90 (90)pN1 + pN2 10 (10)

MetastasesSynchronous 17 (17)

Metachronous 18 (18)Tumor necrosis

Yes 10 (10)No 90 (90)

SSIGN ScoreTotal 880 - 1 19 (22)

2 13 (15)3 3 (3)4 14 (16)5 13 (15)6 6 (7)7 7 (8)8 5 (6)9 3 (3)

≥ 10 5 (6)VHL mutation

Sequencing Completed 37Yes 16 (43)

Missense 4 (11)Deletion 11 (30)

Indel 1 (3) No 21 (57)

33

included age, sex, 2002 tumor-node-metastasis (TNM) classification, Fuhrman grade, ECOG performance data, primary tumor size, presence of tumor necrosis and SSIGN score. Tumor stage and grade were determined by one pathologist (TGNJ). Patients were evaluated from time of diagnosis to 10 year follow-up. OS was defined as time from nephrectomy till date of death or last clinical follow-up.

Tissue microarray constructionFormalin-fixed paraffin embedded (FFPE) renal tumor material was obtained from the Biobank of the UMCU after approval of the UMCU Institutional Review Board in accordance with Dutch medical ethical guidelines. A tissue microarray (TMA) was constructed by taking 3 cores (1 mm diameter) from each of the 100 cancer specimens and 3 cores from benign renal parenchyma surrounding the tumor and arranging them in a new composite paraffin block using an arrayer (Beecher instruments, Sun Prairy, USA).

ImmunohistochemistryTMA sections (5 µm) were deparaffinised and rehydrated. Immunohistochemistry procedures included antigen retrieval using citrate buffer (pH 6) (FIH, VHL, PHD1 and PHD3) or ethylenediaminetetraacetic acid buffer (pH 9) (PHD2, HIF-1α and HIF-2α). The primary antibodies and their respective dilutions were as follows: VHL (Clone Ig32, BD Biosciences, Temse, Belgium, 1:100), PHD1 (Abcam, Cambridge, UK, 1:100), PHD2 (Abcam, Cambridge, UK, 1:100), PHD3 (Novus Biologicals, Cambridge, UK, 1:400), HIF-1α (BD Biosciences, Temse, Belgium, 1:50), HIF-2α (Abcam, Cambridge, UK, 1:500) and FIH (Novus Biologicals, Cambridge, UK, 1:100). PowerVision (Immunologic, Duiven, the Netherlands) was used as secondary antibody. All reactions were visualized using diaminobenzidine/H2O2. Finally, array sections were counterstained with haematoxylin.

scoring methodsProtein expression was analyzed by a pathologist (PJvD) blinded to other data. Cytoplasmic staining (PHD1, 2, 3, VHL), percentage of positively stained nuclei (PHD1, 2, 3, FIH, HIF-1α, HIF-2α) and nuclear intensity (FIH), were assessed in a semiquantitative fashion. For cytoplasmic and nuclear staining tumors were scored as 0, no intensity; 1, weak; 2; moderate; 3, strong intensity. Percentages of positively stained nuclei were scored as 0, <15%; 1, 15-75%; 2, >75%. The mean score of three cores of a tumor was used for statistical analysis.

DnA extraction and amplificationTumor tissue was separated from normal kidney on paraffin tissue sections by scratching, guided by H&E staining of adjacent paraffin sections, lysed in Tris/HCl buffer, pH 8.0 with Tween-40 mg/ml proteinase K (2mg/ml) at 56oC o/n. The lysate was boiled for 10’ and subsequently cooled down on ice. After 2’ centrifugation at 14680 rpm, the DNA concentration of the supernatant was measured using the Isogen Nanodrop Spectrophotometer ND-1000. The DNA was subsequently stored at -20oC. Amplification of patient tumor DNA was carried out in 25 ml reactions using 1 mg of tumor DNA. Primers used for amplification are listed in Table 2, exon 1 was amplified in two parts.



DnA purification and VHl gene sequencingResidual primers and single-stranded DNA were removed from PCR products using exonuclease I and Shrimp Alkaline Phosphatase at 37oC for 30 minutes followed by 20’ termination at 80oC. Sequencing was carried using the Big Dye system and the primers listed in Table 2. Sequencing runs were performed with the Gene-amp PCR system 9600 (Applied biosystems). Sequence reaction products were purified using sephadex columns prior to running them on a 3130x/genetic analyzer (Applied Biosystems, Foster City, USA). Results were analyzed using Mutation Surveyor software (SoftGenetics, LLC., State College, PA, USA v3.24).

Table 2

PCR Primers Sequence

Exon 1A Forward 5’ GGT-GGT-CTG-GAT-CGC-GGA-GGG-A 3’

Exon 1A Reverse 5’ CGC-GAG-TTC-ACC-GAG-CGC-AGC-A 3’

Exon 1B Forward 5’ AAC-TGG-GCG-CCG-AGG-AGG-AGA-T 3’

Exon 1B Reverse 5’ GGG-CTT-CAG-ACC-GTG-CTA-TCG 3’

Exon 2 Forward 5’ TTC-ACC-ACG-TTA-GCC-AGG-AC 3’

Exon 2 Reverse 5’ GGT-CTA-TCC-TGT-ACT-TAC-CA 3’

Exon 3 Forward 5’ AGC-CTC-TTG-TTC-GTT-CCT 3’

Exon 3 Reverse 5’ GGA-ACC-AGT-CCT-GTA-TCT 3’

Sequence Primers Sequence

Exon 1A Forward 5’ GAT-CGC-GGA-GGG-AAT-GCC 3’

Exon 1A Reverse 5’ CCCG-CCC-GGC-CTC-CAT-CTC 3’

Exon 1B Forward 5’ GCC-GAG-GAG-GAG-ATG-GAG 3’

Exon 1B Reverse 5’ TTC-AGA-CCG-TGC-TAT-CGT 3’

Exon 2 Forward 5’ CAG-GAC-GGT-CTT-GAT-CTC 3’

Exon 2 Reverse 5’ TTA-CCA-CAA-CAA-CCT-TAT-CT 3’

Exon 3 Forward 5’ TGT-TCG-TTC-CTT-GTA-CTG 3’

Exon 3 Reverse 5’ACT-CAT-CAG-TAC-CAT-CAA-AA 3’

statistical analysisTMA marker expression was correlated to clinicopathological markers using Spearman rank correlation for non-parametric measures. Marker expression was correlated to overall survival (OS) using Kaplan-Meier survival curves followed by log rank analysis to estimate differences in levels of the analyzed variables. Marker independence for prediction of survival was determined by univariable- followed by multivariable Cox regression analysis.

35

R e s u lTs

PatientsComprehensive clinicopathological features of 100 ccRCC patients are depicted in Table 1. Median OS was 55 months (IQR 22-93). At the time of final analysis, 25 patients died because of their disease and 41 patients were still alive, having a median survival time of 82 months (IQR 53-100). 17 patients presented metastases at time of nephrectomy, 18 patients developed metastases within follow-up. Of patients with metastasized disease 7 were treated with immunotherapy, 2 with a combination of immunotherapy and metastasectomy and 4 with metastasectomy. 13 patients received adjuvant radiotherapy for their bone metastases.

sequencing of the VHl geneQuality of the DNA derived from formalin fixed paraffin embedded tissue was sufficient for sequencing of the VHL gene in 37 patients (Table 1). In total, 16 mutations (43%) were identified of which 1 was located in an intron. Mutations were equally distributed along the gene. Most mutations were single or double-basepair deletions (11x) or point mutations (4x), one insertion-deletion was detected. Twelve mutations resulted in frameshifts. The presence or absence of a mutation in the VHL gene did not correlate with clinicopathological parameters, survival of the patients or expression of any of the markers under investigation in this study. expression of HIfMost tumors showed nuclear presence of HIF-1α (94%) and HIF-2α (82%). Median percentages of positive nuclei within a tumor were 27.5% (IQR 6.7-62.5%) for HIF-1α and 2.7% (IQR 0.5-13.5%) for HIF-2α. Expression of HIF-1α and HIF-2α was not associated with any of the clinicopathological parameters considered in this study. HIF-1α was positively correlated with HIF-2α (p<0.0001), PHD1 (p=0.024), PHD2 (p<0.0001), PHD3 (p=0.004), FIH (p<0.0001) and VHL (p=0.031) (Spearman rank correlation). HIF-2α levels were significantly associated with FIH (p<0.0001) and PHD2 (p=0.0155).

expression of prolyl hydroxylasesPHD1, 2 and - were localized to the nuclei of RCC cells. PHD1 was detected in on average 39% of all nuclei, PHD2 in 63% and PHD3 in 84%. Although no association between levels of PHD proteins and OS of patients was identified, a significant correlation (PHD1 p=0.024, PHD2 p=0.0067, PHD3 p=0.0012, Kruskal-Wallis) between Fuhrman grade and expression of all three PHD proteins was established (figure 2). The percentage of nuclei staining positive for PHD 1, 2 and 3 was high in tumors with Fuhrman grade 1.

expression of fIH is associated with overall survivalIn normal kidney tissue, FIH was primarily localized to the cytoplasm. In ccRCC cells, however, FIH staining was predominantly nuclear. The percentage of RCC nuclei with positive FIH staining ranged widely from 1 to 100%. In addition to the percentage of positive nuclei, the intensity of FIH staining varied between patients (figure 3). Therefore, both the percentage of positive cells and the intensity of FIH staining (score 1, 2 or 3) were recorded. A low (<15%) percentage of FIH-positive nuclei and a low intensity of FIH staining (score 1)



figure 2 The percentage of PHD1 (A), PHD2 (B) and PHD3 (C) positive nuclei is highest in lesions with Fuhrman grade 1. Fuhrman grade 1: 12 patients, Fuhrman grade 2: 45 patients, Fuhrman grade 3&4: 31 patients.

figure 3 (A) Expression of FIH in normal kidney, FIH is predominantly present in the cytoplasm of tubuli. (B) Low expression of FIH in ccRCC. (C) High nuclear expression of FIH in ccRCC.

figure 2 - PHD1, 2 and 3 expression levels by Fuhrman grade

figure 3 - FIH immunohistochemistry

figure 4 Kaplan–Meier curves showing the overall survival of ccRCC patients with low versus high percentages of nuclear FIH (A) and low versus high intensity of nuclear FIH staining (B).

figure 4 - Overall survival

37

Table 3

univariate AnalysisVariable HR (95% Cl) Significance

Tumor ClassificationT1 1.00

T2 1.28 (0.46 - 3.55) 0.631

T3 3.13 (1.51 - 6.48) 0.002

T4 5.11 (0.64 - 40.7) 0.123

Fuhrman Nuclear Grade1 1.00

2 2.25 (0.77 - 6.59) 0.138

3 2.45 (0.79 - 7.55) 0.120

4 12.01 (2.11 - 68.42) 0.005

Tumor Diameter 1.08 (1.01 - 1.15) 0.026

Distant Metastases 4.09 (2.14 - 7.83) 0.000

Lymph Node Involvement 1.17 (0.64 - 2.15) 0.616

Necrosis 1.14 (0.47 - 2.73) 0.772

FIH (% positive nuclei) 0.99 (0.98 - 1.02) 0.141

FIH (Intensity) 0.45 (0.23 - 0.90) 0.023

Multivariate Cox Regression AnalysisVariable HR (95% CI) Significance

Tumor ClassificationT1 1.00

T2 0.43 (0.05 - 3.58) 0.437

T3 2.08 (0.82 - 5.26) 0.121

T4 7.60 (1.82 - 31.64) 0.005

Distant Metastases 4.45 0.000

FIH (Intensity) 0.40 0.009

were both significantly correlated with a shorter OS of patients (p=0.001 and p=0.003 log rank) (figure 4). FIH was positively associated with tumor diameter (p=0.016, Spearman rank) but not with any of the other clinicopathological parameters including stage and grade. We evaluated the following variables for prognostic value: tumor stage, tumor grade, tumor diameter, presence of metastatic disease, lymph node involvement, presence of necrosis and expression of FIH (Table 3). Of these, stage, grade, diameter, metastatic disease and intensity of FIH showed a statistically significant (p=0.023) correlation with OS in univariable analysis. Inclusion of these factors in a multivariable backwards conditional Cox regression analysis resulted in a model with metastatic disease, stage and FIH intensity as independent predictors for OS (p=0.009) (Table 3).



DI sCus s Ion

In this study nuclear FIH expression in the primary tumor was shown to have a significant independent prognostic value for ccRCC patients. FIH inhibits HIF-α in an oxygen-dependent manner. However, FIH remains active unless severe hypoxia occurs (19). This suggests that FIH may have an important function as one of the final checks on HIF-α transcriptional activity. Although HIF is not the only target for hydroxylation by FIH, it is the most intensively studied and best characterised to date. The presence of FIH has been investigated in various normal and neoplastic human tissues, in which the intensity and subcellular localisation is very heterogeneous. Although in normal human tissues FIH is predominantly cytoplasmic, nuclear expression of FIH can be relatively strong in certain neoplasms (20). Moreover, FIH has provided variable prognostic values in several tumor types (16, 21). In pancreatic endocrine tumors (PETs) cytoplasmic FIH levels were significantly higher in more malignant PETs, but were not associated with survival. Nuclear FIH did not correlate with any histopathologic variables in this study (16). In invasive breast cancer, both cytoplasmic FIH expression and absence of nuclear FIH were independent prognostic factors for a shorter disease-free survival (21). Our study showed for the first time in ccRCC patients that low expression of nuclear FIH is a significant independent predictor for worse OS. In RCC, FIH can be detected both in the nucleus and cytoplasm, and the specific subcellular lozalisation varies between different RCC patients (20). This was similar in our study (Fig. 3). The mechanism and function of nuclear FIH in RCC, and the reason why low nuclear FIH levels have such strong prognostic value is currently unknown. The absence of nuclear FIH in more aggressive phenotypes could be explained by increasing gene mutations within these tumors, including FIH gene mutations. This is, however, unlikely since FIH gene mutations have not been found in RCC (22). It is therefore possible that FIH is actively retained in the cytoplasm or exported out of the nucleus in tumors associated with worse prognosis. Although VHL gene mutations are considered an early and important event in the development of ccRCC, studies that investigated VHL alterations and survival have provided contradictory results (13, 23-26). Our data suggests that VHL status is not correlated to common clinical prognostic parameters or survival of ccRCC patients. Both HIF-1α and HIF-2α, which are central proteins of the HIF pathway, failed to show a correlation to survival and the most important clinicopathological parameters in our study. In ccRCC, both HIF-1α and HIF-2α are important regulators of angiogenesis and cell proliferation (27, 28). HIF-1α, but not HIF-2α, has prognostic value in colorectal cancer, lymph node negative breast carcinoma and ovarian carcinoma (29-31). In ccRCC HIF-1α did correlate with good prognosis in a study using Western blot, however, immunohistochemical staining of the same population could not confirm this finding (32, 33). Another study reported prognostic value for HIF-1α, but only in metastatic RCC patients (34). HIF-2α has been related to worse survival in several cancer types (35-37). In ccRCC, evidence has accumulated revealing the importance of HIF-2α in tumorigenesis (28, 38, 39). However, although HIF-2α has shown to be inversely correlated to TNM and nuclear grade in RCC patients, we and others have shown no evidence for a correlation of HIF-2α expression with survival (40). HIF-1α and HIF-2α levels did correlate positively to expression of PHD1, 2 and 3 and FIH. Under hypoxic conditions both HIF-1α and HIF-2α upregulate PHD protein expression. This may serve as a negative feedback to decrease HIF-α activity (41, 42). Since hypoxia is common in RCC and many other solid tumors (43), a correlation between molecules participating in the hypoxia response pathway is not surprising and indicates the intricate interplay between these proteins. Alternatively, FIH may exert

39

its effects on tumor biology in a HIF-independent fashion. In addition to the interaction of FIH with the hypoxia pathway, it was recently discovered that in the presence of oxygen HIF competes with many ankyrin repeat domain-containing proteins (ARDs) for hydroxylation by FIH. Many ARDs have a proposed function in cell proliferation and differentiation. Due to this competition hypoxia or a decreased presence of FIH may lead to accumulation of non-hydroxylated ARDs. Although the exact mechanisms have not yet been elucidated, downstream effects of this accumulation may lead to increased tumor aggressiveness (44, 45). PHD proteins have been studied in several other tumor types as potential prognostic markers (16, 17). In these studies high nuclear PHD levels were associated with tumor aggressiveness and worse survival. For ccRCC, as far as we know no information was present for PHD expression in relation to patient survival. We show that there is no correlation between PHD1, 2, 3 and RCC patient survival. Surprisingly, in this study high nuclear PHD levels were significantly correlated to a low Fuhrman grade, indicating that in contrast to other tumor types, increased nuclear PHD levels are primarily present in less aggressive (grade 1) ccRCC. Here we show that FIH is a very promising prognostic marker for ccRCC. However, for its definitive clinical value a prospective clinical trial will be needed. Like most described prognostic markers in literature this study was performed on tissue obtained in the time that immunotherapy was first line therapy for RCC. In the current era of targeted therapy correlation of these prognostic tissue markers to survival is therefore not completely certain, especially when markers have a function in tumor specific immunology. However, since only a minority of our patient population underwent immunotherapy and the here described markers are not known to have a function in the immune system, it is highly unlikely that the prognostic significance will be altered under current treatment standards. Furthermore, prospective clinical trials are needed to determine the additional clinical significance and application of FIH to the currently available prognostic models for non-metastasized disease.

In conclusion, this study shows that nuclear FIH is an important indicator of prognosis in ccRCC patients, with low nuclear FIH expression predicting a poor OS. Unlike previous studies focussing on HIF and its downstream targets, the results presented here highlight the importance of upstream regulatory proteins of the HIF pathway as independent prognostic factors for ccRCC. We anticipate that integration of FIH in established prognostic models could result in a more accurate prediction of survival.


40 Part I Molecular prognosticators

R e f e R e nCe s

(1) Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin. 2005

Mar-Apr;55(2):74-108.

(2) Zisman A, Pantuck AJ, Wieder J, Chao DH, Dorey F, Said JW, et al. Risk group assessment and clinical

outcome algorithm to predict the natural history of patients with surgically resected renal cell carcinoma.

J Clin Oncol. 2002 Dec 1;20(23):4559-66.

(3) Motzer RJ, Hutson TE, Tomczak P, Michaelson MD, Bukowski RM, Oudard S, et al. Overall survival and updated

results for sunitinib compared with interferon alfa in patients with metastatic renal cell carcinoma. J Clin Oncol.

2009 Aug 1;27(22):3584-90.

(4) Zisman A, Pantuck AJ, Dorey F, Said JW, Shvarts O, Quintana D, et al. Improved prognostication of renal cell

carcinoma using an integrated staging system. J Clin Oncol. 2001 Mar 15;19(6):1649-57.

(5) Frank I, Blute ML, Cheville JC, Lohse CM, Weaver AL, Zincke H. An outcome prediction model for patients with

clear cell renal cell carcinoma treated with radical nephrectomy based on tumor stage, size, grade and necrosis:

the SSIGN score. J Urol. 2002 Dec;168(6):2395-400.

(6) Fuhrman SA, Lasky LC, Limas C. Prognostic significance of morphologic parameters in renal cell carcinoma.

Am J Surg Pathol. 1982 Oct;6(7):655-63.

(7) Ficarra V, Novara G, Galfano A, Brunelli M, Cavalleri S, Martignoni G, et al. The ‘Stage, Size, Grade and

Necrosis’ score is more accurate than the University of California Los Angeles Integrated Staging System for

predicting cancer-specific survival in patients with clear cell renal cell carcinoma. BJU Int. 2009 Jan;103(2):

165-70.

(8) Patard JJ, Kim HL, Lam JS, Dorey FJ, Pantuck AJ, Zisman A, et al. Use of the University of California Los Angeles

integrated staging system to predict survival in renal cell carcinoma: an international multicenter study.

J Clin Oncol. 2004 Aug 15;22(16):3316-22.

(9) Zigeuner R, Hutterer G, Chromecki T, Imamovic A, Kampel-Kettner K, Rehak P, et al. External Validation of the

Mayo Clinic Stage, Size, Grade, and Necrosis (SSIGN) Score for Clear-Cell Renal Cell Carcinoma in a Single

European Centre Applying Routine Pathology. Eur Urol. Jan;57(1):102-11.

(10) Wolff AC, Hammond ME, Schwartz JN, Hagerty KL, Allred DC, Cote RJ, et al. American Society of Clinical

Oncology/College of American Pathologists guideline recommendations for human epidermal growth factor

receptor 2 testing in breast cancer. J Clin Oncol. 2007 Jan 1;25(1):118-45.

(11) Kaelin WG, Jr. The von Hippel-Lindau tumour suppressor protein: O2 sensing and cancer. Nat Rev Cancer.

2008 Nov;8(11):865-73.

(12) Schofield CJ, Ratcliffe PJ. Oxygen sensing by HIF hydroxylases. Nat Rev Mol Cell Biol. 2004 May;5(5):343-54.

(13) Kim WY, Kaelin WG. Role of VHL gene mutation in human cancer. J Clin Oncol. 2004 Dec 15;22(24):

4991-5004.

(14) Bui MH, Seligson D, Han KR, Pantuck AJ, Dorey FJ, Huang Y, et al. Carbonic anhydrase IX is an independent

predictor of survival in advanced renal clear cell carcinoma: implications for prognosis and therapy.

Clin Cancer Res. 2003 Feb;9(2):802-11.

(15) Klatte T, Seligson DB, LaRochelle J, Shuch B, Said JW, Riggs SB, et al. Molecular signatures of localized clear

cell renal cell carcinoma to predict disease-free survival after nephrectomy. Cancer Epidemiol Biomarkers Prev.

2009 Mar;18(3):894-900.

(16) Couvelard A, Deschamps L, Rebours V, Sauvanet A, Gatter K, Pezzella F, et al. Overexpression of the oxygen

sensors PHD-1, PHD-2, PHD-3, and FIH Is associated with tumor aggressiveness in pancreatic endocrine

tumors. Clin Cancer Res. 2008 Oct 15;14(20):6634-9.

41

(17) Jokilehto T, Rantanen K, Luukkaa M, Heikkinen P, Grenman R, Minn H, et al. Overexpression and nuclear

translocation of hypoxia-inducible factor prolyl hydroxylase PHD2 in head and neck squamous cell carcinoma

is associated with tumor aggressiveness. Clin Cancer Res. 2006 Feb 15;12(4):1080-7.

(18) Jans J, van Dijk JH, van Schelven S, van der Groep P, Willems SH, Jonges TN, et al. Expression and Localization

of Hypoxia Proteins in Prostate Cancer: Prognostic Implications After Radical Prostatectomy. Urology.

2009 Oct 23.

(19) Stolze IP, Tian YM, Appelhoff RJ, Turley H, Wykoff CC, Gleadle JM, et al. Genetic analysis of the role of the

asparaginyl hydroxylase factor inhibiting hypoxia-inducible factor (HIF) in regulating HIF transcriptional

target genes. J Biol Chem. 2004 Oct 8;279(41):42719-25.

(20) Soilleux EJ, Turley H, Tian YM, Pugh CW, Gatter KC, Harris AL. Use of novel monoclonal antibodies to determine

the expression and distribution of the hypoxia regulatory factors PHD-1, PHD-2, PHD-3 and FIH in normal and

neoplastic human tissues. Histopathology. 2005 Dec;47(6):602-10.

(21) Tan EY, Campo L, Han C, Turley H, Pezzella F, Gatter KC, et al. Cytoplasmic location of factor-inhibiting

hypoxia-inducible factor is associated with an enhanced hypoxic response and a shorter survival in invasive

breast cancer. Breast Cancer Res. 2007;9(6):R89.

(22) Morris MR, Maina E, Morgan NV, Gentle D, Astuti D, Moch H, et al. Molecular genetic analysis of FIH-1, FH,

and SDHB candidate tumour suppressor genes in renal cell carcinoma. J Clin Pathol. 2004 Jul;57(7):706-11.

(23) Parker AS, Cheville JC, Lohse CM, Igel T, Leibovich BC, Blute ML. Loss of expression of von Hippel-Lindau

tumor suppressor protein associated with improved survival in patients with early-stage clear cell renal cell

carcinoma. Urology. 2005 Jun;65(6):1090-5.

(24) Patard JJ, Rioux-Leclercq N, Masson D, Zerrouki S, Jouan F, Collet N, et al. Absence of VHL gene alteration

and high VEGF expression are associated with tumour aggressiveness and poor survival of renal-cell carcinoma.

Br J Cancer. 2009 Oct 20;101(8):1417-24.

(25) Schraml P, Hergovich A, Hatz F, Amin MB, Lim SD, Krek W, et al. Relevance of nuclear and cytoplasmic von

hippel lindau protein expression for renal carcinoma progression. Am J Pathol. 2003 Sep;163(3):1013-20.

(26) Smits KM, Schouten LJ, van Dijk BA, Hulsbergen-van de Kaa CA, Wouters KA, Oosterwijk E, et al. Genetic and

epigenetic alterations in the von hippel-lindau gene: the influence on renal cancer prognosis. Clin Cancer Res.

2008 Feb 1;14(3):782-7.

(27) Qing G, Simon MC. Hypoxia inducible factor-2alpha: a critical mediator of aggressive tumor phenotypes. Curr

Opin Genet Dev. 2009 Feb;19(1):60-6.

(28) Shinojima T, Oya M, Takayanagi A, Mizuno R, Shimizu N, Murai M. Renal cancer cells lacking hypoxia inducible

factor (HIF)-1alpha expression maintain vascular endothelial growth factor expression through HIF-2alpha.

Carcinogenesis. 2007 Mar;28(3):529-36.

(29) Bos R, van der Groep P, Greijer AE, Shvarts A, Meijer S, Pinedo HM, et al. Levels of hypoxia-inducible factor-

1alpha independently predict prognosis in patients with lymph node negative breast carcinoma. Cancer. 2003

Mar 15;97(6):1573-81.

(30) Osada R, Horiuchi A, Kikuchi N, Yoshida J, Hayashi A, Ota M, et al. Expression of hypoxia-inducible factor

1alpha, hypoxia-inducible factor 2alpha, and von Hippel-Lindau protein in epithelial ovarian neoplasms and allelic

loss of von Hippel-Lindau gene: nuclear expression of hypoxia-inducible factor 1alpha is an independent

prognostic factor in ovarian carcinoma. Hum Pathol. 2007 Sep;38(9):1310-20.

(31) Rasheed S, Harris AL, Tekkis PP, Turley H, Silver A, McDonald PJ, et al. Hypoxia-inducible factor-1alpha and

-2alpha are expressed in most rectal cancers but only hypoxia-inducible factor-1alpha is associated with

prognosis. Br J Cancer. 2009 May 19;100(10):1666-73.

(32) Lidgren A, Hedberg Y, Grankvist K, Rasmuson T, Bergh A, Ljungberg B. Hypoxia-inducible factor 1alpha

expression in renal cell carcinoma analyzed by tissue microarray. Eur Urol. 2006 Dec;50(6):1272-7.


42

(33) Lidgren A, Hedberg Y, Grankvist K, Rasmuson T, Vasko J, Ljungberg B. The expression of hypoxia-inducible

factor 1alpha is a favorable independent prognostic factor in renal cell carcinoma. Clin Cancer Res. 2005

Feb 1;11(3):1129-35.

(34) Klatte T, Seligson DB, Riggs SB, Leppert JT, Berkman MK, Kleid MD, et al. Hypoxia-inducible factor 1 alpha in

clear cell renal cell carcinoma. Clin Cancer Res. 2007 Dec 15;13(24):7388-93.

(35) Giatromanolaki A, Koukourakis MI, Sivridis E, Turley H, Talks K, Pezzella F, et al. Relation of hypoxia inducible

factor 1 alpha and 2 alpha in operable non-small cell lung cancer to angiogenic/molecular profile of tumours and

survival. Br J Cancer. 2001 Sep 14;85(6):881-90.

(36) Griffiths EA, Pritchard SA, McGrath SM, Valentine HR, Price PM, Welch IM, et al. Hypoxia-associated markers

in gastric carcinogenesis and HIF-2alpha in gastric and gastro-oesophageal cancer prognosis. Br J Cancer.

2008 Mar 11;98(5):965-73.

(37) Holmquist-Mengelbier L, Fredlund E, Lofstedt T, Noguera R, Navarro S, Nilsson H, et al. Recruitment of HIF-

1alpha and HIF-2alpha to common target genes is differentially regulated in neuroblastoma: HIF-2alpha

promotes an aggressive phenotype. Cancer Cell. 2006 Nov;10(5):413-23.

(38) Carroll VA, Ashcroft M. Role of hypoxia-inducible factor (HIF)-1alpha versus HIF-2alpha in the regulation of HIF

target genes in response to hypoxia, insulin-like growth factor-I, or loss of von Hippel-Lindau function: implications

for targeting the HIF pathway. Cancer Res. 2006 Jun 15;66(12):6264-70.

(39) Kondo K, Kim WY, Lechpammer M, Kaelin WG, Jr. Inhibition of HIF2alpha is sufficient to suppress pVHL-

defective tumor growth. PLoS Biol. 2003 Dec;1(3):E83.

(40) Sandlund J, Ljungberg B, Wikstrom P, Grankvist K, Lindh G, Rasmuson T. Hypoxia-inducible factor-2alpha

mRNA expression in human renal cell carcinoma. Acta Oncol. 2009;48(6):909-14.

(41) Aprelikova O, Chandramouli GV, Wood M, Vasselli JR, Riss J, Maranchie JK, et al. Regulation of HIF prolyl

hydroxylases by hypoxia-inducible factors. J Cell Biochem. 2004 Jun 1;92(3):491-501.

(42) Henze AT, Riedel J, Diem T, Wenner J, Flamme I, Pouyseggur J, et al. Prolyl hydroxylases 2 and 3 act in gliomas

as protective negative feedback regulators of hypoxia-inducible factors. Cancer Res. Jan 1;70(1):357-66.

(43) Kim Y, Lin Q, Glazer PM, Yun Z. Hypoxic tumor microenvironment and cancer cell differentiation. Curr Mol Med.

2009 May;9(4):425-34.

(44) Cockman ME, Webb JD, Ratcliffe PJ. FIH-dependent asparaginyl hydroxylation of ankyrin repeat domain-

containing proteins. Ann N Y Acad Sci. 2009 Oct;1177:9-18.

(45) Meteoglu I, Erdogdu IH, Meydan N, Erkus M, Barutca S. NF-KappaB expression correlates with apoptosis and

angiogenesis in clear cell renal cell carcinoma tissues. J Exp Clin Cancer Res. 2008;27:53.

Part I Molecular prognosticators

43Nuclear FIH independently predicts overall survival of ccRCC patients Chapter 3


Chapter 4 Regulation of e2f1 by the von Hippel-lindau tumor suppressor protein predicts survival in renal cell cancer patients

Joost S Vermaat*, Dorus A Mans*, Bart G Weijts, Ellen van Rooijen, Jeroen van Reeuwijk, Karsten Boldt, Laura GM Daenen, Petra van der Groep, Benjamin Rowland, Ronald Roepman, Emile E Voest, Paul J van Diest, Marianne C Verhaar, Alain de Bruin, and Rachel H Giles

* Authors contributed equally

Submitted



Biallelic mutations of the von Hippel-Lindau (VHL) gene are the most common cause of renal cell carcinoma (RCC). We demonstrate here that the VHL gene product (pVHL) inhibits E2F1 expression at both mRNA and protein level in zebrafish, human cells and patient kidney tissue. Furthermore, increased expression of E2F1 in RCC cells upon VHL inactivation resulted in a marked delay of cell cycle progression. High nuclear E2F1 expressions of primary RCC tumors were associated with a smaller tumor diameter and with a favorable American Joint Committee on Cancer, in a large cohort of 138 patients with primary RCC who underwent nephrectomy. Nuclear E2F1 expression significantly predicted disease-free survival (DFS) as well as for overall survival (OS), where low E2F1 was associated with poor survival. Accordingly, E2F1 expression was significantly correlated with p27 expression, indicating that increased expression of E2F1 in RCC (e.g. resulting from VHL inactivation) induces cell senescence via p27 and suggests that low expression of E2F1 in RCC is a novel risk factor for poor survival. In conclusion, this study demonstrates that E2F1 expression is inhibited by pVHL in vitro and in vivo, is independent of HIF1α, and affects cell cycle progression of RCC cells. Furthermore, E2F1 expression is a novel predictor for RCC patients’ survival and might be potentially useful for clinical management.

47Regulation of E2F1 by VHL tumor suppressor protein predicts survival in RCC patients Chapter 4

AB sTRACT

Biallelic mutations of the von Hippel-Lindau (VHL) gene are the most common cause of sporadic and inherited renal cell carcinoma (RCC). Loss of VHL has been reported to affect proliferation by deregulating cell cycle associated proteins. We report here that the VHL gene product (pVHL) inhibits E2F1 expression at both mRNA and protein level in zebrafish, human cells and patient kidney tissue. Increased expression of E2F1 in RCC cells upon VHL inactivation resulted in a marked delay of cell cycle progression. Increased expression of E2F1 is observed in RCC samples compared to matched normal kidney tissue from the same individual. RCCs from all eight von Hippel-Lindau patients in our cohort with known germline VHL mutations expressed significantly more E2F1 compared to sporadic RCC sections with either a clear-cell (cc) or non-cc histology. Analysis of 138 primary RCCs revealed that E2F1 is expressed significantly higher in tumors with a diameter ≤7 cm and with a favorable American Joint Committee on Cancer (AJCC) stage. Cox regression analysis showed significant prediction of E2F1 expression for disease-free survival (DFS) as well as for overall survival (OS). E2F1 expression in RCC significantly correlates with p27 expression, indicating that increased expression of E2F1 in RCC (e.g. resulting from VHL inactivation) induces cell senescence via p27 and suggests that low expression of E2F1 in RCC is a novel risk factor for poor survival. AB B R eVIATIon s

(cc)RCC; (clear cell) Renal Cell Carcinoma, mRCC; metastatic Renal Cell Carcinoma, AJCC stage; American Joint Committee on Cancer stage, VHL; von Hippel-Lindau, HIF; hypoxia inducible factor, MCMs; mini-chromosome maintenance proteins, OS; overall Survival, DFS; disease free survival, TAP; tandem affinity purification, MS; Mass Spectrometry, FFPE; formalin-fixed paraffin embedded, TMA; tissue micro array, IQR; Interquartile Range, CI; Confidence Interval,


I nTRoDuCTIon

Renal cell carcinoma (RCC) accounts for more than 57,000 newly diagnosed patients in the United States and 3% of all cancer-associated deaths (1). RCC consists of several histological subtypes, of which the clear-cell (cc) variant is the most predominant, accounting for up to 75% (2). Mutations inactivating the von Hippel-Lindau (VHL) tumor suppressor gene are present in the majority of sporadic ccRCCs (2) and are known to cause VHL disease (MIM608537), which is characterized by familial predisposition to ccRCCs, multiple cysts in pancreas and kidney, pheochromocytomas and hemangioblastomas (2). The pVHL protein encoded by VHL is an E3 ubiquitin ligase, and physically interacts with Elongin B/C, Cullin-2, and Rbx-1, to form a functional protein complex in (poly-)ubiquitinating targets, like prolyl hydroxylated hypoxia-inducible factor 1 and 2 alpha (HIF1a, HIF2a), thereby targeting these proteins for proteasomal degradation. Inactivation of VHL results in the stabilization of HIF1/2a, which leads to an increased transactivation of HIF1/2a downstream effectors involved in angiogenesis, cell cycle progression and cellular transformation, resembling a hypoxic gene expression signature (2). Besides being an active E3 ligase component, pVHL is involved in extracellular matrix assembly, primary cilia function, stability of peripheral microtubules and cell growth control (2). Many of the identified cell cycle components differentially expressed in VHL-deficient cells result from increased stability of HIF1/2a, however some do not, such as the increased expression of p27, which leads to senescence (2-4). The senescence checkpoint response of cells is known to be regulated by the transcription factor E2F1 (5). This protein is a key regulatory participant in cell cycle progression, and can either promote or suppress tumorigenesis. E2F1 is expressed in late G1-phase of the cell cycle and is required for the expression of downstream components important for G1/S transition including mini-chromosome maintenance proteins (MCMs), which are components of the DNA-replication initiation complex (6). DNA damage enhances the stability of E2F1, resulting in the transactivation of genes involved in DNA repair (e.g. BRCA1, RAD51) or apoptosis (e.g. TP73, APAF1) (5;7). Increased expression of E2F1 has been observed in a variety of cancers, however the correlation with prognosis varies. In breast, thyroid, lung and pancreatic cancer, high E2F1 expression predicts poor outcome, while in transitional cell bladder carcinomas low E2F1 expression correlates with a worse prognosis, suggesting a tumor suppressive role for E2F1 in this cancer type (8). In this study, we validate data from an established zebrafish model for VHL (9) to uncover a signature for E2F1 activity in VHL-deficient RCC cell lines, and primary tumor material from RCC patients. Conversely, re-introduction of wildtype and mutant-pVHL into VHL-deficient RCC cell lines results in a marked downregulation of E2F1 expression and activity in a dose-dependent and HIFa-independent manner. Cell cycle defects attributed to VHL-loss in RCC cells are comparable with ectopic overexpression of E2F1 in pVHL-proficient RCC cells. Finally, we performed retrospective immunohistochemistry on a large cohort of RCC patients to investigate whether expression of E2F1 primary kidney tumors is associated with clinical- and histopathological characteristics. Our data suggest that E2F1 regulation is a novel HIFa-independent function for pVHL whose expression in renal cell carcinoma can predict disease-free survival (DFS) and overall survival (OS) of patients with RCC.

49

PATI e nTs & M eTHoDs

ZebrafishExperiments were conducted in accordance with the Dutch guidelines for the care and use of laboratory animals, with the approval of the Animal Experimentation Committee (DEC) of the Royal Netherlands Academy of Arts and Sciences (KNAW). Mutant alleles vhlhu2117(p.Q23X) and vhlhu2081(p.C31X) have been described previously (10). Transheterozygote embryos (vhlhu2117/vhlhu2081) were used in experimental assays. Whole-mount in situ hybridization was performed for mcm2, mcm5 and vegfa as described earlier (10).

PlasmidsVHL expression constructs pVHL30 (pVHL isoform 1; UniprotID: P40337-1), pVHL19 (pVHL isoform 3; UniprotID: P40337-3), pVHL-ΔF76, pVHL-Y112D, pVHL-Δ95-123 and pVHL-F119S were cloned into the pBabe-puro plasmid containing a N-terminal HA-tag. Vsv-pVHL30 has been previously described(11). The Gateway Entry clone for pVHL isoform 2 (kindly provided by Dr. Nico Katsanis, Duke University, Durham, USA; UniprotID: P40337-2) was used for TAP-cloning using Gateway LR-clonase (Invitrogen). E2F1-HA was kindly provided by Dr. Gustavo Leone (Ohio State University, Columbus, USA).

Cell culture and transfections Cell lines RCC10 and RCC10 stably transfected with full-length pVHL30 (clone 90 (12)) were obtained from Dr. Patrick Maxwell (Imperial College, London, UK). RCC10 stably transfected with HA-pVHL30 (clone 3D10), HA-pVHL19, HA-pVHL-F119S, HA-pVHL-DF76 or pBabe-HA alone (Mock), and 786-0 stably transfected with either Vsv-pVHL30 or pcDNA3-Vsv alone (Mock) were generated by standard electroporation and puromycin (2 mg/ml) selection. Stable clones of 786-0 and RCC10 cells, E2f7/8 double knockout mouse embryonic fibroblasts (MEFs) and HEK293T cells were cultured in DMEM supplemented with antibiotics and 10% FCS. ON-TARGET SMARTpool siRNAs targeting HIF1a (13) and USP30 were purchased from Dharmacon and transfected using Dharmafect (Dharmacon) according to the manufacturer’s protocol. Target sequences of all siRNAs used are listed in supplementary Table 1. All siRNA transfections were performed for 48 hours. E2f7/8 double knockout MEFs were transfected using Superfect (Qiagen). HEK293T cells were transfected using polyethelenimine (PEI; Polysciences) or calcium phosphate precipitation. MG132 (5 mM; Sigma-Aldrich) was used for 16 hours to inhibit proteasomal degradation. RCC cells were incubated in a hypoxia workstation (Ruskinn, 1% O2) for 24 hours.

lentiviral transductionMISSION shRNAs targeting VHL or MISSION Non-targeting control shRNAs (Sigma, supplementary Table 3) were transfected to generate lentiviral particles in 293T packaging human embryonic kidney cells with a three-plasmid expression system and the pLKO.1-Puro containing MISSION shRNAs by calcium phosphate precipitation. Viral supernatants were collected 48 hours after transfection, filtered through 0.22-µm pore nitrocellulose filters, concentrated by ultracentrifugation at 50,000×g for 140 minutes at RT and stored at -80°C. RCC10 clone 3D10 cells (1x106 cells/10 cm dish) were infected for 8 hours with lentiviral supernatants containing 6 µg/ml of polybrene (Millipore). Cells were then recovered in complete fresh medium for an additional 48 hours.

Regulation of E2F1 by VHL tumor suppressor protein predicts survival in RCC patients Chapter 4


Cell cycle analysisCells blocked at the G1/S transition by two sequential 24-hour blocks in thymidine (2.5 mM; Sigma), before being released for the time indicated. Nocodazole (250 ng/ml; Sigma) was added after 6 hours of release to catch cells in mitosis. Cells were washed in PBS and fixed overnight in 70% ethanol at -20°C. Fixed cells were washed in 1% BSA/PBS, treated with RNase (1:100; Sigma-Aldrich) and stained with propidium iodide (1:100; Sigma-Aldrich). Cell cycle analysis was performed on the FACS Calibur (BD-Biosciences). At least 10,000 gated events were scored per experiment. All experiments were performed at least twice independently.

Western blottingWestern blots were incubated with anti-E2F1 (KH95, 1:500; Santa Cruz Biotechnology), anti-HIF1a (1:500; Novus Biologicals), anti-VHL (Ig32, 1:500; BD Pharmingen), anti-VHL (clone 1B3B11 (14); 1:3), anti-b-catenin (1:500; BD-Biosciences), anti-b-actin (1:10,000; Abcam), anti-γ-tubulin (ab11317, 1:1000; Abcam), and anti-HA (mouse hybridoma clone 12CA5 supernatant, 1:3).

RT-PCRTotal RNA was isolated using Trizol (Invitrogen). cDNA was made using 2 mg of total RNA and reverse transcribed using oligo dT primer (Invitrogen) and Superscript RT II enzyme (Invitrogen). MCM3 and 18S primers were used for the semi-quantitative RT-PCR. Triplicate real-time PCRs were performed (BioRad single color detection system with Biorad’s SYBR green) and normalized to GAPDH. Primers used for RT-PCR and qPCR are shown in supplementary Table 2.

e2f1 luciferase reporterE2f1 reporter assays were performed in triplicate using 200 ng E2f1-luciferase reporter (15), 20 ng renilla and 0, 100, 500 or 1000 ng Vsv-pVHL30 in E2f7/8 knockout MEFs (16) using the DUAL luciferase system (Promega), as previously described (14).

strep/flAg tandem affinity purification (TAP) and Mass spectrometric (Ms) analysis HEK293T cells were transfected for 48 hours with pVHL (isoform 2) fused to an N-terminal SF-TAP-tag using PEI (PolySciences) as a transfection reagent, before being lysed in 30 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Nonidet-P40, protease inhibitor cocktail (Roche) and phosphatase inhibitor cocktail 2 and 3 (Sigma-Aldrich) for 20 minutes at 4°C. The Streptavidin- and FLAG-based tandem affinity purification steps were performed as previously described (17). 5% of the final eluate samples was separated by SDS-PAGE and stained with silver according to standard protocols. The remaining sample was protein-precipitated with chloroform and methanol and stored at -80°C. Further processing of protein precipitates, MS analysis and peptide identification was carried out as reported previously (18). Proteins identified in control SF-TAP experiments, were manually removed from the analysis.

RCC patient tumor collection and immunohistochemistryFormalin-fixed paraffin embedded tumor tissue from 138 RCCs dating from January 1994 to September 2006 were collected from the Pathology archives of the University Medical Center Utrecht (UMCU) after authorization of the UMCU Institutional Review Board in accordance with Dutch medical ethical guidelines. Patients diagnosed with oncocytoma were excluded. Patient-matched normal adjacent kidney tissue was included in the analyses, allowing for comparative analyses of tumor versus matched normal kidney. A tissue microarray (TMA) was constructed

51

using material of these 138 cases to study the nuclear expression of E2F1, and its upstream and downstream targets HIF-1α, GLUT1, Ki67 and p27 in various histological subtypes of RCC. 4 mm sections of each paraffin block were mounted on silane-coated glass slides, dewaxed in xylene, rehydrated through graded ethanol concentrations and used for immunohistochemistry. Standard H&E sections of each TMA block were reassessed to independently confirm diagnosis and the presence of representative areas of normal and tumor tissue (19). Stainings were performed using anti-E2F1 (KH95, 1:100; Santa Cruz Biotechnology), anti-p27 (1:200; R&D Systems), anti-GLUT1 (ab652, 1:200; Abcam), anti-Ki67 (1:200; Dako) and anti-HIF1a (1:50; Becton Dickinson). An experienced pathologist (PJvD) together with an independent researcher (RHG) blindly scored three cores per tumor containing approximately 2,000-5,000 cells each. For subsequent analysis the mean score of these three cores was used.

statistical Analysis. Student’s T-tests were performed to evaluate differences between E2F1 expression and clinico-pathological features. Pearson correlation coefficient analysis demonstrates correlation between E2F1 expression and tumor diameter/other scored data. Cox Proportional Hazard (PH) regression was used to analyze the association of E2F1 expression to clinic-pathological characteristics with DFS and OS. To dichotomize this variable for appropriate Kaplan-Meier estimation an objective cut-off point of 25% E2F1 positivity was chosen. All statistical analyses were performed using SPSS (version 15.0) and p<0.05 (all reported two-sided) was considered to be significant.

Patient Characteristics. This retrospective study of 138 had a median age of 62.3 years (Interquartile Range (IQR); 53.2–72.2). Detailed patient characteristics are depicted in Table 1. DNA sequencing of VHL of 117 of these tumors with a clear-cell histology, including eight known VHL patients, verified which of these tumors carried somatic VHL mutations. In the other 21 RCC (“non-cc” group) the histological subtype was mainly papillary (N=11) or a mixed variant of papillary and cc subtype (N=4), four patients had a sarcomatoid and two patients a chromophobe subtype. At time of nephrectomy, 12 patients (8%) were synchronously diagnosed with metastatic disease and were excluded from analysis of disease-free survival. According to the AJCC (American Joint Committee on Cancer) staging system, 53 patients were classified with AJCC stage I disease, whereas 22, 51 and 12 patients were categorized with respectively AJCC stage II, III and IV disease. During clinical follow-up 33 patients (23%) developed metastases, most of them (N=19 patients) within one year. In total 83 patients (58%) completed 5-years survival, 34 patients (24%) died due to the disease and 25 patients (18%) were censored as they had not reached 5-year survival at time of analysis or died from other causes. Overall survival (OS) was described as time from nephrectomy until date of death or last clinical follow-up. Clinical follow up was ended when patients were disease-free for at least five years. Disease-free survival (DFS) was defined as the time from primary diagnosis until the date of presence of first metastasis or recurrence of RCC disease. The mean DFS and OS were 3.6 years (95% Confidence Interval (CI): 3.3 - 4.0 years) and 4.1 years (95% CI: 3.8-4.4 years) respectively. The 1-, 2- and 3-year cumulative DFS were 78%, 74% and 71% and for OS 90%, 84% and 79%, respectively. At the moment of final analysis 94 patients (68%) were still alive. Eleven patients confronted with metastasized disease received interferon-based therapy. Sunitinib was given to two patients. One patient was treated with interleukin-2 and two patients underwent metastasectomy with curative intention. The other 29 RCC patients with metastasized disease were treated with palliative best supportive care.



Function Gene NCBI Reference Fold changeKnown to be E2F regulated mcm6 NM_001082849 1.93

rrm2 NM_131450 3.11tk1 NM_199832 1.44

zgc:110727 NM_214782 1.47rrm1 NM_131455 2.18tyms NM_131760 2.02pola2 NM_199581 1.65mcm5 NM_178436 2.34prim1 NM_201448 1.64

DNA and chroma�n metabolism top2a NM_001003834 2.56lmnb1 NM_152972 1.51

si:dkeyp-35b8.5 NM_001126388 1.91h2afx NM_201073 1.54

mthfd2 NM_001002181 2.59Receptors and signal transduc�on stka NM_212566 1.41

vegfab NM_001044855 2.70smc2 NM_199542 2.16

si:ch73-60e21.1 XM_001920548 1.41Cell cycle ccnb1 NM_131513 2.03

Transcrip�on factors ldb3 NM_201505 1.45Miscellaneous zgc:110064 NM_001020565 1.77

zgc:63734 NM_199675 1.58kifc1 NM_131206 1.73

RCC10786-0- + - + pVHL

E2F1

b-actin

pVHL

0

0.5

1.0

1.5

Fold

incr

ease

/dec

reas

e E

2F1

exp

ress

ion

vs G

AP

DH

MCM3

18S

RCC10 786-0- + - + pVHL

E2f

1 lu

c. v

s re

nilla

0 100 500 1000 ng pVHL

RCC10RCC10 + pVHL786-0786-0 + pVHL *

**

Figure 1

A

B C D

E F

sibling

Vhl mutant

0

5

10

15

20

25

30


rrm2 NM_131450 3.11tk1 NM_199832 1.44


DNA and chromatin metabolism top2a NM_001003834 2.56lmnb1 NM_152972 1.51


mthfd2 NM_001002181 2.59Receptors and signal transduction stka NM_212566 1.41

vegfab NM_001044855 2.70smc2 NM_199542 2.16


Transcription factors ldb3 NM_201505 1.45Miscellaneous zgc:110064 NM_001020565 1.77

zgc:63734 NM_199675 1.58kifc1 NM_131206 1.73

50

25

50

kDa

**

RCC10786-0- + - + pVHL

E2F1

b-actin

pVHL

0

0.5

1.0

1.5

Fold

incr

ease

/dec

reas

e E

2F1

exp

ress

ion

vs G

AP

DH

MCM3

18S

RCC10 786-0- + - + pVHL

E2f

1 lu

c. v

s re

nilla

0 100 500 1000 ng pVHL

RCC10RCC10 + pVHL786-0786-0 + pVHL *

**

Figure 1

A

B C D

E F

sibling

Vhl mutant

0

5

10

15

20

25

30


rrm2 NM_131450 3.11tk1 NM_199832 1.44





vegfab NM_001044855 2.70smc2 NM_199542 2.16



zgc:63734 NM_199675 1.58kifc1 NM_131206 1.73

50

25

50

kDa

**

RCC10786-0- + - + pVHL

E2F1

b-actin

pVHL

0

0.5

1.0

1.5

Fold

incr

ease

/dec

reas

e E

2F1

exp

ress

ion

vs G

AP

DH

MCM3

18S

RCC10 786-0- + - + pVHL

E2f

1 lu

c. v

s re

nilla

0 100 500 1000 ng pVHL

RCC10RCC10 + pVHL786-0786-0 + pVHL *

**

Figure 1

A

B C D

E F

sibling

Vhl mutant

0

5

10

15

20

25

30


rrm2 NM_131450 3.11tk1 NM_199832 1.44





vegfab NM_001044855 2.70smc2 NM_199542 2.16



zgc:63734 NM_199675 1.58kifc1 NM_131206 1.73

50

25

50

kDa

**

RCC10786-0- + - + pVHL

E2F1

b-actin

pVHL

0

0.5

1.0

1.5

Fold

incr

ease

/dec

reas

e E

2F1

exp

ress

ion

vs G

AP

DH

MCM3

18S

RCC10 786-0- + - + pVHL

E2f

1 lu

c. v

s re

nilla

0 100 500 1000 ng pVHL

RCC10RCC10 + pVHL786-0786-0 + pVHL *

**

Figure 1

A

B C D

E F

sibling

Vhl mutant

0

5

10

15

20

25

30


rrm2 NM_131450 3.11tk1 NM_199832 1.44





vegfab NM_001044855 2.70smc2 NM_199542 2.16



zgc:63734 NM_199675 1.58kifc1 NM_131206 1.73

50

25

50

kDa

**

B

e f

DC

A

figure 1 - pVHL negatively regulates E2F1 and E2F1 downstream targets

53

R e s u lTs

pVHl suppresses E2F1 expression. Zebrafi sh embryos with homozygous or trans-heterozygous vhl truncating mutations recapitulate most aspects of VHL disease, including a renal cell regulation defect(20). We previously described microarray analysis of vhl mutant embryos (7 dpf) (10). Exploring patterns of transcription factor activity suggests that an underlying E2F1 activity signature is prevalent in vhl mutant zebrafi sh. E2F1-specifi c transcriptional targets (21), including mcm5, mcm6, rrm1, rrm2 or top2a, are upregulated in vhl mutant embryos 1.41-3.11 fold (figure 1A). Our previous validation of this dataset (10) had shown these values to be signifi cant. We confi rmed increased mcm5 RNA expression by in situ hybridization on whole mount vhl mutant and sibling embryos (figure 1B, N=10 embryos). E2F1 upregulates expression of MCMs and is required for DNA-replication (6). Semi-quantitative RT-PCR in two VHL-/- RCC cell lines (RCC10 and 786-0), stably transfected with either empty vector or full-length pVHL30, demonstrates lower expression of MCM3 in RCC10 and 786-0 cells after reconstitution with pVHL30 (figure 1C), supporting vhl zebrafi sh data. To determine whether MCM mRNA changes refl ect E2F1 protein levels, we performed western blots of E2F1 in these cell lines and observed striking downregulation of E2F1 in both cell lines when reconstituted with pVHL30 (figure 1D). RNA from these cell lines was analyzed by qPCR (in triplicate) for E2F1, E2F2, E2F3, E2F7, E2F8 and GAPDH mRNA expression levels. Reintroduction of pVHL signifi cantly decreased expression of E2F1 mRNA in both RCC cell lines (figure 1e), while this difference was not observed for any other E2F family members tested (not shown). One of E2F1’s more sensitive targets is its own E2F1 promoter, and validated E2F1 luciferase reporter assays are regularly used to test E2F1 transactivity (15). To study the effect of pVHL on endogenous E2F1 transactivation, we used E2f7/E2f8 double knockout mouse embryonic fi broblasts which express high endogenous levels of E2f1 (16) and which can be transfected effi ciently, in contrast to RCC cells. Transfecting increasing amounts of exogenous Vsv-pVHL30 balanced with empty vector to normalize total DNA transfected, a signifi cant dose-dependent decrease in E2f1 reporter luciferase activity was measured, using a two-tailed Student’s T-test (p=0.007 for 100 ng Vsv-pVHL; figure 1f). If pVHL affects protein turnover of E2F1 that would also explain the observed drop in E2F1 mRNA levels. We transfected HEK293T cells with E2F1-HA in combination with empty vector or Vsv-pVHL30, and used MG132 to block proteasomal degradation. Concordant with our previous

figure 1 (A): Microarray expression analysis of homozygous vhl mutant zebrafi sh embryos (7 dpf) compared to wildtype siblings (7 dpf). All genes listed are signifi cantly upregulated in the absence of vhl and are known to be regulated specifi cally by E2F1(21). (B): Photos of the tail of a wildtype sibling and a vhl mutant embryo after in situ hybridization of mcm5 on whole mount embryos (7.5 dpf; N=10). (C): Semi-quantitative RT-PCR of MCM3 in unsynchronized RCC cells with or without pVHL30 expression. 18S expression measured as input control. (D): Lysates of unsynchronized RCC cells with or without pVHL30 expression analyzed for E2F1 expression (upper panel). b-actin used as loading control (middle panel); note that pVHL (lower panel) runs at different sizes due to different tags. (e): Analysis of E2F1 mRNA expression in RCC cells with or without pVHL30 expression via qPCR. Each measurement is done in triplicate. mRNA expression of E2F1 is normalized using GAPDH expression. A two-tailed Student’s T-test was used to determine statistical signifi cance, indicated with asterisks (*). (f): Analysis of increasing amounts of pVHL30 on the activity of E2F1 using a E2F1 promoter luciferase (luc.) assay in E2f7/8 double knockout MEFs. Renilla used for normalization. A two-tailed Student’s T-test was used to determine statistical signifi cance, indicated with asterisks (*).



findings, exogenous pVHL30 decreased E2F1 protein levels in cells untreated with MG132 (figure 2A). Exposure to MG132 for 16 hours resulted in the well-characterized accumulation of endogenous HIF1a (figure 2A) (22). We detected only a slightly increased pool of (poly-)ubiquitinated E2F1 in these cell lines regardless of whether they expressed exogenous pVHL30 or not (Fig. 2A). It is noteworthy that blocking proteasomal degradation of E2F1 by MG132 treatment in cells transfected with pVHL30 increases E2F1 levels about three-fold, suggesting that pVHL has a role in E2F1 turnover.

Regulation of e2f1 expression by VHl occurs in a HIf1a-independent mannerpVHL’s best characterized function is as an E3 ubiquitin ligase, regulating HIF1/2a proteasomal degradation in normoxic conditions (2;3;23). In hypoxic conditions however, these transcription factors escape pVHL recognition, resulting in an increased stability of (nuclear) HIF1/2a and subsequent increase of downstream targets of HIF1/2a. Since E2F1 expression has been shown by others to be increased as a result of hypoxia (24), we wondered whether upregulation of E2F1 upon VHL inactivation is mediated through HIF1/2a. We incubated RCC10 and 786-0 cells stably transfected with either empty vector or pVHL30 for 24 hours in normoxia (21% O2) or hypoxia (1% O2). Indeed, E2F1 is upregulated in hypoxic VHL-null RCC cells. However, reintroduction of pVHL30 overrides this effect (figure 2B), suggesting that the turnover of E2F1 protein levels by pVHL30 is independent of HIF1/2a. We further tested the involvement of HIF1a by depleting cellular levels of HIF1a in RCC10 cells using ON-TARGET SMARTpools targeting HIF1a and used ON-TARGET SMARTpools targeting USP30 as control in this experiment. While effective in downregulating the expression of the HIF1a, knockdown of HIF1a did not result in aberrant E2F1 expression (figure 2C). We conclude that the observed downregulation of E2F1 upon re-introduction of pVHL30 in RCC10 cells is not mediated by HIF1a. We examined E2F1 expression in subconfluent RCC10 clones stably expressing a panel of VHL-variant alleles. pVHL19, a normally occurring pVHL isoform (isoform 3) originating from an alternative translation start site and capable of downregulating HIF1a (25;26), decreased E2F1 expression, like pVHL30 (figure 2D). Interestingly, both VHL disease-associated alleles

figure 2 (A): Western blot analysis of E2F1 in HEK293T cells transfected with either HA-tagged E2F1 in combination with empty vector or Vsv-pVHL30. Where indicated, cells were exposed to MG132 for 16 hours to block proteasomal degradation. HIF1α was used as control for proteasomal degradation block upon MG132 treatment (upper panel). pVHL was used as expression control (third panel) and b-actin was used as loading control (lower panel). Samples were loaded separately on the same SDS-PAGE gel, but not adjacent as indicated with black vertical lines. (B): RCC cell lines 786-0 and RCC10 with or without pVHL30 expression were incubated for 24 hours in normoxia (21% O2) or hypoxia (1% O2), where indicated. Lysates were analyzed on western blot for E2F1 expression (upper panel). b-actin used here as loading control (second panel) and pVHL used as expression control (lower panel). (C): RCC10 cells with or without pVHL30 were transfected with ON-TARGET SMARTpools targeting USP30 (negative control) or HIF1α, where indicated. Cells were lysed 48 hours after siRNA transfection and analyzed for E2F1 (upper panel), HIF1a (third panel) and pVHL (lower panel). b-catenin was used as loading control (second panel). Sample loaded separately on the same SDS-PAGE gel, is indicated with a black vertical line. (D): Subconfluent RCC cells, stably transfected with pVHL30, pVHL19 or VHL-disease variants (pVHL-ΔF76, VHL disease type 1; pVHL-F119S, VHL disease type 2C), were analyzed for E2F1 protein expression by western blot (upper panel). b-actin was used in this experiment as loading control (lower panel). (e): On the left, analysis of the influence of a panel of HA-tagged pVHL variants on the activity of E2f1 using a E2f1 promoter luciferase (luc.) assay in E2f7/8 double knockout MEFs. Renilla used here for normalization. A two-tailed Student’s T-test was used to determine statistical significance, indicated with asterisks (*). On the right, western blot analysis of the expression of all pVHL variants tested in the E2f1 promoter luciferase assay. Anti-HA antibody was used to detect HA-tagged pVHL (upper panel). Gamma tubulin antibody was used as input control (lower panel). Samples loaded separately on the same SDS-PAGE gel, are indicated with a black vertical line.

55

pVHL

USP30

-siRNA

HIF1α-si

RNA

USP30-si

RNA

HIF1α

β-catenin

E2F1

pVHL- - +- + - +- +- + pVHL

E2F1

pVHL

β-actin

21% 1% 21% 1% oxygen

786-0 RCC10

E2F1

pVHL3

0 (clo

ne 3D

10)

pVHL3

0 (clo

ne 90

)

pVHL1

9

Mock

pVHL-∆

F76

pVHL-F

119S

1 2C

β-actin

Figure 2

A B C

D E

0

5

10

15

20

E2f

1 lu

c. v

s re

nilla

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

*

**

**

HIF1α

E2F1

E2F1 +

pVHL3

0

E2F1 +

pcDNA3

MG132E2F

1 + pV

HL30

E2F1 +

pcDNA3

- - + +

β-actin

pVHL

VHL disease type

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

γ-tubulin

pVHL

50

kDa

25

50

75

100

150

50

50

kDa

25

50

25

100

100

kDa

50

50

kDa

26

50kDa

34

pVHL

USP30

-siRNA

HIF1α-si

RNA

USP30-si

RNA

HIF1α

β-catenin

E2F1

pVHL- - +- + - +- +- + pVHL

E2F1

pVHL

β-actin

21% 1% 21% 1% oxygen

786-0 RCC10

E2F1

pVHL3

0 (clo

ne 3D

10)

pVHL3

0 (clo

ne 90

)

pVHL1

9

Mock

pVHL-∆

F76

pVHL-F

119S

1 2C

β-actin

Figure 2

A B C

D E

0

5

10

15

20E

2f1

luc.

vs

reni

lla

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

*

**

**

HIF1α

E2F1

E2F1 +

pVHL3

0

E2F1 +

pcDNA3

MG132E2F

1 + pV

HL30

E2F1 +

pcDNA3

- - + +

β-actin

pVHL

VHL disease type

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

γ-tubulin

pVHL

50

kDa

25

50

75

100

150

50

50

kDa

25

50

25

100

100

kDa

50

50

kDa

26

50kDa

34

pVHL

USP30

-siRNA

HIF1α-si

RNA

USP30-si

RNA

HIF1α

β-catenin

E2F1

pVHL- - +- + - +- +- + pVHL

E2F1

pVHL

β-actin

21% 1% 21% 1% oxygen

786-0 RCC10

E2F1

pVHL3

0 (clo

ne 3D

10)

pVHL3

0 (clo

ne 90

)

pVHL1

9

Mock

pVHL-∆

F76

pVHL-F

119S

1 2C

β-actin

Figure 2

A B C

D E

0

5

10

15

20

E2f

1 lu

c. v

s re

nilla

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

*

**

**

HIF1α

E2F1

E2F1 +

pVHL3

0

E2F1 +

pcDNA3

MG132E2F

1 + pV

HL30

E2F1 +

pcDNA3

- - + +

β-actin

pVHL

VHL disease type

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

γ-tubulin

pVHL

50

kDa

25

50

75

100

150

50

50

kDa

25

50

25

100

100

kDa

50

50

kDa

26

50kDa

34

pVHL

USP30

-siRNA

HIF1α-si

RNA

USP30-si

RNA

HIF1α

β-catenin

E2F1

pVHL- - +- + - +- +- + pVHL

E2F1

pVHL

β-actin

21% 1% 21% 1% oxygen

786-0 RCC10

E2F1

pVHL3

0 (clo

ne 3D

10)

pVHL3

0 (clo

ne 90

)

pVHL1

9

Mock

pVHL-∆

F76

pVHL-F

119S

1 2C

β-actin

Figure 2

A B C

D E

0

5

10

15

20

E2f

1 lu

c. v

s re

nilla

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

*

**

**

HIF1α

E2F1

E2F1 +

pVHL3

0

E2F1 +

pcDNA3

MG132E2F

1 + pV

HL30

E2F1 +

pcDNA3

- - + +

β-actin

pVHL

VHL disease type

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

γ-tubulin

pVHL

50

kDa

25

50

75

100

150

50

50

kDa

25

50

25

100

100

kDa

50

50

kDa

26

50kDa

34

pVHL

USP30

-siRNA

HIF1α-si

RNA

USP30-si

RNA

HIF1α

β-catenin

E2F1

pVHL- - +- + - +- +- + pVHL

E2F1

pVHL

β-actin

21% 1% 21% 1% oxygen

786-0 RCC10

E2F1

pVHL3

0 (clo

ne 3D

10)

pVHL3

0 (clo

ne 90

)

pVHL1

9

Mock

pVHL-∆

F76

pVHL-F

119S

1 2C

β-actin

Figure 2

A B C

D E

0

5

10

15

20

E2f

1 lu

c. v

s re

nilla

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

*

**

**

HIF1α

E2F1

E2F1 +

pVHL3

0

E2F1 +

pcDNA3

MG132E2F

1 + pV

HL30

E2F1 +

pcDNA3

- - + +

β-actin

pVHL

VHL disease type

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

γ-tubulin

pVHL

50

kDa

25

50

75

100

150

50

50

kDa

25

50

25

100

100

kDa

50

50

kDa

26

50kDa

34

pVHL

USP30

-siRNA

HIF1α-si

RNA

USP30-si

RNA

HIF1α

β-catenin

E2F1

pVHL- - +- + - +- +- + pVHL

E2F1

pVHL

β-actin

21% 1% 21% 1% oxygen

786-0 RCC10

E2F1

pVHL3

0 (clo

ne 3D

10)

pVHL3

0 (clo

ne 90

)

pVHL1

9

Mock

pVHL-∆

F76

pVHL-F

119S

1 2C

β-actin

Figure 2

A B C

D E

0

5

10

15

20

E2f

1 lu

c. v

s re

nilla

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

*

**

**

HIF1α

E2F1

E2F1 +

pVHL3

0

E2F1 +

pcDNA3

MG132E2F

1 + pV

HL30

E2F1 +

pcDNA3

- - + +

β-actin

pVHL

VHL disease type

pcDNA3

pVHL1

9

pVHL3

0

pVHL-F

119S

pVHL-Y

112D

pVHL-∆

95-12

3

pVHL-∆

F76

γ-tubulin

pVHL

50

kDa

25

50

75

100

150

50

50

kDa

25

50

25

100

100

kDa

50

50

kDa

26

50kDa

34

figure 2 - E2F1 regulation by pVHL is HIF-independent

A

C

e

B

D

f



Fig

ure

3

A B

E2F

1

clone

3D10

clone

3D10

+ sh

VHL#4

clone

3D10

+ sh

VHL#5

clone

3D10

+ sh

VHL#6

β-ac

tin

pVH

L

clone

D10

clone

D10

+ E2F

1

RCC10

E2F

1

β-ca

teni

n

G1/

S

G2/

M

RC

C10

clon

e 3D

10 c

lone

3D

10

+ s

hVH

L#4

clo

ne 3

D10

+

shV

HL#

6 c

lone

3D

10

+ s

hVH

L#5

hrs

afte

r re

leas

e0

624

120

624

120

624

120

624

120

624

120

100 80 60 40 20% of cells

RC

C10

clon

e 3D

10cl

one

3D10

+ E

2F1

G1/

S

G2/

M

Unsyn

ch.

04.

58

11hr

s af

ter r

elea

se

0

100 80 60 40 20% of cells

Unsyn

ch.

04.

58

11hr

s af

ter r

elea

se

Unsyn

ch.

04.

58

11hr

s af

ter r

elea

se

kDa502550 5075

kDa

figu

re 3

- :e

ffec

ts o

f pVH

l on

the

cell

cycl

e ar

e e

2f1-

depe

nden

t

57

pVHL-DF76(incapable of downregulating HIF1/2a(26)) and pVHL-F119S (still capable of downregulating HIF1/2a(27)) reduce E2F1 expression compared to a Mock control, which indicates that the regulation of E2F1 by pVHL is independent of HIF1/2a (figure 2D). Next, we checked whether transient overexpression of pVHL30 and VHL variant alleles in E2f7/8 double knock-out MEFs did affect the transactivation of E2f1 as measured via an E2f1 reporter assay. Apart from pVHL-DF76, all alleles tested reduced E2f1 transactivation significantly (figure 2e). Expression of all constructs was validated by western blot analysis (figure 2e). The physical association between E2F1 and pVHL was interrogated by performing a tandem affinity purification in HEK293T cells of SF-TAP-tagged pVHL isoform 2, followed by mass spectrometric analysis. Known pVHL-E3 ubiquitin ligase complex proteins Cullin-2, Elongin B, Elongin C, Rbx1 (2;3;28) and members of the COP9 signalosome complex (29) were co-purified with the pVHL protein complex; however E2F1 was not detected in this analysis (supplementary Table 4) implying that E2F1 is not directly bound to pVHL or that the interaction is transient. Interestingly, known transcriptional repressors of E2F1, TRIM28, SETDB1 and NuRD complex proteins, RBBP4 and RBBP7 (30) were found in this analysis (Supplementary Table 4) suggesting that regulation of E2F1 by pVHL could be mediated on a transcriptional level via TRIM28-SETDB1-RBBP4/7.

e2f1 regulation by pVHl affects cell cycle progression in RCC cellsBecause increased E2F1 levels upon VHL inactivation in RRC cells may reflect altered cell cycle profiles, we analyzed cell cycle profiles of subconfluently cultured RCC10 cells, with or without pVHL30 (figure 3A). Cell cycles were synchronized at the G1/S transition by a double thymidine block, and then fixed for FACS analysis at 0, 6, 12, and 24 hours after release from thymidine in the presence of nocodazole to catch mitotic cells. We did not observe significant changes in cell cycle profiles after thymidine synchronization (t=0; figure 3A). However, after release of thymidine block, pVHL-proficient RCC10 cells (clone 3D10) were enriched in the G2/M-phase of the cell cycle compared to VHL-null RCC10 cells after 6 hrs (p=0.006). Interestingly, at t=6hrs this effect was (significantly) reversible upon three independent shRNA-mediated knockdown of the coding sequence of exogenous pVHL (figure 3A, shVHL#4 p=0.05, shVHL#5 p<0.001; shVHL#6 p=0.04) indicating that this effect is not due to clonal drift and is fundamental to pVHL action. We next used this cell system to interrogate the role for E2F1 on the observed effect of pVHL on the cell cycle. We transiently expressed exogenous E2F1 in the RCC10 clone 3D10 (reconstituted with pVHL30); again, we observed accelerated entry into G2/M-phase in clone

figure 3 (A): Left panel: western blot analysis of E2F1 using lysates of subconfluently cultured RCC10 cells stably transfected with pVHL30 (clone 3D10), with or without VHL knockdown. Three independent shRNAs (#4, #5 and #6) targeting the coding sequence of VHL were included. pVHL was used as expression control and for validation of the shRNAs (middle panel), b-actin was used as loading control (lower panel). Samples loaded separately on the same SDS-PAGE gel, are indicated with a black vertical line. Right panel: cell cycle profiles of sub-confluently cultured RCC10 cells and clone 3D10 cells, with or without VHL shRNAs (#4, #5 and #6). Cells were synchronized to G1/S transition by a double thymidine block, and then fixed for FACS analysis at 0, 6, 12, and 24 hours after release from thymidine block in the presence of nocodazole. (B): On the left, western blot analysis of E2F1 using lysates of sub-confluently cultured RCC10 cells, clone 3D10 and lysates of clone 3D10 cells expressing exogenous E2F1 (upper panel). b-catenin was used as loading control. On the right, cell cycle profiles of sub-confluently cultured RCC10 cells, clone 3D10 cells, and clone 3D10 cells expressing exogenous E2F1. Cells were synchronized to G1/S transition by a double thymidine block, and then fixed for FACS analysis at 0, 4.5, 8, and 11 hours after release from thymidine block in the presence of nocodazole.



039 041 061

Fold

indu

ctio

n E

2F1

vs

GA

PD

H e

xpre

ssio

n

0

1

1

VHL mutation# Comment

2

3

4

5

6

7

8

Missense mutation

Missense mutation

Missense mutation

Missense mutation

Splice site mutation, previously known as IVS1+1G>A

Missense mutation

Genomic deletions of exons 1 and 2, exact breakpoints unknown


c.509T>A; p. Val170Asp




c.500G>A; p. Arg167Gln

c.340+1G>A

c.1-?_463+?

c.1-?_463+?

Nuclear E2F1 Nuclear HIF1a

90% 65%

35% 100%

90%90%

90% 50%

50% 10%

50% 90%

50% 90%

35% 100%

sample-id

normaltumor

(#3)(#7)Anti-E2F1

Tumor

Normal

Figure 4

A

B C

D

G

PTG

Clear-c

ell (C

C; N=1

09)

Non-C

C (N=2

1)

VHL (N=8

)

20

40

60

80

0

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or *

**4

3

2

039 041 061

Fold

indu

ctio

n E

2F1

vs

GA

PD

H e

xpre

ssio

n

0

1

1


2

3

4

5

6

7

8

Missense mutation

Missense mutation

Missense mutation

Missense mutation


Missense mutation








c.340+1G>A

c.1-?_463+?

c.1-?_463+?


90% 65%

35% 100%

90%90%

90% 50%

50% 10%

50% 90%

50% 90%

35% 100%

sample-id

normaltumor

(#3)(#7)Anti-E2F1

Tumor

Normal

Figure 4

A

B C

D

G

PTG

Clear-c

ell (C

C; N=1

09)

Non-C

C (N=2

1)

VHL (N=8

)

20

40

60

80

0

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or *

**4

3

2

039 041 061

Fold

indu

ctio

n E

2F1

vs

GA

PD

H e

xpre

ssio

n

0

1

1


2

3

4

5

6

7

8

Missense mutation

Missense mutation

Missense mutation

Missense mutation


Missense mutation








c.340+1G>A

c.1-?_463+?

c.1-?_463+?


90% 65%

35% 100%

90%90%

90% 50%

50% 10%

50% 90%

50% 90%

35% 100%

sample-id

normaltumor

(#3)(#7)Anti-E2F1

Tumor

Normal

Figure 4

A

B C

D

G

PTG

Clear-c

ell (C

C; N=1

09)

Non-C

C (N=2

1)

VHL (N=8

)

20

40

60

80

0

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or *

**4

3

2

039 041 061

Fold

indu

ctio

n E

2F1

vs

GA

PD

H e

xpre

ssio

n

0

1

1


2

3

4

5

6

7

8

Missense mutation

Missense mutation

Missense mutation

Missense mutation


Missense mutation








c.340+1G>A

c.1-?_463+?

c.1-?_463+?


90% 65%

35% 100%

90%90%

90% 50%

50% 10%

50% 90%

50% 90%

35% 100%

sample-id

normaltumor

(#3)(#7)Anti-E2F1

Tumor

Normal

Figure 4

A

B C

D

G

PTG

Clear-c

ell (C

C; N=1

09)

Non-C

C (N=2

1)

VHL (N=8

)

20

40

60

80

0

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or *

**4

3

2

figure 4 - Nuclear E2F1 expression is high in RCCs and dependent on VHL

B

C D

A

59

3D10 cells as compared to RCC10 VHL-null cells. Overexpression of E2F1 in clone 3D10 cells significantly reduced this effect (at 8 hours release p<0.001, figure 3B). We conclude that the entry and progression of RCC cells through G2 in vitro may be largely explained by increased E2F1 expression subsequent to bi-allelic inactivation of VHL, with possible consequences to the cell cycle machinery of renal cells in vivo.

e2f1 is upregulated in VHl-associated renal cell carcinomaDevelopment of ccRCC in patients with germline inactivating VHL mutations is associated with loss of function of the second (wildtype) VHL allele, resulting in the stability of HIF1/2a (2;31). To investigate a relationship between expression of E2F1 and loss of VHL, we analyzed E2F1 and HIF1a expression using a TMA of paraffin-embedded tumor and normal renal tissue from 138 primary RCCs, eight of which are known to carry germline mutations in VHL (figure 4A) (32). The H&E stained TMA slides showed sufficient representative tumor and normal tissue in all cases. E2F1 immunoreactivity was seen in the nucleus of tumor cells with a labeling index (LI) of 0% to 90% (figure 4A), with a mean percentage of 31.6% (SD=28.8%) and a median percentage of 20% (IQR; 6.6–50%) as represented in Table 2. For both HIF1a and E2F1, a sample was determined to be histologically positive when more than 25% of all nuclei showed unambiguous staining. In normal kidney samples, some E2F1 staining was observed in glomeruli. Proximal tubules often showed some cytoplasmic E2F1 staining, while distal tubules did not generally stain for E2F1 (figure 4B). We never observed more than 25% nuclear positivity for either E2F1 or HIF1a in any of the normal tissues from all 138 patients examined. In primary tumor tissue of almost all VHL patients, we observed high nuclear HIF1a and E2F1 staining (figure 4A, B); in a single VHL patient RCC (figure 4A, sample #5), we observed high nuclear E2F1, but low positivity for HIF1a, suggesting that expression of E2F1 and HIF1a in RCC do not always correlate. Four individuals from a single VHL family with the same germline mutation in VHL (c.509T>A) overlapped with respect to HIF1a and E2F1 expression in their RCCs. No significant differences in E2F1 expression were found between cc and non-cc tumor histology (figure 4C). However in RCCs from VHL patients E2F1 expression was significantly increased (figure 4C; Student’s T-test, p<0.05 (two-tailed) for both cc and non-cc).

e2f1 expression is increased sporadic renal cancersIn three sporadic kidney tumors matched with normal tissue from the same patient, freshly acquired post-operation material was used to quantify the differences in E2F1 mRNA levels by qPCR. All three patient tumors exhibited increased E2F1 mRNA expression compared to normal kidney tissue, when normalized to GAPDH. In two samples, the increased expression of E2F1 in tumor

figure 4 (A): Inherited VHL mutations of eight VHL patients with clear-cell RCCs, combined with the results of immunohistochemical staining of the associated RCCs for nuclear E2F1 and HIF1a. Shown are the percentages of scored nuclei postive for either E2F1 or HIF1a. (B): Photos of paraffin-embedded tumor and normal material of the kidney from two patients with VHL disease, stained for E2F1. G, glomerulus; PT, proximal tubules. Bar=100 um. (C): Nuclear E2F1 expression analyzed in paraffin-embedded tumor renal tissue from 138 RCC patients, eight of which are known VHL patients. Chi-square statistical analysis of nuclear E2F1 expression (≥25%) between clear-cell RCCs derived from VHL patients and 109 sporadic clear-cell or 21 sporadic non-clear-cell RCCs with unknown VHL status revealed that E2F1 is significantly higher expressed in the nucleus in VHL-associated RCCs (p=0.04), compared to sporadic RCCs. Asterisk (*) indicates statistical significance. (D): qPCR for E2F1 in three sporadic renal tumors, matched with normal kidney tissue for each patient. E2F1 mRNA expression normalized vs. GAPDH. A two-tailed Student’s T-test was used to determine statistical significance. Asterisks (*) indicate statistical significance.



versus normal was significant (figure 4D; two-tailed Student’s T-test, sample 041, p=0.029 and sample 061, p=0.038).

e2f1 expression associated with tumor diameter, metastasis, Dfs and os. Given the increased expression of E2F1 in VHL-/- RCC cells and in RCCs, we hypothesized that E2F1 expression in RCCs would correlate with clinical-pathological features. Indeed, high E2F1 expression was associated with a smaller tumor diameter (N=138, two-tailed Pearson Correlation=-0.3, p=0.001). Using the AJCC staging system criterion of 7 cm tumor diameter as a cut-off showed a significant difference in E2F1 expression between tumors with a small and large diameter (figure 5A; two-tailed Student’s T-test, p=0.006), suggesting that E2F1 has a tumor suppressive role in RCCs. The expression of p27, which was included as well, showed a significant correlation with E2F1 (N=130, two-tailed Pearson Correlation=-0.3, p=0.001), while Ki67, GLUT-1 or HIF1a did not show a significant correlation with E2F1 (Table 2). The correlation between E2F1 and p27 indicates that E2F1 expression in RCCs is likely linked to senescence, which is known for E2F1 when overexpressed in human fibroblasts(5). These data also fit with previous data supporting a HIF-independent, but pRb-dependent senescence in kidney tissue upon VHL loss (4). In addition, AJCC stage III and AJCC stage IV RCCs showed significant diminished E2F1 expression compared to AJCC stage I RCCs (figure 5B; Students’ T-test, two-tailed, p=0.003 and p=0.04, respectively). Since the AJCC staging system is related to patients’ prognosis, we examined the association of E2F1 expression with DFS and OS. E2F1 expression as a continuous variable significantly predicted DFS and OS using a Cox PH regression analysis (p=0.02 and p=0.01 respectively) and indicated that a lower E2F1 expression appointed poor DFS and OS (Table 1). Expression of HIF-1α, GLUT1, Ki67 and p27 as a continuous variable did not predict DFS or OS (Table 2). To stratify RCC patients into ‘favorable’ and ‘poor’ risk groups for subsequent Kaplan-Meier analysis we chose 25% nuclear E2F1 expression as an objective cut-off point, because (a) <25% was observed in all examined normal cells, (b) 25% was the median of all observed cases in our patient cohort and (c) has been used before as a cut-off (33). Patients were then categorized into a group predicting favorable survival (≥25% nuclear E2F1 expression) and a group of poor risk patients (E2F1 <25%). Kaplan-Meier curves demonstrated strongly discriminative outcomes for DFS and OS (Log Rank test, p=0.02 and p=0.01, figure 5C, D, respectively). The favorable risk group had a mean DFS of 4.4 years (95% CI: 4.0–4.7 years), with 1-, 2- and 3-year DFS rates of 94%, 89% and 86%, respectively. The poor risk group had a mean DFS of 3.6 years (95% CI: 3.1–4.1), with 1-, 2- and 3-year survival rates of 76%, 73% and 69%, respectively. For OS, the mean survival times for the favorable and poor prognostic risk groups were 4.5 and 3.7 years (95% CI: 4.1-4.8 and 3.3-4.2 years respectively) with 1-, 2- and 3-year OS rates of 93%, 88% and 87% for the favorable risk group and 87%, 74% and 67% for the poor prognostic group.

figure 5 (A): Correlation of nuclear E2F1 expression to tumor diameter in RCC patients. A tumor diameter of 7 cm was used as cut-off. Significance of nuclear E2F1 expression between tumors with a diameter smaller than or equal to 7 cm and tumors with a diameter larger than 7 cm was calculated using a two-tailed Student’s T-test. Asterisk (*) indicates statistical significance. (B): Correlation of nuclear E2F1 expression to AJCC staging system (stage I-IV)in RCC patients. Significance of nuclear E2F1 expression between AJCC stage I and AJCC stages II-IV was calculated using a Student’s T-test. Asterisks (*) indicate statistical significance. (C): Correlation of nuclear E2F1 expression to disease free survival and overall survival. (D): RCC patients were categorized into a group predicting favorable survival (≥ 25% nuclear E2F1 expression, 67 patients) and a group of poor risk patients (E2F1<25%, 71 patients) for Kaplan-Meier analysis. Significance was determined using Log Rank tests.

61

Figure 5

A

0

10

20

30

40

50

Tumor

diamete

r

≤ 7 cm

(N=8

1)

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or

*

Tumor

diamete

r

> 7 cm

(N=5

7)

B

C D

0

10

20

30

40

50

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or

***

AJCC st

age I

(N=5

3)

AJCC st

age I

I (N=2

2)

AJCC st

age I

V (N=1

2)

AJCC st

age I

II (N=5

1)

0 1 2 3 4 5

Pro

porti

on d

isea

se fr

ee

0.0

0.5

0.6

0.7

0.8

0.9

1.0

Years0 1 2 3 4 5

Pro

porti

on s

urvi

ved

0.00.40.6

0.7

0.8

0.9

1.0

Years

Disease free survival Overall survival

< 25% nuclear E2F1 (62 patients; 40 disease free)≥ 25% nuclear E2F1 (64 patients; 53 disease free) ≥ 25% nuclear E2F1 (67 patients; 57 alive)

< 25% nuclear E2F1 (71 patients; 47 alive)

Figure 5

A

0

10

20

30

40

50

Tumor

diamete

r

≤ 7 cm

(N=8

1)

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or

*

Tumor

diamete

r

> 7 cm

(N=5

7)

B

C D

0

10

20

30

40

50

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or

***

AJCC st

age I

(N=5

3)

AJCC st

age I

I (N=2

2)

AJCC st

age I

V (N=1

2)

AJCC st

age I

II (N=5

1)

0 1 2 3 4 5

Pro

porti

on d

isea

se fr

ee

0.0

0.5

0.6

0.7

0.8

0.9

1.0

Years0 1 2 3 4 5

Pro

porti

on s

urvi

ved

0.00.40.6

0.7

0.8

0.9

1.0

Years




Figure 5

A

0

10

20

30

40

50

Tumor

diamete

r

≤ 7 cm

(N=8

1)

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or

*

Tumor

diamete

r

> 7 cm

(N=5

7)

B

C D

0

10

20

30

40

50

% n

ucle

ar E

2F1

stai

ning

i

n pr

imar

y R

CC

tum

or

***

AJCC st

age I

(N=5

3)

AJCC st

age I

I (N=2

2)

AJCC st

age I

V (N=1

2)

AJCC st

age I

II (N=5

1)

0 1 2 3 4 5

Pro

porti

on d

isea

se fr

ee

0.0

0.5

0.6

0.7

0.8

0.9

1.0

Years0 1 2 3 4 5

Pro

porti

on s

urvi

ved

0.00.40.6

0.7

0.8

0.9

1.0

Years




figure 5 - Nuclear E2F1 expression in RCCs predicts survival


A

C

D

B

62 Part I Molecular prognosticators

DI sCus s Ion

Our study shows that E2F1 expression is negatively regulated by the von Hippel-Lindau tumor suppressor protein in RCC cells. Introduction of pVHL into VHL-null RCC cells decreases E2F1 mRNA and protein levels and expression of known downstream targets of E2F1, like MCMs, independent of HIF1a. This effect in RCC cells is in line with the observed upregulation of mcm expression in vhl mutant zebrafish (10) or worm (34). Our data indicate that the regulation of E2F1 expression by pVHL occurs via transcriptional and post-translational mechanisms. Blocking proteasomal degradation of E2F1 by MG132 treatment in cells transfected with pVHL increased E2F1 protein levels about three-fold, suggesting that pVHL is (in)directly involved in E2F1 turnover. In addition, TAP-analysis of pVHL showed interaction with TRIM28, SETDB1 and NuRD complex proteins, RBBP4 and RBBP7, all known transcriptional repressors of E2F1. Future studies will reveal the contribution of the pVHL-E3 uiquitin ligase complex in regulating E2F1 expression. Re-introduction of pVHL into VHL-null RCC cells did slightly increase the content of cells in G1-phase of the cell cycle. This effect is known for pVHL when VHL-proficient RCC cells have been serum-starved (35) or have been exposed to DNA damage (36) and overlaps with the diminished MCM expression upon restoration of pVHL in RCC cells. Surprisingly, we observed that pVHL enhanced the entrance and progression of RCC cells through G2-phase of the cell cycle after release from thymidine block. This effect of pVHL on transition through G2 was blocked after introducing high levels of E2F1, which indicates that the regulation of E2F1 by pVHL is linked to cell cycle progression of RCC cells and may represent a novel function for pVHL. We identified that E2F1 expression is increased in renal tumors, corroborating and extending recent data describing E2F1 expression in renal cell carcinoma (37). Moreover, we observed that ccRCCs with germline mutations in VHL express significant more E2F1 than sporadic tumors of unknown VHL status, suggesting that VHL loss increases E2F1 expression in RCC cells in vivo. Although E2F1 is generally thought to be a positive regulator of the cell cycle, E2f1 knockout mice are predisposed to cancer, reflecting a more complex role for E2F1 in vivo (38). Our human data similarly makes the link between E2F1 expression and tumor suppression. Because lack of VHL mutations in combination with low expression levels of HIF-target CAIX is linked to increased tumor aggressiveness and poor survival (39), our data suggests that VHL-deficient kidney tumors are less aggressive due to E2F1-mediated tumor suppression. Accordingly, VHL mutations are thought to be an important contributor to RCC-initiation but are not correlated with progression (32). The zebrafish vhl mutant expression microarray data reflect the very earliest changes in kidney cell proliferation and transformation; the fact that full-blown carcinomas retain increased E2F1 expression suggests a growth advantage, for at least early AJCC stages of RCC. The most aggressive tumors (AJCC stages III and IV) have lost nuclear E2F1 expression to a certain extent and might reflect a molecular switch triggering cellular cascade events leading to a more aggressive tumor type. This is supported by the fact that E2F1 accumulation plays a role in the DNA damage response by prolonging G1-arrest which leads to senescence of cells(40). Loss of the E2F1 locus on chromosome 20q11 has not been described in RCC; however genomic instability in later carcinomas might reflect loss of E2F1. We demonstrate that in RCCs, E2F1 significantly correlates to its known target p27(30). Therefore it was not surprising that high p27 expression was borderline significant with regards to a smaller tumor diameter (Pearson Correlation =-0.2, p=0.06). Collectively, these data might indicate that E2F1 prevents tumor progression through p27-mediated senescence. Conventionally, parameters such as AJCC stage, histological Fuhrman grade and patient

63

performance status have been used to determine the prognosis of patients with primary RCC. TMAs are excellent tools to investigate expression patterns of molecular markers and their possible value in predicting patient outcome. An improved prognostic model would be useful for patient counseling, planning follow-up and selecting patients for additional treatment (38). In this study we show that primary RCCs with decreased nuclear E2F1 expression is associated with a larger tumor diameter, a poorer AJCC stage, reduced DFS and OS. Primary RCC tumors with <25% nuclear E2F1 could be used as a poor risk factor in our study, whereas ≥25% was protective. Accordingly, in clinical studies high expression of E2F1 is associated with increased DFS and OS in other distinct cancer types, such as squamous cell carcinoma of the tongue (41), oesophageal adenocarcinoma (33), bladder (42) and gastric cancer (43). In contrast, high E2F1 expression is correlated with decreased DFS and/or poor OS in melanoma (44), glioblastomas (45) and gastrointestinal stromal tumors (46). This apparent contradiction might be attributable to the more squamous cell subtype, the heterogeneity of the used patient cohorts, and the low number of included patients. It is likely that the regulation of E2F1 expression, which is orchestrated by multiple factors, is cell-type specific and determines the balance of the proliferation/apoptotic program (47). In conclusion, this study demonstrates that E2F1 expression is inhibited by pVHL in vitro and in vivo, is independent of HIF1a, and affects cell cycle progression of RCC cells. Furthermore we show that E2F1 expression is a novel predictor for RCC patients’ survival and might be potentially useful for clinical management.

ACK noW le D g M e nTs

We would like to thank Maryam Saghafiam, Emine Bolat, Minh Nguyen, Sabrina Elshof, Sander Basten, Susanne van Schelven, Sylvia E.C. van Beersum and Cristel Snijckers for technical assistance, Maartje Los for collecting renal cell carcinoma tissue, and Marius Ueffing and René Medema for helpful discussions.

g RAnT s u P P oRT

This work is supported by the Netherlands Organization for Scientific Research (VIDI award 016.066.354 to R.H. Giles), Alexandre Suerman UMC Stipendiums, the Dutch Cancer Society (KWF UU2006-3565 to E.E. Voest ), UMC Utrecht Focus & Massa (to M.C. Verhaar and A. de Bruin), and the European Community’s Seventh Framework Programme FP7/2009 under grant agreement no. 241955; SYSCILIA (to R. Roepman. and R.H. Giles).


64


(1) Jemal A, Siegel R, Xu J, Ward E. Cancer statistics, 2010. CA Cancer J Clin 2010 September;60(5):277-300.

(2) Kaelin WG, Jr. The von Hippel-Lindau tumour suppressor protein: O2 sensing and cancer. Nat Rev Cancer

2008 November;8(11):865-73.

(3) Hsu T. Complex cellular functions of the von Hippel-Lindau tumor suppressor gene: insights from model

organisms. Oncogene 2011 September 26.

(4) Young AP, Schlisio S, Minamishima YA, Zhang Q, Li L, Grisanzio C et al. VHL loss actuates a HIF-independent

senescence programme mediated by Rb and p400. Nat Cell Biol 2008 March;10(3):361-9.

(5) Dimri GP, Itahana K, Acosta M, Campisi J. Regulation of a senescence checkpoint response by the E2F1

transcription factor and p14(ARF) tumor suppressor. Mol Cell Biol 2000 January;20(1):273-85.

(6) Ohtani K, Iwanaga R, Nakamura M, Ikeda M, Yabuta N, Tsuruga H et al. Cell growth-regulated expression

of mammalian MCM5 and MCM6 genes mediated by the transcription factor E2F. Oncogene 1999 April

8;18(14):2299-309.

(7) Tsantoulis PK, Gorgoulis VG. Involvement of E2F transcription factor family in cancer. Eur J Cancer 2005

November;41(16):2403-14.

(8) Biswas AK, Johnson DG. Transcriptional and Nontranscriptional Functions of E2F1 in Response to DNA

Damage. Cancer Res 2012 January 1;72(1):13-7.

(9) van RE, Voest EE, Logister I, Bussmann J, Korving J, van Eeden FJ et al. von Hippel-Lindau tumor suppressor

mutants faithfully model pathological hypoxia-driven angiogenesis and vascular retinopathies in zebrafish. Dis

Model Mech 2010 May;3(5-6):343-53.

(10) van RE, Voest EE, Logister I, Korving J, Schwerte T, Schulte-Merker S et al. Zebrafish mutants in the von

Hippel-Lindau tumor suppressor display a hypoxic response and recapitulate key aspects of Chuvash

polycythemia. Blood 2009 June 18;113(25):6449-60.

(11) Lolkema MP, Mans DA, Snijckers CM, van NM, van BM, Voest EE et al. The von Hippel-Lindau tumour

suppressor interacts with microtubules through kinesin-2. FEBS Lett 2007 October 2;581(24):4571-6.

(12) Krieg M, Haas R, Brauch H, Acker T, Flamme I, Plate KH. Up-regulation of hypoxia-inducible factors HIF-

1alpha and HIF-2alpha under normoxic conditions in renal carcinoma cells by von Hippel-Lindau tumor

suppressor gene loss of function. Oncogene 2000 November 16;19(48):5435-43.

(13) Sowter HM, Raval RR, Moore JW, Ratcliffe PJ, Harris AL. Predominant role of hypoxia-inducible transcription

factor (Hif)-1alpha versus Hif-2alpha in regulation of the transcriptional response to hypoxia. Cancer Res 2003

October 1;63(19):6130-4.

(14) Giles RH, Lolkema MP, Snijckers CM, Belderbos M, van der GP, Mans DA et al. Interplay between VHL/

HIF1alpha and Wnt/beta-catenin pathways during colorectal tumorigenesis. Oncogene 2006 May

18;25(21):3065-70.

(15) Araki K, Nakajima Y, Eto K, Ikeda MA. Distinct recruitment of E2F family members to specific E2F-binding sites

mediates activation and repression of the E2F1 promoter. Oncogene 2003 October 23;22(48):7632-41.

(16) Li J, Ran C, Li E, Gordon F, Comstock G, Siddiqui H et al. Synergistic function of E2F7 and E2F8 is essential

for cell survival and embryonic development. Dev Cell 2008 January;14(1):62-75.

(17) Coene KL, Mans DA, Boldt K, Gloeckner CJ, van RJ, Bolat E et al. The ciliopathy-associated protein homologs

RPGRIP1 and RPGRIP1L are linked to cilium integrity through interaction with Nek4 serine/threonine kinase.

Hum Mol Genet 2011 September 15;20(18):3592-605.

(18) Boldt K, Mans DA, Won J, van RJ, Vogt A, Kinkl N et al. Disruption of intraflagellar protein transport

in photoreceptor cilia causes Leber congenital amaurosis in humans and mice. J Clin Invest 2011 June

1;121(6):2169-80.


65

(19) Kroeze SG, Vermaat JS, van BA, van Melick HH, Voest EE, Jonges TG et al. Expression of nuclear

FIH independently predicts overall survival of clear cell renal cell carcinoma patients. Eur J Cancer 2010

December;46(18):3375-82.

(20) van RE, Santhakumar K, Logister I, Voest E, Schulte-Merker S, Giles R et al. A Zebrafish Model for VHL and

Hypoxia Signaling. Methods Cell Biol 2011;105:163-90.

(21) Ma Y, Croxton R, Moorer RL, Jr., Cress WD. Identification of novel E2F1-regulated genes by microarray. Arch

Biochem Biophys 2002 March 15;399(2):212-24.

(22) Salceda S, Caro J. Hypoxia-inducible factor 1alpha (HIF-1alpha) protein is rapidly degraded by the ubiquitin-

proteasome system under normoxic conditions. Its stabilization by hypoxia depends on redox-induced changes.

J Biol Chem 1997 September 5;272(36):22642-7.

(23) Maxwell PH, Wiesener MS, Chang GW, Clifford SC, Vaux EC, Cockman ME et al. The tumour

suppressor protein VHL targets hypoxia-inducible factors for oxygen-dependent proteolysis. Nature 1999 May

20;399(6733):271-5.

(24) O’Connor DJ, Lu X. Stress signals induce transcriptionally inactive E2F-1 independently of p53 and Rb.

Oncogene 2000 May 11;19(20):2369-76.

(25) Iliopoulos O, Ohh M, Kaelin WG, Jr. pVHL19 is a biologically active product of the von Hippel-Lindau gene

arising from internal translation initiation. Proc Natl Acad Sci U S A 1998 September 29;95(20):11661-6.

(26) Schoenfeld A, Davidowitz EJ, Burk RD. A second major native von Hippel-Lindau gene product, initiated

from an internal translation start site, functions as a tumor suppressor. Proc Natl Acad Sci U S A 1998 July

21;95(15):8817-22.

(27) Eng C, Crossey PA, Mulligan LM, Healey CS, Houghton C, Prowse A et al. Mutations in the RET

proto-oncogene and the von Hippel-Lindau disease tumour suppressor gene in sporadic and syndromic

phaeochromocytomas. J Med Genet 1995 December;32(12):934-7.

(28) Kamura T, Koepp DM, Conrad MN, Skowyra D, Moreland RJ, Iliopoulos O et al. Rbx1, a component of the VHL

tumor suppressor complex and SCF ubiquitin ligase. Science 1999 April 23;284(5414):657-61.

(29) Miyauchi Y, Kato M, Tokunaga F, Iwai K. The COP9/signalosome increases the efficiency of von Hippel-

Lindau protein ubiquitin ligase-mediated hypoxia-inducible factor-alpha ubiquitination. J Biol Chem 2008 June

13;283(24):16622-31.

(30) Wang C, Rauscher FJ, III, Cress WD, Chen J. Regulation of E2F1 function by the nuclear corepressor KAP1. J

Biol Chem 2007 October 12;282(41):29902-9.

(31) Kim WY, Kaelin WG. Role of VHL gene mutation in human cancer. J Clin Oncol 2004 December

15;22(24):4991-5004.

(32) Nordstrom-O’Brien M, van der Luijt RB, van RE, van den Ouweland AM, Majoor-Krakauer DF, Lolkema MP et

al. Genetic analysis of von Hippel-Lindau disease. Hum Mutat 2010 May;31(5):521-37.

(33) Evangelou K, Kotsinas A, Mariolis-Sapsakos T, Giannopoulos A, Tsantoulis PK, Constantinides C et al. E2F-

1 overexpression correlates with decreased proliferation and better prognosis in adenocarcinomas of Barrett

oesophagus. J Clin Pathol 2008 May;61(5):601-5.

(34) Bishop T, Lau KW, Epstein AC, Kim SK, Jiang M, O’Rourke D et al. Genetic analysis of pathways regulated by

the von Hippel-Lindau tumor suppressor in Caenorhabditis elegans. PLoS Biol 2004 October;2(10):e289.

(35) Pause A, Lee S, Lonergan KM, Klausner RD. The von Hippel-Lindau tumor suppressor gene is required for cell

cycle exit upon serum withdrawal. Proc Natl Acad Sci U S A 1998 February 3;95(3):993-8.

(36) Roe JS, Kim H, Lee SM, Kim ST, Cho EJ, Youn HD. p53 stabilization and transactivation by a von Hippel-

Lindau protein. Mol Cell 2006 May 5;22(3):395-405.

(37) Turowska O, Nauman A, Pietrzak M, Poplawski P, Master A, Nygard M et al. Overexpression of E2F1 in clear

cell renal cell carcinoma: a potential impact of erroneous regulation by thyroid hormone nuclear receptors.

Thyroid 2007 November;17(11):1039-48.


66

(38) Yamasaki L, Jacks T, Bronson R, Goillot E, Harlow E, Dyson NJ. Tumor induction and tissue atrophy in mice

lacking E2F-1. Cell 1996 May 17;85(4):537-48.

(39) Patard JJ, Fergelot P, Karakiewicz PI, Klatte T, Trinh QD, Rioux-Leclercq N et al. Low CAIX expression and

absence of VHL gene mutation are associated with tumor aggressiveness and poor survival of clear cell renal

cell carcinoma. Int J Cancer 2008 July 15;123(2):395-400.

(40) Pickering MT, Stadler BM, Kowalik TF. miR-17 and miR-20a temper an E2F1-induced G1 checkpoint to

regulate cell cycle progression. Oncogene 2009 January 8;28(1):140-5.

(41) Kwong RA, Nguyen TV, Bova RJ, Kench JG, Cole IE, Musgrove EA et al. Overexpression of E2F-1 is

associated with increased disease-free survival in squamous cell carcinoma of the anterior tongue. Clin

Cancer Res 2003 September 1;9(10 Pt 1):3705-11.

(42) Rabbani F, Richon VM, Orlow I, Lu ML, Drobnjak M, Dudas M et al. Prognostic significance of transcription

factor E2F-1 in bladder cancer: genotypic and phenotypic characterization. J Natl Cancer Inst 1999 May

19;91(10):874-81.

(43) Lee J, Park CK, Park JO, Lim T, Park YS, Lim HY et al. Impact of E2F-1 expression on clinical outcome of

gastric adenocarcinoma patients with adjuvant chemoradiation therapy. Clin Cancer Res 2008 January

1;14(1):82-8.

(44) Alla V, Engelmann D, Niemetz A, Pahnke J, Schmidt A, Kunz M et al. E2F1 in melanoma progression and

metastasis. J Natl Cancer Inst 2010 January 20;102(2):127-33.

(45) Alonso MM, Fueyo J, Shay JW, Aldape KD, Jiang H, Lee OH et al. Expression of transcription factor E2F1 and

telomerase in glioblastomas: mechanistic linkage and prognostic significance. J Natl Cancer Inst 2005

November 2;97(21):1589-600.

(46) Haller F, Gunawan B, von HA, Schwager S, Schulten HJ, Wolf-Salgo J et al. Prognostic role of E2F1

and members of the CDKN2A network in gastrointestinal stromal tumors. Clin Cancer Res 2005 September

15;11(18):6589-97.

(47) Hallstrom TC, Mori S, Nevins JR. An E2F1-dependent gene expression program that determines the balance

between proliferation and cell death. Cancer Cell 2008 January;13(1):11-22.


supplementary Table 1: Overview of siRNAs used in this studysupplementary Table 2: Overview of primers used for RT-qPCR analysissupplementary Table 3: Overview of VHL shRNAs used in this studysupplementary Table 4: Mass spectrometric analysis and Scaffold analysis of SF-TAP-tagged pVHL (isoform 2). Shown are all unique proteins that were identified in 2 VHL-TAP analyses. For all proteins, Entrez gene symbol, Entrez gene full name, sequence coverage per TAP analysis, number of unique peptides identified per TAP analysis and a sum of unique peptide identified in both analyses, UniProt ID, UniProt Function and the number of TAPs in which the proteins have been identified are shown. Proteins identified in control SF-TAP experiments, were manually removed from this analysis.

supplementary tables will be published online in the near future.

67Regulation of E2F1 by VHL tumor suppressor protein predicts survival in RCC patients Chapter 4


Chapter 5 Two-protein signature of novel serological markers apolipoprotein-A2 and serum amyloid alpha predicts prognosis in patients with metastatic renal cell cancer and improves the currently used prognostic survival models

Joost S Vermaat, Ingeborg van der Tweel, Niven Mehra, Stefan Sleijfer, John B Haanen, Jeanine M Roodhart, Judith Y Engwegen, Catharina M Korse, Marlies H Langenberg, Wim Kruit, Gerard Groenewegen, Rachel H Giles, Jan H Schellens, Jos H Beijnen, Emile E Voest

Annals of Oncology 2010 July; 21(7):1472-81

70 Part II Proteomics


Our manuscript demonstrates that combining proteomics-based screening with subsequent validation by conventional and commercially available protein quantification methods in serum of a large cohort of mRCC patients can yield novel biomarkers associated to survival. The impact of our findings is highly clinically relevant as novel proteins, Apolipoprotein A2 (ApoA2) and Serum Amyloid Alpha (SAA), accurately predict patients’ survival, yielding a two-protein signature that significantly better classifies patients into three prognostic groups differing in survival rates than the currently used Motzer (MSKCC) criteria. However, when the two-protein signature is combined with the MSKCC criteria, an even better prognostic model is obtained shifting 38% of the mRCC patients into another risk group compared to the original MSKCC assignment. The quantification of these serological proteins ApoA2 and SAA can easily and inexpensively be implemented in any well-equipped hospital in an objective, quantitative and non-invasive manner.

71ApoA2 and SAA predicts prognosis in mRCC patients (and improves the currently used risk models) Chapter 5

AB sTRACT

Background and objectiveIn metastatic renal cell cancer (mRCC) the Memorial Sloan Kettering Cancer Center (MSKCC) risk model is widely used for clinical trial design and patient management. To improve prognostication, we applied proteomics to identify novel serological proteins associated with overall survival (OS).

Patients and MethodsSera from 114 metastastic RCC patients were screened by SELDI-TOF mass spectrometry (MS). Identified proteins were related to OS. Three proteins were subsequently validated with ELISAs and immunoturbidimetry. Prognostic models were statistically bootstrapped to correct for overestimation.

ResultsSELDI-TOF MS detected ten proteins associated with OS. Of these, apolipoprotein A-II (ApoA2), serum amyloid alpha (SAA), and transthyretin were validated for their association with OS (p = 5.5x10-9, p = 1.1x10-7 and p = 0.0004, respectively). Combining ApoA2 and SAA yielded a prognostic two-protein signature (AIC = 732, p = 5.2x10-7). Including previously identified prognostic factors, multivariable Cox regression analysis revealed ApoA2, SAA, LDH, performance status and number of metastasis sites as independent factors for survival. Using these five factors, categorization of patients into three risk groups generated a novel protein-based model predicting patient prognosis (AIC = 713, p = 4.3x10-11) more robustly than the MSKCC model (AIC = 729, p = 1.3x10-7). Applying this protein-based model instead of the MSKCC model would have changed the risk group in 38% of the patients.

Conclusion Proteomics and subsequent validation yielded two novel prognostic markers and survival models which improved prediction of OS in mRCC patients over commonly used risk models. Implementation of these models has the potential to improve current risk stratification, although prospective validation will still be necessary.

72

AB B R eVIATIon s (m)RCC; (metastatic), Renal Cell Cancer, MSKCC; Memorial Sloan-Kettering Cancer Center,OS; overall survival, PFS; Progression Free Survival, ORR; Objective Response Rate, MS; Mass Spectrometry, SELDI-TOF; Surface Enhanced Laser Desorption Ionization Time-Of-Flight, CI; Confidence Interval, IQR; InterQuartile Range, ApoA2; Apolipoprotein A-II, SAA; Serum Amyloid Alpha, TTR; Transthyretin, ELISA; Enzyme Linked Immunosorbent Assay, Da; Dalton, AIC; Akaike’s Information Criteria, KPS; Karnofsky Performance Status, mTOR; mammalian, Target Of Rapamycin, VEGF; Vascular Endothelial Growth Factor, TDT; time from diagnosis to start treatment, LDH; Lactate Dehydrogenase

I nTRoDuCTIon

The worldwide incidence of renal cell carcinoma (RCC) is 209,000, with newly diagnosed cases in 2006 (1). Metastasized disease is present in 25-30% of RCC patients (mRCC) at diagnosis (1). Interferon-alpha and interleukin-2 have been used as first-line mRCC treatment for decades (2), but with modest patient benefit (1). Recently, new treatment modalities including tyrosine kinase inhibitors (TKIs) targeting the Vascular Endothelial Growth Factor (VEGF)-receptor, mammalian target of rapamycin (mTOR)-inhibitors and the addition of bevacizumab (humanized monoclonal VEGF-antibody) to interferon-alpha have improved patient outcomes, although associated with substantial toxicities (3-6). Given the highly variable disease course and various treatment drawbacks, it is important to a-priori select patients who clearly benefit from therapy. To date, the Memorial Sloan-Kettering Cancer Center (MSKCC) risk model (2) is the most widely used prognostic classification system for patient management and clinical trial design. This model originally included five prognosticators of poor survival: low Karnofsky performance status (KPS <80%), elevated lactate dehydrogenase (LDH), decreased hemoglobin, raised ‘corrected’ calcium and absence of prior nephrectomy (7). Using these factors, patients can be categorized into three prognostic groups (favorable, intermediate or poor) differing in overall survival. In 2002, Motzer et al. refined this model interchanging absence of prior nephrectomy for time between diagnosis and start of systemic treatment and are currently widely used as the MSKCC criteria for mRCC patient prognostication (2). At the Cleveland Clinic, these MSKCC criteria were further validated and expanded with the number of metastatic sites (≤1 versus ≥2) and prior radiotherapy (8). Although, recent years have demonstrated the value of MSKCC criteria in patient management, this model could be refined. The objective of this large retrospective cohort study among mRCC patients was to associate novel serum proteins with survival by applying a proteomic-based approach, Surface-Enhanced Laser Desorption Ionization Time-Of-Flight Mass Spectrometry (SELDI-TOF MS). Proteomic results were validated using ELISA’s and immunoturbidimetry. We identified a protein signature that could accurately predict survival. Finally, we investigated whether addition of novel proteins significantly improved the predictive accuracy of the traditional risk factors in the MSKCC model.

Part II Proteomics

73


sample Collection and survivalSerum samples from 114 consecutive RCC patients with metastasis were collected (2001-2006) at three Dutch cancer institutes: the University Medical Center Utrecht (n = 63), the Netherlands Cancer Institute (n = 29) and the Erasmus Medical Center Rotterdam (n = 22). Each patient provided informed consent before blood withdrawal as stipulated by the three institutional ethical boards and were enrolled in different study protocols. Blood was obtained by venapuncture, coagulated at room temperature (0.5-6 hours), centrifuged (1500-1900g) and frozen (71% at -80oC; 29% at -30oC). In 16% of the study cohort (18 mRCC patients) blood samples were collected before primary tumour resection. Blood of 81 (71%) patients was drawn at start of first line systemic treatment. In 7% of the cases (8 patients) blood samples were gathered at start of second line therapy. Clinical features, disease characteristics and baseline biochemical parameters were gathered from medical records, including MSKCC risk factors. Overall survival was defined as time from blood collection to date of death or last follow-up.

screening and Validation Methods: serum Protein Profiling Serum was screened to discover protein profiles using the semi-quantitative SELDI-TOF MS (Ciphergen Biosystems Inc., Freemont, CA, USA) as reported by Won et al. (9) and validated in our laboratory (10). Detailed assay procedures are described in Appendix-A. Subsequently, apolipoprotein A-II (ApoA2) was quantified by immunoturbidimetry using the nephelometer BN-II system (Dade Behring Inc, Newark, USA). ELISA kits quantified serum amyloid alpha (SAA; Tridelta Development, Kildare, Ireland) and transthyretin (TTR; Immundiagnostik AG, Bensheim, Germany) levels according to manufacturer’s instructions.

statistics and Bioinformatics ProteinChip Software package, version 3.1 (Ciphergen Biosystems), was used to analyze entire protein profiles. Spectra were baseline-subtracted and normalized to the total ion current. Spectra with normalization factors >2.00 or <0.50 were excluded. SELDI-TOF MS intensities were authenticated as mass peaks when signal to noise ratio ≥5, mass accuracy window ≤1%, and presence ≥30% of all spectra. Mass peak intensities were used as continuous variables for association with survival. Statistical significance was considered as p<0.05. Univariable Cox Proportional Hazards (PH) regression analysis was applied to associate validated proteins and MSKCC criteria to survival. Subsequently, to derive a model based on validated protein-signatures, all variables were dichotomized and multivariable Cox PH regression analysis was applied with backward stepwise selection. Multivariable Cox PH regression analysis was used to test whether novel prognostic proteins improved the accuracy of the MSKCC model and subsequently a backward stepwise selection procedure was applied to design a novel protein-based model. Akaike’s Information Criteria (AIC) and R2 were used to compare the accuracy of survival prediction of the MSKCC model with our novel protein-based models (11;12). The lower the AIC the better the predictive model fits the data. Kaplan-Meier curves described the predictive accuracy of various prognostic models. To prevent over-optimistic predictive accuracy we applied bootstrapping techniques to internally validate and estimate the optimism R2 in our prognostic models to correct for overestimation

ApoA2 and SAA predicts prognosis in mRCC patients (and improves the currently used risk models) Chapter 5

74

(11;13), repeating the entire estimation process in each bootstrap sample 500 times. R2 is a measure for the proportion variability explained by the model and the predictive strength of the model. The optimism of our novel models is the bias due to overfitting of the data. The R2 of the original model is adjusted by the optimism to obtain the corrected R2, which better reflects the estimated probability of our findings when applied in a subsequent validation studies. Statistical analysis was performed using SPSS (version 15.0) and R software (version 2.3.1).

R e s u lTs

Patient CharacteristicsDetailed characteristics of 114 patients are depicted in Table 1. All patients were planned to receive interferon-based treatment in clinical trials. However, some patients were only screened, but did not receive systemic treatment due to ineligibility or performance status. In total 89 patients (78%) received first-line systemic therapy for metastatic RCC, principally interferon-based therapy (83 patients, 73%). Blood collection was generally performed before start of first systemic therapy (81 patients, 71%); however in 8 patients blood was sampled after first-line treatment with interferon but before second line therapy. The other 25 patients (22%) received either palliative care (21 patients) or were followed without interventions (4 patients) because of stable disease. After first-line treatment, 39 patients (34%) received second-line systemic therapy consisting of small molecules. Table 2 shows the various treatment modalities. The median overall survival time

Table 2 - Treatment Modalities

No. (%)

first line systemic therapyInterferon-based * 83 (73%)Sunitinib 2 (2%)Interleukin-2 2 (2%)Tubulin-antagonist: ABT-751 2 (2%)Total 89 (79%)

second line systemic therapy Sunitinib 11 (10%)Sorafenib 8 (7%)Interleukin-2 6 (5%)Cediranib * 5 (4%)Tubulin-antagonist: ABT-751 4 (3%)Anti-EGFR 2 (2%)Angiostatin 1 (1%)Telatinib + Bevacizumab 1 (1%)Anti-Interleukin-6 1 (1%)Total 39 (34%)

* With or without Gefinitib† With or without Thalidomide

Part II Proteomics

75

Table 1 - Patients Characteristics: Categorical

Patients Median Survival

Factor No. % Death (%) (months) 95% CI p*

overall 114 100 81 15.4 11.3 - 19.6 ―

sexMale 76 67 85 14.9 10.2 - 19.6 ―Female 36 33 71 20.3 5.3 - 35.3 0.2

Karnofsky Performance status

> 80 83 73 77 18.2 13.9 - 22.2 ―≤ 80 31 27 90 5.4 3.7 - 7.1 <0.001

Prior nephrectomy †

Yes 78 68 78 17.2 12.7 - 21.8 ―No 36 32 86 10.2 7.6 - 12.8 0.2

Prior radiotherapyNo 107 94 80 15.7 10.9 - 20.5 ―Yes 7 6 86 11.2 1.8 - 20.7 0.6

lung metastasisNo 40 35 80 11.2 6.4 - 16.1 ―Yes 74 64 81 17.0 13.7 - 20.3 0.07

lymph node metastasis ‡No 63 55 73 20.4 14.6 - 26.2 ―Yes 51 45 90 10.6 6.2 - 15.0 0.001

Hepatic metastasisNo 85 75 79 17.2 12.4 - 22.1 ―Yes 29 25 86 9.0 0.8 - 17.3 0.06

osseous metastasisNo 74 65 76 16.1 12.1 - 20.1 ―Yes 40 35 90 11.6 3.7 - 19.6 0.1

Cns metastasisNo 111 97 80 15.7 11.1 - 20.4 ―Yes 3 3 100 6.5 0 - 13.0 0.1

Pancreatic metastasisNo 106 93 80 15.7 11.3 - 20.1 ―Yes 8 7 87 11.1 7.0 - 15.2 0.6

Adrenal metastasisNo 101 89 81 15.4 10.3 - 20.5 ―Yes 13 11 77 16.1 0 - 42.2 0.4

other metastatic sites §No 107 94 80 16.1 11.3 - 21.0 ―Yes 7 6 100 6.5 5.5 - 19.0 0.2

HistologyClear Cell 89 76 76 17.4 13.2 - 21.6 ―Other ―― 16 14 100 8.5 3.4 - 13.5 0.02Unknown ¶ 9 8 89 3.4 3.3 - 3.5 ―

* Log Likelihood Ratio test. † According to trial design 16% patients underwent nephrectomy in combination with IFN treatment within 1 months after blood collection. ‡ Primarily retroperitoneal lymph nodes, as well as mediastinal lymph nodes. § Abdominal and skin metastasis. II Papillary,chromofobe, sarcomatoid and mixed types. ¶ 6 patients diagnosed based on clinical evidence and 3 patients diagnosed with renal cell cancer, but histology was not further specified. # Comparison of Clear Cell versus other and unknown histologies.


76

(continued) Table 1 - Patients Characteristics: Continuous

Patients Median Survival(months)Factor No. % Death (%) 95% CI p*

Age at trial entry, years

Mean 60.0

SD 10.1

Median 60.4 0.9

IQR 54 - 68

≤ 60 years 55 48 76 12.2 5.3 - 19.1

> 60 years 59 52 85 16.1 12.6 - 19.7 0.8

Time Diagnosis to trial entry, months

Mean 27.6

SD 47.2

Median 6.9 0.07

IQR 1.4 - 32.4

> 12 months

50 44 76 19.7 13.5 - 29.1

≤ 12 months 64 56 84 11.1 10.0 - 21.5 0.07

no. of metastatic sites

Median 2

Range 1 - 4

0 or 1 36 32 61 31.0 8.6 - 53.4

≥ 2 78 68 90 11.5 11.3 - 19.6 < 0.001

Albumin, g/dl †

Mean 3.7

SD 0.7

Median 3.9 < 0.001

IQR 3.4 - 4.2

< LLN 35 31 89 5,5 3.3 - 7.7

≥ LLN 79 69 77 20.3 16.4 - 24.2 < 0.001

lactate Dehydrogenase, Iu/l ‡

Mean 490

SD 480

Median 399 < 0.001

IQR 319 - 537

≤ 1.5 x ULN 105 92 79 17.0 14.0 - 20.1

> 1.5 x ULN 9 8 100 3.3 3.2 - 3.3 < 0.001

Part II Proteomics

77

(continued) Table 1 - Patients Characteristics: Continuous

Patients Median Survival(months)

Factor No. % Death (%) 95% CI p*

Corrected Calcium, mg/dl §

Mean 9.4

SD 1.3

Median 9.2 0.1

IQR 8.3 - 9.4

≤ 10mg/dL 104 91 79 17.0 13.6 - 20.4

> 10 mg/dL 10 9 100 4.0 0 - 12.5 0.007

Hemoglobin, g/dl wMales

Mean 12.8

SD 2.2

Median 13.2 0.001

IQR 11.0 - 14,3

< LLN 33 43 94 9.6 5.1 - 14.1

≥ LLN 43 57 79 18.0 14.7 - 21.4 0.008

females

Mean 12.3

SD 1.9

Median 12.5 0.01

IQR 10.8 - 13.9

< LLN 19 50 89 5.7 3.1 - 8.4

≥ LLN 19 50 53 27.7 8.9 - 32.1 0.003

Total

Mean 12.6

SD 2.1

Median 12.7 < 0.001

IQR 11.0 - 14.2

< LLN 52 46 92 7.0 3.1 - 10.9

≥ LLN 62 54 71 20.4 11.3 - 19.6 < 0.001

Abbreviations: SD, standard deviation; IQR, Interquartile Range; LLN, lower limit of laboratory’s reference range; ULN, upper limit of laboratory’s reference range.

* Log-rank test for categorical form of the variable and Cox proportional hazards model for continuous form of the variable.† Albumin: lower limit of reference range was 3.5 g/dL.‡ LDH: 450 IU/L was used as ULN.§ Corrected Calcium = total calcium - 0.707 (albumin - 3.4).ІІ Hemoglobin: lower limits of reference range for men and women were 13.5 and 12.0 g/dL, respectively.


78

was 15.4 months (95% CI: 11.3 – 19.6) with 1-, 2- and 3-year cumulative survival proportions of 57%, 33% and 24%, respectively. At fi nal analysis, 22 patients (19%) were still alive.

sceening Method: Protein Profiles Associated with survivalProtein profi les of all patients were generated in two independent experiments and individual serum samples were assessed fi ve times. The fi rst experiment was performed in duplicate and resulted in 25 distinct mass peaks. To determine the robustness of the obtained data, the complete sample set was re-analyzed in triplicate. Freshly thawed serum samples were measured within three weeks to minimize laser intensity variation. This second experiment detected 31 distinct mass peaks (supplemental Table 1) in which 25 mass peaks were identical to the fi rst experiment, indicating that the data obtained by SELDI-TOF MS were highly reproducible. Of these detected mass peaks 17 are known proteins (14). Univariable Cox regression analysis for all 31 mass peaks identifi ed 10 proteins related to survival. For illustration, representative protein spectra of mass peaks (zoomed-in) are shown for two patients with poor and favourable survival (figure 1). Mass peaks with corresponding mass/charge (M/z) ratios of 8.6, 11.7 and 13.7 kDa were selected for further investigation in the subsequent analysis, because; (1) they were strongly associated to survival, (2) they are known proteins (14;15), namely apolipoprotein-A2 (ApoA2), serum amyloid

figure 1 Two patients are shown with poor survival and two patients with long-term survival. The three depicted ion signals were validated by conventional techniques. ApoA2 and SAA were both incorporated into novel protein-based risk models.

figure 1 - Zoomed-in regions of mass spectrometry profi les generated by the screening method SELDI-TOF MS

InTe

ns

ITY

M/z M/z

Part II Proteomics

79

Tabl

e 3

- Va

lidat

ion

Met

hod

(ELI

SA

s &

Imm

unot

urbi

dim

etry

): U

niva

riat

e S

urvi

val A

naly

sis

of N

ovel

Pro

tein

s

Con

tinuo

us F

orm

Cat

egor

y Fo

rm

Par

amet

er

Est

imat

eS

Ep*

Cut

poin

t Use

d A

ICP

*H

azar

d R

atio

95%

CI

Apo

lipro

tein

A-I

I (A

poA

2)C

ontin

uous

For

m―0

.007

0.00

013

5.5

x 10

-9C

ateg

ory:

Ter

tiles

(mg/

L)73

92.

2 x

10-5

Mea

n (m

g/L)

277

Low

est T

ertil

e<

24

83.

52.

1 - 6

.1S

D91

Inte

rmed

iate

Ter

tile

248

- 30

92.

01.

2 - 3

.4M

edia

n27

7H

ighe

st T

ertil

e †

> 3

09

――

IQR

228

- 324

ser

um A

myl

oid

Alp

ha (

sA

A)

Con

tinuo

us F

orm

0.00

160.

0002

71.

1 x

10-7

Cat

egor

y: T

ertil

es(n

g/m

L)72

91.

2 x

10-7

Mea

n (n

g/m

L)26

5Lo

wes

t Ter

tile

†<

19.

2―

―S

D3

80

Inte

rmed

iate

Ter

tile

19.2

- 21

72.

01.

2 - 3

.4M

edia

n47

Hig

hest

Ter

tile

> 2

174.

62.

7 - 7

.9IQ

R16

- 4

66

Tran

sthy

retin

(TT

R)

Con

tinuo

us F

orm

―0.0

40.

013

0.00

04

Cat

egor

y: T

ertil

es(n

g/m

L)76

80.

0003

Mea

n (n

g/m

L)17

.0Lo

wes

t Ter

tile

< 1

2.2

2.8

1.7

- 4.7

SD

9.9

Inte

rmed

iate

Ter

tile

12.2

- 19

.71.

50.

9 - 2

.6M

edia

n16

.0H

ighe

st T

ertil

e †

> 1

9.7

――

IQR

11.3

- 22

.1*

Like

lihoo

d ra

tio te

st fo

r uni

varia

ble

surv

ival

ana

lysi

s.†

Tert

ile u

sed

as re

fere

nce.


80

alpha (SAA) and transthyretin (TTR) respectively, and (3) because conventional quantifi cation techniques for these proteins were available for subsequent validation.

Validation Method: ApoA2, sAA and TTR Associated with survivalConcentrations of ApoA2, SAA and TTR were measured by conventional antibody-directed quantifi cation methods. The continuous variables ApoA2, SAA and TTR signifi cantly predicted patient survival (p = 5.5 x 10-9, 1.1 x 10-7 and 0.0004, respectively, Table 3). The positive regression coeffi cient for SAA indicated that higher SAA levels were associated with reduced survival, whereas negative regression coeffi cients for ApoA2 and TTR designated prolonged survival for higher concentrations.

figure 2 – (A) two-protein signature based on ApoA2 and SAA, (B) the MSKCC risk model and (C) the novel protein-based risk model comprising KPS (≤80%), ≥2 metastatic sites, LDH (>1.5x ULN), ApoA2 (<309 mg/L) and SAA (>19.2 ng/mL) as risk factors. Both novel risk models predicted patients’ survival more accurately than the MSKCC model.

figure 2 - Survival of 114 advanced RCC patients described by risk models

Part II Proteomics

A B

C

81

Two-Protein signature Accurately Predicts survivalIncluding the three validated proteins in multivariable Cox PH regression analysis revealed a prognostic model with ApoA2 and SAA. TTR was excluded because it did not significantly improve the ApoA2/SAA model. When SAA levels were not elevated, low quantities of ApoA2 were highly prognostic for poor survival. Therefore we used the combination of the lowest two tertiles for ApoA2 and the highest two tertiles for SAA as risk factors for poor survival. Subsequently, ApoA2 and SAA were stratified into three risk groups, a favourable risk group with no risk factors (ApoA2>309 mg/L and SAA<19.2 ng/mL), an intermediate group with one risk factor (ApoA2≤309 mg/L or SAA≥19.2 ng/mL), and a poor survival group with both risk factors (Table 3). Accordingly, combining ApoA2 and SAA as a two-protein signature significantly predicted survival (figure 2A, AIC = 732, R2 = 0.23, p = 5.2 x 10-10). Median survival times for favourable, intermediate and poor prognostic groups were 58.3, 22.3 and 8.8 months with overall survival rates of 52%, 21% and 7%, respectively. sAA and ApoA2 improve the MsKCC model We examined the relation between several pretreatment features and survival with univariable Cox regression analysis (8). Ten risk factors were significantly associated with survival (Table 1). Poor KPS, elevated LDH, low hemoglobin, reduced albumin, presence of metastasis, ≥2 metastasis sites, histology, increased corrected calcium and the presence of lymph node metastasis contributed to poor survival. These results support earlier findings (8). Prior nephrectomy, time from diagnosis to start treatment (TDT), presence of liver or skeletal metastasis and prior radiotherapy were not significant. The currently used MSKCC risk model (2) was also prognostic for survival in our dataset (figure 2B, AIC = 729, R2 = 0.24, p = 1.3 x 10-7). Median survival times for MSKCC favourable, intermediate and poor prognostic groups were 38.0, 15.7 and 4.0 months with overall survival rates of 38%, 19% and 0%, respectively. Multivariable Cox regression analysis revealed that expanding the MSKCC model with dichotomized novel proteins ApoA2 and SAA further improved predictive accuracy of the MSKCC model (Table 5A, AIC = 724, R2 = 0.27, p = 1.3 x 10-8).

A novel Protein-Based Prognostic Model and PerformanceAs the MSKCC model benefitted from the expansion of our proteins, it was probable that some of the included traditional risk factors did not significantly contribute to the predictive accuracy of this extended MSKCC model. Therefore we investigated whether we could develop a novel protein-based model and included as poor risk factors the MSKCC criteria; KPS (≤80%), LDH (>1.5x ULN), hemoglobin (<LLN), ‘corrected’ calcium (>ULN), TDT (≤12 months), together with prognostic proteins ApoA2 (>309 mg/L) and SAA (<19.2 ng/mL) and ≥2 metastasis sites in a multivariable Cox PH regression model. As suggested by Mekhail et al. (8), only the number of metastases sites was fitted in the analysis as this variable surpassed the independent predictors of

Table 4 - Results of Multivariate Analysis

Factor Poor Prognostic Category Parameter Estimate

SE P Hazard Ratio

95% CI

Karnofsky Performance Status ≤ 80% 0.83 0.26 0.00047 2.3 1.4 - 3.7

Lactate Dehydrogenase > 1.5x upper limit of reference range 1.28 0.26 0.00073 3.6 1.7 - 7.5

Number of Metastatic Sites 2 or 3 0.56 0.38 0.032 1.8 1.1 - 2.9

Apolipoprotein A-II ≤ highest tertile (≤309 mg/L) 0.65 0.24 0.014 1.9 1.1 - 3.2

Serum Amyloid Alpha > lowest tertile (>19.2 ng/mL) 0.71 0.26 0.0063 2.0 1.2 - 3.4

All variables included AIC = 712 P = 1.4 x 10-10


82

Tabl

e 5a

- N

ovel

pro

tein

-bas

ed m

odel

s co

mpa

red

to p

rior

defi

ned

cate

gori

zed

surv

ival

mod

els

Sur

viva

l Mod

elR

isk

Fact

ors

AIC

PR

isk

Fact

ors

No.

P

atie

nts

Med

ian

OS

(m

onth

s)H

azar

d R

atio

95%

CI

Mot

zer

et a

l., J

CO

19

99

Hb,

LD

H, c

orre

cted

Cal

cium

, K

PS

and

prio

r Sur

gery

732

7.0

x 10

-7

0 †

34

27.2

――

1 or

26

612

.22.

31.

4 - 3

.8≥

314

4.7

7.3

3.6

- 14.

7

Mot

zer

et a

l., J

CO

200

2H

b, L

DH

, cor

rect

ed C

alci

um, K

PS

and

Tim

e fro

m D

iagn

osis

to S

tudy

Ent

ry72

91.

3 x

10-7

0

†24

38.

0―

―(M

SK

CC

Ris

k M

odel

)1

or 2

7015

.72.

01.

2 - 3

.6

≥ 3

204.

07.

73.

9 - 2

5.4

Mek

hail

et a

l., J

CO

200

5H

b, L

DH

, cor

rect

ed C

alci

um, T

ime

from

D

iagn

osis

to S

tudy

Ent

ry, P

rior R

T an

d N

umbe

r of M

etas

tatic

Site

s

721

3.7

x 10

-9

0 or

1 †

135

8.3

――

25

618

.22.

61.

5 - 4

.3≥

345

6.5

8.4

4.4

- 16.

1Tw

o-P

rote

in S

igna

ture

Apo

A2

and

SA

A73

25.

2 x

10-7

0

†21

58.

3―

―1

3322

.32.

11.

0 - 4

.32

60

8.8

4.7

2.4

- 9.3

MS

KC

C M

odel

exp

ande

d w

ith A

poA

2 an

d S

AA

Hb,

LD

H, c

orre

cted

Cal

cium

, KP

S, T

ime

from

D

iagn

osis

to S

tudy

Ent

ry, A

poA

2 an

d S

AA

724

1.3

x 10

-8

0 or

1 †

2742

.8―

―2

or 3

46

17.4

2.5

1.4

- 4.6

≥ 4

415.

25.

93.

1 - 1

1.0

Nov

el P

rote

in-B

ased

Mod

elLD

H, K

PS

, Num

ber o

f Met

asta

tic S

ites,

SA

A

and

Apo

A2

713

4.3

x 10

-11

0

or 1

†31

54.

2―

―2

or 3

6112

.73.

21.

8 - 5

.6≥

422

3.3

11.2

5.6

- 22.

4

Abb

revi

atio

ns: A

IC; A

kaik

e’s

Info

rmat

ion

Crit

eria

, Hb;

Hem

oglo

bin,

LD

H; L

acta

te D

ehyd

roge

nase

, KP

S; K

arno

fsky

Per

form

ance

Sco

re, R

T; R

adio

ther

apy

† R

isk

grou

p us

ed a

s re

fere

nce.

Part II Proteomics

83

Tabl

e 5b

- C

ompa

riso

n of

MS

KC

C r

isk

mod

el a

nd T

wo-

Pro

tein

Sig

natu

re

Two-

Pro

tein

Sig

natu

re R

isk

Gro

ups

Tota

lFa

vora

ble

Inte

rmed

iate

Poo

rM

SK

CC

Ris

k G

roup

sP

atie

nts

(No.

)(%

) S

urvi

ved

Med

ian

Sur

viva

l (m

onth

s)

Pat

ient

s (N

o.)

(%)

Sur

vive

dM

edia

n S

urvi

val

(mon

ths)

Pat

ient

s (N

o.)

(%)

Sur

vive

dM

edia

n S

urvi

val

(mon

ths)

Pat

ient

s (N

o.)

(%)

Sur

vive

dM

edia

n S

urvi

val

(mon

ths)

Favo

rabl

e24

38

38.

09

675

8.3

103

03

8.0

50

17.4

Inte

rmed

iate

7019

15.7

1242

20.5

2218

18.3

36

1110

.5P

oor

200

4.0

0―

―1

04.

019

04.

4To

tal

114

2152

58.

333

2122

.36

07

8.8

64

patie

nts

(56%

) sw

itche

d be

twee

n ris

k gr

oups

Tabl

e 5c

- C

ompa

riso

n of

MS

KC

C r

isk

mod

el a

nd N

ovel

Pro

tein

-Bas

ed M

odel

Nov

el P

rote

in-B

ased

Ris

k G

roup

s

Tota

lFa

vora

ble

Inte

rmed

iate

Poo

rM

SK

CC

Ris

k G

roup

sP

atie

nts

(No.

)(%

) S

urvi

ved

Med

ian

Sur

viva

l (m

onth

s)

Pat

ient

s (N

o.)

(%)

Sur

vive

dM

edia

n S

urvi

val

(mon

ths)

Pat

ient

s (N

o.)

(%)

Sur

vive

dM

edia

n S

urvi

val

(mon

ths)

Pat

ient

s (N

o.)

(%)

Sur

vive

dM

edia

n S

urvi

val

(mon

ths)

Favo

rabl

e24

38

38.

013

625

8.3

119

18.2

0―

―In

term

edia

te70

1915

.718

39

27.7

44

1412

.28

05.

4P

oor

200

4.0

0―

―6

04.

414

02.

4To

tal

114

314

85

4.4

6112

12.7

220

3.3

43 p

atie

nts

(38%

) sw

itche

d be

twee

n ris

k gr

oups

12 188

11 6110

64

patie

nts

(56%

) sw

itche

d be

twee

n ris

k gr

oups

43 p

atie

nts

(38%

) sw

itche

d be

twee

n ris

k gr

oups

5 36


84

liver, bone and lymph metastasis. Because radiotherapy had limited effect on survival, we excluded this parameter. Stepwise backward Cox PH regression modelling resulted in five independent risk factors; ApoA2, SAA, KPS, LDH and number of metastases sites, providing a significantly improved protein-based model (Table 4, AIC = 712, R2 = 0.38 p, = 1.4 x 10-10). Based on these five risk factors, patients were categorized into three groups predicting good (0 or 1 risk factor), intermediate (2-3 factors) and poor risk patients (≥4 factors). Kaplan-Meier curves showed powerfully discriminative outcomes (figure 2C, AIC = 713, p = 4.3 x 10-11). The favourable risk group (27% of patients) had a median survival of 54.4 months, with 1-, 2- and 3-year survival rates of 90%, 71% and 58%, respectively. The intermediate risk group (54%) had a median survival of 12.7 months, with 1-, 2- and 3-year survival rates of 54%, 25% and 15%, respectively. The median survival of the poor risk group (19%) was 3.3 months, with 1-, 2- and 3-year survival rates of 18%, 5% and 0%, respectively. Comparison to MsKCC ModelTo interrogate the value of each of the models, we performed AIC-based inter-model comparisons (Table 5a). Compared to the MSKCC model, 59 patients (52%) shifted between prognostic groups using the two-protein signature (Table 5b). The largest modification was the reallocation of 36 patients from the intermediate MSKCC group to the poor prognostic group. The two-protein signature had similar survival prediction (AIC = 732) compared to the MSKCC model (AIC = 729). Interestingly, applying our protein-based model, 43 patients (38%) switched risk groups, as compared to the MSCKK model (Table 5C). Thirty-one patients classified into the favourable risk group by the protein-based model comprised 18 patients who were assigned to the intermediate group by the MSKCC model. Likewise, the MSKCC model allocated 11 patients to the favourable group and 6 patients to the poor group, who were categorized into the intermediate group by the protein-based model. In our novel model the poor risk group comprised 22 patients, of which 8 patients with poor survival had an intermediate risk according to the MSKCC model. As median survival times of the reallocated patients appropriately suited to their corresponding novel risk category, justified the switch in group of these patients, thereby improving the predictive accuracy of the risk model (Table 5C). Conclusively, our novel protein-based model was remarkably more discriminative and predictive than the MSKCC model (Table 5A; AIC = 713 versus 729, respectively).

Results in Perspective: Validation by Bootstrapping Ideally we would verify these findings with an external dataset. However, since the samples used in this study were collected, new targeted therapies have replaced interferon-based treatment, making sera of a large similar patient cohort unavailable. Consequently, we decided to correct overestimation of our exploratory results with statistical bootstrapping to strengthen our findings (13). For our two-protein signature and the novel protein-based prognostic model, R2 was equal to 0.23 and 0.38 respectively. Both models were bootstrapped to correct optimism and showed comparable R2 values equal to 0.22 and 0.34 respectively. With an optimism of respectively 0.01 and 0.04, we concluded that our proposed models proved to be internally valid, though the predictive accuracy was overestimated approximately 10%.

Part II Proteomics

85

DI sCus s Ion

This study shows that validated MS-based proteomic profiling is able to successfully and reproducibly identify proteins that predict patients’ survival. Quantification of SAA and ApoA2 showed significant association with survival, revealing novel promising prognostic markers. These proteins improved patient prognostic categorization and suggest that they may be used either as a stand-alone test or as an extension to the commonly used MSKCC criteria. The latter yielded a novel protein-based model, comprising five factors; LDH, KPS, number of metastases, SAA and ApoA2, which enabled the classification of advanced RCC patients into three categories with distinct survival rates. Classifying patients accurately into survival categories is crucial for clinical management, outcome evaluation and best supportive care. Equally, patient prognosis increasingly culminates in therapeutic choice, as the National Comprehensive Cancer Network guidelines recommend to treat patient with poor prognosis differently compared to intermediate and favourable risk groups (16). The median OS time of the favourable MSKCC risk group differs from that of the original report (2), 38.0 months versus 29.6 months respectively. This difference is most likely explained by the fact that a considerable number of patients (42%, 10/24) of this favourable MSKCC group received TKI therapy as second line treatment. The median OS time of the intermediate and poor MSKCC risk group found in our study was similar to the original report (2); for the intermediate risk group 15.7 versus 13.8 months and for the poor risk group 4.0 versus 4.9 months. Consequently, it is increasingly imperative to assign patients to best suitable therapeutic strategy by applying risk models. Our protein-based model shifted 38% patients to another risk group and suggested that these novel markers contribute to discriminate patients better in terms of prognosis than the MSKCC model. Finally, ApoA2 and SAA contribute to prognostic models, and potentially improve existing risk models, thereby facilitating patient management. SELDI-TOF MS is used as a cancer diagnostic device (17) and to predict relapse or therapeutic response (18). Previously, ApoA2 and SAA have been independently described in several diagnostic studies being discriminatory between cancer patients and healthy controls. ApoA2 is a major apolipoprotein of high-density lipoproteins, modulating cholesterol transport(15). ApoA2 levels are decreased in pancreatic (19), colorectal (20), and ovarian cancer (21). SAA is predominantly produced in the liver, macrophages and adipocytes and has proinflammatory and lipolytic functions (22). SAA is upregulated in RCC (9;10), ovarian (23), neuroblastoma (24), prostate (18), pancreatic (25), and hepatocellular cancer (26). Our findings are supported in literature; SAA has been implicated as a marker to determine cancer dissemination and response to treatment, irrespective of the exact tumour type (27;28). In general, cancer patients with metastatic disease become catabolic due to high-energy consuming tumors and loss of appetite, resulting in disturbed lipid metabolism, acute phase response, nutrient losses and altered hepatic protein production. Consequently, lipid and protein spectra will change. It is therefore tempting to speculate that these two proteins might also predict prognosis in other advanced malignancies. Accordingly, studies describe reduced ApoA2 levels in advanced cancers (21) and elevated SAA in disseminated cancers (18;27;28) confirming that our two-protein signature might be closely related to disease status and corresponding physiologic response of patients. Accordingly, SAA has very recently been described as highly predictive for PFS and OS in large cohorts of renal cell cancer (29;30), breast cancer (31) and melanoma


86

patients (32). Therefore, we hypothesize that the quantity of these predictive serological proteins is an accurate reflection of actual disease status of patients and that changes over time will predict progressive disease regardless given therapy. A limitation of our study is that samples were mainly collected prior to the approval of novel agents such as TKIs, mTOR- and VEGF-inhibitors. However, it seems that the risk factors constituting the MSKCC model are also applicable to these compounds (3;4;6). As the MSKCC model is based on a population of untreated mRCC patients receiving Interferon-α treatment, there might be a bias in the results of comparing the MSKCC model with our novel prediction models as 22% of patients in this study did not receive first line therapy. Accordingly, 19 patients (17%) received TKI-based treatment as second-line treatment. TKI therapy has shown increased objective response rates in large cohorts of mRCC patients (3;4), which also might interfere with prognosis. At most, to position our exploratory findings in perspective and prevent premature conclusions, we have employed bootstrapping to correct for optimism of our data, showing just 10% overestimation. Nevertheless, we strongly suggest that these findings require further validation in a large multicenter prospective study in a homogenous untreated mRCC patient cohort, before being applied to practice. Thereby we hope to conclude that these novel markers and corresponding models may improve assignment of patients to suitable risk-directed therapies. The present study demonstrates that combining proteomics-based screening with subsequent validation by protein quantification methods can yield novel biomarkers. The impact of these findings is highly clinically relevant. First, we have identified several novel proteins that predict survival from serum samples obtained and preserved at three different institutions. The heterogeneous manner in which these samples were collected indicates that the identified proteins are truly prognostic in a broad population of RCC patients. Second, when all quality controls are included, SELDI-TOF MS can be an effective method to screen for novel prognostic proteins. Third, the prognostic value of three proteins was confirmed by commonly used protein quantification methods, supporting the reproducibility of the SELDI-TOF MS. Fourth, our two-protein signature seemed to outperform the prognostic value of the MSKCC model and may even be used as an independent prognostic stand-alone test. Most important, two of these serological proteins can be easily and inexpensively implemented in any well-equipped hospital in an objective, quantitative and non-invasive manner, improving the predictive accuracy of traditional clinical prognostic factors, yielding novel risk models. By reallocating about 1 in 3 patients into a different prognostic group, this model suggests to classify patients more accurately than the MSKCC model. If this is confirmed by subsequent prospective validation studies, these novel prognostic markers ApoA2 and SAA and their associated models have the major potential to contribute to more tailored therapeutic approaches in practice.

ACK noW le D g e M e nTs

The authors thank all medical doctors, researchers and technicians who were involved in patient management and blood processing. We are also grateful for the assistance of all affiliated chemical laboratories and we thank Professor Karel Moons (Julius Center, Utrecht) for constructive discussions.

Part II Proteomics

87

AP P e n DI X

Screening Method: Serum Protein Profiling through SELDI-TOF Mass Spectrometry. Briefly, 2 ll of serum was denatured with 3 ll of 20 mM Tris-HCl, containing 9.5 M urea, 2% 3-[(3-cholamideopropyl)dimethylammonio]-1-propanesulfate, and 2% dithiotreitol. After mixing, 45 ll of 50 mM sodium acetate (pH 4.0), containing 0.05% Triton X-100 (binding buffer), was added. CM10 chips were equilibrated twice with 350 ll of binding buffer on a shaking platform for 5 min. Fifty microliter of the diluted serum was added to each well and incubated for 1h. The array was washed with 350 ll of binding buffer, rinsed with a 1 : 20 dilution of binding buffer and air-dried; 0.5 ll of a 50% solution of sinapinic acid in 50% acetonitrile + 0.5% trifluoroacetic acid was applied twice to the spots for energy absorption. A standard peptide mixture (Ciphergen Biosystems) containing vasopressin (1084.3 Da), somatostatin (1637.9 Da), dynorphin (2147.5 Da), ACTH (2933.5 Da), insulin b-chain (bovine, 3495.9 Da), insulin (human recombinant, 5807.7 Da), and hirudin recombinant (6963.5 Da) externally calibrated M/z values. PBS-IIC ProteinChipReader (Ciphergen Biosystems) analyzed all CM10 chips in a mass range of 1500-100 000 Da and the resulting data were processed by Ciphergen ProteinChip Software 3.1.


88



renal cell carcinoma (mRCC): a literature review. Cancer Treat Rev 2008 May;34(3):193-205.


trials of new therapies against advanced renal cell carcinoma. J Clin Oncol 2002 January 1;20(1):289-96.

(3) Escudier B, Eisen T, Stadler WM, Szczylik C, Oudard S, Siebels M et al. Sorafenib in advanced clear-cell

renal-cell carcinoma. N Engl J Med 2007 January 11;356(2):125-34.

(4) Motzer RJ, Hutson TE, Tomczak P, Michaelson MD, Bukowski RM, Rixe O et al. Sunitinib versus interferon alfa

in metastatic renal-cell carcinoma. N Engl J Med 2007 January 11;356(2):115-24.

(5) Hudes G, Carducci M, Tomczak P, Dutcher J, Figlin R, Kapoor A et al. Temsirolimus, interferon alfa, or both for

advanced renal-cell carcinoma. N Engl J Med 2007 May 31;356(22):2271-81.

(6) Escudier B, Pluzanska A, Koralewski P, Ravaud A, Bracarda S, Szczylik C et al. Bevacizumab plus interferon

alfa-2a for treatment of metastatic renal cell carcinoma: a randomised, double-blind phase III trial. Lancet 2007

December 22;370(9605):2103-11.

(7) Motzer RJ, Mazumdar M, Bacik J, Berg W, Amsterdam A, Ferrara J. Survival and prognostic stratification of 670

patients with advanced renal cell carcinoma. J Clin Oncol 1999 August;17(8):2530-40.

(8) Mekhail TM, bou-Jawde RM, Boumerhi G, Malhi S, Wood L, Elson P et al. Validation and extension of the

Memorial Sloan-Kettering prognostic factors model for survival in patients with previously untreated metastatic

renal cell carcinoma. J Clin Oncol 2005 February 1;23(4):832-41.

(9) Won Y, Song HJ, Kang TW, Kim JJ, Han BD, Lee SW. Pattern analysis of serum proteome distinguishes renal

cell carcinoma from other urologic diseases and healthy persons. Proteomics 2003 December;3(12):2310-6.

(10) Engwegen JY, Mehra N, Haanen JB, Bonfrer JM, Schellens JH, Voest EE et al. Validation of SELDI-TOF MS

serum protein profiles for renal cell carcinoma in new populations. Lab Invest 2007 February;87(2):161-72.

(11) Frank E.Harrell. Regression Modeling Strategies With Applications to Linear Models, Logistic Regression,

and Survival Analysis. New York: Springer; 2001.

(12) Akaike H. Information theory and an extension of the maximum likelihood principle. Budapest, Hungary:

Akademiai Kaiado; 1973.

(13) Steyerberg EW, Bleeker SE, Moll HA, Grobbee DE, Moons KG. Internal and external validation of predictive

models: a simulation study of bias and precision in small samples. J Clin Epidemiol 2003 May;56(5):441-7.

(14) Engwegen JY, Gast MC, Schellens JH, Beijnen JH. Clinical proteomics: searching for better tumour markers

with SELDI-TOF mass spectrometry. Trends Pharmacol Sci 2006 May;27(5):251-9.

(15) Martin-Campos JM, Escola-Gil JC, Ribas V, Blanco-Vaca F. Apolipoprotein A-II, genetic variation on chromosome

1q21-q24, and disease susceptibility. Curr Opin Lipidol 2004 June;15(3):247-53.

(16) National Comprehensive Cancer Network, Practice guidelines in oncology: Kidney cancer (v.1.2009). Accessed

2009 March 21

(17) Petricoin EF, Ardekani AM, Hitt BA, Levine PJ, Fusaro VA, Steinberg SM et al. Use of proteomic patterns in

serum to identify ovarian cancer. Lancet 2002 February 16;359(9306):572-7.

(18) Le L, Chi K, Tyldesley S, Flibotte S, Diamond DL, Kuzyk MA et al. Identification of serum amyloid A as a biomarker

to distinguish prostate cancer patients with bone lesions. Clin Chem 2005 April;51(4):695-707.

(19) Ehmann M, Felix K, Hartmann D, Schnolzer M, Nees M, Vorderwulbecke S et al. Identification of potential

markers for the detection of pancreatic cancer through comparative serum protein expression profiling. Pancreas

2007 March;34(2):205-14.

Part II Proteomics

89

(20) Glojnaric I, Casl MT, Simic D, Lukac J. Serum amyloid A protein (SAA) in colorectal carcinoma. Clin Chem Lab

Med 2001 February;39(2):129-33.

(21) Wang J, Zhang X, Ge X, Guo H, Xiong G, Zhu Y. Proteomic studies of early-stage and advanced ovarian cancer

patients. Gynecol Oncol 2008 August 12.

(22) Yang RZ, Lee MJ, Hu H, Pollin TI, Ryan AS, Nicklas BJ et al. Acute-phase serum amyloid A: an inflammatory

adipokine and potential link between obesity and its metabolic complications. PLoS Med 2006 June;3(6):e287.

(23) Zhang H, Kong B, Qu X, Jia L, Deng B, Yang Q. Biomarker discovery for ovarian cancer using SELDI-TOF-MS.

Gynecol Oncol 2006 July;102(1):61-6.

(24) Sandoval JA, Turner KE, Hoelz DJ, Rescorla FJ, Hickey RJ, Malkas LH. Serum protein profiling to identify high-risk

neuroblastoma: preclinical relevance of blood-based biomarkers. J Surg Res 2007 October;142(2):268-74.

(25) Orchekowski R, Hamelinck D, Li L, Gliwa E, vanBrocklin M, Marrero JA et al. Antibody microarray profiling

reveals individual and combined serum proteins associated with pancreatic cancer. Cancer Res 2005 December

1;65(23):11193-202.

(26) He QY, Zhu R, Lei T, Ng MY, Luk JM, Sham P et al. Toward the proteomic identification of biomarkers for the

prediction of HBV related hepatocellular carcinoma. J Cell Biochem 2008 February 15;103(3):740-52.

(27) Rosenthal CJ, Sullivan LM. Serum amyloid A to monitor cancer dissemination. Ann Intern Med 1979

September;91(3):383-90.

(28) Biran H, Friedman N, Neumann L, Pras M, Shainkin-Kestenbaum R. Serum amyloid A (SAA) variations in patients

with cancer: correlation with disease activity, stage, primary site, and prognosis. J Clin Pathol 1986

July;39(7):794-7.

(29) Ramankulov A, Lein M, Johannsen M, Schrader M, Miller K, Loening SA et al. Serum amyloid A as indicator of

distant metastases but not as early tumor marker in patients with renal cell carcinoma. Cancer Lett 2008

September 28;269(1):85-92.

(30) Kimura M, Tomita Y, Imai T, Saito T, Katagiri A, Ohara-Mikami Y et al. Significance of serum amyloid A on the

prognosis in patients with renal cell carcinoma. Cancer 2001 October 15;92(8):2072-5.

(31) Pierce BL, Ballard-Barbash R, Bernstein L, Baumgartner RN, Neuhouser ML, Wener MH et al. Elevated

biomarkers of inflammation are associated with reduced survival among breast cancer patients. J Clin Oncol

2009 July 20;27(21):3437-44.

(32) Findeisen P, Zapatka M, Peccerella T, Matzk H, Neumaier M, Schadendorf D et al. Serum amyloid A as a

prognostic marker in melanoma identified by proteomic profiling. J Clin Oncol 2009 May 1;27(13):2199-208.



Chapter 6 Validation of serum Amyloid Alpha as an Independent Biomarker for Progression free- and overall survival in Metastatic Renal Cell Cancer Patients

Joost S Vermaat, Frank L Gerritse, Astrid A van der Veldt, Wijnand M Roessingh, Tatjana M Niers, Sjoukje F Oosting, Stefan Sleijfer, Jeanine M Roodhart, Jos H Beijnen, Jan H Schellens, Jourik A Gietema, Epie Boven, Dick J Richel, John B Haanen, Emile E. Voest

European Urology, 2012 Jan 23. [Epub ahead of print]

92


In metastatic renal cell carcinoma (mRCC), the Memorial Sloan-Kettering Cancer Center (MSKCC) prognostic model is used to stratify patients for appropriate risk-directed therapy. This study validated our previous findings demonstrating that Apolipoprotein-A2 (ApoA2) and Serum Amyloid Alpha (SAA) levels significantly and independently predicted overall survival (OS) in mRCC patients. A training cohort consisting of 114 mRCC patients treated with interferon and a validation cohort comprising 151 mRCC patients treated with tyrosine kinase inhibitors (TKIs) were included. Multivariate analysis identified SAA –but not ApoA2- in both independent patient sets as a robust and independent prognosticator for both progression free survival (PFS) and OS. Applying tertiles as objective cut-off values for SAA-levels, mRCC patients were categorized in three risk groups with favourable, intermediate and poor survival times, demonstrating accurate risk prognostication. Importantly, SAA as single biomarker showed equal prognostic accuracy when compared to the complex multi-factorial MSKCC risk model. Furthermore, applying SAA as additional risk factor the predictive accuracy of the MSKCC model improved significantly in both the training and validation cohort, yielding enhanced discrimination of patient survival times for every MSKCC risk group. In conclusion, SAA is a robust and independent prognostic factor for PFS and OS and improved the MSKCC predictive accuracy and can be incorporated into clinical patient management and clinical trials.

Part II Proteomics

93Validation of SAA as an Independent Biomarker for Survival in mRCC Patients Chapter 6

AB sTRACT

Background and objective Recently, we identified Apolipoprotein-A2 (ApoA2) and Serum Amyloid Alpha (SAA) as independent prognosticators in metastatic renal cell cancer (mRCC) patients, thereby improving the Memorial-Sloan Kettering Cancer Center (MSKCC) model accuracy. To validate these results prospectively in a separate cohort of tyrosine kinase inhibitors (TKIs)-treated mRCC patients.

Patients and methods For training we used 114 interferon-treated mRCC patients (inclusion 2001 - 2006). For validation we studied 151 TKI-treated mRCC patients (inclusion 2003 - 2009). Using Cox PH regression analysis SAA and ApoA2 were associated with progression-free survival (PFS) and overall survival (OS). In 72 TKI-treated patients SAA-levels were analyzed longitudinally as potential early marker for treatment effect.

Results Baseline ApoA2- and SAA-levels significantly predicted PFS and OS in the training- and validation cohort. Multivariate analysis identified SAA in both separate patient sets as a robust and independent prognosticator for PFS and OS. In contrast to our previous findings ApoA2 interacted with SAA in the validation cohort and therefore did not contribute to a better predictive accuracy then SAA alone and was therefore excluded from further analysis. According to the tertiles of SAA-levels patients were categorized in three risk groups, demonstrating accurate risk prognostication. Importantly, SAA as single biomarker showed equal prognostic accuracy when compared to the multi-factorial MSKCC risk mode. Using ROC-analysis SAA-levels >71 ng/ml was designated as optimal cut-off value in the training cohort which was confirmed for its significant sensitivity and specificity in the validation cohort. Applying SAA >71 ng/ml as additional risk factor significantly improved the predictive accuracy of MSKCC model in both independent cohorts. Changes in SAA-levels after 6-8 weeks of TKI-treatment had no value in predicting treatment outcome.

Conclusion SAA – but not ApoA2- demonstrated to be a robust and independent prognosticator for PFS and OS in mRCC patients. When incorporated in the MSKCC model SAA showed additional prognostic value for patient management.

94

AB B R eVIATIon s (m)RCC; (metastatic) Renal Cell Carcinoma, MSKCC; Memorial Sloan-Kettering Cancer Center, OS; Overall Survival, PFS; Progression Free Survival, PS; performance status, ApoA2; Apolipoprotein A-II, SAA; Serum Amyloid Alpha, ELISA; Enzyme Linked Immunosorbent Assay, AIC; Akaike’s Information Criteria, TDT; time from diagnosis to start of treatment, LDH; Lactate Dehydrogenase, TKI; tyrosine kinase inhibitor, ECOG; Eastern Cooperative Oncology Group Performance Status, AP; Alkaline Phosphatase, TDT; time between diagnosis and start of systemic treatment

I nTRoDuCTIon

Therapeutic options for metastatic Renal Cell Cancer (mRCC) patients have significantly expanded over the past years. Interferon-alpha (IFN) and interleukin-2 have been surpassed by targeted agents as first-line treatment (1-3) including Vascular Endothelial Growth Factor (VEGF)-receptor tyrosine-kinase-inhibitors (TKIs) sorafenib, sunitinib and pazopanib, the mTOR (mammalian-Target-of-Rapamycin)-inhibitors temsirolimus/everolimus) and bevacizumab targeting VEGF-monoclonal-antibody (4-8). Because not all patients equally benefit from targeted therapies and can induce severe toxicity, the identification of useful prognosticators to categorize patients with respect to outcome is an unmet need (9). Accordingly, studies to determine the value of targeted agents in combination with nephrectomy in mRCC patients are ongoing (10). The Memorial Sloan-Kettering Cancer Center (MSKCC) risk model was developed to discriminate between mRCC patients with a good, intermediate and poor prognosis and is widely used for treatment allocation and stratification in clinical trials (1;11). The first MSKCC model, described by Motzer et al. (12), was further refined and currently comprehends: performance status (PS), lactate dehydrogenase (LDH), hemoglobin, ‘corrected’ calcium and time between diagnosis and start of systemic treatment (TDT) as prognostic factors for overall survival (OS) (1;13). Risk stratification is used for selecting the most appropriate therapy for individual mRCC patients (14-17), but has its limitations such as altered first-line and additional systemic therapies. It remains a challenge to optimize this stratification, thereby improving a solid a-priori patient selection for ‘risk-directed’ targeted therapy (9). Recent investigations identified trombocytosis, leukophilia, neutrophilia and elevated alkaline phosphatise (AP) as additional poor risk factors (8;18-21). We have identified Apolipoprotein-A2 (ApoA2) and Serum Amyloid Alpha (SAA) as prognosticators (22), which independently improved the MSKCC prognostic accuracy. We composed a protein-based model, including ApoA2, SAA, LDH, PS and ≥2 metastatic sites, which outperformed the MSKCC prognostic accuracy. Comparing both models revealed that 38% of patients switched between risk groups. An important limitation of our previous study was that the findings were not further validated in mRCC patients receiving VEGFR-TKIs. Furthermore, progression free survival (PFS) is considered an important primary study end-point (23). Therefore prognosticators also need to be evaluated for their correlation with PFS (20;24;25). The first aim of the present study was to associate ApoA2 and SAA with PFS and OS in our training cohort (114 IFN-based treated mRCC patients) and to subsequently validate these results

Part II Proteomics

95

in a new validation cohort of 151 TKI-treated mRCC patients. Next, SAA prognostic accuracy as a single biomarker was compared with the multi-factorial MSKCC model. We determined whether an optimal cut-off value for SAA-levels could improve the MSKCC model’s prognostic accuracy. Finally, we investigated whether repetitive SAA measurements could be used as an early marker for disease progression.


Patient Populations and sample Collection This investigation included two independent patient groups -training and validation cohort- which were separately evaluated for ApoA2- and SAA-correlation with survival. First, a retrospective training cohort of 114 mainly interferon-based treated mRCC patients was collected from three Dutch cancer institute’s (inclusion 2001 - 2006); University Medical Center Utrecht (n = 63), Netherlands Cancer Institute (n = 29) and Erasmus Medical Center Rotterdam (n = 22). This cohort was previously reported (22) but for this study OS was updated, included PFS zand incorporated the recently recognized prognosticators AP, leukocyte- and platelet counts (8;18-20;24). Neutrophil counts were not available for the majority of these patients. Patients were treated with either IFN-based therapy (83 patients), with interleukin-2, sunitinib or a tubulin-antagonist (n = 6, Table 1) or received no systemic treatment at all because of contra-indications, poor PS or lack of progression as a consequence of which no treatment was initiated (n = 25). Blood withdrawal generally occurred at start of first-line therapy (n = 81). From 8 patients blood was collected after IFN-therapy. Following initial treatment 39 patients received additional second-line systemic therapy consisting of sunitinib, sorafenib or interleukin-2. Blood samples of 18 patients were gathered immediately before primary tumor resection. Secondly, a validation cohort of 151 mRCC patients treated with TKIs (inclusion 2003 - 2009) was assembled by prospectively collecting blood samples in five Dutch cancer institute’s; University Medical Center Utrecht (n = 24), Netherlands Cancer Institute (n = 30), Amsterdam Medical Center (n = 41), VU Medical Center (n = 44) and University Medical Center Groningen (n = 12). Patients received either sunitinib (n = 103) or sorafenib (n = 48) and sera was collected at treatment initiation. For 82 patients this TKI-therapy was first-line treatment. Sixty-three patients were previously treated with IFN (Table 1). After TKI-treatment, 72 patients received sunitinib (n = 18), sorafenib (n = 33) or everolimus/temsirolimus (n = 21) as additional treatment. Besides baseline levels, SAA was also quantified at 6 weeks for sunitinib (n = 31) and at 8 weeks for sorafenib (n = 41) treated patients. Informed consent was signed by every individual according to all institutional ethical review boards and patients were registered in different study protocols. For all participating hospitals PFS was described according to investigator-assessed criteria. OS was defined as time from blood collection to date of death, last follow-up or when patients accomplished 48 months of follow-up. Baseline characteristics were assembled from patient medical histories, comprehending MSKCC factors and other prognosticators. Of note, 8 patients overlapped between training- and validation cohorts. These 8 patients were first evaluated when interferon treated (training) and subsequently considered as an independent entity for a second time when receiving TKI-therapy (validation). In general, blood was coagulated at room temperature (0.5 - 6hrs), centrifuged (1500 - 1900g) and stored in freezer. ApoA2 (Dade Behring, Newark, USA) and SAA-concentrations (Tridelta Development, Kildare, Ireland) were determined with conventional antibody-directed enumeration assays.

Validation of SAA as an Independent Biomarker for Survival in mRCC Patients Chapter 6

96

Table 1 - Patient Characteristics and Treatment Modalities

Training Cohort Validation Cohort

overallno.114

(%)(100%)

no.151

(%)(100%)

PATIenTs CHARACTeRIsTICs

Males 76 (67%) 107 (71%)Age (Mean Years)SD

60.0 60.210.1 10.1

eCog Performance status0 83 (73%) 93 (62%)≥1 31 (27%) 58 (38%)Prior nephrectomyYes 78 (68%) 113 (75%)No 36 (32%) 38 (38%)no. of Metastasic sites1 36 (32%) 28 (19%)≥2 78 (68%) 123 (81%)Time from Diagnosis to start Therapy

>12 months 50 (44%) 79 (52%)≤12 months 64 (56%) 72 (48%)PathologyClear Cell 91 (80%) 137 (91%)Other 16 (14%) 14 (9%)Clinical Evidence 7 (6%) –

TReATMenT MoDAlITIes

systemic Therapy (blood withdrawal at initiation)Interferon-based 83 (73%) –Sunitinib 2 (2%) 103 (68%)Sorafenib – 48 (32%)Interleukin-2 2 (2%) –Tubulin-antagonist: ABT-751 2 (2%) –Best Supportive care 25 (22%) –

Total 114 (100%) 151 (100%)PretreatmentInterferon-based † 8 (7%) 63 (42%)Sorafenib – 4 (3%)Interleukin-2 – 2 (1%)

Total 8 (7%) 69 (46%)Additional systemic TherapySunitinib 11 (10%) 18 (12%)Sorafenib 8 (7%) 33 (22%)Everolimus – 20 (13%)Interleukin-2 6 (5%) –Cediranib * 5 (4%) –Tubulin-antagonist: ABT-751 4 (3%) –Anti-EGFR 2 (2%) –Angiostatin 1 (1%) –Telatinib + Bevacizumab 1 (1%) –Temsirolimus – 1 (<1%)Anti-Interleukin-6 1 (1%) –

Total 39 (34%) 72 (48%)

† With or without Thalidomide * With or without Gefinitib

Part II Proteomics

97

statistical Analysis All data was normally distributed and differences between the training and validation cohort were evaluated with Student’s T-test. To associate the prognosticators with survival univariate Cox Proportional Hazards (PH) regression analysis was employed. Receiver operating characteristic (ROC) curve-analysis was utilized to investigate the sensitivity and specificity of prognosticators. Multivariate Cox PH regression analysis was applied with backward-stepwise selection to investigate whether variables independently predicted survival. To assess the exactitude of prognostication of sovereign risk models the Akaike’s Information Criteria (AIC) and R2 were applied as previously performed (22;26;27). Conclusively, the lower the AIC the better the predictive model fits the data. Kaplan-Meier curves illustrated the predictive precision of various risk models. Statistical analyses were performed using SPSS (version 15.0) and R-software (version 2.3.1).

R e s u lTs

Patient survival outcomes for independent cohorts In the training cohort median PFS time was 5.9 months (95% CI: 4.2 – 7.7). However, of 8 patients PFS was censored at 3 months as for these patients only limited follow-up time was accessible. Median OS time was 15.7 months (95% CI: 11.0 – 20.2). At the moment of final analysis, 24 patients were still alive. In the validation cohort median PFS time was 7.0 months (95% CI: 5.1 – 8.8). Median OS time was 15.0 months (95% CI: 12.4 – 17.7). At final analysis, 22 patients were still alive. Median PFS times for sunitinib and sorafenib were respectively 6.0 (95% CI: 3.8 – 8.6) and 7.4 months (95% CI: 4.4 – 10.4), which was not statistically different (p = 0.3). Accordingly, median OS times for sorafenib and sunitinib treatment were also similar (p = 0.6), respectively 14.9 (95% CI: 10.9 – 19.0) and 15.0 months (95% CI: 11.9 – 18.0).

univariate Analysis Table 2A displays clinical characteristics of the training and validation cohort related to PFS and OS. ECOG, TDT, prior nephrectomy, LDH, calcium, hemoglobin, albumin, AP, platelet count, leukocytes, neutrophils, number of metastatic sites and histology were significantly (or borderline) associated to PFS and OS. Mean SAA concentrations were 265 ng/ml (SD; 380 ng/ml), and 259 ng/ml (SD; 303 ng/ml) for respectively the training and validation cohort, which were statistically not different (p = 0.87). SAA as continuous variable significantly predicted PFS (p = 3.9 x 10-11 and 1.0 x 10-5)

and OS (p = 1.0 x 10-10 and 3.7 x 10-7) for both cohorts. The positive parameter estimate for SAA indicated prolonged survival for lower concentrations. Mean ApoA2 levels in the training cohort were 278 mg/L (SD; 91 mg/L), and 270 mg/L (SD; 85 mg/L) in the validation cohort (ApoA2-levels not determined in 19 patients), which was not statistically different between both cohorts (p = 0.49). ApoA2 was significantly associated with PFS and OS in the training cohort (Table 2B, p = 6.2 x 10-6 and p = 2.2 x 10-7, respectively). The negative parameter estimate for ApoA2 indicated reduced survival for lower concentrations. Although, in the validation cohort its predictive accuracy diminished, ApoA2 was still prognostic for PFS and OS (p = 0.02 and P = 0.01).

Multivariate Analysis As described previously (22), ApoA2 and SAA had significant independent prognostic


Table 1 - Patient Characteristics and Treatment Modalities

Training Cohort Validation Cohort

overallno.114

(%)(100%)

no.151

(%)(100%)

PATIenTs CHARACTeRIsTICs

Males 76 (67%) 107 (71%)Age (Mean Years)SD

60.0 60.210.1 10.1

eCog Performance status0 83 (73%) 93 (62%)≥1 31 (27%) 58 (38%)Prior nephrectomyYes 78 (68%) 113 (75%)No 36 (32%) 38 (38%)no. of Metastasic sites1 36 (32%) 28 (19%)≥2 78 (68%) 123 (81%)Time from Diagnosis to start Therapy

>12 months 50 (44%) 79 (52%)≤12 months 64 (56%) 72 (48%)PathologyClear Cell 91 (80%) 137 (91%)Other 16 (14%) 14 (9%)Clinical Evidence 7 (6%) –

TReATMenT MoDAlITIes

systemic Therapy (blood withdrawal at initiation)Interferon-based 83 (73%) –Sunitinib 2 (2%) 103 (68%)Sorafenib – 48 (32%)Interleukin-2 2 (2%) –Tubulin-antagonist: ABT-751 2 (2%) –Best Supportive care 25 (22%) –

Total 114 (100%) 151 (100%)PretreatmentInterferon-based † 8 (7%) 63 (42%)Sorafenib – 4 (3%)Interleukin-2 – 2 (1%)

Total 8 (7%) 69 (46%)Additional systemic TherapySunitinib 11 (10%) 18 (12%)Sorafenib 8 (7%) 33 (22%)Everolimus – 20 (13%)Interleukin-2 6 (5%) –Cediranib * 5 (4%) –Tubulin-antagonist: ABT-751 4 (3%) –Anti-EGFR 2 (2%) –Angiostatin 1 (1%) –Telatinib + Bevacizumab 1 (1%) –Temsirolimus – 1 (<1%)Anti-Interleukin-6 1 (1%) –

Total 39 (34%) 72 (48%)

† With or without Thalidomide * With or without Gefinitib

98

Tabl

e 2

A -

Uni

vari

ate

anal

ysis

of b

asel

ine

char

acte

rist

ics

pred

ictiv

e fo

r P

rogr

essi

on F

ree

Sur

viva

l and

Ove

rall

Sur

viva

l

Pro

gres

sion

-fre

e su

rviv

alo

vera

ll s

urvi

val

Trai

ning

Coh

ort N

=11

4Va

lidat

ion

Coh

ort N

=15

1Tr

aini

ng C

ohor

t N=

114

Valid

atio

n C

ohor

t N=

151

Varia

ble

Haz

ard

Rat

ioP

-val

ueH

azar

d R

atio

P-v

alue

Haz

ard

Rat

ioP

-val

ueH

azar

d R

atio

P-v

alue

EC

OG

sta

tus

* (0

ver

sus

1)2.

00.

003

1.3

0.2

2.3

<0.

001

1.4

0.05

Tim

e fro

m d

iagn

osis

to tr

eatm

ent *

(≥

12 v

ersu

s <

12 m

onth

s)1.

40.

11.

30.

21.

70.

016

1.5

0.03

Prio

r nep

hrec

tom

y (y

es v

ersu

s no

)1.

50.

051.

40.

071.

50.

08

1.5

0.0

4

LDH

leve

l *1.

0<

0.00

11.

0<

0.00

11.

0<

0.00

11.

0<

0.00

1

Cor

rect

ed C

alci

um le

vel *

,§1.

20.

005

1.5

<0.

001

1.1

0.0

91.

6<

0.00

1

Hem

oglo

bin

leve

l *0.

7<

0.00

10.

80.

003

0.7

<0.

001

0.7

<0.

001

Alb

umin

leve

l0.

4<

0.00

10.

5<

0.00

10.

4<

0.00

10.

5<

0.00

1

Alk

alin

e ph

osph

atas

e le

vel

1.0

0.00

41.

0<

0.00

11.

0<

0.00

11.

0<

0.00

1

Abs

olut

e le

ucoc

yte

coun

t1.

10.

021.

2<

0.00

11.

10.

002

1.2

<0.

001

Abs

olut

e ne

utro

phil

coun

t–

–1.

0<

0.00

1–

–1.

0<

0.00

1

Pla

tele

t cou

nt1.

00.

001

1.0

0.00

21.

00.

003

1.0

0.00

1

≥2 m

etas

tatic

site

s (n

o ve

rsus

yes

)2.

10.

002

1.6

0.0

42.

30.

001

1.8

0.02

His

tolo

gy

(Cle

ar C

ell v

s no

n-C

C)

2.2

0.00

61.

60.

11.

70.

021.

80.

03

Prio

r im

mun

othe

rapy

(Ifn

ver

sus

none

)1.

20.

61.

00.

80.

90.

91.

10.

5

Not

e: L

og-r

ank

test

for c

ateg

oric

al fo

rm o

f the

var

iabl

e an

d C

ox p

ropo

rtio

nal h

azar

ds m

odel

for c

ontin

uous

form

of t

he v

aria

ble.

For c

ontin

uous

var

iabl

es, a

haz

ard

ratio

>1

= ri

sk re

duct

ion

whe

n th

e va

lue

decr

ease

s an

d a

haza

rd ra

tio <

1 =

risk

redu

ctio

n w

hen

the

valu

e in

crea

se;

For b

inar

y va

riabl

es (c

ateg

orie

s in

bol

d), a

haz

ard

ratio

>1

= ri

sk re

duct

ion

for t

he fi

rst c

ateg

ory

and

a ha

zard

ratio

<1

= ri

sk re

duct

ion

for t

he s

econ

d ca

tego

ryFo

r sev

eral

labo

rato

ry m

arke

rs, h

azar

d ra

tios

are

clos

e to

1 d

ue to

the

cont

inuo

us s

cale

of t

he m

arke

r.Th

is ta

ble

is in

ana

logy

with

the

tabl

e pu

blis

hed

by P

atil

et a

l. [2

0]A

bbre

viat

ions

; EC

OG

; Eas

tern

Coo

pera

tive

Onc

olog

y G

roup

, LD

H; l

acta

te d

ehyd

roge

nase

* P

ogno

stic

fact

ors

incl

uded

in M

emor

ial S

loan

-Ket

terin

g C

ance

r Cen

ter r

isk

grou

ps [1

]§

Cor

rect

ed C

alci

um =

tota

l cal

cium

- 0.

707

(alb

umin

- 3.

4).

Part II Proteomics

99

Tabl

e 2

B -

Uni

vari

ate

anal

ysis

of S

erum

Am

yloi

d A

lpha

(S

AA

) an

d A

polip

opro

tein

A2

(Apo

A2)

Pro

gres

sion

-fre

e su

rviv

alo

vera

ll s

urvi

val

Trai

ning

Coh

ort

Valid

atio

n C

ohor

tTr

aini

ng C

ohor

tVa

lidat

ion

Coh

ort

Varia

ble

Cha

ract

eris

tics

P-v

alue

*C

hara

cter

istic

sP

-val

ue*

Cha

ract

eris

tics

P-v

alue

*C

hara

cter

istic

sP

-val

ue*

Apo

lipro

tein

A-I

I (A

poA

2)n

=11

4n

=13

2n

=11

4n

=13

2

Con

tinuo

us F

orm

6.2

x 1

0-60.

022.

2 x

10-7

0.01

Par

amet

er E

stim

ate

–0.0

06

0–0

.002

4–0

.007

1–0

.002

7

Mea

n (m

g/L)

278

270

278

270

SD

9185

9185

Med

ian

277

260

277

260

IQR

228

- 324

223

- 329

228

- 324

223

- 329

ser

um A

myl

oid

Alp

ha (

sA

A)

n=

114

n=

151

n=

114

n=

151

Con

tinuo

us F

orm

3.9

x 10

-11

1.0

x 10

-51.

0 x

10-1

03.

7 x

10-7

Par

amet

er E

stim

ate

0.00

140.

0012

0.00

160.

0014

Mea

n (n

g/m

L)26

525

926

525

9

SD

38

03

033

80

303

Med

ian

48

107

48

107

IQR

16 -

46

610

7 - 5

04

16 -

46

610

7 - 5

04

* C

ox P

ropo

rtio

nal H

azar

ds R

egre

ssio

n A

naly

sis

was

use

d fo

r con

tinuo

us fo

rm o

f the

var

iabl

e.


Tabl

e 2

A -

Uni

vari

ate

anal

ysis

of b

asel

ine

char

acte

rist

ics

pred

ictiv

e fo

r P

rogr

essi

on F

ree

Sur

viva

l and

Ove

rall

Sur

viva

l

Pro

gres

sion

-fre

e su

rviv

alo

vera

ll s

urvi

val

Trai

ning

Coh

ort N

=11

4Va

lidat

ion

Coh

ort N

=15

1Tr

aini

ng C

ohor

t N=

114

Valid

atio

n C

ohor

t N=

151

Varia

ble

Haz

ard

Rat

ioP

-val

ueH

azar

d R

atio

P-v

alue

Haz

ard

Rat

ioP

-val

ueH

azar

d R

atio

P-v

alue

EC

OG

sta

tus

* (0

ver

sus

1)2.

00.

003

1.3

0.2

2.3

<0.

001

1.4

0.05

Tim

e fro

m d

iagn

osis

to tr

eatm

ent *

(≥

12 v

ersu

s <

12 m

onth

s)1.

40.

11.

30.

21.

70.

016

1.5

0.03

Prio

r nep

hrec

tom

y (y

es v

ersu

s no

)1.

50.

051.

40.

071.

50.

08

1.5

0.0

4

LDH

leve

l *1.

0<

0.00

11.

0<

0.00

11.

0<

0.00

11.

0<

0.00

1

Cor

rect

ed C

alci

um le

vel *

,§1.

20.

005

1.5

<0.

001

1.1

0.0

91.

6<

0.00

1

Hem

oglo

bin

leve

l *0.

7<

0.00

10.

80.

003

0.7

<0.

001

0.7

<0.

001

Alb

umin

leve

l0.

4<

0.00

10.

5<

0.00

10.

4<

0.00

10.

5<

0.00

1

Alk

alin

e ph

osph

atas

e le

vel

1.0

0.00

41.

0<

0.00

11.

0<

0.00

11.

0<

0.00

1

Abs

olut

e le

ucoc

yte

coun

t1.

10.

021.

2<

0.00

11.

10.

002

1.2

<0.

001

Abs

olut

e ne

utro

phil

coun

t–

–1.

0<

0.00

1–

–1.

0<

0.00

1

Pla

tele

t cou

nt1.

00.

001

1.0

0.00

21.

00.

003

1.0

0.00

1

≥2 m

etas

tatic

site

s (n

o ve

rsus

yes

)2.

10.

002

1.6

0.0

42.

30.

001

1.8

0.02

His

tolo

gy

(Cle

ar C

ell v

s no

n-C

C)

2.2

0.00

61.

60.

11.

70.

021.

80.

03

Prio

r im

mun

othe

rapy

(Ifn

ver

sus

none

)1.

20.

61.

00.

80.

90.

91.

10.

5

Not

e: L

og-r

ank

test

for c

ateg

oric

al fo

rm o

f the

var

iabl

e an

d C

ox p

ropo

rtio

nal h

azar

ds m

odel

for c

ontin

uous

form

of t

he v

aria

ble.

For c

ontin

uous

var

iabl

es, a

haz

ard

ratio

>1

= ri

sk re

duct

ion

whe

n th

e va

lue

decr

ease

s an

d a

haza

rd ra

tio <

1 =

risk

redu

ctio

n w

hen

the

valu

e in

crea

se;

For b

inar

y va

riabl

es (c

ateg

orie

s in

bol

d), a

haz

ard

ratio

>1

= ri

sk re

duct

ion

for t

he fi

rst c

ateg

ory

and

a ha

zard

ratio

<1

= ri

sk re

duct

ion

for t

he s

econ

d ca

tego

ryFo

r sev

eral

labo

rato

ry m

arke

rs, h

azar

d ra

tios

are

clos

e to

1 d

ue to

the

cont

inuo

us s

cale

of t

he m

arke

r.Th

is ta

ble

is in

ana

logy

with

the

tabl

e pu

blis

hed

by P

atil

et a

l. [2

0]A

bbre

viat

ions

; EC

OG

; Eas

tern

Coo

pera

tive

Onc

olog

y G

roup

, LD

H; l

acta

te d

ehyd

roge

nase

* P

ogno

stic

fact

ors

incl

uded

in M

emor

ial S

loan

-Ket

terin

g C

ance

r Cen

ter r

isk

grou

ps [1

]§

Cor

rect

ed C

alci

um =

tota

l cal

cium

- 0.

707

(alb

umin

- 3.

4).

100

value regarding OS in the training cohort. In contrast, when applying multivariate analysis for continuous variables, ApoA2 interacted with SAA in the validation cohort and therefore did not contribute to a better predictive accuracy then SAA alone. Therefore, ApoA2 was excluded from further analysis where other known prognosticators were included. In the training cohort multivariate analysis indentified SAA, ECOG and albumin as independent prognosticators for PFS and SAA, ECOG, LDH, TDT, ≥2 metastatic sites for OS (Table 3). In the validation cohort, SAA, LDH and leukocytes were determined to be independent prognosticators for PFS and SAA, LDH, leukocytes and calcium for OS. Remarkably, in both cohorts only SAA was an independent, robust prognosticator for both PFS and OS (Table 3). sAA as single marker compared to the MsKCC prediction model As expected, the MSKCC model (1) significantly predicted OS in both cohorts (Table 4, AIC = 717 (p = 3.0 x 10-9) and 1083 (p = 2.2 x 10-5)) respectively for the training and validation cohort). In the training cohort median OS times for MSKCC favourable, intermediate and poor prognostic groups were 39.1, 15.7 and 4.0 months (figure 1A). For the validation cohort median OS times for these risk groups were similar respectively 31.0, 16.7 and 6.6 (figure 1B). Interestingly, adopting tertiles as objective cut-off values for SAA-levels for both cohorts, patients were categorized in a favorable, intermediate and poor risk group, with corresponding survival times. Median OS times for the training cohort were 34.3, 16.1 and 5.2 months, respectively (figure 1C). For the validation cohort the median OS times were respectively 22.3, 15.2 and 10.6 months (figure 1D). For both independent cohorts the measurement of only SAA was equally accurate in predicting OS as the multi-factorial MSKCC model, with AIC = 715 (p = 3.0 x 10-9) and AIC = 717 (p = 4.6 x 10-9) for the training cohort and respectively AIC = 1084 (p = 2.2 x 10-5) and AIC = 1083 (p = 2.2 x 10-5) for the validation cohort for SAA and MSKCC model (Table 4). sAA improves the prognostic accuracy of the MsKCC modelThe MSKCC model is based on dichotomized risk factors. We therefore determined the optimal predictive cut off value for SAA-levels using ROC-analysis. The ROC-curve for SAA in the training cohort significantly demonstrated OS (figure 1e, p<0.001, AUC 0.79 (95% CI: 0.68 - 0.89)). Based on this ROC-curve SAA-levels were dichotomized using 71 ng/ml as an optimal significant cut-off value (sensitivity and specificity of respectively 54% and 88%). For the validation cohort the ROC-curve for SAA considerably predicted OS as well (figure 1f, p<0.001, AUC 0.73 (95% CI: 0.63 – 0.83)). The cut-off value of 71 ng/ml was validated in the validation cohort with a remarkably similar sensitivity and specificity of respectively 59% and 77%. Applying these dichotomized SAA-levels for Kaplan-Meier analysis demonstrated that SAA was consistently predictive for PFS in either cohorts (p = 3.6 x 10-7 and p = 1.3 x 10-7, respectively, figure 1 g). Similar PFS times were found for the favorable (14.1 and 12.6 months) and poor risk group (2.8 and 5.0 months), respectively for training and validation cohort. Importantly, using 71 ng/ml as cut-off value SAA also significantly predicted OS (Table 4, AIC = 726, p = 1.8 x 10-6 and AIC = 1080, p = 1.8x10-6,respectively) and comparable median OS times were observed in the training and validation cohort (figure 1H), respectively 27.3 months (95% CI: 23.0 – 31.5) and 24.6 months (95% CI: 21.9 – 29.2) for the favorable group and respectively 13.2 months (95% CI: 9.1 – 16.7) and 14.4 months (95% CI: 11.7 – 17.1) for the poor risk category. Including dichotomized SAA (> 71ng/ml) as additional poor risk factor the prognostic accuracy of the MSKCC model in both cohorts improved significantly (Table 4, AIC = 713 (p = 4.3 x 10-12) and AIC = 1073 (p = 7.4 x 10-8) for respectively training and validation cohort).

Part II Proteomics

101

Tabl

e 3

- R

esul

ts o

f Mul

tivar

iabl

e A

naly

sis

for

Pro

gres

sion

-Fre

e S

urvi

val a

nd O

vera

ll S

urvi

val

Pro

gres

sion

-fre

e su

rviv

alo

vera

ll s

urvi

val

Trai

ning

Coh

ort N

=11

4Va

lidat

ion

Coh

ort N

=1

51

Tra

inin

g C

ohor

t N=

114

Va

lidat

ion

Coh

ort N

=1

51

Varia

ble

Haz

ard

ratio

(9

5% C

I)P

-val

ueH

azar

d ra

tio

(95%

CI)

P-v

alue

Haz

ard

ratio

(95

% C

I)P

-val

ueH

azar

d ra

tio (

95%

CI)

P-v

alue

SA

A1.

001

(1

.000

-1.0

02)

0.00

21.

001

(1

.000

-1.0

02)

<0.

001

1.00

(0

.99

9-1.

001)

0.00

21.

00

(0.9

99-

1.00

1)0.

002

EC

OG

(0 v

ersu

s ≥1

)1.

7

(1.0

-2.7

)0.

04

––

2.2

(1

.3-3

.7)

0.00

3–

–

LDH

––

1.00

1

(1.0

00-1

.002

)<

0.00

11.

00

(0.9

99-

1.00

1)<

0.00

11.

00

(0.9

99-

1.00

1)<

0.00

1

TDT

(≥1

vers

us <

1 ye

ar)

––

––

1.7

(1

.0-2

.7)

0.03

––

≥2 m

etas

tatic

Site

sno

ver

sus

yes

––

––

1.8

(1.1

-3.1

)0.

02–

–

Alb

umin

0.6

(0.4

-0.8

)0.

005

––

––

––

Leuk

ocyt

es–

–1.

13

(1.0

5-1.

21)

0.00

2–

–1.

1

(1.0

1-1.

18)

0.02

7

Cor

rect

ed C

alci

um–

––

––

–1.

4

(1

.05-

1.74

)0.

018

Not

e: F

or c

ontin

uous

var

iabl

es, a

haz

ard

ratio

>1

= ri

sk re

duct

ion

whe

n th

e va

lue

decr

ease

s an

d a

haza

rd ra

tio <

1 =

risk

redu

ctio

n w

hen

the

valu

e in

crea

se;

For b

inar

y va

riabl

es (c

ateg

orie

s in

bol

d), a

haz

ard

ratio

>1

= ri

sk re

duct

ion

for t

he fi

rst c

ateg

ory

and

a ha

zard

ratio

<1

= ri

sk re

duct

ion

for t

he s

econ

d ca

tego

ryFo

r sev

eral

labo

rato

ry m

arke

rs, h

azar

d ra

tios

are

clos

e to

1 d

ue to

the

cont

inuo

us s

cale

of t

he m

arke

r.Th

is ta

ble

is in

ana

logy

with

the

tabl

e pu

blis

hed

by P

atil

et a

l. [2

0]


102

figure 1 Kaplan-Meier survival analysis of the MSKCC risk model for the training cohort (A) (AIC = 717, p = 3.0 x 10-9) and the validation set (B) (AIC = 1083, p = 2.2 x 10-5) and according to the SAA risk model (C and D) (AIC = 715, p = 4.6 x 10-9 and AIC = 1084, p = 2.2 x 10-5, respectively) based on tertiles as objective cut off values, demonstrated identical prognostic accuracy as the MSKCC model. ROC curves for SAA of the training cohort (e) elucidated an AUC of 0.79 (95% CI: 0.68 – 0.89, p<0.001), showing a sensitivity and specificity or our selected cut-off value (71 ng/mL) was respectively 54% and 88%. This was confirmed in the validation set (f) demonstrating a similar AUC of 0.73 (95% CI: 0.63 – 0.83, p<0.001) and when applying the predefined cut-off value (>71 ng/ml) a remarkably comparable sensitivity and specificity of 59% and 77% was observed. Accordingly, increased SAA-levels (>71 ng/ml) were associated with worse PFS (g) in both the Training and Validation cohort, (p = 3.6 x 10-7 and 1.3 x 10-7), which also held true for poor OS (H) (p = 1.8 x 10-6 and 1.8 x 10-6, respectively). Expanding the MSKCC model with SAA (>71ng/mL) improved the prognostic accuracy in both the training (I) (AIC = 713, p = 4.3 x10-12) and validation (J) (AIC = 1073, p = 7.4 x10-8) cohort.

figure 1 - Kaplan-Meier survival analysis

Part II Proteomics

A

C

B

D

103Validation of SAA as an Independent Biomarker for Survival in mRCC Patients Chapter 6

e

g

I

f

H

J

104

Kaplan-Meier curves displayed vigorously discriminative consequences with improvement of OS times for every MSKCC risk group (Table 4, figure 1 I and 1 J), implicating that applying SAA to the MSKCC model a substantial number of patients might switch between risk groups with a consequence for choice of treatment strategy. Importantly, incorporating SAA as continuous variable also improved the MSKCC model (data not shown). longitudinal sAA-levels early Predicts Disease Progression Using our prior defined cut-off value, patients were categorized into four groups according to longitudinal alternating SAA-levels (figure 2A);i) consistently-elevated SAA-levels (>71 ng/mL) (n = 29); ii) increased SAA-baseline levels (>71 ng/mL), but declined at 6-8 weeks (n = 7); iii) low SAA-baseline quantities (<71 ng/mL) (n = 8) with increased SAA-levels at 6-8 weeks; iv) sustained low SAA-concentration (<71 ng/mL) for both time points (n = 28). These categorized groups were significantly associated with PFS (figure 2B, p = 0.02). Significant differences in survival times were observed between the consistently-elevated and sustained low SAA-levels (median PFS time 5.1 versus 11.0 months). However, changes in SAA-levels during treatment did not have sufficient value to predict treatment outcome.

figure 2A) Patients categorized by longitudinal alternating SAA-levels during TKI-treatment.B) SAA-levels significantly predicted 1-Year-PFS in an early stage after 6-8 weeks at 2nd evaluation.

figure 2 - Longitudinal SAA-levels did not demonstrate an early prediction for disease progression

Part II Proteomics

A

B

105

Tabl

e 4

- Va

lidat

ion

of R

isk

Mod

els

(Ove

rall

Sur

viva

l)

Trai

ning

Coh

ort;

n=

114

Valid

atio

n C

ohor

t; n

=15

1

Sur

viva

l Mod

elR

isk

Fact

ors

AIC

P-V

alue

Ris

k Fa

ctor

sN

o.

Pat

ient

sM

edia

n O

S

(mon

ths)

AIC

P-V

alue

Ris

k Fa

ctor

sN

o.

Pat

ient

sM

edia

n O

S

(mon

ths)

MS

KC

C R

isk

Mod

elM

otze

r et a

l, [1

]H

b, L

DH

, cor

rect

ed

Cal

cium

, EC

OG

an

d Ti

me

from

D

iagn

osis

to S

tudy

E

ntry

717

3.0

x 10

-9

0 †

243

9.1

1083

2.2

x 10

-5

0 †

1831

.0

1 or

26

915

.71

or 2

8317

.2

≥ 3

214.

4≥

35

06.

6

Ser

um A

myl

oid

Alp

haTe

rtile

s71

54.

6 x

10-9

Low

est †

38

34.

310

84

2.2

x 10

-5Lo

wes

t †5

022

.3

Mid

dle

38

16.1

Mid

dle

5115

.2

Hig

hest

38

5.2

Hig

hest

50

10.6

Ser

um A

myl

oid

Alp

haO

ptim

al C

ut-o

ff Va

lue

726

1.8

x 10

-6<

71

ng/m

l †62

27.3

108

01.

8 x

10-6

< 7

1 ng

/ml †

69

25.6

> 7

1 ng

/ml

5213

.2>

71

ng/m

l82

14.4

MS

KC

C M

odel

E

xpan

ded

with

SA

AH

b, L

DH

, cor

rect

ed

Cal

cium

, EC

OG

, Ti

me

from

Dia

gnos

is

to S

tudy

Ent

ry,

and

SA

A

713

4.3

x 10

-12

0

†20

not r

each

ed10

737.

4 x

10-8

0

†16

34.

3

145

17.3

145

19.2

2 or

329

10.3

2 or

35

013

.9

≥ 4

204.

0≥

44

06.

1

Abb

revi

atio

ns: A

IC; A

kaik

e’s

Info

rmat

ion

Crit

eria

, Hb;

Hem

oglo

bin,

LD

H; L

acta

te D

ehyd

roge

nase

, EC

OG

; E

aste

rn C

oope

rativ

e O

ncol

ogy

Gro

up, S

AA

; Ser

um A

myl

oid

Alp

ha, A

poA

2; A

polip

opro

tein

-A2

† R

isk

grou

p us

ed a

s re

fere

nce.


106

DI sCus s Ion

The present study validated SAA – but not ApoA2- as a robust and independent predictor for PFS and OS in two independent mRCC patients cohorts. When SAA-levels were objectively trichotomized, a prognostic model was constructed with identical prognostic accuracy when compared to the MSKCC model in separate patient sets. Interestingly, measuring only SAA-levels was equally effective in predicting OS as the commonly used multi-factorial MSKCC model. For accurate prognostication this might implicate that the multi-factorial MSKCC model could be replaced by one simple SAA determination, as accurate SAA measurements can straightforwardly be implemented in any hospital. Moreover, incorporating SAA in the MSKCC model improved its prognostic significance demonstrating that a substantial number of patients might switch between risk groups. Expanding the MSKCC model with SAA further improved the discrimination of particularly the intermediate risk category. The importance of this finding is illustrated by the significant differences in clinical outcome of anti-VEGF treated mRCC patients who were mainly categorized in the intermediate risk group (27;28). Therefore baseline SAA-levels might contribute to a better a-priori selection for risk-directed therapy, thereby avoiding long-term schemes of toxic treatment modalities for patients who will probably not profit (9). Changes in SAA-levels during TKI-treatment did not reveal SAA as an early marker for disease progression. Since this analysis was performed in a relatively small number of patients receiving two different treatment modalities no definitive conclusion can be drawn at this stage. Importantly, ApoA2 could not be validated as independent biomarker. As reviewed (30), SAA is either produced by hepatocytes, tumor or inflammatory cells, contributing to inflammation, tumor evolution and dissemination. Recently, it was illustrated that SAA was predictive for OS in large patient cohorts of several cancers(RCC, lung, breast and melanoma) (31-35), supporting our outcomes. However, this is the first investigation that demonstrated the significant relation of SAA with PFS and that expanding the MSKCC with SAA improved its predictive accuracy. Previously a correlation between SAA and C-reactive protein – another acute-phase protein- was found (36-38), however SAA was independently of C-reactive protein prognostic for OS in melanoma and RCC patients (34;35), which also holds true for a subset of our patients (n = 76, data not shown). As hypothesized (22;30), SAA reflects actual disease status and might be considered as a pan-biomarker for disease aggressiveness irrespective of tumor type or treatment modality. Our study used two independent cohorts of mRCC patients representative for the general population of patients with this disease. The survival of both separate cohorts was comparable and importantly, was similar to that reported in previous studies (20;21). The present study also has several limitations. In the training cohort patients received mainly upfront interferon-based treatment, whereas in the validation cohort TKIs were initially given at time of biomarker analysis. For both patient sets differences in pretreatment and additional therapies were described. The heterogeneity regarding treatment might influence patients’ outcome. Additionally PFS was not assessed using well-defined RECIST criteria.

Part II Proteomics

107

C onClus Ion

In mRCC patients SAA is a robust and independent prognosticator for PFS and OS, regardless of treatment modality. Using SAA-levels, a prognostic risk model was constructed consisting of three risk categories with identical predictive accuracy when compared to the MSKCC model. Expanding the MSKCC model with SAA showed additional prognostic value for patient management. The value of measuring SAA during TKI treatment as an early marker for response was not established. It remains questionable whether there will be a single biomarker which might sufficiently identify treatment responders and outcome. Consequently, SAA -either alone or as part of the MSKCC model- may be used as an objective prognostic marker in future clinical trials in mRCC patients and may improve the a-priori selection of appropriate risk-directed therapy for every individual.

TAK e HoM e M e s sAg e

SAA may be used (alone or as part of the MSKCC model) as an objective prognostic marker in future clinical trials in mRCC patients and may improve the a-priori selection of appropriate risk-directed therapy for every individual.


The authors are grateful for the assistance of all medical doctors, researchers, technicians and all affiliated chemical laboratories who were involved in patient management and blood processing.


108



trials of new therapies against advanced renal cell carcinoma. J Clin Oncol 2002 January 1;20(1):289-96.


renal cell carcinoma (mRCC): a literature review. Cancer Treat Rev 2008 May;34(3):193-205.

(3) Di LG, Autorino R, Sternberg CN. Metastatic renal cell carcinoma: recent advances in the targeted therapy era

Eur Urol 2009 December;56(6):959-71.

(4) Motzer RJ, Hutson TE, Tomczak P, Michaelson MD, Bukowski RM, Rixe O et al. Sunitinib versus interferon alfa

in metastatic renal-cell carcinoma. N Engl J Med 2007 January 11;356(2):115-24.

(5) Hudes G, Carducci M, Tomczak P, Dutcher J, Figlin R, Kapoor A et al. Temsirolimus, interferon alfa, or both for

advanced renal-cell carcinoma. N Engl J Med 2007 May 31;356(22):2271-81.

(6) Escudier B, Pluzanska A, Koralewski P, Ravaud A, Bracarda S, Szczylik C et al. Bevacizumab plus interferon

alfa-2a for treatment of metastatic renal cell carcinoma: a randomised, double-blind phase III trial. Lancet 2007

December 22;370(9605):2103-11.

(7) Sternberg CN, Davis ID, Mardiak J, Szczylik C, Lee E, Wagstaff J et al. Pazopanib in locally advanced or

metastatic renal cell carcinoma: results of a randomized phase III trial. J Clin Oncol 2010 February 20;28(6):1061-8.

(8) Motzer RJ, Escudier B, Oudard S, Hutson TE, Porta C, Bracarda S et al. Phase 3 trial of everolimus for

metastatic renal cell carcinoma : final results and analysis of prognostic factors. Cancer 2010 September

15;116(18):4256-65.

(9) Rini BI. Metastatic renal cell carcinoma: many treatment options, one patient. J Clin Oncol 2009 July 1;27(19):3225-34.

(10) Bex A, Jonasch E, Kirkali Z, Mejean A, Mulders P, Oudard S et al. Integrating surgery with targeted therapies for

renal cell carcinoma: current evidence and ongoing trials. Eur Urol 2010 December;58(6):819-28.

(11) Patard JJ, Pignot G, Escudier B, Eisen T, Bex A, Sternberg C et al. ICUD-EAU International Consultation on

Kidney Cancer 2010: treatment of metastatic disease. Eur Urol 2011 October;60(4):684-90.

(12) Motzer RJ, Mazumdar M, Bacik J, Berg W, Amsterdam A, Ferrara J. Survival and prognostic stratification of 670

patients with advanced renal cell carcinoma. J Clin Oncol 1999 August;17(8):2530-40.

(13) Mekhail TM, bou-Jawde RM, Boumerhi G, Malhi S, Wood L, Elson P et al. Validation and extension of the

Memorial Sloan-Kettering prognostic factors model for survival in patients with previously untreated metastatic

renal cell carcinoma. J Clin Oncol 2005 February 1;23(4):832-41.

(14) Atkins MB, Choueiri TK, Cho D, Regan M, Signoretti S. Treatment selection for patients with metastatic renal

cell carcinoma. Cancer 2009 May 15;115(10 Suppl):2327-33.

(15) Escudier B, Albiges L, Blesius A, Loriot Y, Massard C, Fizazi K. How to select targeted therapy in renal cell

cancer. Ann Oncol 2010 October;21 Suppl 7:vii59-vii62.

(16) Hudes GR, Carducci MA, Choueiri TK, Esper P, Jonasch E, Kumar R et al. NCCN Task Force report: optimizing

treatment of advanced renal cell carcinoma with molecular targeted therapy. J Natl Compr Canc Netw 2011

February;9 Suppl 1:S1-29.

(17) Motzer RJ, Agarwal N, Beard C, Bhayani S, Bolger GB, Carducci MA et al. Kidney cancer. J Natl Compr Canc

Netw 2011 September 1;9(9):960-77.

(18) Bamias A, Karadimou A, Lampaki S, Lainakis G, Malettou L, Timotheadou E et al. Prognostic stratification of

patients with advanced renal cell carcinoma treated with sunitinib: comparison with the Memorial Sloan-Kettering

prognostic factors model. BMC Cancer 2010;10:45.

(19) Donskov F, von der MH. Impact of immune parameters on long-term survival in metastatic renal cell carcinoma.

J Clin Oncol 2006 May 1;24(13):1997-2005.

(20) Patil S, Figlin RA, Hutson TE, Michaelson MD, Negrier S, Kim ST et al. Prognostic factors for progression-free

Part II Proteomics

109

and overall survival with sunitinib targeted therapy and with cytokine as first-line therapy in patients with

metastatic renal cell carcinoma. Ann Oncol 2011 February;22(2):295-300.

(21) Manola J, Royston P, Elson P, McCormack JB, Mazumdar M, Negrier S et al. Prognostic Model for Survival in

Patients with Metastatic Renal Cell Carcinoma: Results from the International Kidney Cancer Working Group.

Clin Cancer Res 2011 August 15;17(16):5443-50.

(22) Vermaat JS, van dT, I, Mehra N, Sleijfer S, Haanen JB, Roodhart JM et al. Two-protein signature of novel

serological markers apolipoprotein-A2 and serum amyloid alpha predicts prognosis in patients with metastatic

renal cell cancer and improves the currently used prognostic survival models. Ann Oncol 2010 July;21(7):1472-81.

(23) Heng DY, Xie W, Bjarnason GA, Vaishampayan U, Tan MH, Knox J et al. Progression-free survival as a predictor

of overall survival in metastatic renal cell carcinoma treated with contemporary targeted therapy. Cancer 2010

November 18.

(24) Motzer RJ, Bukowski RM, Figlin RA, Hutson TE, Michaelson MD, Kim ST et al. Prognostic nomogram for sunitinib

in patients with metastatic renal cell carcinoma. Cancer 2008 October 1;113(7):1552-8.

(25) Karakiewicz PI, Sun M, Bellmunt J, Sneller V, Escudier B. Prediction of progression-free survival rates after

bevacizumab plus interferon versus interferon alone in patients with metastatic renal cell carcinoma: comparison

of a nomogram to the motzer criteria. Eur Urol 2011 July;60(1):48-56.

(26) Frank E.Harrell. Regression Modeling Strategies With Applications to Linear Models, Logistic Regression, and

Survival Analysis. New York: Springer; 2001.

(27) Akaike H. Information theory and an extension of the maximum likelihood principle. Budapest, Hungary:

Akademiai Kaiado; 1973.

(28) Heng DY, Mackenzie MJ, Vaishampayan UN, Bjarnason GA, Knox JJ et al. Primary anti-vascular endothelial

growth factor (VEGF)-refractory metastatic renal cell carcinoma: clinical characteristics, risk factors, and

subsequent therapy Ann of Oncol 2011; Nov 5. [Epub ahead of print]

(29) Heng DY, Xie W, Regan MM, Warren MA, Golshayan AR, Sahi C, et al. Prognostic factors for overall survival

in patients with metastatic renal cell carcinoma treated with vascular endothelial growth factor-targeted agents:

results from a large, multicenter study. J of Clin Oncol 2009; Dec 1;27(34):5794-9

(30) Malle E, Sodin-Semrl S, Kovacevic A. Serum amyloid A: an acute-phase protein involved in tumour pathogenesis.

Cell Mol Life Sci 2009 January;66(1):9-26.

(31) Ramankulov A, Lein M, Johannsen M, Schrader M, Miller K, Loening SA et al. Serum amyloid A as indicator of

distant metastases but not as early tumor marker in patients with renal cell carcinoma. Cancer Lett 2008

September 28;269(1):85-92.

(32) Kimura M, Tomita Y, Imai T, Saito T, Katagiri A, Ohara-Mikami Y et al. Significance of serum amyloid A on the

prognosis in patients with renal cell carcinoma. Cancer 2001 October 15;92(8):2072-5.

(33) Pierce BL, Ballard-Barbash R, Bernstein L, Baumgartner RN, Neuhouser ML, Wener MH et al. Elevated

biomarkers of inflammation are associated with reduced survival among breast cancer patients. J Clin Oncol

2009 July 20;27(21):3437-44.

(34) Wood SL, Rogers M, Cairns DA, Paul A, Thompson D, Vasudev NS et al. Association of serum amyloid A protein

and peptide fragments with prognosis in renal cancer. Br J Cancer 2010 June 29;103(1):101-11.

(35) Findeisen P, Zapatka M, Peccerella T, Matzk H, Neumaier M, Schadendorf D et al. Serum amyloid A as a

prognostic marker in melanoma identified by proteomic profiling. J Clin Oncol 2009 May 1;27(13):2199-208.

(36) Yamada T. Serum amyloid A (SAA): a concise review of biology, assay methods and clinical usefulness. Clin

Chem Lab Med 1999 April;37(4):381-8.

(37) Manley PN, Ancsin JB, Kisilevsky R. Rapid recycling of cholesterol: the joint biologic role of C-reactive protein

and serum amyloid A. Med Hypotheses 2006;66(4):784-92.

(38) Raynes JG, Cooper EH. Comparison of serum amyloid A protein and C-reactive protein concentrations in cancer

and non-malignant disease. J Clin Pathol 1983 July;36(7):798-803.



Chapter 7 Circulating cell-free mitochondrial DnA as prognostic but not predictive marker for various types of cancer

Joost S Vermaat, Jeanine M Roodhart, Frank L Gerritse, Marije G Gerritsen, Wijnand M Roessingh, Patrick H van Zon, Laura G M Daenen, Niven Mehra, Hans Kristian Ploos van Amstel, Rene H Medema, and Emile E Voest

Submitted

112 Circulating cell-free mitochondrial DNA as prognostic but not predictive marker for various types of cancer Chapter 7Part III MITOX


Previous studies have demonstrated circulating (genomic) cell-free nucleic acids as promising diagnostic and prognostic markers in cancer patients. In this study we have investigated the prognostic and predictive signifi cance of mitochondrial-(mt)DNA as pan-tumor marker in different human cancers. A robust qPCR assay which stringently determined mtDNA-levels in plasma of cancer patients was developed. In a large training cohort comprising of 310 patients with different cancer types we demonstrated that mtDNA accurately predicted Overall Survival (OS) for several tumor types, however not for every tumor type. A subsequent prospective validation cohort including 124 patients with diverse tumor types confi rmed the association of mtDNA with OS for various cancer types. However, in this validation set mtDNA had no value in predicting response to treatment. Conclusively, this is the fi rst study that showed that mtDNA can be considered as a pan-tumor prognostic indicator.

113Circulating cell-free mitochondrial DNA as prognostic but not predictive marker for various types of cancer Chapter 7

AB sTRACT

Background and objectivePrevious studies have demonstrated circulating cell-free nucleic acids as promising diagnostic and prognostic markers in cancer patients. Here, we investigated the prognostic and predictive significance of mitochondrial-(mt)DNA as pan-tumor marker in different human cancers.

Patients and methodsIn a training cohort of 310 patients with multiple cancer types, we studied the prognostic value of mtDNA. Subsequently, a prospective validation cohort of 124 patients with different tumor types and various chemotherapeutic modalities, were recruited to confirm mtDNA as prognosticator and to determine its predictive value for therapy response. mtDNA was isolated from plasma and quantified by a developed and robust quantitative PCR. mtDNA-levels were associated with Progression Free Survival (PFS) and Overall Survival (OS) using univariate regression analysis.

Results Comparable median baseline mtDNA concentrations between the training and validation cohorts were obtained, 2515 and 2541 molecules/µL, respectively. No differences in mtDNA-levels between the tumor types were detected. In the training cohort mtDNA significantly predicted OS (p<0.001), but not for every tumor type. Using tertiles as objective cut-off values, patients were categorized into a favorable (<1644 molecules/µL), intermediate (1644-3575 molecules/µL) and poor risk group (>3575 molecules/µL), demonstrating significant discrimination in OS times between risk groups using Kaplan-Meier analysis (p<0.001). Applying these pre-defined mtDNA cut-off values to our validation cohort, Kaplan-Meier analysis critically distinguished the favorable and intermediate from the poor risk group (p=0.004). However, mtDNA-levels were not predictive for response to chemotherapy in the validation set. ConclusionCirculating mtDNA-levels in plasma of cancer patients, determined by a straightforward qPCR, accurately predicted OS for several tumor types.

114 Circulating cell-free mitochondrial DNA as prognostic but not predictive marker for various types of cancer Chapter 7

AB B R eVIATIon s

cfNAs; circulating cell-free nucleic acids, NA; nucleic acids, gDNA; genomic DNA, mtDNA; mitochondrial DNA, qPCR; quantitative Polymerase Chain Reaction, OS; overall survival PFS; progression free survival, mt-COI; mitochondrial Cytochrome Oxidase Subunit I, mt-RNR1; mitochondrial 12S RNA, IQR; Interquartile ranges, 95%CI; 95% Confidence Interval CNAPs; circulating cell-free nucleic acids in plasma or serum

I nTRoDuCTIon

Circulating-free nucleic acids (cfNAs) in blood can be used as potential diagnostic and prognostic cancer markers (1;2). cfNAs cover all in plasma circulating nucleic acids (NA), including genomic DNA (gDNA), mitochondrial DNA (mtDNA), (m)RNA, microRNAs, microsatellites, viruses and nucleosomes. Low cfNAs-levels are present in healthy individuals, originating from hematopoietic blood cells (3). In cancer patients, increased cfNAs-levels were determined originating from apoptosis, necrosis, lysis, and spontaneous secretion by tumor cells (4-8). Elevated cfNA-levels are also found in other (bio)pathological processes such as inflammation, autoimmune disorders, ischemia or trauma (1). Interestingly, mutations present in tumor DNA were also identified as similar variants in genomic cfNAs in the same patient, indicating that these elevated cfNAs (partly) originate from tumor cells (1;2;9-12). This highlights the possibility to identify tumor-specific alterations in a non-invasive way (13-17). Initially cfNAs have been proposed as a diagnostic cancer marker, unfortunately with conflicting results showing large variations in cfNAs-levels between different cancer types (18) and discrepancies in sensitivities and specificities (1;2;19). This limits the use of well-defined cut-off values for specific cancer types in diagnosis/prognostication and prohibits broad applicability. Furthermore, genomic cfNAs have been studied as prognosticator and as follow-up marker during treatment in cancer patients (20;21). In various malignancies increased cfNAs-levels prior to treatment were associated with poor prognosis, and decreased cfNAs-levels after treatment were related with response (22-26) and/or associated with improved disease-free and overall survival (OS) (24;27). Besides the genomic cfNAs, extrachromosomal mitochondrial cfNAs have been investigated (28). Using real-time amplification, plasma mtDNA-levels of healthy subjects (29) and cancer patients have been quantified (30). mtDNA-quantification might yield increased sensitivity or specificity over genomic cfNAs, as in general a single cell differs in number of mitochondria NAs from several to thousands copies of mtDNA, compared to invariably 2 copies in gDNA (20). Remarkably, when compared to healthy controls both decreased and elevated mtDNA-levels are found in patients with various malignancies (31-35). In prostate cancer patients increased mtDNA-levels correlated with poor prognosis after radical prostatectomy (30;36). Accordingly, we showed that mtDNA-quantification had increased sensitivity and specificity as diagnostic and prognostic marker over genomic cfNAs (30). However this observation could not be confirmed for breast cancer (34). It is unclear whether the prognostic findings in prostate cancer (30) may be translated to other tumor types, and whether mtDNA may be used as a pan-tumor marker.

Part III MITOX

115

Therefore, we embarked on a study to evaluate mtDNA as a prognostic and predictive marker for multiple cancer types. NAs were extracted from plasma samples and a quantitative Polymerase-Chain-Reaction (qPCR) assay for accurate determination of mtDNA-levels was developed. In a training cohort including 310 patients with different tumor types, elevated mtDNA-levels significantly predicted poor OS. Next, mtDNA-levels were confirmed for their significant prognostication in a validation set of 124 various cancer patients. However, in this validation cohort mtDNA-levels were not predictive for chemotherapy response. PATI e nTs An D M eTHoDs

Characterization of Training and Validation CohortsTo investigate the prognostic value of plasma mtDNA in cancer patients, we generated a training cohort of consecutive patients visiting the outpatient clinic of the University Medical Utrecht Cancer Center (UMCU), the Netherlands. Between 2003-2007 citrate blood samples were collected from 310 untreated, new or relapsed cancer patients with various tumor types. Each patient received standard treatment according to their pathology and disease stage. Follow-up finished when patients died from disease or when they completed 48 months. To validate the results of our training cohort, we prospectively assembled a validation set comprehending 124 consecutive patients with diverse tumor types, receiving maximum tolerated dose (MTD) chemotherapy in a three weekly schedule either as (neo-)adjuvant or as chemotherapy for metastatic disease. Because we hypothesized that tumor cell death induced by chemotherapy leads to increased release of mtDNA, tumor type was considered to be of less importance, allowing a cohort with different tumor types. Besides prognostication, the predictive significance of mtDNA-levels for therapy response, marked by progression free survival (PFS), was also explored. Patients with previous chemotherapy or surgery within 4 weeks were excluded. Patients were recruited between 2006-2010 in UMCU and follow-up ended June 2011. Blood sampling was performed before the 1st cycle of chemotherapy. In a subgroup plasma was collected 2 and 4 hours after the start of chemotherapy initiation, 7 days thereafter and immediately before the 2nd cycle (day 21). This study was approved by the hospital Ethics Committee and written informed consent was obtained from all patients. Response evaluation was performed after the 3rd cycle of chemotherapy according to RECIST criteria. PFS and OS were defined as time from start chemotherapy respectively to date of tumor progression and at time of patients’ tumor-related death or when follow-up ended after 48 months. Patient characteristics were recorded at time of first blood collection.

Isolation and purification of specific mitochondrial genes Using MyCycler Thermal Cycler (BioRad, Hercules, CA) gDNA of a healthy subject was amplified with specific primers for the mitochondrial genes Cytochrome Oxidase Subunit-I (mt-COI) and mitochondrial 12S-RNA (mt-RNR1) (Hs02596864_g1 and Hs02596859_g1 respectively, Applied Biosystems (AB, Carlsbad, CA) and Taq DNA-polymerase (NewEngland Biolabs, Ipswich, MA, #M0267S). Next, DNA amplicons were size-selected (length 100-150bp) and purified (MinElute Gel Extraction Kit (Qiagen, Valencia, CA)). DNA templates were ligated into the PGEM-T Easy Vector (Promega,Madison,WI), transformed in DH5α Escheria Coli cells (Invitrogen Life Technologies, Carlsbad, CA) and incubated with Lysine-Buffer and Ampicillin on

Circulating cell-free mitochondrial DNA as prognostic but not predictive marker for various types of cancer Chapter 7


deze ga ik opnieuw maken

IPTG/X-gal agar plates. After plasmid DNA purification the mitochondrial inserts were secluded using restriction enzymes (Promega, Madison, WI). mtDNA inserts were checked on gel for contamination and sequenced using Sanger Sequencing to affirm its specific identity using software program VECTOR NTI suite-9 and verified with the NCBI database.

standard curveDNA-concentrations were determined using the Qubit® fluorometer (Invitrogen Life Science, Carlsbad, CA). The number of molecules per µL was calculated using the formula: molecules per microliter = (( x * A ) / ( l * 109

* 650)); where x is the amount of DNA (ng/µL), A constant of Avogadro, l vector length (bp) (obtained from Applied Biosystems (AB)). Using serial dilutions a standard curve for both genes was constructed.

Plasma isolation After blood collection with CPT Vacutainer tubes with sodium citrate (BD Biosciences, Mountain View, CA), plasma was centrifuged within two hours at 1600 RCF for 30 minutes. The plasma was carefully removed without disturbing the buffy coat, transferred to a 15 ml tube, and centrifuged a second time at 3200 RCF for 15 minutes at 4°C to remove possible cellular contamination, large aggregates, and eliminate platelets from plasma (defined as two-spin or platelet-poor plasma). Isolated plasma was stored at –80°C.

DnA nucleic Acid IsolationPlasma samples were thawed on ice and briefly spun down. Plasma NA were isolated using QIAamp DNA Mini Kit (Qiagen, Valencia, CA) following manufacturer’s protocol. Briefly, 400µL plasma was incubated with protein-kinase and AL-buffer. Ethanol was added and the entire mixture was applied to the silica-membrane-based column and spun down. After washing the column twice, DNA was eluted with 100 µL purified enzyme-free H20 and stockpiled in -20°C freezer.

Quantitative Polymerase Chain Reaction (qPCR)All samples and standard curves were measured with Abi Prism 7000 Sequence Detecting System (AB) using previous described primers for mt-COI and mt-RNR1. As positive control standard curves were taken along every measurement and purified enzyme-free water was used as negative control. In total 40-50 amplification rounds (2 minutes at 50°C followed by 10minutes at 95°C, and thereafter 15 seconds at 95°C and 1 minute at 60°C) were performed to generate Cycle threshold (Ct)-values and subsequently used to calculate mtDNA copy numbers.

statistical Analysis Raw qPCR data were analyzed using ABI Prism 7000 Sequence Detection System (SDS) software (AB) and Ct-values were computed. Student’s T-test and ANOVA-test evaluated differences in mtDNA-levels between two or more groups, respectively. Spearman Correlation associated mtDNA-quantities with other patients’ characteristics. Uni/multivariate Cox Proportional Hazard (PH) regression correlated mtDNA-levels with PFS and OS. mtDNA-levels were objectively dichotomized to classify risk groups for Kaplan-Meier estimation. Log-rank statistics assessed deviations in survival times between categorized risk groups. Results were analyzed using SPSS software (version 15.0). Error bars depicted standard errors of the mean. P-values below 0.05 (two-sided) were considered significant.

Part III MITOX

117

R e s u lTs

mtDnA-quantification by a developed qPCR-assayTo develop an accurate qPCR gDNA from a healthy subject was augmented specific primers for mtCOI and mtRNR1. After plasmid mtDNA isolation and restriction mtDNA amplicons were checked on agarose gel (supplemental figure 1A). Plasmid DNA of the recombinants for mtCOI and mtRNR1 were linearized using SpeI (New England Biology, Ipswich, MA) show-ing inserts with a fragment length of 113 and 135bp for mtCOI and mtRNR1, respectively with subsequent nucleotide sequence (TCAGCAGTCCTACTTCTCCTATCTCTCCCAGTCCTAGCT-GCTGGCATCACTATACTACTAACAGACCGCAACCTCAACACCACCTTCTTCGACCCA and TGGCACGAAATTGACCACCCTGGGGTTAGTATAGCTTAGTTAAACTTTCGTTTATTGCTA-AAGGTTAATCACTGCTGTTTCCCGTGGGGGTGTGGCTAGGCTAAGCGTTTTGAGCTGCATA)(Hs02596864_g1 and Hs02596859_g1 respectively, Applied Biosystems). mtDNA-concentra-tions were 0,77 and 0,96 ng/µL and correlated with 627,000 and 658,000 molecules/µL for mt-COI and mtRNR1. Standard curves were constructed using serial dilutions (10-2x, 10-4x, 10-6x, 10-

7x and 10-8x dilutions) and were taken along for every measurement showing a small inter-variability between determinations (all analysis R2 ≥ 0.97, supplemental figure 1B/1C). All quantifications were assessed in triplicate. Of 18 samples only duplo measurements were available as one PCR failed. The mean coefficient of variation was 7.5% with a standard deviation of 7.4%, demonstrat-ing a small intra-variability between triplicate samples. To explore the reproducibility of our devel-oped qPCR-test, we repeated the analysis for eleven samples, showing a Pearson-correlation of 0.96 (p<0.001 two-tailed, supplemental figure 1D). By employing these quality assessments we confirmed that qPCR technique was a robust assay for accurate mtDNA-determination.

Patient characteristics: training and validation cohortBy using NASBA (Nucleic Acid Sequence Based Amplification) as a different methodology for mtDNA-quantification we previously showed the prognostic value of mtDNA in prostate cancer patients (30). To confirm that our novel qPCR-based assay gave similar results we analyzed a total of 310 cancer patients as a training cohort, including breast (N=39), cervix (N=8), colorectal (N=45), Ear/Nose/Throat (ENT, N=36), ovarian (N=22), prostate (N=86) and Renal Cell Cancer (RCC, N=74) patients. Table 1 depicts patient characteristics. The majority of patients had metastatic disease (83%) and received anti-cancer therapy (90%), according to local guidelines. At time of analysis 251 patients died from cancer whereas the other 59 patients were censored. The median OS time was 17.2 months (95% CI: 14.0 – 20.4 months). Breast cancer patients had a significant longer survival (mean OS; 41.2 months). In the training cohort median and mean levels for the mtCOI gene were 2541 and 5607 (IQR: 1348-5040) molecules/µL respectively (figure 1A). No differences in baseline mtCOI-levels between tumor types were detected (p=0.13, Table 1 and figure 1B). We found no significant differences in mtCOI-levels in metastatic versus non-metastatic disease (figure 1C, p=0.21). To validate the association of mtDNA with OS and to evaluate the potential predictive importance of mtDNA for therapy response, a validation cohort of 124 consecutive patients with various malignancies (including breast (n=32), cervix (n=5), colon (n=24), ENT (n=24), ovarian (n=16), prostate (n=14) and other (N=11) cancer types)) was assembled. All patients were treated with various chemotherapeutic modalities or in case of ENT with chemo-radiation.



Tabl

e 1

- P

atie

nts

Cha

ract

eris

tics

of th

e Tr

aini

ng C

ohor

t

Can

cer

Type

Bre

ast

Cer

vix

Col

orec

tal

EN

TO

vari

anP

rost

ate

RC

CO

vera

llS

tatis

tical

Diff

eren

ces

betw

een

Can

cer

Type

sN

o. O

f Pat

ient

s3

98

453

622

86

7431

0%

133

1512

728

2410

0%

Dea

th26

6382

788

687

8577

Sex

Fem

ale

39

811

622

−22

108

Mal

e−

−3

43

0−

86

5220

2

Age

Mea

n Ye

ars

53.2

49.

35

9.5

59.

35

9.2

68.

261

.261

.2S

D12

4.5

9.1

8.2

9.4

7.9

9.7

10.5

Met

asta

sis

Yes

133

4229

2081

7025

8N

o26

53

72

54

52Th

erap

yC

hem

othe

rapy

plat

inum

bas

ed−

529

2418

−−

76an

thra

cycl

in b

ased

5−

−−

−−

−5

taxa

ne b

ased

−−

−−

−−

−−

FEC

21−

−−

−−

−21

Oth

er−

−4

22

36

125

6Im

mun

e Th

erap

y−

−−

−−

−53

53H

orm

onal

−−

−−

−6

8−

68

TOTA

L27

9m

tCO

-IM

ean

259

831

08

8274

9221

456

643

245

88

45

607

P=

0.13

*M

edia

n20

4220

3122

053

426

2478

2124

2903

2541

IQR

1087

- 31

2192

6-43

9351

2 - 6

192

194

6 - 6

438

1337

- 53

88

1153

- 4

950

150

9 - 6

021

134

8 - 5

04

0P

rogn

ostic

for

surv

ival

†P

=0.

69P

=0.

14P

=0.

064

P=

0.12

3P

=0.

87P

=0.

035

P=

0.00

4P

<0.

001

ove

rall

sur

viva

lM

edia

n S

urvi

val

not r

each

ed22

.312

.612

.69.

717

.812

.717

.2P

<0.

001

‡95

% C

I−

0 - 4

6.0

7.8

- 17.

410

.1 -

15.1

2.7

- 16.

713

.8 -

21.8

8.3

- 17.

114

.0 -

20.4

* A

NO

VA-te

st u

sed.

† C

ox P

H R

egre

ssio

n w

as a

pplie

d.‡

Log-

rank

test

.

Part III MITOX

119

figure 1a. No differences in mtCOI-levels between the training and validation cohorts were identified (P=0.9). b. No statistical variations in mtCOI-levels between the diverse tumor types in both the training and the validation cohort (ANOVA respectively P=0.13 and P=0.86). c. No differences in metastastic disease or patients without metastasis were found in both cohorts.

figure 1 - No differences in mtCO-1 levels between training and validation cohort


C

B

A

Tabl

e 1

- P

atie

nts

Cha

ract

eris

tics

of th

e Tr

aini

ng C

ohor

t

Can

cer

Type

Bre

ast

Cer

vix

Col

orec

tal

EN

TO

vari

anP

rost

ate

RC

CO

vera

llS

tatis

tical

Diff

eren

ces

betw

een

Can

cer

Type

sN

o. O

f Pat

ient

s3

98

453

622

86

7431

0%

133

1512

728

2410

0%

Dea

th26

6382

788

687

8577

Sex

Fem

ale

39

811

622

−22

108

Mal

e−

−3

43

0−

86

5220

2

Age

Mea

n Ye

ars

53.2

49.

35

9.5

59.

35

9.2

68.

261

.261

.2S

D12

4.5

9.1

8.2

9.4

7.9

9.7

10.5

Met

asta

sis

Yes

133

4229

2081

7025

8N

o26

53

72

54

52Th

erap

yC

hem

othe

rapy

plat

inum

bas

ed−

529

2418

−−

76an

thra

cycl

in b

ased

5−

−−

−−

−5

taxa

ne b

ased

−−

−−

−−

−−

FEC

21−

−−

−−

−21

Oth

er−

−4

22

36

125

6Im

mun

e Th

erap

y−

−−

−−

−53

53H

orm

onal

−−

−−

−6

8−

68

TOTA

L27

9m

tCO

-IM

ean

259

831

08

8274

9221

456

643

245

88

45

607

P=

0.13

*M

edia

n20

4220

3122

053

426

2478

2124

2903

2541

IQR

1087

- 31

2192

6-43

9351

2 - 6

192

194

6 - 6

438

1337

- 53

88

1153

- 4

950

150

9 - 6

021

134

8 - 5

04

0P

rogn

ostic

for

surv

ival

†P

=0.

69P

=0.

14P

=0.

064

P=

0.12

3P

=0.

87P

=0.

035

P=

0.00

4P

<0.

001

ove

rall

sur

viva

lM

edia

n S

urvi

val

not r

each

ed22

.312

.612

.69.

717

.812

.717

.2P

<0.

001

‡95

% C

I−

0 - 4

6.0

7.8

- 17.

410

.1 -

15.1

2.7

- 16.

713

.8 -

21.8

8.3

- 17.

114

.0 -

20.4

* A

NO

VA-te

st u

sed.

† C

ox P

H R

egre

ssio

n w

as a

pplie

d.‡

Log-

rank

test

.


Tabl

e 2

- Pat

ient

s C

hara

cter

istic

s of

the

Valid

atio

n C

ohor

t

Can

cer T

ype

Bre

ast

Cer

vix

Col

orec

tal

EN

TO

varia

nPr

osta

teO

ther

§O

vera

llS

tatis

tical

Diff

eren

ces

betw

een

Can

cer t

ypes

No.

Of P

atie

nts

325

2424

1614

912

4

%26

419

1913

117

100

% D

eath

3840

5042

7586

5665

Sex

Fem

ale

325

85

16−

571

Mal

e−

−16

19−

144

53

Age

Mea

n (Y

ears

)54

.250

.059

.461

.763

.169

.253

.260

.9

SD

7.9

6.5

10.6

11.2

6.9

7.9

12.3

52.0

- 65

.9

Met

asta

sis

No

132

514

5−

342

Yes

193

1910

1114

682

Ther

apy

plat

inum

bas

ed−

522

2014

−3

64

anth

racy

clin

bas

ed19

−−

−−

−3

22

taxa

ne b

ased

13−

−1

114

130

othe

r−

−2

31

−2

8

mtC

O-I

Mea

n59

9050

0478

9424

8547

1538

6926

7755

89

Med

ian

1704

3132

4038

2075

4017

2832

2106

2724

IQR

960

- 361

928

36 -

8111

1826

- 72

8323

54 -

3682

1603

- 62

7212

30 -

5547

1261

- 41

0331

28p=

0.67

*

Prog

nost

ic fo

r sur

viva

l †0.

070.

540.

710.

006

0.47

0.02

0.72

0.11

Pred

ictiv

e fo

r pro

gres

sion

-fr

ee s

urvi

val †

0.31

0.54

0.73

0.03

0.30

0.41

0.28

0.71

Ove

rall

Sur

viva

l

Med

ian

Sur

viva

l33

.8 #

30.4

#20

.330

.2 #

12.2

14.0

14.0

23.9

95%

CI

27.5

- 40

.111

.5 -

49.3

10.2

- 30

.521

.8 -

38.6

9.2

- 15.

29.

4 - 1

8.6

12.9

- 15

.214

.9 -

32.9

p=0.

006

‡

Prog

ress

ion-

free

Sur

viva

l

Med

ian

Sur

viva

l22

.0 #

23.3

#12

.815

.85.

66.

36.

5 10

.6

95%

CI

15.5

- 28

.69.

5 - 3

7.1

0.5

- 25.

10.

1 - 3

1.5

2.5

- 8.7

3.1

- 9.6

2.9

- 10.

28.

9 - 1

2.2

p=0.

001

‡

* AN

OVA

-test

use

d.

† C

ox P

H R

egre

ssio

n w

as a

pplie

d.

‡ Lo

g-ra

nk te

st.

§ In

clud

ing

five

sarc

oma’

s, tw

o st

omac

h, o

ne m

elan

oma

and

one

men

ingi

oma.

#

Mea

n S

urviv

al ti

mes

in s

tead

of M

edia

n.

Part III MITOX

121

Table 2 summarizes patient characteristics. Compared to the training cohort, the validation cohort did notcontain RCC patients and included 9 other cancer types. Importantly, 43 patients (11 patients receiving adjuvant- and 32 patients neoadjuvant chemotherapy) were excluded from PFS analysis. At time of final analysis, 65 patients had died from metastatic disease, 72 patients had progressive disease and 59 patients were censored. The median OS time was 23.9 months (95% CI: 14.9 – 32.9 months). The median PFS time was 10.6 months (95% CI: 8.9 – 12.2 months). Patients with breast-, cervix- or ENT cancer had significant longer survival times compared to other cancer types (p<0.01 for PFS and OS). The mean and median mtCOI-levels were 4982 and 2515 molecules/µL (IQR: 1338-4663) respectively. No differences in baseline mtCOI-concentrations between both independent patient cohorts were detected (p=0.67, figure 1A). Similar to the training cohort results, no significant differences in mtCOI-levels were found between the tumor types (P=0.77) in the validation cohort, suggesting that mtCOI might be used as a pan-tumor marker (Table 2, figure B). Again, mtCOI-concentrations did not statistically differ between patients with metastasized disease and non-metastatic disease in this validation cohort (p=0.27, figure 1C).

Training Cohort: mtDnA prognostic for overall survivalIncluding all 310 cancer patients univariate analysis showed a significant association of mtCOI-levels (as continuous variable) with OS (p<0.001). High mtCOI-levels predicted poor prognosis and low concentrations were associated with prolonged survival. Colorectal- and ENT cancer showed respectively borderline significance (P=0.06) and a weak association (p=0.12), where RCC (P=0.008) and prostate cancer (P=0.003, which confirmed our previous results generated with NASBA methodology (30)) contributed at most to this correlation. Importantly, mtCOI-levels for breast-, cervix- and ovarian cancer were not prognostic for OS (Table 1). Accordingly, multivariate regression analysis showed that mtCOI significantly predicted OS, independent of tumor type (Table 3A). As for 246 patients LDH-quantities were available, multivariate regression showed independent prognostication for mtCOI and LDH with OS (Table 3B). Next, mtCOI-levels were dichotomized for Kaplan-Meier estimation. Using tertiles as objective cut-off points patients were categorized into a favorable (<1643 molecules/µL), an intermediate(1643-3575 molecules/µL) and a poor risk group (>3575 molecules/µL), displaying significant discrimination for OS times between risk groups (Log Rank: χ2=24 and p<0.001, figure 2A). The median OS times were 27.1 months (95% CI: 20.7 - 33.6), 17.8 months (95% CI: 12.6 - 23.0), and 9.0 months (95% CI: 4.2 - 13.8) for respectively the favorable, intermediate and poor risk groups.

Table 3A: Multivariate Cox Regression analysis of mtCO-I and Tumor Types

Variable (n=310 patients) p-ValueTumor Type 0.001mtCO-I 0.002

Table 3B: Multivariate Cox Regression analysis of mtCO-I and LDH

Variable (n=246 patients) p-ValueLDH 0.001mtCO-I 0.003



figure 2a. Using tertiles as objective cut off points for mtCOI-levels Kaplan-Meier estimation significantly discriminated between survival times for the three risk groups (Log Rank: χ2=23 and p<0.001). b. A second mitochondrial gene mtRNR1, confirmed the prognostic value for OS, showing distinct survival times when using tertiles as objective cut off points (Log Rank: χ2=24 and p<0.001). c. The cut-off point of the upper tertile (>3575 mols/ul) was verified as poor prognosticator in the validation cohort and discriminated between patients with a favourable and a poor OS (Log Rank: χ2=11 and p<0.001).

figure 2 Survival Analysis of mtDNA-levels of both patients cohorts

Part III MITOX

CB

A

123

To further support our hypothesis that mitochondrial genes in general could predict OS, we tested a second mitochondrial gene –mtRNR1- in an at randomly selected group of 126 cancer patients. The mean and median mtRNR1-levels were 4395 and 1876 molecules/µL respectively (IQR: 1006 - 4893). Levels of both mitochondrial genes mtCOI and mtRNR1 were significantly correlated (R2 = 0.873, P<0.001). mtRNR1 also predicted OS using univariate regression analysis(P=0.002). Categorizing patients in three risk groups, using again tertiles as objective cut-off points, resulted in discriminative prognostication of OS for each risk group (Log Rank: χ2

= 24 and p<0.001, figure 2B). The OS times were 31.7 (mean) months (95% CI: 26.4 – 37.1), 12.2 (median) months (95% CI: 3.4 - 21.1), and 5.7 (median) months (95% CI: 2.3 - 9.8) for respectively the favorable (<1193 molecules/µL), intermediate (1194 - 3742 molecules/µL) and poor risk (>3742 molecules/µL) groups. Compared to mtCOI classified patients, similar patients were categorized in various risk groups.

Prospective Validation Cohort: mtDnA prognostic for os, but not predictive for therapy responseSubsequently, we confirmed our findings in an independent prospective validation set of 124 patients. The previously defined cut-off values for mtCOI-levels from the training cohort were applied to this validation cohort. Interestingly, Kaplan-Meier estimation significantly distinguished a favorable (<1643 molecules/µL), intermediate (1643-3575 molecules/µL) and poor risk group(>3757 molecules/µL)(Log Rank: χ2 = 11 and p=0.004, figure 2C). Compared to the intermediate and favorable group, the poor risk group displayed significant inferior survival with a median OS time of 13.5 months (95% CI: 11.7 - 15.3). However, no discrimination in survival was found between the favorable and intermediate risk groups as both groups presented similar OS times, 30.8 (95% CI: 24.7 - 36.8) and 33.2 (95% CI: 26.9 - 39.4) months respectively.Additionally, as continuous variables mtCOI-levels at 7 days (N=82) and 21 days (N=83) after chemotherapy initiation were also predictive for OS (p=0.01 and p=0.02, respectively). As an exploratory analysis we evaluated whether mtCOI-levels could be used as a predictive marker for chemotherapy response in metastasized patients in the validation cohort. First, we investigated whether post-chemotherapy mtCOI-levels varied between chemotherapy modalities. We found no significant differences in mtCOI-levels at the different time points after chemotherapy between various chemotherapy modalities (p>0.2) (figure 3A). Therefore we performed our analysis for the entire group of metastasized patients (N=82) regardless of chemotherapeutic modality and evaluated baseline mtCOI-levels, and longitudinally time points after chemotherapy initiation for their predictive value. Using univariate regression analysis we found no correlation between mtCOI-levels and PFS (p>0.2), at any time points of blood collection. To examine whether changes in mtCOI-levels at different time points compared to baseline were predictive, we calculated the mtCOI-levels as percentage-change compared to baseline for each time point. When analyzing the kinetics of the mtCOI-levels after chemotherapy, we found a slight increase in the first hours after chemotherapy (mean increase 1.7 fold), after which the mtCOI-levels returned almost to baseline. However, these fold-increase variables were not predictive for PFS (p>0.2) (figure 3B). Furthermore, when comparing the kinetics between patients with a good prognosis, defined as a PFS longer than 4 months, compared to patients with a poor prognosis, no differences in the kinetics were observed (figure 3C).



figure 3a. The mtDNA-levels 7 and 21 days after start of the chemotherapy did not differ significantly between the different types of chemotherapy (Kruskal-Wallis p>0.2). b. Evaluating the kinetics of mtDNA after chemotherapy revealed a small increase in the first hours after chemotherapy (mean increase 1.7 fold), after which the mtDNA-levels returned almost to baseline. c. The kinetics of mtDNA after chemotherapy did not differ between patients with a long PFS (>4 months) and a short PFS (<4 months).

figure 3 mtDNA-levels are not predictive for response of chemotherapy

Part III MITOX

A

B

C

125

DI sCus s Ion

Here, we successfully developed a simple qPCR as a robust assay to accurately quantify circulating mtDNA-levels in human plasma. In a large training- and prospective validation cohort, both comprising patients with various cancer types, baseline mtCOI-levels significantly predicted OS, independent of tumor type. Elevated mtCOI-levels implicated an inferior OS compared to prolonged OS for low mtCOI-levels. These findings were confirmed for a second mitochondrial gene mtRNR1. Conclusively, mtDNA-levels can be used as a prognostic biomarker for multiple tumor types, however not every tumor type. In patients with breast-, ovarian and cervix cancer this correlation was not identified possibly due to lower patient numbers and fewer patients with metastasized disease. As in only woman-related tumor types mtDNA-levels did not correlate with OS, gender might influence mtDNA-levels. Our finding that mtDNA was less powerful predictive for OS in the validation cohort, may be explained by several factors. Compared to the training cohort the validation set comprised; a) fewer patients with metastasized disease with as a consequence a longer OS time (p=0.023), b) smaller patient numbers per tumor type, c) dissimilar treatment modalities and d) shorter clinical follow up. Unfortunately, mtDNA-levels at various time points did not predict treatment response. These results are in contrast with a study with 27 breast cancer patients treated with anthracyclins, where high baseline mtDNA-levels correlated with inferior disease-free survival (37). Although, we found a rapid increase in mtDNA-levels 2-4 hours after chemotherapy initiation no significant correlation with PFS was found between changes in mtDNA-levels. This could be influenced by heterogeneity and small patient numbers per tumor and therapy type. Furthermore, our chosen time points might not be optimal as ‘early time points’ are too premature to expect cellular damage and mtDNA-release and ‘later time points’ are too early to expect a mtDNA-decrease due to diminished tumor volume. An important issue in cfDNA-use is the sample preparation. In literature, dissimilarities in sample handlings, plasma isolation, storage and quantification techniques have been described (30;33;38-45). These differences contributed to conflicting cfNA-levels detected in different cancer patients (28;46-49). Clearly, this may greatly influence the validity of cfNAs as prognostic and diagnostic markers (1;2;19;21;43). Standardization of sample collection and technical procedures are essential for conclusive comparisons between studies and therefore our developed straightforward qPCR might be an interesting application. Another important item is the centrifugation of plasma to remove large particles (e.g. platelets and complexes that contain NA). Previous reports indicate that for optimal discrimination between cases and controls, plasma subjected to only one spin is inferior than plasma which is spun twice (29;46). Consequently we banked plasma of patients participating in this study according to a two-spin protocol. As shown for CRC, a qualitative classification of mtDNA might also improve the predictive significance of our quantitative approach of mtDNA-determination (50). With only a quantitative approach the percentage of contamination with mtDNA from normal cells could not be revealed. In conclusion, circulating mtDNA quantified by a straightforward qPCR, can be used as prognostic marker for OS regardless of tumor type, however not for all cancer types. mtDNA-levels were not predictive for response to chemotherapy. We anticipate that further prospective studies with large homogenous patient cohorts per tumor type are necessary to address the clinical significance in broader extent of circulating mtDNA as prognostic tumor marker for survival.




The authors thank all medical doctors, researchers, technicians and all affiliated chemical laboratories for their assistance in patient management and blood processing. This study was supported by a grant of the Dutch Cancer Society (Number: UU 2007-3924). None of the contributing authors received any research funding or had any conflicts of interest to declare.

supplemental figure 1

Part III MITOX

A

A

C

D

Co-1

RnR-1


supplemental figure 1A. Mitochondrial DNA inserts were loaded on a 2% normal agarose gel and affirmed for its fragments lengths. B. Standard qPCR curves for respectively mtCOI and mtRNR1 C. Standard curves were accurately and robustly analyzed with our developed qPCR for the mitochondrial genes mtCOI and mtRNR1 (mean R2=0.98). D. The developed qPCR was highly reproducible showing a Pearson Correlation of 0.96 (p<0.001, N=11 patients).

Bs

TAn

DA

RD

Cu

RV

e m

tRn

R1

sTA

nD

AR

D C

uR

Ve

mtC

o-I

B



(1) Schwarzenbach H, Hoon DS, Pantel K. Cell-free nucleic acids as biomarkers in cancer patients.

Nat Rev Cancer 2011 June;11(6):426-37.

(2) Fleischhacker M, Schmidt B. Circulating nucleic acids (CNAs) and cancer--a survey. Biochim Biophys Acta

2007 January;1775(1):181-232.

(3) Lui YYN, Chik K W, Chiu RWK, Ho CY, Lam CWK, Lo YMD. Predominant Hematopoietic Origin of Cell-free

DNA in Plasma and Serum after Sex-mismatched Bone Marrow Transplantation. Clin Chem 2002 March

1;48(3):421-7.

(4) Stroun M, Maurice P, Vasioukhin V, Lyautey J, Lederrey C, LEFORT F et al. The Origin and Mechanism of

Circulating DNA. Ann NY Acad Sci 2000 April 1;906(1):161-8.

(5) Fournie GJ, Courtin JP, Laval F, Chale JJ, Pourrat JP, Pujazon MC et al. Plasma DNA as a marker of cancerous

cell death. Investigations in patients suffering from lung cancer and in nude mice bearing human tumours.

Cancer Lett 1995 May 8;91(2):221-7.

(6) Giacona MB, Ruben GC, Iczkowski KA, Roos TB, Porter DM, Sorenson GD. Cell-free DNA in human blood

plasma: length measurements in patients with pancreatic cancer and healthy controls. Pancreas 1998

July;17(1):89-97.

(7) Anker P, Stroun M, Maurice PA. Spontaneous release of DNA by human blood lymphocytes as shown in an in

vitro system. Cancer Res 1975 September;35(9):2375-82.

(8) Wang BG, Huang HY, Chen YC, Bristow RE, Kassauei K, Cheng CC et al. Increased plasma DNA integrity in

cancer patients. Cancer Res 2003 July 15;63(14):3966-8.

(9) Anker P, Lefort F, Vasioukhin V, Lyautey J, Lederrey C, Chen XQ et al. K-ras mutations are found in DNA

extracted from the plasma of patients with colorectal cancer. Gastroenterology 1997 April;112(4):1114-20.

(10) Mulcahy HE, Lyautey J, Lederrey C, qi C, X, Anker P, Alstead EM et al. A prospective study of K-ras mutations

in the plasma of pancreatic cancer patients. Clin Cancer Res 1998 February;4(2):271-5.

(11) Sorenson GD, Pribish DM, Valone FH, Memoli VA, Bzik DJ, Yao SL. Soluble normal and mutated DNA

sequences from single-copy genes in human blood. Cancer Epidemiol Biomarkers Prev 1994;3(1):67-71.

(12) Jahr S, Hentze H, Englisch S, Hardt D, Fackelmayer FO, Hesch RD et al. DNA Fragments in the Blood Plasma

of Cancer Patients: Quantitations and Evidence for Their Origin from Apoptotic and Necrotic Cells. Cancer Res

2001 February 1;61(4):1659-65.

(13) Frickhofen N, Muller E, Sandherr M, Binder T, Bangerter M, Wiest C et al. Rearranged Ig heavy chain DNA is

detectable in cell-free blood samples of patients with B-cell neoplasia. Blood 1997 December 15;90(12):4953-60.

(14) Mayall F, Jacobson G, Wilkins R, Chang B. Mutations of p53 gene can be detected in the plasma of patients

with large bowel carcinoma. J Clin Pathol 1998 August;51(8):611-3.

(15) Silva JM, Dominguez G, Garcia JM, Gonzalez R, Villanueva MJ, Navarro F et al. Presence of tumor DNA in

plasma of breast cancer patients: clinicopathological correlations. Cancer Res 1999 July 1;59(13):3251-6.

(16) Vasioukhin V, Anker P, Maurice P, Lyautey J, Lederrey C, Stroun M. Point mutations of the N-ras gene in the

blood plasma DNA of patients with myelodysplastic syndrome or acute myelogenous leukaemia. Br J Haematol

1994;86(4):774-449.

(17) Wong IHN, Lo YMD, Zhang J, Liew CT, Ng MHL, Wong N et al. Detection of Aberrant p16 Methylation in the

Plasma and Serum of Liver Cancer Patients. Cancer Res 1999 January 1;59(1):71-3.

(18) Wu TL, Zhang D, Chia JH, Tsao KH, Sun CF, Wu JT. Cell-free DNA: measurement in various carcinomas and

establishment of normal reference range. Clin Chim Acta 2002 July;321(1-2):77-87.

(19) Pathak AK, Bhutani M, Kumar S, Mohan A, Guleria R. Circulating cell-free DNA in plasma/serum of lung cancer

patients as a potential screening and prognostic tool. Clin Chem 2006 October;52(10):1833-42.

Part III MITOX

129

(20) Bremnes RM, Sirera R, Camps C. Circulating tumour-derived DNA and RNA markers in blood: a tool for early

detection, diagnostics, and follow-up? Lung Cancer 2005 July;49(1):1-12.

(21) Goebel G, Zitt M, Zitt M, Muller HM. Circulating nucleic acids in plasma or serum (CNAPS) as prognostic and

predictive markers in patients with solid neoplasias. Dis Markers 2005;21(3):105-20.

(22) Holdenrieder S, Stieber P, von Pawel J, Raith H, Nagel D, Feldmann K et al. Circulating nucleosomes predict the

response to chemotherapy in patients with advanced non-small cell lung cancer. Clin Cancer Res 2004

September 15;10(18 Pt 1):5981-7.

(23) Sozzi G, Conte D, Mariani L, Lo VS, Roz L, Lombardo C et al. Analysis of circulating tumor DNA in plasma at

diagnosis and during follow-up of lung cancer patients. Cancer Res 2001 June 15;61(12):4675-8.

(24) Gautschi O, Bigosch C, Huegli B, Jermann M, Marx A, Chasse E et al. Circulating Deoxyribonucleic Acid As

Prognostic Marker in Non-Small-Cell Lung Cancer Patients Undergoing Chemotherapy. J Clin Oncol 2004

October 15;22(20):4157-64.

(25) Anker P, Mulcahy H, Stroun M. Circulating nucleic acids in plasma and serum as a noninvasive investigation for

cancer: time for large-scale clinical studies? Int J Cancer 2003 January 10;103(2):149-52.

(26) Frattini M, Gallino G, Signoroni S, Balestra D, Battaglia L, Sozzi G et al. Quantitative analysis of plasma DNA in

colorectal cancer patients: a novel prognostic tool. Ann N Y Acad Sci 2006 September;1075:185-90.

(27) Kamat AA, Baldwin M, Urbauer D, Dang D, Han LY, Godwin A et al. Plasma cell-free DNA in ovarian cancer: an

independent prognostic biomarker. Cancer 2010 April 15;116(8):1918-25.

(28) Barker PE, Murthy M. Biomarker Validation for Aging: Lessons from mtDNA Heteroplasmy Analyses in Early

Cancer Detection. Biomark Insights 2009;4:165-79.

(29) Chiu RW, Chan LY, Lam NY, Tsui NB, Ng EK, Rainer TH et al. Quantitative analysis of circulating mitochondrial

DNA in plasma. Clin Chem 2003 May;49(5):719-26.

(30) Mehra N, Penning M, Maas J, van DN, Giles RH, Voest EE. Circulating mitochondrial nucleic acids have

prognostic value for survival in patients with advanced prostate cancer. Clin Cancer Res 2007 January 15;13

(2 Pt 1):421-6.

(31) Ellinger J, Muller DC, Muller SC, Hauser S, Heukamp LC, von RA et al. Circulating mitochondrial DNA in serum:

A universal diagnostic biomarker for patients with urological malignancies. Urol Oncol 2010 September 25.

(32) Zachariah RR, Schmid S, Buerki N, Radpour R, Holzgreve W, Zhong X. Levels of circulating cell-free nuclear

and mitochondrial DNA in benign and malignant ovarian tumors. Obstet Gynecol 2008 October;112(4):843-50.

(33) Kohler C, Radpour R, Barekati Z, Asadollahi R, Bitzer J, Wight E et al. Levels of plasma circulating cell free

nuclear and mitochondrial DNA as potential biomarkers for breast tumors. Mol Cancer 2009;8:105.

(34) Xia P, An HX, Dang CX, Radpour R, Kohler C, Fokas E et al. Decreased mitochondrial DNA content in blood

samples of patients with stage I breast cancer. BMC Cancer 2009;9:454.

(35) Ellinger J, Albers P, Muller SC, von RA, Bastian PJ. Circulating mitochondrial DNA in the serum of patients with

testicular germ cell cancer as a novel noninvasive diagnostic biomarker. BJU Int 2009 July;104(1):48-52.

(36) Ellinger J, Muller SC, Wernert N, von RA, Bastian PJ. Mitochondrial DNA in serum of patients with prostate

cancer: a predictor of biochemical recurrence after prostatectomy. BJU Int 2008 August 5;102(5):628-32.

(37) Hsu CW, Yin PH, Lee HC, Chi CW, Tseng LM. Mitochondrial DNA content as a potential marker to predict

response to anthracycline in breast cancer patients. Breast J 2010 May;16(3):264-70.

(38) Chiu RW, Lui WB, El Sheikhah A, Chan AT, Lau TK, Nicolaides KH et al. Comparison of protocols for extracting

circulating DNA and RNA from maternal plasma. Clin Chem 2005 November;51(11):2209-10.

(39) Chiu RW, Poon LL, Lau TK, Leung TN, Wong EM, Lo YM. Effects of blood-processing protocols on fetal and

total DNA quantification in maternal plasma. Clin Chem 2001 September;47(9):1607-13.

(40) Jung M, Klotzek S, Lewandowski M, Fleischhacker M, Jung K. Changes in concentration of DNA in serum and

plasma during storage of blood samples. Clin Chem 2003 June;49(6 Pt 1):1028-9.



(41) Lam NY, Rainer TH, Chiu RW, Lo YM. EDTA is a better anticoagulant than heparin or citrate for delayed blood

processing for plasma DNA analysis. Clin Chem 2004 January;50(1):256-7.

(42) Sozzi G, Roz L, Conte D, Mariani L, Andriani F, Verderio P et al. Effects of prolonged storage of whole plasma

or isolated plasma DNA on the results of circulating DNA quantification assays. J Natl Cancer Inst 2005

December 21;97(24):1848-50.

(43) Radpour R, Fan AX, Kohler C, Holzgreve W, Zhong XY. Current understanding of mitochondrial DNA in breast

cancer. Breast J 2009 September;15(5):505-9.

(44) Ellinger J, Muller DC, Muller SC, Hauser S, Heukamp LC, von RA et al. Circulating mitochondrial DNA in serum:

A universal diagnostic biomarker for patients with urological malignancies. Urol Oncol 2010 September 25.

(45) Ellinger J, Muller SC, Wernert N, von RA, Bastian PJ. Mitochondrial DNA in serum of patients with prostate

cancer: a predictor of biochemical recurrence after prostatectomy. BJU Int 2008 August 5;102(5):628-32.

(46) Boddy JL, Gal S, Malone PR, Harris AL, Wainscoat JS. Prospective Study of Quantitation of Plasma DNA

Levels in the Diagnosis of Malignant versus Benign Prostate Disease. Clin Cancer Res 2005

February 15;11(4):1394-9.

(47) Gal S, Fidler C, Lo YM, Taylor M, Han C, Moore J et al. Quantitation of circulating DNA in the serum of breast

cancer patients by real-time PCR. Br J Cancer 2004 March 22;90(6):1211-5.

(48) Jakupciak JP, Dakubo GD, Maragh S, Parr RL. Analysis of potential cancer biomarkers in mitochondrial DNA.

Curr Opin Mol Ther 2006 December;8(6):500-6.

(49) Vlassov VV, Laktionov PP, Rykova EY. Circulating nucleic acids as a potential source for cancer biomarkers.

Curr Mol Med 2010 March;10(2):142-65.

(50) Frattini M, Gallino G, Signoroni S, Balestra D, Lusa L, Battaglia L et al. Quantitative and qualitative characterization

of plasma DNA identifies primary and recurrent colorectal cancer. Cancer Lett 2008 May 18;263(2):170-81.

Part III MITOX


Chapter 8 Primary colorectal cancers and their subsequent hepatic metastases are genetically different: implications for selection of patients for targeted treatment

Joost S Vermaat*, Isaac J Nijman*, Marco J Koudijs, Frank L Gerritse, Stefan J Scherer, Michal Mokry, Wijnand M Roessingh, Nico Lansu, Ewart de Bruijn, Richard van Hillegersberg, Paul J van Diest, Edwin E Cuppen, Emile E Voest


Clinical Cancer Research 2012 Feb 1;18(3):688-99.

134 Part IV Genomics


This is the first study which comprehensively compared the genetic constitution of 1,264 genes –involved in the most clinically relevant cancer-associated signaling pathways and processes- in 21 primary colorectal adenocarcinomas (CRC) and their subsequent hepatic metastases by employing targeted deep-sequencing on formalin fixed, paraffin embedded tissue. The differences in potentially clinically relevant genetic variations between the primary CRC tumor and hepatic metastases in important pathways are of such magnitude that an impact on treatment outcome is realistic. This indicates that genetic analysis of the metastasis may have more predictive power when patients are selected for specific treatment modalities, thus allowing for further refinement of treatment algorithms.

135 Primary colorectal cancers and their hepatic metastases are genetically different Chapter 8

AB sTRACT

IntroductionIn the era of DNA-guided personalized cancer treatment, it is essential to perform predictive analysis on the tissue that matters. Here we analyzed genetic differences between primary colorectal adenocarcinomas (CRCs) and their respective metastasis.

Patients and MethodsThe primary CRC and the subsequent hepatic metastasis of 21 CRC patients were analyzed using targeted deep-sequencing of DNA isolated from formalin fixed paraffin embedded archived material.

ResultsWe have interrogated the genetic constitution of a designed “Cancer Mini-Genome” consisting of all exons of 1,264 genes associated with pathways relevant to cancer. In total, 6,696 known and 1,305 novel variations were identified in 1,174 and 667 genes respectively, including 817 variants that potentially altered protein function. On average 83 (SD 69) potentially function-impairing variations were gained in the metastasis and 70 (SD 48) variations were lost, showing that the primary tumor and hepatic metastasis are genetically significantly different. Besides novel and known variations in genes such as KRAS, BRAF, KDR, FLT1, PTEN, and PI3KCA, aberrations in the up-/downstream genes of EGFR/PI3K/VEGF-pathways and other pathways (mTOR, TGFb, etc.) were also detected, potentially influencing therapeutic responsiveness. Chemotherapy between removal of the primary tumor and the metastasis (n = 11) did not further increase the amount of genetic variation.

ConclusionOur study indicates that the genetic characteristics of the hepatic metastases are different from those of the primary CRC tumor. As a consequence, the choice of treatment in studies investigating targeted therapies should ideally be based on the genetic properties of the metastasis rather than on those of the primary tumor.

136

AB B R eVIATIon s NGS; Next Generation Sequencing, (m)CRC; (metastasized) colorectal cancer, FFPE; Formalin-Fixed Paraffin-Embedded, SNV; Single Nucleotide Variation, CAN-genes; candidate cancer genes, indels; small insertions and/or deletions, EGFR; Epidermal Growth Factor Receptor, KRAS; v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog, PTEN; phosphatase and tensin homolog, PIK3CA; phosphoinositide-3-kinase, catalytic, alpha polypeptide, TGFb; transforming growth factor beta, mTOR; mammalian Target of Rapamycin, VEGF; Vascular Endothelial Growth Factor, HRAS; v-Ha-ras Harvey rat sarcoma viral oncogene homolog, NRAS; neuroblastoma RAS viral (v-ras) oncogene homolog, FLT1; fms-related tyrosine kinase 1 (vascular endothelial growth factor/vascular permeability factor receptor), KDR; kinase insert domain receptor (a type III receptor tyrosine kinase)

I nTRoDuCTIon

The main challenge for the future of cancer treatment is to provide every individual patient with the most effective drug tailored to their specific cancer. Personalized cancer treatment is still in the very early stages, but several recent studies have shown the potential of this approach. For example, trastuzumab (Genentech, CA) is approved for patients with breast and gastric cancer that have HER2-amplified tumors (1). For patients with metastatic colorectal cancer (mCRC), the anti-epidermal growth factor receptor (EGFR)-antibodies panitumumab (Amgen, CA) and cetuximab (Merck, MA) are approved for patients with wild-type (wt)KRAS tumors (2;3); mCRC patients with a mutated KRAS gene do not benefit from treatment with these antibodies (4-8). Unfortunately, treating patients with anti-EGFR antibodies while they express (wt)KRAS remains beneficial to only a subset of patients. In addition, it is becoming increasingly clear that patients with loss-of-function mutations in PTEN or BRAF are also unresponsive to cetuximab (2;3). These examples illustrate that solitary gene analyses in personalized anti-cancer treatment have limitations (9;10) and emphasize the need for more complete genetic profiling of relevant pathways within the tumor to optimize treatment strategies (11). Additionally, it is important to determine which tumor tissue will best predict treatment outcome. Current clinical practice is to use archived material of the primary tumor to determine the constitution of the molecular target in order to select patients for treatment (2-6). However, there are several biological reasons why this may not be optimal. First, genomic instability is a hallmark of cancer, and caused by constant selection pressure tumors rapidly change their genetic make-up over time. Secondly, specific populations of tumor cells may be more prone to metastasize than others, which is likely to result in an enrichment of these cells and consequently their genetic aberrations in the metastases. Thirdly, systemic treatment may induce selection pressure toward a specific genetic phenotype or induce additional genetic changes. Taken together, tumors are genetically dynamic, which suggests that selecting patients for targeted treatments based on the characteristics of the primary tumor and not their metastases may not be optimal. It has been suggested that cancer-initiating mutations will be present in every tumor localization but that cancer-driving mutations may be enriched or depleted in the metastasis (12). Sequencing 189 candidate cancer genes (CAN) in breast- and colorectal cancer showed substantial differences within tumors, indicating that each tumor type is heterogeneous and that tumorigenesis is likely tumor-specific

Part IV Genomics

137

(13). Importantly, these analyses were performed on primary tumors and did not address the question of genetic differences between primary tumors and their metastases (14). Next generation sequencing (NGS) is an extremely powerful technology for genetic analysis of complete signaling pathways in large patient cohorts (11). We therefore embarked on a study to comprehensively compare the genetic constitution of the 1,264 genes comprising the most relevant cancer-associated pathways and processes in 21 primary CRCs and their subsequent hepatic metastases by employing targeted deep-sequencing.


Patient selectionWe selected 21 patients with colorectal adenocarcinomas (CRC), from whom both formalin-fixed paraffin-embedded (FFPE) samples of the primary tumor and their sequential hepatic metastasis were available (Table 1). We included only patients with a minimal time of six months between primary tumor resection and metastasectomy. Patients who received no treatment (n = 10, chemo-naïve group) and patients who received chemotherapy (n = 11, chemo-treated group) between removal of the primary tumor and metastasis were included. Eight patients in the chemo-treated group received 5-fluorouracil, oxaliplatin and leucovorin as chemotherapy and three patients received capecitabin, oxaplatin and bevacizumab between the surgical interventions. The primary CRC tumor of the majority of patients were localized in the colon (n = 16), where 5 patients were diagnosed with rectal adenocarcinoma. At time of primary CRC resection patients were classified according to the TNM-staging system (Table 1), demonstrating that all patients were diagnosed with stage II, III or IV disease. In the chemo-treated group, two patients were diagnosed with synchronous hepatic metastasis, the other eight patients with metachronous hepatic metastasis. All procedures were approved by the UMC Utrecht ethics committee.

Tumor- and DnA samplingA pathologist confirmed the colorectal adenocarcinoma (CRC) diagnosis of each sample and demarcated tumor areas from normal tissue. To obtain samples consisting of ≥ 80% tumor cells, we micro-dissected tumor tissue using a laser-capture micro-dissection microscope (PALM, Carl Zeiss, Germany). Dissected tumor cells were incubated with sodium-thiocyanate to dissolve DNA and protein cross-linking. Subsequently, samples were treated overnight at 55°C with 20 mg/ml Proteinase-K and DNA was isolated using the QIAamp DNA Micro Kit (Qiagen, CA). At least 2 µg DNA was isolated per sample, of which >1 µg was stored at -20°C for validation purposes and 1 µg was used for library preparation.

library PreparationSequencing libraries were prepared as described previously (15), with minor modifications. Briefly, 1 µg of tumor DNA was sheared using a Covaris S2 sonicator (Covaris, MA) (duty cycle 20%, intensity 5, 200 cycles/burst for 10 minutes) and end-repaired using End-It DNA end-repair kit (Epicenter Technologies). After ligation of short sequencing adaptors, the library was amplified using a truncated version of sequencing primers and size was selected on 2% agarose gel for a 125–175 base pair fraction.

Primary colorectal cancers and their hepatic metastases are genetically different Chapter 8

138

Designing the “Cancer Mini-genome”To interrogate differences in genetic make-up between primary tumor and metastasis with possible therapeutic consequences, we composed a “Cancer Mini-Genome” consisting of all the exons (ensembl v56) of 1,264 genes (totaling ~7 Mbp), including known oncogenes, tumor suppressor genes, identified colorectal CAN Genes (13), all 518 kinases (16), and important pathways related to tumorigenesis and anti-cancer treatment (for example, angiogenesis; apoptosis and EGFR-, PIK3CA-, TGFb-, mTOR-, and VEGF-pathways were all included) (17). All genes of the “Cancer Mini-Genome” are listed in supplemental Table 1. A custom-made PERL script was used to design 60-mer capture probes for the target regions with a 15bp moving window on both genomic strands. The best probes in each window was selected based on Tm, GC%, and monomer stretches as described previously (18). Probes with more than one additional location in the genome with a similarity over 60% were discarded and the remaining probes were ordered on a custom 1M CGH array (Agilent, CA). A list of sequences of the capture probes used for the Cancer Mini-genome is mentioned in supplemental Table 2. Array-based enrichment for “Cancer Mini-genome” and Massive Parallel sequencingEnrichment was performed as described previously (18). Briefly, size-selected libraries were PCR amplified using ten additional cycles to produce amounts necessary for hybridization, and subsequently purified and hybridized to CGH-capture arrays to enrich for exonic regions of our designed “Cancer Mini-Genome”. After 72 hours of hybridization at 42°C, arrays were extensively washed and specifically hybridized library molecules were eluted and amplified by 13 cycles of PCR. After purification, barcodes (SOLiD barcode primers, Applied Biosystems) unique to each sample were incorporated into each library by performing four additional PCR cycles. Ten pre-barcoded individual libraries were pooled and sequenced on the SOLiD 3+ system (Life Technologies) according to the manufacturer’s instructions. Bioinformatics and statisticsSequence data were mapped against the human reference genome (GRCh37/hg19) using BWA (19). Single Nucleotide Variations (SNVs) were called with a custom PERL script as previously described (15) with the settings mentioned below and the results were further processed by custom-made PERL scripts, with subsequent prediction of the consequence of genetic variations at the amino acid level. Bioinformatic analyses were performed to assess frequency of variants, affected pathways, and functional relevance of identified somatic variations. Normal non-neoplastic tissue from five patients -of whom also the primary CRC tumor and hepatic metastasis- was sequenced to annotate germline variants and to exclude common existing SNVs in accordance with previous studies (20;21). Identified variations in the “Cancer Mini-Genome” were annotated according to existing databases (ensembl59). To limit the influence of SNV caller artifacts, we applied three different SNV callers (custom PERL script (15;18), VarScan (22) and the GATK toolkit (23)) and considered the overlapping variants as high-confidence variants. The following settings were applied:1. Custom PERL filter: coverage between 20-2000x; three reads supporting each allele on each strand, minimum of three independent reads per variant, no strand imbalance greater than 1:10, call quality≥10, reads should map uniquely, clonality filtering: no more than 5 identical reads counted per allele, alleles counted if present ≥ 3x.2. VarScan: Variants with a p ≤ 10-8, minimal coverage of 20x, and two non-reference reads and a variant frequency ≥ 0.20.

Part IV Genomics

139

3. GATK toolkit: Minimal 20x coverage, variant frequency ≥ 0.20. In general, the overlap was 33 ± 10% of all the variable positions called by the SNV callers, indicating significant bias depending on the used SNV caller. In cases where a difference between the tumors was discovered, we examined the raw data of the primary sample for any indication of the variant that might have escaped the strict SNV caller filters. If the allele was detected only minimally (presence of three non-reference reads), this variant was not called as de novo, but as equal between tumors. To predict the possible impact of the identified variation on amino acid structure and thus the protein function, PolyPhen-2 (Polymorphism Phenotyping version2 (24)) was used.

Table 1 - Basic Patient Characteristics

Chemo-Naive Group Chemo-Treated Group Overall

Total number of Patients 10 11 21

Age onset CRC DiagnosisYears 60.0 61.4 60.7

Range 45.1 - 72.5 51.3 - 69.4 45.1 - 72.5

gender

Males : Females 6 ; 4 5 ; 6 11 ; 10

TnM ClassificationT2 2 — 2

T3 7 10 17

T4 1 1 2

N0 6 1 6

N+ 4 10 14

M0 10 8 18

M+ † 0 2 2

Tumor localisationColon 6 10 16

Rectal # 4 1 5

Time between primary tumor resection and metastasectomyMonths 21.9 17.7 19.7

Range 7.5 - 48.6 7.0 - 49.2 7.0 - 49.2

overall survival Time (Years)Years 4.2 4.9 4.6

SD 0.6 0.8 0.6

No. of Patients died from disease 5 6 11

Chemotherapeutic schedule‡5-FU, Oxaliplatin and Leucovorin − 8 8

Capecetabin, Oxaliplatin and Bevacimuab

− 3 3

*No significant differences in patients characteristics(p>0.05) between Chemo-Naive and Chemo-Treated Group were observered. † Two patients were diagnosed with hepatic metastasized disease; however first the primary CRC tumor was resected -followed by chemotherapy- and >6 months later the hepatic metastasis.# From each group one patient with rectal carcinoma was primarily treated with radiotherapy.‡ All 11 patients received at least 4 cycli of chemotherapy between primary CRC resection and hepatic metastasectomy.


140

ValidationTo validate NGS-identified variants with conventional Sanger sequencing, genomic DNA of all samples were whole-genome-amplified using the Repli-g FFPE kit (Qiagen) with 100ng of input DNA according to the company’s protocol. Primers were designed to amplify a ~200bp amplicon containing a random number of putative variants and amplicons were re-sequenced using conventional Sanger sequencing according to standard protocols with WGA-amplified DNA as input. In addition, the samples were analyzed through routine diagnostics on KRAS (HRM and Sanger sequencing in duplo).

Table 2 – General sequencing results per individual tissue

Healthy Tissue

Primary Colorectal Tumor

Liver Metastasis

Overall mean

Overall Range

N=5 N=21 N=21 N=46 N=46 Mean number of reads

29,206,762 34,031,468 38,671,220 35,591,282 21-70 x 106

mapping percentage

59.00 62.71 62.57 62.26 51-80 x 106

Mean number ofmapped reads

14,151,812 22,571,654 26,435,760 23,402,441 12-51 x 106

Percentage on target

48.60 66.33 66.95 64.72 42-85 %

Mean number of reads on target

6,955,220 15,167,653 18,504,941 15,785,119 5-43 x 106

Mean target sequence (Mbp)

348 758 925 789 250-2150

Mean coverage of requested target

97.60 156.76 183.86 162.57 81-359

Median coverage of requested target

63.00 85.43 94.14 86.94 52-217

Percentage of requested target covered

98.94 98.68 98.69 98.71 97-99 %

Percentage of designed target covered

94.17 93.62 93.58 93.66 91-96 %

Percentage of requested target covered > =20x

78.68 79.03 79.70 79.29 71-88 %

Percentage of designed target covered > =20x

85.38 85.74 86.45 86.02 77-95 %

R e s u lTs

Targeted Resequencing and Identification of Variants Basic patient characteristics are shown in Table 1 and a schematic representation of followed sequencing procedure is presented in figure 1. The number of sequenced nucleotides of all included samples was greater than 100 GB. The mean percentage of mapped reads was 62%, of which 65% on average was on target for our “Cancer Mini-Genome.” Per sample, an average of

Part IV Genomics

141

about 789 Mbp (range 250–2150) of the target genes were sequenced, resulting in an average mean and median coverage of 163 and 87, respectively (Table 2). In the exons of our “Cancer Mini-Genome” of all 42 tumor samples combined, 8,405 high confidence genetic variations were detected when compared to the reference genome(hg37), including 8,001 SNVs (96%) and 404 (4%) small insertions or deletions (figure 2). Of these identified SNVs, 84% (n = 6,696) were known variants when compared to ensembl59 (1,174 genes), and 1,305 were novel (667 genes). The complete catalog of all scored variations is shown in supplemental Table 3. To obtain an estimate of the inherited background of the genetic variations, we sequenced ‘normal’ non-neoplastic colorectal tissue of five patients of which also the paired primary CRC tumor and hepatic metastasis was analyzed. A total of 192 variants (2%) were found to be common in all these samples and were therefore excluded from the comparison. To further validate the NGS results, we assigned a random subset of candidate genetic variations for conventional sequencing techniques. In total, 108 variations in 32 genes were validated with a true-positive rate of 84%. supplemental Table 4 lists all re-validated variants using Sanger sequencing and the results of the diagnostic sequencing of KRAS.

figure 1 Schematic representation of the NGS-based Cancer Mini-Genome variation discovery workflow and analysis pipeline.

Selection of 21 CRC patiens with hepatic metastasis

Selecting Tumor Cells >80% by laser-capture micro-dissection microscope

Capturing exons of 1,264 cancer-associated genes ‘Cancer Mini-Genome’

next generation sequencing SOLiD 3+ system (Life Technolgies)

Bioinformatics Applying three independent SNV-callers

Prediction Protein-damaging impact of snV Polyphen-2

Validation of 118 variants (32 genes) using sanger sequencing

gDnA-isolation and Library Preparation

figure 1 Developed NGS-based analysis pipeline


142

Table 3 - Comparison of SNV’s of CRC genes between primary CRC tumor and subsequent hepatic metastasis currently used in clinical practice

Case BRAF EGFR HRAS KRAS NRAS PIK3CA

Patient #1 Primary CRC tumor wt wt wt wt wt wtPatient #1 Subsequent Livermetastasis wt wt wt wt wt wtPatient #2 Primary CRC tumor wt wt wt wt wt wtPatient #2 Subsequent Livermetastasis wt wt wt wt wt wtPatient #3 Primary CRC tumor wt wt wt wt wt wtPatient #3Subsequent Livermetastasis wt wt wt wt wt wtPatient #4 Primary CRC tumor wt wt wt G12A b wt wtPatient #4 Subsequent Livermetastasis wt wt wt wt wt wtPatient #5 Primary CRC tumor wt wt wt wt wt wtPatient #5 Subsequent Livermetastasis wt wt wt wt wt wtPatient #6 Primary CRC tumor wt wt wt wt wt wtPatient #6 Subsequent Livermetastasis wt wt wt wt wt wtPatient #7 Primary CRC tumor wt wt wt G12D b wt wtPatient #7 Subsequent Livermetastasis wt wt wt G12D b wt wtPatient #8 Primary CRC tumor wt wt wt wt wt wtPatient #8 Subsequent Livermetastasis wt wt wt wt wt wtPatient #9 Primary CRC tumor wt wt wt wt wt wtPatient #9 Subsequent Livermetastasis wt wt wt wt wt wtPatient #10 Primary CRC tumor wt wt wt G12V b wt wtPatient #10 Subsequent Livermetastasis wt wt wt G12V b wt wtPatient #11 Primary CRC tumor wt wt wt wt Q61K c wtPatient #11 Subsequent Livermetastasis wt wt wt wt Q61K c wtPatient #12 Primary CRC tumor wt wt wt wt wt wtPatient #12 Subsequent Livermetastasis wt wt wt G12D b wt wtPatient #13 Primary CRC tumor wt wt wt wt wt wtPatient #13 Subsequent Livermetastasis wt wt wt wt wt wtPatient #14 Primary CRC tumor wt wt wt wt wt wtPatient #14 Subsequent Livermetastasis wt wt wt wt wt wtPatient #15 Primary CRC tumor wt wt wt wt wt wtPatient #15 Subsequent Livermetastasis wt wt wt G12D b wt wtPatient #16 Primary CRC tumor wt wt wt G13D a wt wtPatient #16 Subsequent Livermetastasis wt wt wt G13D a wt wtPatient #17 Primary CRC tumor wt wt wt wt wt wtPatient #17 Subsequent Livermetastasis wt wt wt wt wt wtPatient #18 Primary CRC tumor wt wt wt wt wt wtPatient #18 Subsequent Livermetastasis wt wt wt wt wt wtPatient #19 Primary CRC tumor wt wt wt wt wt wtPatient #19 Subsequent Livermetastasis wt wt wt wt wt wtPatient #20 Primary CRC tumor wt wt wt G13D a wt wtPatient #20 Subsequent Livermetastasis wt wt wt G13D a wt wtPatient #21 Primary CRC tumor wt wt wt wt wt wtPatient #21 Subsequent Livermetastasis wt wt wt wt wt wt

a KRAS diagnostic codon 13 b KRAS diagnostic codon 12 c NRAS diagnostic codon 611. Patient # 3,6, 8-15 and 21 received chemotherapy between primary CRC resection and liver metastasectomy.2. No variants were detected in BRAF codon 600, EGFR codon 790 and 858, HRAS codon 12,13 and 61 , KRAS codons 61 and 146, NRAS codon 12 and 13, and PIK3CA codons 542, 545 and 1047.

Part IV Genomics

143

KNOWN SNV’S (1.174 genes) n=6.695 (84%) Synonymous = 5460 (83%) Essential Splice Site = 4 (<1%) Non-Synonymous = 1137 (17%) Stop Gained/Lost = 14 (<1%)

NOVEL SNV’S (667 genes) n=1.306 (16%) Synonymous = 962 (74%) Essential Splice Site = 1 (<1%) Non-Synonymous = 318 (33%) Stop Gained/Lost = 15 (<1%)

Comparison Between Primary CRC Tumor and Hepatic Metastasis Using stringent quality control, we filtered non-relevant genetic variations and subsequently compared the primary tumor and its metastasis to determine differences in somatic mutational constitution. On an individual basis, on average 83 (SD 69) variants were gained in the metastasis and 70 (SD 48) variants were lost as presented in figure 2. Next, we investigated the mutational status of a selected set of genes including KRAS, HRAS, NRAS, EGFR, PI3KCA, FLT1, KDR, PTEN, and BRAF, which are well-studied in relation to therapeutic responsiveness (7;8;25-30). Table 3 demonstrates a brief overview of the current clinically relevant codons of BRAF, EGFR, HRAS, KRAS, NRAS and PIK3CA. In almost all 21 CRC patients, we found aberrations in KRAS and EGFR (both ≥90%) either in the primary tumor and/or liver metastasis as shown in Table 4, which includes an overview of variants per

chromosome position for each gene. We identified genetic variants in KRAS codon 12 and 13 and in codon 61 of NRAS, which are currently biomarkers of resistance to anti-EGFR therapy. No variants were detected in BRAF codon 600, EGFR codon 790 and 858, HRAS codon 12, 13 and 61, KRAS codons 61 and 146, NRAS codon 12 and 13, and PIK3CA codons 542, 545 and 1047. In the majority of patients (67% and 52%, respectively), PI3KCA and FLT1 genes were affected; in a smaller but substantial part of our patient population, HRAS, NRAS, KDR, PTEN, and BRAF were mutated (10%, 24%, 19%,24%, and 38%, respectively). Subsequently, mutational-status differences per individual gene between primary tumor and liver metastasis were investigated. Taking all variations into account, dissimilarities of the KRAS and EGFR mutational status between both tumor entities were detected in 52% and 86% of the 21 CRC patients, respectively. Modest variability was observed for HRAS (24%), PIK3CA (19%), FLT1 (10%), NRAS (10%) and BRAF (14%). KDR and PTEN showed a more or less stable pattern of mutational status between both tumor identities, with only 5% of patients presenting deviations.

Total Number of Identified Variations on Exonic Regions n=8.405

Indels = 404 (5%)

snV’s = 8.001 (95%)

figure 2 - Overview of all identified variations on exonic regions of the ‘Cancer Mini-Genome.’

Primary colorectal cancers and their (subsequent) hepatic metastases are genetically different Chapter 8

144

figure 3 Genetic differences between primary tumors and their comparative liver metastases. Results are presented as loss or gain of relevant variants per individual patient. Patients were divided in two groups: with or without chemotherapy between surgical resection of the primary tumor and the metastasis. Chemotherapy did not influence the total number of gained or lost variants (p >0.2 for both).

figure 3 - Genetic differences between primary tumors and their comparative liver metastases

Part IV Genomics

-200 0 200Mean 70 snV’s Mean 83 snV’s

Number of Genetic Variations

lost Mutations gained Mutations

Che

mo-

Trea

ted

N=

11 P

atie

nts

Che

mo-

nai

veN

=10

Pat

ient

s

145

Table 4 Comparison of SNV’s of relevant CRC genes between primary CRC tumor and subsequent hepatic metastasis

Chromosome

Position dbSNP entry

COSMIC entry

Gene Name

No. of Patients with SNV

in the Primary Tumor


in the Liver Metastasis


Differences between Pimary tumor and Liver

metastasis

7 140.449.150 yes BRAF 4 7 37 140.453.154 no yes BRAF 1 1 —7 55.177.608 yes EGFR 1 3 27 55.187.053 yes EGFR 7 5 27 55.229.255 yes EGFR 13 14 37 55.233.089 yes EGFR 11 11 87 55.238.087 yes EGFR 1 3 27 55.238.253 yes EGFR 2 6 67 55.249.057 no EGFR 3 6 37 55.268.346 no EGFR 3 8 57 55.268.916 yes EGFR 9 11 67 55.274.084 yes EGFR 2 5 57 55.275.482 yes EGFR 3 6 37 55.275.910 yes EGFR 1 1 —7 55.276.094 yes EGFR 6 6 —7 55.276.144 yes EGFR 1 2 17 55.276.280 yes EGFR 2 3 17 55.277.751 yes EGFR 2 2 27 55.278.852 yes EGFR 8 11 3

13 28.875.434 yes FLT1 5 3 213 28.875.789 yes FLT1 5 5 —13 28.893.642 no FLT1 9 10 313 28.896.979 yes FLT1 2 2 —13 28.964.198 no FLT1 4 4 —11 537031 yes HRAS 11 11 211 534242 yes HRAS 17 17 44 55.972.946 yes KDR 1 1 —4 55.979.558 yes KDR 1 2 14 55.991.717 no KDR 1 1 —

12 25.359.227 yes KRAS 1 1 —12 25.359.841 yes KRAS 3 3 —12 25.360.559 yes KRAS 1 2 112 25.361.142 yes KRAS 7 9 212 25.361.646 yes KRAS 13 14 712 25.361.756 yes KRAS 2 3 312 25.362.552 yes KRAS 12 11 312 25.362.777 no KRAS 7 8 112 25.398.262 no yes KRAS 3 3 412 25.398.281 no yes KRASa 2 2 —12 25.398.284 no yes KRASb 3 4 31 115256530 no yes NRASc 1 1 —1 115249843 yes NRAS 4 3 23 178.954.702 no PIK3CA 1 2 13 178.957.783 yes PIK3CA 10 13 3

10 89.729.772 yes PTEN 1 1 —10 89.731.297 yes PTEN 2 1 1

Notes:a KRAS diagnostic codon 13 b KRAS diagnostic codon 12 c NRAS diagnostic codon 611. No variants were detected in BRAF codon 600, EGFR codon 790 and 858, HRAS codon 12,13 and 61, KRAS codons 61 and 146, NRAS codon 12 and 13, and PIK3CA codons 542, 545 and 1047.2. For the genes BRAF, FLT1, KDR, KRAS, NRAS, and PTEN one or more probable protein-damaging variant(s) were detected.3. For all these mentioned genes, KRAS codon 12 and 13 were re-sequenced using Sanger Methodology.4. Supplemental Table-5 presents on overview of all variant information for these genes per patients’ primary tumor and hepatic metastasis.


146

Tabl

e 5

- N

ovel

and

kno

wn

iden

tified

mut

atio

ns c

ateg

oriz

ed fo

r C

RC

rel

evan

t gen

es a

nd th

erap

eutic

targ

ets

with

pre

dict

ion

on p

rote

in fu

nctio

n

Cat

egor

yTh

erap

eutic

Ta

rget

(s)

Up/

dow

nstr

eam

or

asso

ciat

ed g

enes

Pro

babl

y/P

ossi

bly

Dam

agin

g P

rote

in F

unct

ion

Tum

or S

uppr

esso

r Gen

esA

PC

; DC

C; A

PC

2; P

TEN

; STK

11; T

P53

; SM

AD

4A

PC

*; D

CC

; PTE

N*;

TP

53*

Onc

ogen

esC

DK

8; B

RA

F; H

RA

S; K

RA

S; N

RA

S; R

ETB

RA

F*; K

RA

S*;

NR

AS

*; R

ET

Cel

l Cyc

leA

UR

K, P

LKA

UR

KA

-C; C

CN

A/B

/D/E

/H; C

DC

14A

/B; C

DC

25A

-C; C

DK

1/2/

6;

CD

KN

1A/B

; CD

KN

2C/D

; E2F

1-8;

GA

DD

45; H

IPK

1-4;

INC

EN

P; M

CL1

; M

CM

2-7;

PLK

1-; S

TAT3

; SK

P2;

WE

E1

CC

NH

; CD

C14

A; C

DK

N1B

*; E

2F2*

/8*;

HIP

K4;

P

LK4

Apo

ptos

isA

LPK

1-3;

APA

F1; A

RA

F; A

TM; B

AX

; BC

L2L1

4; B

CL9

; BID

; CA

D;

CA

SP

1-11

; CFL

AR

; CH

EK

1/2;

DIA

BLO

; DFF

B; F

AD

D; F

AS

; FA

STK

; IR

AK

2/3;

ITC

H; M

AP

3K5;

MD

M2/

4; R

OC

K2;

TN

FRS

F8; T

NFR

SF1

B;

TNFR

SF-

10A

/D; T

P53

BP

1,2;

TR

AD

D; T

RA

F2; T

RA

F3IP

2; T

RR

AP

ALP

K-1,

2*,3

; ATM

, BC

L2L1

4*; B

CL9

; BID

*;

CA

SP

10; I

RA

K2,

3; IT

CH

*; T

NFR

SF1

0-A

/D*;

TP

53B

P1,

2; T

RA

DD

*

DN

A m

ism

atch

Rep

air

ER

CC

1-6;

MKS

1; M

LH-1

/3; M

SH

-2/6

; MU

TYH

, RA

D9A

; RA

D50

/51/

52E

RC

C-4

,5; R

AD

52*

Epi

gene

tic G

enes

HD

AC

1-11

; MLH

1H

DA

C6*

,7*,

10*

EG

FR-p

athw

ayE

GFR

, ER

BB

CR

K, E

GF,

EG

FL6,

ELK

1; E

RB

B2-

4, G

RB

2; J

AK

2,3;

JU

N, J

UN

B, H

RA

S;

KR

AS

; KS

R1/

2; M

AP

2K1/

2/4;

MA

P3K

1; M

AP

KA

PK

2/3;

MA

PK

3/8;

MYC

; R

AF1

; RA

P1A

; RA

P1G

AP

; RA

PG

EF2

/3/4

; RA

SA

1/2/

3; R

AS

GR

F2; S

HC

-1/

2; S

RF;

SO

S1,

STA

T-5A

/B

ER

BB

2/3*

; KS

R1;

MA

P3K

1*; M

OS

; PLC

G1*

; R

AF1

*; R

AP

GE

F3*/

4*

PI3

K-pa

thw

ayP

I3K

AK

T1-3

; AK

T1S

1; C

DK

N1A

/B; C

TNN

B1;

CTN

NB

IP1;

FLC

N; F

OXO

1/3/

4;

GS

K3B

; IR

S1/

2; M

OS

; NR

AS

; PA

K4;

PD

K1;

PD

PK

1; P

IK3C

2-A

/B/G

; P

IK3C

3; P

IK3C

-A/B

/D/G

; PIK

3R-1

/2/4

/5; P

DK-

1/2;

PR

KA

A2;

SG

K1/

2

FOX

O4*

; IR

S-1

*/2*

; PIK

3C2B

/G*;

PK

D1;

S

GK

1

mTO

R-p

athw

aym

TOR

DD

IT4;

EIF

4A2;

EIF

4B/E

; MTO

R; R

HE

B; R

ICTO

R; R

PS

6KB

1/2;

STR

AD

B;

TSC

1/2

MO

S; M

TOR

; RIC

TOR

*

TGFB

-pat

hway

TGFB

RLT

BP

1; M

AP

3K7;

SM

AD

2-4/

7; T

FG; T

GFB

1; T

GFB

R1/

2LT

BP

1; T

GFB

1/2*

VE

GF-

path

way

KD

R, F

ltC

DC

42; C

DC

42B

PA/G

; Flt-

1,3,

4; H

SP

90A

A1;

KD

R; M

AP

K14

; MA

P-

KA

PK

2-3;

NFA

TC1/

3; P

TK2-

B; P

XN

; PTK

2-B

; SR

C; V

EG

FAC

DC

42B

PG

; Flt-

3/4;

KD

R; N

FATC

1; P

LCG

1*

Rec

epto

r (Ty

rosi

ne K

inas

es)

ALK

, EP

HA

/B,

IGFR

, MS

TR,

NTR

K, P

DG

FR,

TLR

ALK

; EP

HA

3; E

PH

B6;

ELK

4; F

BX

W7;

IGF2

; IG

F1/2

R; I

GS

F9; I

L1A

/B;

INS

R; I

RS

1/2;

KIT

; ME

RTK

; MET

; MS

T1R

; NTR

K1-

3; P

DG

FRA

, PD

G-

FRB

1/2;

PD

GFR

L; T

LR1-

5

ALK

*; E

PH

A1*

/A5*

; IG

F2R

*; IG

SF9

*; IL

1A*;

IN

SR

*; IR

S1*

; KIT

; MS

T1R

*; N

TRK

1; P

DG

-FR

B*;

TLR

1*/3

/4/5

*

* In

clud

es n

ovel

iden

tified

SN

V’s

for t

hat s

peci

fic g

ene

Part IV Genomics

147

supplemental Table 5 presents on overview of all variant information for these genes per patient. To investigate the impact of chemotherapy on the mutational status of the metastasis, CRC patients were categorized into two equally matched groups of chemotherapy-naive (=10) and chemotherapy-treated (n=11) patients. In this relatively small data set we could not show a significantly increased number of variants as a result of chemotherapy. Accordingly, a significant relation with genetic burden and time between primary tumor and metastasis resection was not observed.

Probability scores of Protein Impact of non-synonymous Variations with emphasis on the egfR PathwayPrediction of possible impact on protein function of all non-synonymous variations (1,455), identified 41 nonsense-, splice site variants and indels (small insertions/deletions). Of all other 1,414 non-synonymous variants, 281 were predicted as probably damaging (p>0,85, using polyphen2). Moreover, 170 were classified as possibly damaging (0.20>>p<<0.85), 740 as benign (p<0.2), and 261 not classified at all. The entire catalog of all genes potentially affected at the protein level is listed in supplemental Table 3. Novel potentially protein-changing mutations were identified throughout our cancer mini-genome which contains genes that are relevant for a variety of cellular functions and responses (i.e. mTOR-, TGFb-pathways, tyrosine kinases, apoptosis, etc.) (Table 5). Given the current significance of KRAS status for treatment decision making, we specifically addressed the EGFR pathway (25). First, we analyzed the variants in codons 12 and 13 by conventional sequencing and found mutations in either codon 12 or 13 in six out of 21 patients (supplemental Table 4). One patient converted to wild-type KRAS in the metastasis. These data were identical to the data generated by our NGS platform. Next, we analyzed individual components of the EGFR pathway. None of the 21 CRC patients had a predicted protein-changing EGFR mutation in the primary tumor or liver metastasis. Ten of the 21 CRC patients had a relevant genetic variation in BRAF or KRAS relevant for possible therapeutic responsiveness. Interestingly, every patient showed one or more protein-altering genetic aberration in genes downstream of EGFR. These genes included MAP3K1 (MEKK1), MAPK10 (JNK3), PLCG1 (PLCγ), PIK3CG (PI3K), PRKCB (PKC), and RAF1 (Table 5). Furthermore, we also found KSR1, MOS, NF1, and RAPGEF-3/4—all potentially relevant variants in other genes strongly associated to the EGFR-pathway.

D I sCus s Ion

The purpose of the study was to determine which tissue best reflects the presence of a target for therapy: the primary CRC or their metastasis. The most important finding was that substantial genetic differences existed between the primary tumor and its metastasis. Significant numbers of either loss or gain of functionally relevant genetic variations were found, suggesting that metastases may provide a better predictive window for targeted therapy than the primary cancer. In the era of targeted therapy and personalized cancer treatment, individual genetic mutations in the primary tumor drives patient selection and therapeutic strategies (7;8;31). If we focus on the codon 12 and 13 KRAS mutations commonly used to determine whether metastatic CRC patients will benefit from treatment with antibodies against EGFR, our data are in line with recent reports where a concordance of >95% for the 12G mutation was found between the primary tumor and metastasis (27;29;32). However, we analyzed all exons of the KRAS gene and found a



substantial number of additional genetic changes outside codon 12 and 13. Their clinical relevance has yet to be determined but our findings underline the importance of having a complete overview of the genetic alterations. However, the power of the NGS analysis used in this study is that it provides a complete overview of all known components of the targeted pathway. Using this technology we demonstrated that, in addition to the well known diagnostic KRAS codon 12 and 13 mutations, several variants that could potentially alter protein function were present in other components of the pathway. This may explain why the clinical benefit of anti-EGFR antibodies remains limited, even in preselected patients based on codon 12/13 KRAS analysis. We also clearly showed a substantial loss or gain of variants from the primary tumor to its metastasis in our selected set of 1,264 genes. The finding that chemotherapy did not change the genetic profile significantly was rather surprising, but can have several explanations. This study was not primarily designed to investigate chemotherapy induced changes and therefore lacked sample power to draw any definitive conclusions. Metastases are generally heterogeneous, and clonality may be a confounder in the analysis (33-36). In CRCs, the cancer-initiating genes of the primary tumor were investigated in the metastasis; this demonstrated heterogeneity of identified variants in the primary tumor and its paired metastasis, which supports our findings (34;37). Similarly, in lobular breast cancer, the mutational evolution between the metastasis and its primary origin was studied in a single patient, illuminating 19 novel non-synonymous coding mutations of the 32 variations discovered in the metastasis (38). In a patient with basal-like breast cancer, the metastasis displayed two de novo mutations and an additional large deletion (39). These numbers are lower than observed in our analysis, which may be explained by the enrichment for specific genes and higher sequencing coverage in our study and the stringency of the analysis that allows the detection of lower frequency mutations. Recently, the genomes of the primary origin and metastasis have been sequenced in pancreatic cancer patients in two independent investigations (40;41). Both studies showed that clonal expansions found in the metastasis primarily originated from the primary tumor, demonstrating that clones evolve over time. Mutations in cancer-driving genes may occur later and help the tumors to become more malignant under a certain selection pressure. Equally, Evgenieva et al. showed that multipoint microsampling elucidated discrepancies in genetic variations in APC, KRAS, TP53 between primary CRC tumors and its subsequent hepatic metastasis and also demonstrated subpopulations for these genes in most individual tumors (55). These studies show the potentially very important contribution of genetic heterogeneity. Future prospective studies including the analysis of multiple specimen of the same tumor may provide a better perspective of the relevance of genetic heterogeneity to clinical decision making. This holds true for both current diagnostic testing of tumors and the use of NGS. In our study, chemotherapy did not further increase the number of genetic variations; however, the number of patients analyzed is too small to draw any firm conclusions. Furthermore, the selective pressure does not necessarily act through quantity of mutations, but rather their quality. In the present study we went to great lengths to make sure that the quality of the sequence data generated was high. Samples were sequenced until we reached a predetermined level of 40x mean coverage, which is generally accepted in the field to allow reliable heterozygote variant detection. Using this strategy we were able to cover our intended target regions for more than 83% with sufficient coverage to allow for confident variant calling. Our study does reveal some future challenges. First, differentiation between low- and high-frequency variations was not our primary aim, but could potentially identify “initiating” and “driving” variants (12;42). Complicating factors as tumor heterogeneity, genome or gene copy number status,

149

quantitative sequencing need also to be addressed in further studies. Second, to acquire a basic idea of the fraction of germline variants we sequenced normal non-neoplastic tissue of five patients to determine an approximate level of germline contribution. However, for clinical interventions of metastatic disease, only the differences in genetic make-up between primary CRC tumor and hepatic metastasis are therapeutically relevant. Third, for target-enriched next-generation sequencing we used genomic DNA from selected tumor-rich areas (>80%) from FFPE material. It is debated whether the FFPE material procedure introduces variation by the depurination or fragmentation. However, Solid-sequencing is not limited by the shorter fragment sizes in a FFPE DNA sample. Schweiger et al. compared FFPE tissue with snap-frozen tissue and showed identical results for copy number and single nucleotide variants (56). We compared the mutational spectra with non-FFPE material and did not observe significant difference as well (data not shown). However, as DNA yields are generally low and highly variable, systematic exploitation of large collections of archived material for retrospective studies will still be challenging. Fourth, functional confirmation of the novel identified variants is necessary to discern real pathogenic from benign variants and to validate bioinformatic prediction algorithms. Systematic collection of these large scale data in open source databases enables the possibility to address these issues. Over the next few years, enormous amounts of genetic data will likely be generated. Improved tools to analyze cellular pathways and network analyses will become readily available (43-49). Linking clinical outcome to these genetic data and enforcing open-source data-sharing will help to accelerate the development of clinical decision-making algorithms (50;51). Taken together, our study indicates that obtaining a biopsy from the metastasis and re-evaluating the genetic makeup at the time of treatment may be preferred over an archived primary tumor. This requires a different mindset for oncologists because biopsies from metastases are not commonly taken at the start of treatment. Obviously, the clinical benefits need to outweigh the risk that patients take by undergoing a biopsy. However, in the metastatic setting, the complication rate of ultrasound-guided biopsies is 0–5% depending on tumor location (52-54). Biopsies should be guided by stringent quality control, and items such as the percentage tumor cells in the biopsy, tumor inflammation, and necrosis should be noted. If patients are allocated to clinical studies with targeted agents, a NGS readout is currently not sufficient and the target should be validated by other certified methods. Similarly, drug approval by the regulatory authorities will also require stringent quality control measures. Regarding therapeutic responsiveness, it is also crucial to investigate differences in genetic profiles between primary tumors and synchronous metastases. In conclusion, this study demonstrates that the differences in potentially relevant genetic variations between the primary CRC and hepatic metastases in important pathways are of such magnitude that an impact on treatment outcome is realistic. This indicates that genetic pathway-analysis of the metastasis may have more predictive power when patients are selected for specific treatment modalities, thus allowing for further refinement of treatment algorithms.


This study was supported by unrestricted grants from Roche and the Dutch ‘Barcode for Life’ Foundation.

supplementary tables have been published online: http://clincancerres.aacrjournals.org/content/18/3/688.long


150


(1) Romond EH, Perez EA, Bryant J, Suman VJ, Geyer CE, Jr., Davidson NE et al. Trastuzumab plus adjuvant

chemotherapy for operable HER2-positive breast cancer. N Engl J Med 2005 October 20;353(16):1673-84.

(2) Loupakis F, Pollina L, Stasi I, Ruzzo A, Scartozzi M, Santini D et al. PTEN expression and KRAS mutations on

primary tumors and metastases in the prediction of benefit from cetuximab plus irinotecan for patients with

metastatic colorectal cancer. J Clin Oncol 2009 June 1;27(16):2622-9.

(3) Di NF, Martini M, Molinari F, Sartore-Bianchi A, Arena S, Saletti P et al. Wild-type BRAF is required for response

to panitumumab or cetuximab in metastatic colorectal cancer. J Clin Oncol 2008 December 10;26(35):5705-12.

(4) Lievre A, Bachet JB, Le CD, Boige V, Landi B, Emile JF et al. KRAS mutation status is predictive of response to

cetuximab therapy in colorectal cancer. Cancer Res 2006 April 15;66(8):3992-5.

(5) Tol J, Koopman M, Cats A, Rodenburg CJ, Creemers GJ, Schrama JG et al. Chemotherapy, bevacizumab, and

cetuximab in metastatic colorectal cancer. N Engl J Med 2009 February 5;360(6):563-72.

(6) Garm Spindler KL, Pallisgaard N, Rasmussen AA, Lindebjerg J, Andersen RF, Cruger D et al. The importance

of KRAS mutations and EGF61A>G polymorphism to the effect of cetuximab and irinotecan in metastatic

colorectal cancer. Ann Oncol 2009 May;20(5):879-84.

(7) Bardelli A, Siena S. Molecular mechanisms of resistance to cetuximab and panitumumab in colorectal cancer.

J Clin Oncol 2010 March 1;28(7):1254-61.

(8) Hawkes E, Cunningham D. Relationship between colorectal cancer biomarkers and response to epidermal

growth factor receptor monoclonal antibodies. J Clin Oncol 2010 October 1;28(28):e529-e531.

(9) Potti A, Dressman HK, Bild A, Riedel RF, Chan G, Sayer R et al. Genomic signatures to guide the use of

chemotherapeutics. Nat Med 2006 November;12(11):1294-300.

(10) Ginsburg GS, Willard HF. Genomic and personalized medicine: foundations and applications. Transl Res 2009

December;154(6):277-87.

(11) Chin L, Gray JW. Translating insights from the cancer genome into clinical practice. Nature 2008 April

3;452(7187):553-63.

(12) Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med 2004 August;10(8):789-99.

(13) Sjoblom T, Jones S, Wood LD, Parsons DW, Lin J, Barber TD et al. The consensus coding sequences of

human breast and colorectal cancers. Science 2006 October 13;314(5797):268-74.

(14) Nguyen DX, Bos PD, Massague J. Metastasis: from dissemination to organ-specific colonization. Nat Rev

Cancer 2009 April;9(4):274-84.

(15) Nijman IJ, Mokry M, van BR, Toonen P, de BE, Cuppen E. Mutation discovery by targeted genomic enrichment

of multiplexed barcoded samples. Nat Methods 2010 November;7(11):913-5.

(16) Manning G, Whyte DB, Martinez R, Hunter T, Sudarsanam S. The protein kinase complement of the human

genome. Science 2002 December 6;298(5600):1912-34.

(17) Sjoblom T. Systematic analyses of the cancer genome: lessons learned from sequencing most of the annotated

human protein-coding genes. Curr Opin Oncol 2008 January;20(1):66-71.

(18) Mokry M, Feitsma H, Nijman IJ, de BE, van der Zaag PJ, Guryev V et al. Accurate SNP and mutation detection

by targeted custom microarray-based genomic enrichment of short-fragment sequencing libraries. Nucleic

Acids Res 2010 June;38(10):e116.

(19) Li H, Durbin R. Fast and accurate long-read alignment with Burrows-Wheeler transform. Bioinformatics 2010

March 1;26(5):589-95.

(20) Jones S, Zhang X, Parsons DW, Lin JC, Leary RJ, Angenendt P et al. Core signaling pathways in human

pancreatic cancers revealed by global genomic analyses. Science 2008 September 26;321(5897):1801-6.

Part IV Genomics

151

(21) Wood LD, Parsons DW, Jones S, Lin J, Sjoblom T, Leary RJ et al. The genomic landscapes of human breast and

colorectal cancers. Science 2007 November 16;318(5853):1108-13.

(22) Koboldt DC, Chen K, Wylie T, Larson DE, McLellan MD, Mardis ER et al. VarScan: variant detection in massively

parallel sequencing of individual and pooled samples. Bioinformatics 2009 September 1;25(17):2283-5.

(23) McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A et al. The Genome Analysis

Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res 2010

September;20(9):1297-303.

(24) Adzhubei IA, Schmidt S, Peshkin L, Ramensky VE, Gerasimova A, Bork P et al. A method and server for

predicting damaging missense mutations. Nat Methods 2010 April;7(4):248-9.

(25) Chen J, Huang XF, Katsifis A. Activation of signal pathways and the resistance to anti-EGFR treatment in

colorectal cancer. J Cell Biochem 2010 December 1;111(5):1082-6.

(26) Dasari A, Messersmith WA. New strategies in colorectal cancer: biomarkers of response to epidermal growth

factor receptor monoclonal antibodies and potential therapeutic targets in phosphoinositide 3-kinase and

mitogen-activated protein kinase pathways. Clin Cancer Res 2010 August 1;16(15):3811-8.

(27) Knijn N, Mekenkamp LJ, Klomp M, Vink-Borger ME, Tol J, Teerenstra S et al. KRAS mutation analysis: a

comparison between primary tumours and matched liver metastases in 305 colorectal cancer patients. Br J

Cancer 2011 March 15;104(6):1020-6.

(28) Laurent-Puig P, Cayre A, Manceau G, Buc E, Bachet JB, Lecomte T et al. Analysis of PTEN, BRAF, and EGFR

status in determining benefit from cetuximab therapy in wild-type KRAS metastatic colon cancer. J Clin Oncol

2009 December 10;27(35):5924-30.

(29) Park JH, Han SW, Oh DY, Im SA, Jeong SY, Park KJ et al. Analysis of KRAS, BRAF, PTEN, IGF1R,

EGFR intron 1 CA status in both primary tumors and paired metastases in determining benefit from cetuximab

therapy in colon cancer. Cancer Chemother Pharmacol 2011 February 22.

(30) Richman SD, Seymour MT, Chambers P, Elliott F, Daly CL, Meade AM et al. KRAS and BRAF mutations in

advanced colorectal cancer are associated with poor prognosis but do not preclude benefit from oxaliplatin or

irinotecan: results from the MRC FOCUS trial. J Clin Oncol 2009 December 10;27(35):5931-7.

(31) Ocana A, Pandiella A. Personalized therapies in the cancer “omics” era. Mol Cancer 2010;9:202.

(32) Mariani P, Lae M, Degeorges A, Cacheux W, Lappartient E, Margogne A et al. Concordant analysis of KRAS

status in primary colon carcinoma and matched metastasis. Anticancer Res 2010 October;30(10):4229-35.

(33) Bell DW. Our changing view of the genomic landscape of cancer. J Pathol 2010 January;220(2):231-43.

(34) Goasguen N, de CC, Brouquet A, Julie C, Prevost GP, Laurent-Puig P et al. Evidence of heterogeneity within

colorectal liver metastases for allelic losses, mRNA level expression and in vitro response to chemotherapeutic

agents. Int J Cancer 2010 September 1;127(5):1028-37.

(35) Marusyk A, Polyak K. Tumor heterogeneity: causes and consequences. Biochim Biophys Acta 2010

January;1805(1):105-17.

(36) Salk JJ, Fox EJ, Loeb LA. Mutational heterogeneity in human cancers: origin and consequences. Annu Rev

Pathol 2010;5:51-75.

(37) Jones S, Chen WD, Parmigiani G, Diehl F, Beerenwinkel N, Antal T et al. Comparative lesion sequencing

provides insights into tumor evolution. Proc Natl Acad Sci U S A 2008 March 18;105(11):4283-8.

(38) Shah SP, Morin RD, Khattra J, Prentice L, Pugh T, Burleigh A et al. Mutational evolution in a lobular breast

tumour profiled at single nucleotide resolution. Nature 2009 October 8;461(7265):809-13.

(39) Ding L, Ellis MJ, Li S, Larson DE, Chen K, Wallis JW et al. Genome remodelling in a basal-like breast cancer

metastasis and xenograft. Nature 2010 April 15;464(7291):999-1005.

(40) Campbell PJ, Yachida S, Mudie LJ, Stephens PJ, Pleasance ED, Stebbings LA et al. The patterns and dynamics

of genomic instability in metastatic pancreatic cancer. Nature 2010 October 28;467(7319):1109-13.


152

(41) Yachida S, Jones S, Bozic I, Antal T, Leary R, Fu B et al. Distant metastasis occurs late during the genetic

evolution of pancreatic cancer. Nature 2010 October 28;467(7319):1114-7.

(42) Velculescu VE. Defining the blueprint of the cancer genome. Carcinogenesis 2008 June;29(6):1087-91.

(43) Hawkins RD, Hon GC, Ren B. Next-generation genomics: an integrative approach. Nat Rev Genet 2010

June 8;11(7):476-86.

(44) Mamanova L, Coffey AJ, Scott CE, Kozarewa I, Turner EH, Kumar A et al. Target-enrichment strategies for next-

generation sequencing. Nat Methods 2010 February;7(2):111-8.

(45) Metzker ML. Sequencing technologies - the next generation. Nat Rev Genet 2010 January;11(1):31-46.

(46) Shendure J, Ji H. Next-generation DNA sequencing. Nat Biotechnol 2008 October;26(10):1135-45.

(47) Swanton C, Caldas C. Molecular classification of solid tumours: towards pathway-driven therapeutics. Br J

Cancer 2009 May 19;100(10):1517-22.

(48) Baudot A, de lT, V, Valencia A. Mutated genes, pathways and processes in tumours. EMBO Rep 2010

October;11(10):805-10.

(49) Chittenden TW, Howe EA, Culhane AC, Sultana R, Taylor JM, Holmes C et al. Functional classification analysis

of somatically mutated genes in human breast and colorectal cancers. Genomics 2008 June;91(6):508-11.

(50) Ein-Dor L, Zuk O, Domany E. Thousands of samples are needed to generate a robust gene list for predicting

outcome in cancer. Proc Natl Acad Sci U S A 2006 April 11;103(15):5923-8.

(51) Ding L, Wendl MC, Koboldt DC, Mardis ER. Analysis of next-generation genomic data in cancer:

accomplishments and challenges. Hum Mol Genet 2010 October 15;19(R2):R188-R196.

(52) Appelbaum L, Kane RA, Kruskal JB, Romero J, Sosna J. Focal hepatic lesions: US-guided biopsy--lessons from

review of cytologic and pathologic examination results. Radiology 2009 February;250(2):453-8.

(53) Terjung B, Lemnitzer I, Dumoulin FL, Effenberger W, Brackmann HH, Sauerbruch T et al. Bleeding complications

after percutaneous liver biopsy. An analysis of risk factors. Digestion 2003;67(3):138-45.

(54) Thanos L, Zormpala A, Papaioannou G, Malagari K, Brountzos E, Kelekis D. Safety and efficacy of percutaneous

CT-guided liver biopsy using an 18-gauge automated needle. Eur J Intern Med 2005 December;16(8):571-4.

(55) Teodora Evgenieva Goranova, Masayuki Ohue, Yutaro Shimoharu and Kikuya Kato

Dynamics of cancer cell subpopulations in primary and metastatic colorectal tumors

Clinical and Experimental Metastasis 2011; 28:427–435

(56) Michal R. Schweiger, Martin Kerick, Bernd Timmermann, Marcus W. Albrecht et al.

Genome-Wide Massively Parallel Sequencing of FFPE Tumor Tissues for Copy-Number- and Mutation-

Analysis. Plos One 2009 May; Volume 4; Issue 5; e5548

Part IV Genomics

153 Primary colorectal cancers and their (subsequent) hepatic metastases are genetically different Chapter 8


Chapter 9 general discussionand future perspectives

156

The major ambition of current clinical cancer research is to increase and optimize ‘personalized’ cancer therapy. This thesis attempt to contribute to this topic by: (Part I) elucidating two novel tissue biomarkers in renal cell cancer (RCC) patients; (Part II) improving prognostication and risk-redirected therapy in metastasized RCC patients with a new blood biomarker; (Part III) developing a qPCR determining prognostic circulating mitochondrial DNA in patients with several cancer types and (Part IV) illustrating that differences in genetic constitution between the primary Colorectal (CRC) tumors and their subsequent hepatic metastases are of such magnitude that an impact on targeted therapies is realistic, based on the genetic profile of the metastasis.

PART – INovel tissue biomarkers improve disease staging in Renal Cell Cancer (RCC)

Chapter 3Clinical and pathologic factors are presently used in various prognostic models for disease staging of Renal Cell Cancer (RCC). The past decade cancer research increasingly elucidated entire molecular pathways and enabled therefore the search for novel tissue biomarkers to improve these disease staging models. The independent prognostic significance of the upstream proteins -prolyl hydroxylases domain proteins (PHD) 1, 2 and 3 and factor inhibiting HIF (FIH)-, that directly influence the hypoxia inducible factor (HIF) pathway, were analyzed in this study for their prognostic significance in patients with clear cell renal cell carcinoma (ccRCC). A tissue microarray, including tumor tissue of 100 ccRCC patients who underwent nephrectomy, was constructed to determine the immunohistochemical expression levels of HIF, FIH and PHD1, 2 and 3. Consecutively the expression levels of these proteins were associated with overall survival (OS) and other clinicopathological factors. Expression levels of HIF-1α were significantly associated with HIF-2α , PHD1, PHD2, PHD3, FIH and VHL. Accordingly, HIF-2a was correlated with FIH and PHD2. Univariate Cox regression analysis demonstrated a significant correlation of tumor stage, grade, diameter, metastastic disease and FIH expression levels with OS. Low nuclear expression of FIH showed poor OS, where high nuclear intensity of FIH was associated with favorable OS. Moreover, with multivariable analysis low nuclear FIH levels remained a strong independent prognostic factor. FIH inhibits HIF-α in an oxygen-dependent manner. Only in severe hypoxia condition FIH becomes inactive (1). This suggests that FIH may have an important function as one of the final checks on HIF-α transcriptional activity. The absence of nuclear FIH in more aggressive phenotypes could be explained by increasing gene mutations within these tumors, including FIH gene mutations. However, this is unlikely, since FIH gene mutations have not been found in RCC (2). It is therefore possible that FIH is actively retained in the cytoplasm or exported out of the nucleus in tumors associated with worse prognosis. In invasive breast cancer, both cytoplasmic FIH expression and absence of nuclear FIH were independent prognostic factors for a shorter disease-free survival(3), which supports our results. Both HIF-1α and HIF-2α, which are central proteins of the HIF pathway, failed to show a correlation to survival and the most important clinicopathological parameters in this study. Under hypoxic conditions both HIF-1α and HIF-2α upregulate PHD protein expression. This may serve as a negative feedback to decrease HIF-α activity (4;5). Since hypoxia is common in RCC and many other solid tumors (6), a correlation between molecules participating in the hypoxia response pathway is not surprising

157General discussion and future perspectives Chapter 9

and indicates the intricate interplay between these proteins. Alternatively, FIH may exert its effects on tumor biology in a HIF-independent fashion. In conclusion, this investigation demonstrated that FIH can be used as a novel independent tissue biomarker for prognostication of OS of ccRCC patients. These results mark the essence to study upstream regulatory proteins of the HIF pathway besides downstream targets. Future prospective studies are warranted to evaluate whether FIH expression levels are of additional clinical importance when incorporated in the currently used prognostic models for disease staging for non-metastasized ccRCC.

Chapter 4Mutations of the von Hippel-Lindau (VHL) gene are the most common cause of sporadic and inherited renal cell carcinoma (RCC). There is increasing evidence that the transcription factor E2F1 is regulated by VHL. This study attempts to reveal this regulation mechanism in an established zebrafish model for VHL and in VHL-deficient RCC cell lines. The second aim is to demonstrate the clinical importance of E2F1 in tumor material from RCC patients. Re-introduction of wildtype and mutant-pVHL into VHL-deficient RCC cell lines resulted in a marked downregulation of E2F1 expression and activity in a dose-dependent and HIFα-independent manner. Cell cycle defects attributed to pVHL dysfunction in RCC cell lines are rescued with ectopic overexpression of E2F1. In zebrafish and RCC cells the MCM3 which is an expression target of E2F1 was regulated by pVHL. At a molecular level, the novel regulation of E2F1 expression by pVHL, might be the result of the demonstrated correlation of pVHL with the TRIM28-SETDB1-RBBP4/7 complex. Using immunohistochemistry on a large retrospective cohort of 138 RCC patients, this study demonstrated that increased nuclear expression of E2F1 is observed in primary RCC tumors when compared to matched normal kidney tissue from the same individual. Patients with germline VHL mutations expressed significantly higher nuclear levels of E2F1 in RCC compared to RCC of patients with either clear-cell (cc) or non-cc histology. Accordingly, higher E2F1 expression (> 25%) was associated with smaller tumor diameter (<7cm) and with a favorable American Joint Committee on Cancer (AJCC) stage. Univariate Cox regression analysis correlated decreased nuclear E2F1 expression (< 25%) with poor disease-free (DFS) and reduced overall survival (OS), and elevated E2F1 expression (> 25%) with favorable DFS and OS, thereby elucidating E2F1 expression in the primary RCC as novel prognosticator. Other studies also demonstrate an association of overexpression of E2F1 with increased DFS and OS in several cancer types (7-9), but in contrary high E2F1 expression is correlated with decreased DFS and/or poor OS (10-12). This apparent contradiction might be attributable to the more squamous cell subtype, the heterogeneity of the used patient cohorts and the low number of included patients. This study identified that ccRCCs with germline VHL mutations display significantly higher nuclear E2F1 expression than sporadic tumors of unknown VHL status, suggesting that VHL loss increases E2F1 expression in vivo. Because lack of VHL mutations is linked to increased tumor aggressiveness (13), these results suggest that VHL-deficient tumors are less aggressive due to E2F1-mediated tumor suppression. Furthermore, expression of E2F1 target MCM3 is regulated by pVHL in zebrafish and RCC cells, suggesting an effect on the cell cycle. Since E2F1 and MCMs promote G1/S cell cycle transition (14), regulation of E2F1 by pVHL might represent a novel role in cell regulation for pVHL. This investigation demonstrates that in RCCs, E2F1 significantly correlates to its known target p27 (15). Therefore it was not

158

surprising that high p27 expression was borderline significant with regards to a smaller tumor diameter. Collectively, these data might indicate that E2F1 prevents tumor progression through p27-mediated senescence. Overall, the regulation of E2F1 is orchestrated by multiple factors including several feedback loops; it is likely that E2F1 regulation is cell-type specific balancing the proliferation/apoptotic program (16). In conclusion, this study demonstrates in vitro and in vivo that E2F1 regulation is a novel HIFa-independent function for pVHL at the transcriptional level. Accordingly, E2F1 is a significant prognosticator for DFS and OS in patients who underwent neprectomy for RCC with curative intention. Prospective studies including multivariate analysis of known risk variables are necessary to evaluate whether this novel risk factor E2F1 is an independent prognosticator and whether E2F1 might be potentially useful for application in clinical management of RCC patients.

PART – I INovel serum proteins improve currently used risk models in metastasized RCC

Chapter 5The Memorial Sloan Kettering Cancer Center (MSKCC) risk model is widely used to allocate patients with metastatic renal cell cancer (mRCC) into three prognostic risk groups (favourable, intermediate or poor). Currently, these risk categories are used to direct mRCC patients to differently targeted therapies. To improve the predictive accuracy of the MSKCC risk model, we applied proteomics to identify novel serological proteins that are associated with overall survival (OS). Baseline sera of 114 interferon-based patients with mRCC were screened by applying Surface-Enhanced-Laser-Desorption-Time of Flight mass spectrometry (SELDI-TOF MS), identifying several proteins which were significantly correlated with overall survival (OS). Of these proteins, apolipoprotein A-II (ApoA2), serum amyloid alpha (SAA), and transthyretin were validated for their association with OS, using conventional and commercially available quantification techniques. With the combination ApoA2 and SAA a novel prognostic two-protein signature was generated. Next, including previously identified prognostic factors, multivariable Cox regression analysis elucidated ApoA2, SAA, Lactate Dehydrogenase (LDH), (ECOG) performance status and a number of metastasis sites as independent factors for OS. Using these five independent variables, a novel protein-based risk model was constructed allocating patients into three risk groups. This protein-based model predicted patient prognosis more robustly than the MSKCC risk model. Applying this protein-based model instead of the original MSKCC assignment would have changed the risk group in 38% of the patients. The proteins ApoA2 and SAA are involved in general physiological mechanisms as lipid metabolism and acute phase response. For example, SAA is either produced by hepatocytes, tumor cells or inflammatory cells and thus contributing to inflammation, tumor evolution and dissemination, as recently has been reviewed (17). Additionally, patients with aggressive and advanced cancer will become catabolic by high energy consuming tumors and loss of appetite, resulting in disturbed common lipid metabolism, acute phase response, nutrient losses and altered hepatic protein production. As a consequence, lipid spectra will change over time during the process of cancer progression. Therefore, protein profiles might predict prognosis in other cancer types and besides fluctuating levels in time, might be predictive for therapy outcome. As this study is based on mRCC patients mainly treated with interferon, further investigation

159

is necessary to demonstrate the prognostic accuracy in mRCC patients treated with Tyrosine Kinase Inhibitors, which are currently first choice of therapy. Other prognostic proteins revealed by SELDI-TOF MS, as Apolipoprotein A1, C1 and C2 for example, may also be of predictive importance and deserve further analysis. In conclusion, this study demonstrates that combining proteomics-based screening with subsequent validation by protein quantification methods can yield novel biomarkers. A novel generated protein-based risk model, including the novel identified prognosticators ApoA2 and SAA, outperformed the predictive accuracy of the currently used MSKCC risk model and classified about 1 in 3 mRCC patients into a different prognostic group. When confirmed by subsequent prospective validation studies, these findings could have major impact on and could contribute to more tailored therapeutic approaches in clinical practice for mRCC patients.

Chapter 6The aim of the next investigation was to validate our previous findings demonstrating that Apolipoprotein-A2 (ApoA2-) and Serum Amyloid Alpha (SAA)-levels significantly and independently predicted overall survival(OS) in mRCC patients (Chapter 5). In this study a training cohort consisting of 114 interferon-treated mRCC patients was included and a validation set comprising 151 mRCC patients treated with tyrosine kinase inhibitors (TKIs). Compared to the previous publication (Chapter 5) the overall and progression-free survival data of the patients in the training cohort were updated. At initiation of systemic therapy ApoA2- and SAA-levels significantly predicted PFS and OS in both the training- and validation cohort. Multivariate analysis identified SAA in both separate patient sets as a robust and independent prognosticator for PFS and OS. In contrast to our previous findings ApoA2 interacted with SAA in the validation cohort and therefore did not contribute to a better predictive accuracy than SAA alone and was therefore excluded from further analysis. Interestingly, applying tertiles as objective cut-off values of SAA-levels, patients were allocated in three risk groups (favourable, intermediate and poor), showing equal prognostic accuracy when compared to the multi-factorial MSKCC risk model in both independent patient sets. Using Receiver Operating Characteristics (ROC)-analysis SAA-levels >71ng/ml was designated as optimal cut-off value in the training cohort. This was subsequently confirmed for its significant sensitivity and specificity in the validation set. Importantly, the MSKCC predictive accuracy considerably improved when expanded with this novel poor risk factor (SAA >71ng/ml) in both patient cohorts, demonstrating enhanced discrimination of patient survival times for every MSKCC risk group. Furthermore, the importance of these results are illustrated by the significant differences in clinical outcome of anti-VEGF treated mRCC patients who were mainly categorized in the intermediate risk group (18;19). Therefore baseline SAA-levels might contribute to a better a-priori selection for risk-directed therapy, thereby avoiding long-term schemes of toxic treatment modalities for patients who will probably not profit (20). However, for both patient sets, differences in pretreatment and additional therapies were described. The heterogeneity regarding treatment might influence patients’ outcome. Supporting the described outcomes, SAA is not only a prognosticator for RCC, as the predictive accuracy for OS has also been shown for other cancer types such as lung, ovarian an breast cancer and melanoma (21-25). Therefore, it has been hypothesized that SAA reflects actual disease status and might be considered as a pan-biomarker for survival prediction irrespective of tumor type or treatment modality (26;27).

General discussion and future perspectives Chapter 9

160

Accordingly, in a subset of patients in the validation cohort, changes in SAA-levels after 6-8 weeks of TKI-treatment had no value in predicting treatment outcome. However, this analysis in based on a relatively small number of patients receiving different treatment modalities. Therefore, it remains questionable whether there will be a single biomarker which might sufficiently identify treatment responders and outcome. In conclusion, regardless of systemic treatment modality, SAA demonstrated to be a robust and independent predictor for PFS and OS in mRCC patients with equal prognostic accuracy when compared to the currently used MSKCC risk model. When incorporated in the MSKCC model SAA showed additional prognostic value for patient management. Consequently, SAA -either alone or as part of the MSKCC model- may be used as an objective prognostic marker in future clinical trials in mRCC patients and may improve the a-priori selection of appropriate risk-directed therapy for every individual.

PART – I I IMitochondrial DNA prognostic for several cancer types

Chapter 7Previous studies have demonstrated that circulating genomic cell-free nucleic acids are promising diagnostic and prognostic markers in cancer patients. The prognostic and predictive implication of mitochondrial- (mt)DNA as a pan-tumor marker in blood of different human cancers was investigated in this study. First a straightforward quantitative Polymerase-Chain-Reaction (PCR) assay of cancer patients was developed, which stringently determined mtDNA-levels isolated from citrate plasma. Next, mtDNA-levels were associated with Progression-Free Survival (PFS) and Overall Survival (OS) using univariate regression analysis. In a large training cohort comprising of 310 patients with diverse tumor types, mtDNA-levels significantly predicted OS for several tumor types, however not for all cancer types. Using tertiles as objective cut-off values, patients were classified into a favorable, intermediate and poor risk group, exhibiting considerable discrimination in OS times between risk groups using Kaplan-Meier analysis. These findings were validated in a prospective cohort of 124 patients with different cancer types. Applying the pre-defined mtDNA cut-off values to our validation cohort, Kaplan-Meier analysis critically distinguished the favorable and intermediate from the poor risk group and confirmed our results of the training set. However, in the validation cohort mtDNA-levels did not predict treatment response. The amount of contradictory literature on circulating nucleic acid levels as tumor marker (28;29) and a variety of methodologies (30-34) clearly indicate that protocols on blood handling, plasma isolation and storage all may greatly influence the validity of circulating nucleic acids as prognostic or diagnostic marker. In this study solid quality assessments with small intra-variability between triplicate samples and small inter-variability between standard curves for this assay were observed. In general, not all cancer types were predictive for OS, possibly due to the fact that the numbers of patients for these groups were small and there were fewer patients with metastasized disease and corresponding improved survival. Additionally, breast-, ovarian- and cervix cancer did not demonstrate an association with survival, suggesting that gender might be of influence. Compared to the training cohort, the validation cohort mtDNA was less powerfully correlated

161

with OS, probably caused by: a) smaller patient numbers with a different distribution and fewer patients with metastasized disease b) correspondingly prolonged overall survival, c) dissimilar treatment modalities and d) a shorter clinical follow up. In conclusion, this is the first study that demonstrates that circulating mtDNA in plasma of cancer patients quantified by a newly developed and straightforward qPCR, can be considered as a pan-tumor prognostic biomarker for OS regardless of tumor type, however not for all cancer types. Additional prospective studies with large homogenous patient cohorts per tumor type are essential to address the clinical consequence of circulating mtDNA as prognostic pan-tumor biomarker for survival in a broader extent.

PART – IVPrimary Colorectal Cancers and Their Subsequent Hepatic metastases are Genetically different

Chapter 8The first strategies in individualized cancer treatment have been adopted in colorectal cancer (CRC). For example, treatment with the antibodies cetuximab or panitumumab directed to the Epidermal Growth Factor Receptor (EGFR) is only considered in patients with wildtype KRAS established in the primary CRC tumor. Obviously, the authority of single-gene analysis is confined as the mutational status throughout the entire EGFR pathway might be responsible for therapy outcome. Another important limitation is that the genetic constitution of the metastasis is unknown and therefore response to therapy is unpredictable. This is the first study which comprehensively compared the genetic constitution of a composed ‘Cancer mini-Genome’ in 21 patients with primary colorectal adenocarcinomas (CRC) and their subsequent hepatic metastases by employing targeted deep-sequencing on formalin fixed, paraffin embedded tissue to demonstrate which tissue best reflects the presence of a target for therapy. First a ‘Cancer Mini-Genome’ was invented consisting of the exons of 1,264 genes, including all relevant kinases, oncogenes, tumor suppressors and pathways important for carcinogenesis (angiogenesis and apoptosis) and targeted treatment (EGFR-, PIK3CA-, TGFb-, mTOR-, and VEGF-pathways). In total, 21 patients with metastatic colorectal cancer (mCRC) who underwent resection of the primary tumor and at a later stage resection of its hepatic metastasis were included. In total, 6,696 known and 1,305 novel variations were identified in 1,174 and 667 genes respectively, including 817 variants that potentially altered protein function. On average 83 (SD 69) potentially function-impairing variations were gained in the metastasis and 70 (SD 48) variations were lost, showing that the primary tumor and hepatic metastasis are genetically significantly dissimilar. Besides novel and known variations in genes such as KRAS, BRAF, KDR, FLT1, PTEN, and PI3KCA, many aberrations in the up-/downstream genes of EGFR/PI3K/VEGF-pathways and other pathways (mTOR, TGFb, etc.) were elucidated. This may clarify why the clinical benefit of anti-EGFR antibodies remains limited, even in preselected patients based on only KRAS and BRAF analysis. Differences in genetic constitution between the primary CRC tumors and hepatic metastases are highly clinically relevant for therapeutic strategy of metastatic disease. This study implicates that for genetic analysis and the subsequent adjustment of targeted therapy a biopsy of the metastasis is preferred over archived primary tumor tissue. This requires a different mindset for oncologists, because biopsies from metastases are not commonly taken at


162

the start of treatment. In conclusion, this investigation indicates that the genetic constitutions of the hepatic metastases are significantly different from those of the primary CRC tumor and are of such magnitude that an impact on treatment outcome is realistic. Therefore, genetic pathway-analysis of the metastasis might further refine the selection of patients for targeted therapies even more than the determination of genetic characteristics of the primary tumor and therefore justifying additional modifications and optimization of currently used treatment algorithms. ‘ngs’ technology will direct future targeted therapiesThe paradigm that cancer evolves over time, suggests that the selective evolutionary process in every cancer cell population in a given patient can develop its own characteristic profile of specific mutated genes. This complex heterogeneity within a specific tumor type is reflected by differences between the primary tumors and associated metastases, as well as by the large disparities observed in mutational profiles between patients. Past experience demonstrates the huge heterogeneity between cancer patients, as well as within one person who might have multiple different clones originating from a single primary tumor. The important question, currently being tackled by NGS researchers, is to determine the extent of heterogeneity between cancer patients with identical tumor types as well as within individual patients. Unfortunately, every metastasized cancer patient will eventually progress during targeted treatment. How a tumor becomes unresponsive to therapy over time is a complex biological mechanism that is different for every patient. Indeed, one of the major reasons for disease progression in the cancer setting is attributable to genetic tumor evolution; mutations may enable the tumor to bypass pathways being targeted. One of the conclusions that can be drawn after this first era of multiple targeted therapeutics is that medical oncologists are increasingly being confronted with progressed cancer patients requiring serially adapted treatments (35). In practice, treating cancer patients with these novel targeted therapies is shifting towards chronic disease management, where tumors consecutively break through targeted therapies. The recurring progression of cancer patients can be seen as a repetitive cycle instigating novel mutational profiles for every new metastatic lesion time by time (figure 1). In addition, NGS can be used to follow tumor progression and evolution over time and to adapt treatment to the evolving cancer entity and to associate mutational profiles to survival data. It is anticipated that every newly diagnosed cancer patient and every recently-progressed patient will enter the cycle of advanced NGS tumor mini-genome analyses and subsequent therapy adjustment according to the current unique genomic alterations of their specific tumor or metastasis. Computational pathway-driven analyses have shown that specific signaling cascades are diversely altered per tumor type indicating subtype preference for carcinogenesis (36). However, crosstalk between pathways might complicate black-and-white conclusions from pathway-driven analyses (37). Overall, it is necessary to integrate genomic data together with proteomic knowledge and basic functional studies to obtain a global overview and comprehensive understanding of tumor evolution and therapy using a systems biology approach (36-38). If not adequately dealt with, this increasing complexity of information could result in an escalating divergence of medical science away from the patient. Because of the complex and multi-dimensional character of these data, computational systems biology methods might be the logical step to do this (39-41).

163

It is essential to realize that the implementation of NGS technology into clinical practice is not an uncomplicated venture. Besides sufficient starting capital, it needs a well-organized infrastructure with enough equipment, well-conducted investigation methodology, a properly trained team (composed of an operational head, a research technician, a physician, a statistician and a bioinformatician) and streamlined validated protocols throughout the entire sequencing pipeline. In conclusion, NGS technology is an exceptionally promising technique, which will unquestionably be implemented in clinical cancer management in the immediate future (42). NGS may reveal novel drugable targets and facilitate diagnosis, prediction of therapeutic responsiveness and thus personalized cancer medicine. However, important technical challenges and clinical questions must be addressed first (43), in order to straddle the canyon between the current screening phase status of NGS technology and proper use in clinical management, ultimately resulting in the reduction of overtreatment and therapeutic cost.

oVe RAll C onClus Ion s

The routine use of ‘personalized cancer medicine’ in clinical settings has become justifiable for three reasons: (a) rapid development of targeted therapies directed to various tumor types, (b) therapeutic efficacy, benefiting from molecular and biological understanding of specific tumor types and (c) improved proteomic and genomic biomarkers stratifying patients to particular strategies. This thesis demonstrates progress on ‘personalized cancer medicine’ approach based on four conclusions: (i) Novel tissue biomarkers FIH and E2F1 in renal cell cancer patients potentially influence tumor staging and gradation.(ii) Newly identified biomarker Serum Amyloid Alpha improves the current risk-directed therapy in mRCC patients. (iii) A simple and elegant qPCR assay determines prognostic mitochondrial DNA in blood of patients with several cancer types. (iv) Differences in genetic constitution between the primary Colorectal (CRC) tumors and their subsequent hepatic metastases are of such magnitude that an impact on targeted therapies is realistic, based on the genetic profile of the metastasis.


164

figure 1 Based on acquired mutational profiles of a tumor mini-genome by Next Generation Sequencing, therapeutic approaches will serially repress acute tumor growth, adapting cancer therapeutic approaches towards chronic disease management. (In cooperation with F.L. Gerritse)

figure 1 - Scheme representing the circle of adaptive personalized cancer medicine

analysis of cancer mini-genome

collect patient material (biopsy/blood/stool etc.)

adapt appropriateindividualized therapy

diseaseprogression

diagnosedcancer patient

165


(1) Stolze IP, Tian YM, Appelhoff RJ, Turley H, Wykoff CC, Gleadle JM et al. Genetic analysis of

the role of the asparaginyl hydroxylase factor inhibiting hypoxia-inducible factor (FIH) in

regulating hypoxia-inducible factor (HIF) transcriptional target genes [corrected]. J Biol Chem

2004 October 8;279(41):42719-25.

(2) Morris MR, Maina E, Morgan NV, Gentle D, Astuti D, Moch H et al. Molecular genetic analysis

of FIH-1, FH, and SDHB candidate tumour suppressor genes in renal cell carcinoma. J Clin

Pathol 2004 July;57(7):706-11.

(3) Tan EY, Campo L, Han C, Turley H, Pezzella F, Gatter KC et al. Cytoplasmic location of factor-

inhibiting hypoxia-inducible factor is associated with an enhanced hypoxic response and a

shorter survival in invasive breast cancer. Breast Cancer Res 2007;9(6):R89.

(4) Aprelikova O, Chandramouli GV, Wood M, Vasselli JR, Riss J, Maranchie JK et al.

Regulation of HIF prolyl hydroxylases by hypoxia-inducible factors. J Cell Biochem 2004 June

1;92(3):491-501.

(5) Henze AT, Riedel J, Diem T, Wenner J, Flamme I, Pouyseggur J et al. Prolyl hydroxylases 2 and

3 act in gliomas as protective negative feedback regulators of hypoxia-inducible factors.

Cancer Res 2010 January 1;70(1):357-66.

(6) Kim Y, Lin Q, Glazer PM, Yun Z. Hypoxic tumor microenvironment and cancer cell differentiation.

Curr Mol Med 2009 May;9(4):425-34.

(7) Evangelou K, Kotsinas A, Mariolis-Sapsakos T, Giannopoulos A, Tsantoulis PK, Constantinides

C et al. E2F-1 overexpression correlates with decreased proliferation and better prognosis in

adenocarcinomas of Barrett oesophagus. J Clin Pathol 2008 May;61(5):601-5.

(8) Rabbani F, Richon VM, Orlow I, Lu ML, Drobnjak M, Dudas M et al. Prognostic significance

of transcription factor E2F-1 in bladder cancer: genotypic and phenotypic characterization. J

Natl Cancer Inst 1999 May 19;91(10):874-81.

(9) Lee J, Park CK, Park JO, Lim T, Park YS, Lim HY et al. Impact of E2F-1 expression on clinical

outcome of gastric adenocarcinoma patients with adjuvant chemoradiation therapy. Clin

Cancer Res 2008 January 1;14(1):82-8.

(10) Alla V, Engelmann D, Niemetz A, Pahnke J, Schmidt A, Kunz M et al. E2F1 in melanoma

progression and metastasis. J Natl Cancer Inst 2010 January 20;102(2):127-33.

(11) Alonso MM, Fueyo J, Shay JW, Aldape KD, Jiang H, Lee OH et al. Expression of transcription

factor E2F1 and telomerase in glioblastomas: mechanistic linkage and prognostic significance.

J Natl Cancer Inst 2005 November 2;97(21):1589-600.

(12) Haller F, Gunawan B, von HA, Schwager S, Schulten HJ, Wolf-Salgo J et al. Prognostic role

of E2F1 and members of the CDKN2A network in gastrointestinal stromal tumors. Clin

Cancer Res 2005 September 15;11(18):6589-97.

(13) Patard JJ, Fergelot P, Karakiewicz PI, Klatte T, Trinh QD, Rioux-Leclercq N et al. Low CAIX

expression and absence of VHL gene mutation are associated with tumor aggressiveness

and poor survival of clear cell renal cell carcinoma. Int J Cancer 2008 July 15

;123(2):395-400.

(14) Ohtani K. Implication of transcription factor E2F in regulation of DNA replication. Front Biosci

1999 December 1;4:D793-D804.

(15) Wang C, Rauscher FJ, III, Cress WD, Chen J. Regulation of E2F1 function by the nuclear

corepressor KAP1. J Biol Chem 2007 October 12;282(41):29902-9.


166

(16) Hallstrom TC, Mori S, Nevins JR. An E2F1-dependent gene expression program that

determines the balance between proliferation and cell death. Cancer Cell 2008

January;13(1):11-22.

(17) Malle E, Sodin-Semrl S, Kovacevic A. Serum amyloid A: an acute-phase protein involved in

tumour pathogenesis. Cell Mol Life Sci 2009 January;66(1):9-26.

(18) Heng DY, Xie W, Regan MM, Warren MA, Golshayan AR, Sahi C et al. Prognostic factors for

overall survival in patients with metastatic renal cell carcinoma treated with vascular

endothelial growth factor-targeted agents: results from a large, multicenter study. J Clin Oncol

2009 December 1;27(34):5794-9.

(19) Heng DY, Mackenzie MJ, Vaishampayan UN, Bjarnason GA, Knox JJ, Tan MH et al. Primary

anti-vascular endothelial growth factor (VEGF)-refractory metastatic renal cell carcinoma:

clinical characteristics, risk factors, and subsequent therapy. Ann Oncol 2011 November 5.

(20) Rini BI. Metastatic renal cell carcinoma: many treatment options, one patient. J Clin Oncol

2009 July 1;27(19):3225-34.

(21) Ramankulov A, Lein M, Johannsen M, Schrader M, Miller K, Loening SA et al. Serum amyloid

A as indicator of distant metastases but not as early tumor marker in patients with renal cell

carcinoma. Cancer Lett 2008 September 28;269(1):85-92.

(22) Kimura M, Tomita Y, Imai T, Saito T, Katagiri A, Ohara-Mikami Y et al. Significance of

serum amyloid A on the prognosis in patients with renal cell carcinoma. Cancer 2001 October

15;92(8):2072-5.

(23) Pierce BL, Ballard-Barbash R, Bernstein L, Baumgartner RN, Neuhouser ML, Wener MH

et al. Elevated biomarkers of inflammation are associated with reduced survival among breast

cancer patients. J Clin Oncol 2009 July 20;27(21):3437-44.

(24) Wood SL, Rogers M, Cairns DA, Paul A, Thompson D, Vasudev NS et al. Association of

serum amyloid A protein and peptide fragments with prognosis in renal cancer. Br J Cancer

2010 June 29;103(1):101-11.

(25) Findeisen P, Zapatka M, Peccerella T, Matzk H, Neumaier M, Schadendorf D et al. Serum

amyloid A as a prognostic marker in melanoma identified by proteomic profiling. J Clin Oncol

2009 May 1;27(13):2199-208.

(26) Vermaat JS, van dT, I, Mehra N, Sleijfer S, Haanen JB, Roodhart JM et al. Two-protein signature

of novel serological markers apolipoprotein-A2 and serum amyloid alpha predicts prognosis

in patients with metastatic renal cell cancer and improves the currently used prognostic

survival models. Ann Oncol 2010 July;21(7):1472-81.

(27) Malle E, Sodin-Semrl S, Kovacevic A. Serum amyloid A: an acute-phase protein involved in

tumour pathogenesis. Cell Mol Life Sci 2009 January; 66(1):9-26.

(28) Boddy JL, Gal S, Malone PR, Harris AL, Wainscoat JS. Prospective Study of Quantitation of

Plasma DNA Levels in the Diagnosis of Malignant versus Benign Prostate Disease. Clin

Cancer Res 2005 February 15;11(4):1394-9.

(29) Gal S, Fidler C, Lo YM, Taylor M, Han C, Moore J et al. Quantitation of circulating DNA in the

serum of breast cancer patients by real-time PCR. Br J Cancer 2004 March 22;90(6):1211-5.

(30) Chiu RW, Poon LL, Lau TK, Leung TN, Wong EM, Lo YM. Effects of blood-processing

protocols on fetal and total DNA quantification in maternal plasma. Clin Chem 2001

September;47(9):1607-13.

(31) Lam NY, Rainer TH, Chiu RW, Lo YM. EDTA is a better anticoagulant than heparin or citrate

for delayed blood processing for plasma DNA analysis. Clin Chem 2004 January;50(1):256-7.

167

(32) Chiu RW, Lui WB, El Sheikhah A, Chan AT, Lau TK, Nicolaides KH et al. Comparison of

protocols for extracting circulating DNA and RNA from maternal plasma. Clin Chem 2005

November; 51(11):2209-10.

(33) Jung M, Klotzek S, Lewandowski M, Fleischhacker M, Jung K. Changes in concentration

of DNA in serum and plasma during storage of blood samples. Clin Chem 2003 June;49 (6 Pt 1):1028-9.

(34) Sozzi G, Roz L, Conte D, Mariani L, Andriani F, Verderio P et al. Effects of prolonged storage

of whole plasma or isolated plasma DNA on the results of circulating DNA quantification

assays. J Natl Cancer Inst 2005 December 21;97(24):1848-50.

(35) Sawyers C. Targeted cancer therapy. Nature 2004 November 18;432(7015):294-7.

(36) Shah SP, Morin RD, Khattra J, Prentice L, Pugh T, Burleigh A et al. Mutational evolution in a lobular breast

tumour profiled at single nucleotide resolution. Nature 2009 October 8;461(7265):809-13.

(37) Ocana A, Pandiella A. Personalized therapies in the cancer “omics” era. Mol Cancer 2010;9:202.

(38) Ding L, Ellis MJ, Li S, Larson DE, Chen K, Wallis JW et al. Genome remodelling in a basal-like breast cancer

metastasis and xenograft. Nature 2010 April 15;464(7291):999-1005.

(39) Metzker ML. Emerging technologies in DNA sequencing. Genome Res 2005 December;15(12):1767-76.

(40) Metzker ML. Sequencing technologies - the next generation. Nat Rev Genet 2010 January;11(1):31-46.

(41) Yu H, Tardivo L, Tam S, Weiner E, Gebreab F, Fan C et al. Next-generation sequencing to generate interactome

datasets. Nat Methods 2011 June;8(6):478-80.

(42) ten B, Jr., Grody WW. Keeping up with the next generation: massively parallel sequencing in clinical diagnostics.

J Mol Diagn 2008 November;10(6):484-92.

(43) Chin L, Gray JW. Translating insights from the cancer genome into clinical practice.

Nature 2008 April 3;452(7187):553-63.



Dutch summary nederlandse samenvatting

169Dutch summary / Nederlandse samenvatting

Doelgerichte therapieën stimulans voor gepersonaliseerde behandeling van kankerKanker is op dit moment de belangrijkste doodsoorzaak in economisch ontwikkelde landen. Toonaangevende Wereldgezondheidsorganisaties verwachten dat de incidentie- en mortaliteitscijfers van kanker zullen stijgen in de komende decennia. Om deze stijgende kankerstatistieken te bedwingen, zijn verbeteringen in de preventie, diagnostiek, voorspelbaarheid van ziektebeloop en nieuwe, doelgerichte therapeutische benaderingen onbetwistbaar van groot belang. Het streven in het huidige onderzoek naar kanker is het uitbreiden en optimaliseren van gepersonaliseerde kankertherapie. De grootste uitdaging voor de behandeling van kanker in de toekomst zal zijn om elke individuele patiënt de meest effectieve medicatie voor zijn unieke kankersoort te verstrekken. Dit proefschrift draagt bij aan dit streven door onderzoek op vier gebieden:Deel-I: beschrijving van twee nieuwe weefsel biomarkers in nierkanker, die niet alleen leiden tot een verbeterde benadering van tumorstadiering, maar ook tot een verbetering van de klinische behandeling.Deel-II: ontdekking van een nieuwe biomarker in bloed, die bijdraagt aan het nauwkeuriger bepalen van de prognose en daardoor aan een verbeterde risico-gemedieerde therapie bij patiënten met gemetastaseerde (uitgezaaide) nierkanker.Deel-III: ontwikkeling van een nieuwe techniek om aan de hand van mitochondriaal DNA in bloed een goede prognose vast te kunnen stellen voor verschillende soorten kanker.Deel-IV: vaststelling van genetisch verschil tussen primaire darmtumor en bijbehorende levermetastasen wat een aanknopingspunt is voor gepersonaliseerde, doelgerichte therapie. Biomarkers verbeteren risico-gemedieerde therapieënOndanks verbetering in behandelstrategieën, blijft het ziekteverloop voor patiënten met kanker onvoorspelbaar en de prognose voor gemetastaseerde ziekte slecht. Daarnaast zijn de afgelopen decennia veel kankerpatiënten blootgesteld aan toxische behandelingen zonder betrouwbare klinische responsevaluatie. Omdat deze zware therapieën veel bijwerkingen kunnen geven, is het belangrijk om alleen patiënten te selecteren die voordeel hebben bij de betreffende behandeling. Biomarkers zijn risicofactoren die de prognose en/of het therapeutische effect van een individuele kankerpatiënt kunnen voorspellen. Deze biomarkers dragen bij aan de selectie van individuele patiënten voor passende therapie op maat. Biomarkers kunnen een belangrijke rol spelen bij de indeling van patiënten in risicogroepen, leidend tot een adequate risico-gemedieerde behandeling, betere klinische uitkomsten en het voorkomen van toxische behandelingen bij patiënten met een slechte prognose, die geen voordeel hebben van de therapie. Het blijft een uitdaging om nieuwe biomarkers te ontdekken die bijdragen aan een verbeterde selectie van patiënten voor risico-gemedieerde therapie. next generation sequencing: een nieuwe en geavanceerde toepassing van genomicsMet de komst van Next Generation Sequencing (NGS) is de ontwikkeling van het genetisch onderzoek van kanker verschoven van de analyse van een op zichzelf staand gen naar onderzoek van honderden genen en complete downstream signaaltransducties tegelijk. NGS maakt de totstandkoming van grote mutatieprofielen mogelijk, wat zou kunnen leiden tot betere richtlijnen en gepersonaliseerde, minder toxische en meer toegepaste therapie. Het in kaart brengen van mutaties draagt tevens bij aan het begrijpen van de evolutie van tumoren. NGS is een uitermate geschikte methode om te kunnen anticiperen op de onmiskenbare behoefte aan analyse van volledige mutatiekaders en onderzoek naar nieuwe biomarkers die richting kunnen geven aan gepersonaliseerde therapie voor kanker.

170

De e l - INieuwe biomarkers in weefsel verbeteren stadiëring van ziekte bij patiënten met niercelcarcinoom

Hoofdstuk 3De onafhankelijke voorspellende betekenis van de upstream eiwitten Prolyl Hydroxylase Domein (PHD) 1, 2 en 3 en FIH (factor remmende Hypoxie Induced Factor (HIF)), die een directe invloed hebben op de HIF- pathway, werd in deze studie geanalyseerd in patiënten met heldercellig niercelcarcinoom. Om de immuunhistologische expressie van HIF, FIH en PHD 1, 2 en 3 te bepalen werd een tissue microarray geconstrueerd, waarbij tumorweefsel van 100 patiënten met heldercellig niercelcarcinoom werd geïncludeerd. Het expressieniveau van deze eiwitten was geassocieerd met totale overleving en andere klinische en pathologische factoren. Het expressieniveau van HIF-1 was significant verbonden met HIF-2a, PHD1, PHD2, PHD3, FIH en VHL (Von Hippel Lindau). Overeenkomstig was HIF-2a gecorreleerd met FIH en PHD2. Univariate Cox regressie analyse liet een significant verband zien tussen tumor stadium en gradatie, diameter, gemetastaseerde ziekte, FIH expressieniveau en totale overleving. Lage expressie van FIH werd gezien bij een slechte/korte totale overleving, terwijl hoge nucleaire intensiteit van FIH werd gezien bij patiënten met een langere totale overleving. Bovendien bleef een laag nucleair FIH niveau ook na multivariate analyse een sterke onafhankelijke prognostische factor. Concluderend laat dit onderzoek zien dat FIH kan worden gebruikt als nieuwe onafhankelijke weefselbiomarker om de prognose van patiënten met heldercellig niercelcarcinoom te kunnen bepalen. Deze resultaten benadrukken de noodzaak om naast de downstream targets ook de upstream regulerende eiwitten van de HIF pathway te bestuderen. In de toekomst is er meer onderzoek nodig om te kunnen bepalen of het FIH expressieniveau, als het wordt toegevoegd aan de huidige gebruikte prognostische modellen, van aanvullende klinische waarde kan zijn om het ziektestadium en de daarop aansluitende behandeling van niet gemetastaseerd niercelcarcinoom te kunnen bepalen. Hoofdstuk 4Dit onderzoek heeft zich gericht om regelmechanisme tussen de transcriptiefactor E2F1 en Von Hippel Lindau (VHL) te verhelderen, met behulp van een bestaand zebravis model en nierkanker cellijnen met een tekort aan VHL. Aansluitend is de klinische waarde van E2F1 in het primaire tumorweefsel van patiënten met niercelcarcinoom bepaald. Herintroductie van wildtype en gemuteerd-pVHL in cellijnen met een tekort aan VHL resulteerde in een dosisafhankelijke downregulatie van E2F1-expressie en activiteit, waarop HIF-α geen invloed had. Defecten in de celcyclus die bijdragen aan pVHL-dysfunctie worden gered door ectopische E2F1 overexpressie. In een zebravis model en nierkankercellijnen werd MCM3, een doeleiwit van E2F1, gereguleerd door pVHL. Op moleculair niveau kan de nieuwe regulatie van E2F1 door pVHL een resultaat zijn van de gedemonstreerde correlatie van pVHL met het TRIM28-SETDB1-RBBP4/7 complex. Deze studie laat door middel van een immunohistologisch retrospectief cohortonderzoek van 138 patiënten met niercelcarcinoom zien dat, in vergelijking met normaal nierweefsel van dezelfde patiënten, verhoogde nucleaire expressie van E2F1 wordt gezien bij primaire niercelkankertumoren. Patiënten met niercelkanker en VHL mutaties in de kiemcellijnen lieten een significant verhoogde expressie van nucleaire E2F1-niveaus zien in vergelijking met niercelkanker van patiënten met een heldercellig of niet-heldercellige histologie. Aansluitend was

171

een verhoogde expressie van E2F1 geassocieerd met een kleinere tumordiameter (< 7 cm) en een gunstig stadium volgens de American Joint Committee on Cancer (AJCC). Univariate Cox regressie analyse liet enerzijds een correlatie zien tussen verlaagde nucleaire E2F1-expressie en een slechte ziektevrije en verminderde totale overleving en anderzijds tussen verhoogde E2F1-expressie en een gunstige ziektevrije en totale overleving. Hiermee werd de nieuwe functie van E2F1 in het voorspellen van prognostische waarde bij primair niercelcarcinoom bevestigd. Dit onderzoek laat zien dat E2F1 regulatie in vitro en in vivo een nieuwe, HIF-α onafhankelijke, functie is voor VHL op transcriptieniveau. Vervolgens werd aangetoond dat E2F1 een significante voorspeller is van ziektevrije en totale overleving bij patiënten die een curatieve nefrectomie ondergingen in verband met niercelcarcinoom. Er zijn prospectieve studies nodig die door middel van multivariate analyse van bekende risicofactoren moeten evalueren of de nieuwe risicofactor E2F1 een onafhankelijke voorspeller is van prognose en of E2F1 daarmee van aanvullende waarde kan zijn in de klinische behandeling van nierkankerpatiënten.

De e l - I I Nieuwe biomarkers in bloed verbeteren de huidige risicomodellen voor gemetastaseerd niercelcarcinoom

Hoofdstuk 5Het Memorial Sloan Kettering Cancer Center (MSKCC) risicomodel wordt wereldwijd gebruikt om patiënten met gemetastaseerde niercelkanker in te delen in drie prognostische risicogroepen met gunstige, gemiddelde en slechte totale overleving. Deze risicogroepen worden momenteel gebruikt om nierkankerpatiënten te behandelen met verschillende doelgerichte therapieën. Om de voorspellende nauwkeurigheid van het MSKCC model te verbeteren, werd proteomics gebruikt om nieuwe serum eiwitten te identificeren die zijn geassocieerd met totale overleving. In het onderzoek werden van 114 nierkankerpatiënten die zijn behandeld met interferon, bloed monsters gescreend waarbij gebruik werd gemaakt van Surface-Enhanced-Laser-Desorption-Time of Flight Massa Spectrometrie (SELDI-TOF MS). Hierbij werden verschillende eiwitten geïdentificeerd die significant gecorreleerd waren met totale overleving.Van deze eiwitten werden Apolipoproteine A-II (Apo-A2), Serum Amyloid Αlpha (SAA) en transthyretine gevalideerd voor hun associatie met totale overleving, waarbij gebruik werd gemaakt van conventionele en commercieel beschikbare gekwantificeerde technieken. Met de combinatie van ApoA2 en SAA werd een nieuwe prognostische vingerafdruk, bestaande uit twee eiwitten, gegenereerd. Vervolgens werden eerder geïdentificeerde prognostische factoren geïncludeerd en kwamen na multivariate Cox regressie analyse ApoA2, SAA, Lactate Dehydrogenase (LDH), algehele lichamelijke toestand (ECOG) en het aantal metastasen als onafhankelijke factoren naar voren in het voorspellen van totale overleving. Met deze vijf onafhankelijke variabelen werd een nieuw op eiwitten gebaseerd risicomodel ontwikkeld, waarbij patiënten in drie risicogroepen werden ingedeeld. Dit op eiwitten gebaseerde risicomodel leverde een nauwkeuriger prognose van patiënten op dan het huidig gebruikte MSKCC model. Door toepassing van dit risicomodel in plaats van het originele MSKCC model werd 38% van de patiënten in een andere risicogroep geplaatst, met als gevolg een andere keuze in therapeutische behandeling.Concluderend laat deze studie zien dat de combinatie van screening door middel van proteomics, gevolgd door eiwit kwantificatiemodellen nieuwe biomarkers kan voortbrengen. Een nieuw,

Dutch summary / Nederlandse samenvatting

172

op eiwitten gebaseerd risicomodel, waarbij de eiwitten ApoA2 en SAA werden geïncludeerd, overtreft de predictieve nauwkeurigheid van het huidig gebruikte MSKCC model en classificeert 1 op de 3 patiënten in een andere risicogroep. Als deze bevindingen door toekomstige prospectieve validatieonderzoeken worden bevestigd, zullen zij mogelijk een grote bijdrage leveren aan meer toegepaste therapeutische benaderingen in de klinische praktijk van patiënten met nierkanker. Hoofdstuk 6Het doel van het beschreven onderzoek was om de eerdere bevindingen (Hoofdstuk 5) te valideren en daarmee de significante onafhankelijke voorspellende waarde van Apolipoproteine-A2 (ApoA2) en Serum Amyloid Α (SAA) voor totale overleving van patiënten met gemetastaseerde nierkanker te bevestigen. In dit onderzoek werd een trainingscohort van 114 patiënten met gemetastaseerde nierkanker hoofdzakelijk behandeld met interferon geïncludeerd. Daarnaast werd een validatiecohort van 151 patiënten met gemetastaseerde nierkanker behandeld met tyrosine kinase remmers (TKIs) geïncludeerd. De totale en progressie-vrije overleving van het trainingscohort die werden gebruikt voor het onderzoek beschreven in hoofdstuk 5, werden geüpdate. Op het moment van aanvang van systemische therapie voorspellen ApoA2 en SAA significant progressie-vrije en totale overleving in zowel de trainings- als het validatiecohort. Multivariate analyse identificeerde SAA in beide afzonderlijke patiëntenverzamelingen als een krachtige en onafhankelijke prognostische factor voor progressievrije en totale overleving. In tegenstelling tot onze eerdere bevindingen, leidde een wisselwerking tussen ApoA2 en SAA in het validatiecohort niet tot een verbeterde predictieve nauwkeurigheid van de combinatie ApoA2 en SAA in vergelijking tot SAA alleen. Hierdoor werd ApoA2 geëxcludeerd uit de verdere analyse. Door het gebruik van tertielen als objectieve afkapwaarde van SAA concentraties, werden patiënten in drie risicogroepen ingedeeld (gunstig, gemiddeld en slecht). Hiermee was de prognostische waarde, interessant genoeg, gelijk aan de prognostische waarde van het multifactoriële MSKCC risicomodel in beide onafhankelijke patiëntengroepen. Met gebruik van de Receiver Operating Characteristics (ROC)-analyse werd een SAA niveau > 71ng/ml aangewezen als optimale afkapwaarde voor het trainingscohort. Voor deze afkapwaarde werden de significante sensitiviteit en specificiteit bevestigd in het validatiecohort. De predictieve nauwkeurigheid van het MSKCC model werd aantoonbaar verbeterd wanneer deze werd uitgebreid met deze nieuwe slecht-risico factor (SAA>71 ng/ml) in zowel het trainings- als het validatiecohort, waarbij een verbeterd onderscheidingsvermogen voor totale overlevingstijd van patiënten voor elke MSKCC risicogroep werd aangetoond. Daarnaast werd in een deelverzameling van het validatiecohort aangetoond dat een verandering in SAA niveau 6-8 weken na behandeling met TKIs geen waarde had in het voorspellen van de uitkomst van behandeling. De conclusie van het onderzoek is dat, ongeacht de systemische therapeutische modaliteit, SAA laat zien dat het een krachtige en onafhankelijke voorspeller is van progressievrije en totale overleving van patiënten met nierkanker, met een aan het huidig gebruikte MSKCC risicomodel gelijkwaardige prognostische nauwkeurigheid. Als SAA wordt opgenomen in het MSKCC model, laat het een toegevoegde prognostische waarde zien voor de behandeling van patiënten. SAA, alleen of in combinatie met het MSKCC model, kan worden gebruikt als een objectieve prognostische marker in toekomstige klinische studies van patiënten met gemetastaseerde nierkanker en zou de a-priori selectie van aansluitende door risico gestuurde therapie voor elk individu kunnen verbeteren.

173

De e l - I I I Mitochondriaal DNA van prognostische waarde voor verschillende vormen van kanker

Hoofdstuk 7Eerdere studies hebben aangetoond dat circulerende genomische celvrije nucleotiden veelbelovende diagnostische en prognostische markers zijn in patiënten met kanker. In deze studie werd de prognostische en predictieve betrokkenheid van mitochondriaal(mt)-DNA als een algemene tumor marker in het bloed van patiënten met verschillende vormen van kanker onderzocht. Allereerst werd er een eenvoudige kwantitatieve Polymerase-Chain-Reaction (qPCR) assay ontwikkeld, die nauwkeurig de hoeveelheid mt-DNA, geïsoleerd uit citraatplasma, bepaalt. Vervolgens werden mt-DNA niveaus geassocieerd met progressievrije en totale overleving, waarbij gebruik werd gemaakt van univariate regressie analyse. In een groot trainingscohort bestaande uit 310 patiënten met diverse typen tumoren, voorspelde het mtDNA-niveau op significante wijze de totale overleving bij verschillende tumortypen, echter niet voor alle kankersoorten. Door middel van het gebruik van tertielen als objectieve afkapwaarde werden patiënten ingedeeld in een gunstige, gemiddeld of slechte risicogroep, waarbij met behulp van Kaplan-Meier analyse, onderscheid werd aangetoond voor totale overleving tussen de verschillende risicogroepen. Deze bevindingen werden gevalideerd in een prospectief cohort bestaande uit 124 patiënten met verschillende typen kanker. De eerder gedefinieerde afkapwaarden werden tevens gebruikt voor het validatiecohort, waarbij Kaplan-Meier analyse doorslaggevend onderscheid liet zien tussen enerzijds de gunstige en gemiddelde risicogroep en anderzijds de slechte risicogroep, waarmee de resultaten van het trainingscohort werden bevestigd. In het validatiecohort voorspelde mt-DNA echter niet het effect op de behandeling. Geconcludeerd mag worden dat dit onderzoek de eerste studie is, die laat zien dat circulerend mt-DNA in plasma van patiënten met kanker gekwantificeerd door een eenvoudige qPCR, kan worden gezien als een algemene prognostische tumormarker voor totale overleving ongeacht het tumor type. Dit geldt echter niet voor alle soorten van kanker. Aanvullende prospectieve studies met grote homogene patiëntencohorten per kankersoort zijn noodzakelijk om de klinische consequentie van circulerend mt-DNA als prognostische algemene tumorbiomarker voor overleving in een bredere context te kunnen bepalen.

Deel - IV Primaire colorectaal kankers en hun corresponderende levermetastasen hebben een verschillende genetische aard

Hoofdstuk 8Een van de eerste strategieën in gepersonaliseerde behandeling van kanker zijn toegepast bij patiënten met colorectaal cancer (CRC). Een voorbeeld hiervan is de behandeling met de antilichamen cetuximab of panitumumab gericht op de Epidermal Growth Factor Receptor (EGFR), die alleen wordt overwogen bij patiënten waarbij wildtype KRAS wordt aangetroffen in de primaire colorectale tumor. Uiteraard is de autoriteit van op zichzelf staande genen beperkt, aangezien de gemuteerde status van de gehele EGFR pathway verantwoordelijk is voor de uitkomst van

Dutch summary / Nederlandse samenvatting

174

behandeling. Een andere belangrijke beperking is dat de genetische constitutie van de metastase onbekend is en de reactie op behandeling om die reden onvoorspelbaar. Dit is de eerste studie die op grootschalige wijze de genetische aard van een samengesteld ‘Kanker Mini-Genome’ bij 21 patiënten met een primair colorectaal adenocarcinoom vergelijkt met de corresponderende levermetastase. Door het gebruik van doelgerichte diepe sequencing wordt aangetoond welk weefsel het beste de aanwezigheid van mutaties in genen weerspiegelt, die een mogelijkheid kunnen zijn voor doelgerichte therapie. Allereerst werd een ‘Kanker Mini-Genoom’ uitgedacht bestaande uit 1264 genen, met inbegrip van alle relevante kinasen, oncogenen, tumor suppressoren en pathways belangrijk voor carcinogenese (angiogenese en apoptose) en doelgerichte behandeling (EGFR-, PIK3CA-, TGFb-, mTOR- and VEGF-) pathways. Eenentwintig patiënten met gemetastaseerd colorectaal kanker (mCRC) die resectie van de primaire tumor en in een later stadium resectie van een levermetastase ondergingen, werden geïncludeerd. In totaal werden 6696 bekende en 1305 nieuwe variaties geïdentificeerd in respectievelijk 114 en 667 genen, waaronder 817 varianten die potentieel de functie van een eiwit kunnen veranderen. Er werden gemiddeld 83 (SD 69) potentieel eiwitfunctie veranderende variaties verkregen in de metastase en 70 (SD 48) variaties gingen verloren. Hiermee werd aangetoond dat de primaire tumor en de levermetastase genetisch significant verschillen. Naast nieuwe en bekende variaties in genen zoals KRAS, BRAF, KDR, FLT1, PTEN and PI3KCA, werden veel genetische veranderingen in de up en downstream genen van EGFR-, PI3K- en VEGF-pathways en andere pathways (mTOR, TGFb etc) duidelijk, die potentieel de therapeutische toegankelijkheid beïnvloeden. Concluderend laat dit onderzoek zien dat de genetische constituties van levermetastasen significant verschillen van de genetische karakteristieken van primaire colorectaal tumoren. Deze verschillen zijn van zo grote omvang dat een invloed op de uitkomst van therapie denkbaar is. Om die reden zou genetische pathyway-analyse van metastasen, meer dan de bepaling van genetische karakteristieken van de primaire tumor, verdere verfijning van selectie van patiënten voor doelgerichte therapie kunnen bewerkstelligen. Dit zou kunnen leiden tot aanvulling cq herziening en optimalisatie van de huidige gebruikte gepersonaliseerde therapeutische strategieën.

175Dutch summary / Nederlandse samenvatting


Contributing authors

177Contributing authors

Jos H. Beijnen

Department of Medical Oncology

The Netherlands Cancer Institute -

Antoni van Leeuwenhoek Hospital

Amsterdam, the Netherlands

Department of Biomedical Analysis

Faculty of Pharmaceutical Sciences, Utrecht University

Utrecht, the Netherlands

Karsten Boldt

Center for Ophthalmic Research

Medical Proteome Center,

Eberhard-Karls University Tuebingen

Tuebingen, Germany

J.l.H. Ruud Bosch

Department of Urology

University Medical Center Utrecht


epie Boven


VU University Medical Center Amsterdam


ewart de Bruijn

Hubrecht Institute

Genome Biology and Bioinformatics Group Utrecht


Alain de Bruin

Department of Pathobiology

Faculty of Veterinary Medicine Utrecht University


Aram van Brussel

Laboratory of Experimental Oncology



edwin e. Cuppen

Hubrecht Institute


Department of Human Genetics



laura g.M. Daenen




Paul J. van Diest

Department of Pathology



Judith Y. engwegen

Department of Pharmacy & Pharmacology

The Netherlands Cancer Institute - Slotervaart Hospital


frank l. gerritse




Marije g. gerritsen




Jourik A. gietema


University Medical Center Groningen,

University of Groningen

Groningen, the Netherlands

Rachel H. giles




gerard groenewegen




Petra van der groep




John B.A.g. Haanen





178

Richard van Hillegersberg

Department of Surgical Oncology



John Hinrichs




Judith J.M. Jans


Laboratory of Experimental Oncology



Trudy g.n. Jonges




Catharina M. Korse

Department of Clinical Chemistry




Harm H.e. van Melick

Department of Urology St. Antonius Hospital

Nieuwegein, the Netherlands

Marco J. Koudijs




stephanie g.C. Kroeze




Wim Kruit


Erasmus Medical Center

Rotterdam, the Netherlands

Marlies H. langenberg




nico lansu

Hubrecht Institute



Dorus A. Mans


Nijmegen Center for Molecular Life Sciences

Radboud University Nijmegen Medical Center

Nijmegen, the Netherlands

Rene H. Medema

Division of Cell Biology

The Netherlands Cancer Institute, Amsterdam


niven Mehra




Michal Mokry

Hubrecht Institute



Tatjana M. niers


Academic Medical Center, University of Amsterdam


Isaac J. nijman

Hubrecht Institute



sjoukje f. oosting


University Medical Center Groningen,

University of Groningen

Groningen, the Netherlands

179

Hans Kristian Ploos van Amstel

Department of Medical Genetics

Wilhelmina Children Hospital



Jeroen van Reeuwijk




Nijmegen, the Netherlands

Dick J. Richel


Academic Medical Center, University of Amsterdam


Ronald Roepman



Institute for Genetic and Metabolic Disease


Nijmegen, the Netherlands.

Wijnand M. Roessingh




Jeanine M. Roodhart




ellen van Rooijen



Hubrecht Institute for Developmental and

Stem Cell Biology


Benjamin Rowland




Jan H. schellens


The Netherlands Cancer Institute - Antoni van

Leeuwenhoek Hospital

Amsterdam, The Netherlands

Department of Biomedical Analysis

Faculty of Pharmaceutical Sciences, Utrecht University

Utrecht, The Netherlands

stefan J. scherer

Hoffman-La Roche Inc. / Genetech

Biomarker Angiogenesis at Roche Pharmaceuticals

Nutley, United States of America

stefan sleijfer


Erasmus Medical Center,

Daniel den Hoed Cancer Center

Rotterdam, the Netherlands

Ingeborg van der Tweel

Julius Center for Health Sciences and Primary Care



Astrid A. van der Veldt


VU University Medical Center Amsterdam


emile e. Voest




Bart g. Weijts

Department of Pathobiology

Faculty of Veterinary Medicine Utrecht University


Patrick H. van Zon

Department of Medical Genetics

Wilhelmina Children Hospital



Contributing authors


Acknowledgements

Dankwoord

181 Acknowledgements / Dankwoord

Allereerst wil ik de patiënten bedanken die aan mijn onderzoek hebben

deelgenomen, ondanks hun ziekte en de daarmee samenhangende

onzekerheid in hun levenssituatie. Zonder hun deelname was dit

onderzoek niet tot stand gekomen.

Dankzij een goede samenwerking met diverse onderzoeksgroepen en personen heeft dit onderzoek kunnen plaatsvinden. Ik bedank hen allen hartelijk voor hun medewerking. Een aantal van hen wil ik nadrukkelijk noemen:

Gewaardeerde promotor professor Voest, beste emile, dank voor ruim een decennium van prettige en vruchtbare samenwerking en wederzijds vertrouwen. Jij hebt mij vanaf mijn tijd als student Farmacie de onderzoeksmogelijkheden en uitdagingen geboden waardoor ik mij verder heb kunnen ontwikkelen als arts en wetenschapper. Het is inspirerend om van dichtbij mee te maken wat jij de afgelopen tien jaar hebt bereikt!

Professor van Diest, beste Paul, dank voor het beoordelen van de vele patiënten samples en het demarceren van tumoren normaal weefsel. Andere medewerkers van de afdeling Pathologie, met name natalie ter Hoeve, ik heb de accuratesse en de snelheid van het uitzoeken van de patiënten samples enorm gewaardeerd.

Professor Cuppen, beste edwin, meerdere malen heb jij gezegd dat ‘technologisch’ alles mogelijk is. Met deze insteek en met dank aan je onderzoeksgroep, onder andere Michal Mokry, Wigard Kloosterman, ewart de Bruijn, nico lansu en Ivo Renkens, is er een robuuste methode ontwikkeld voor verder ‘high-tech’ onderzoek van en voor kankerpatiënten. Ies nijman, dank voor het gedegen ontwerpen en optimaliseren van de analyse scripten van het sequence project, wat heeft geleid tot betrouwbare genomics data.

Professor Beijnen, beste Jos, het voelde voor mij als farmaceut vertrouwd om tussen apothekers mijn onderzoeksperiode te beginnen en dat heeft een mooie basis gelegd voor de verdere onderzoeksjaren.

Professor Haanen, beste John, dank voor de fijne en stimulerende coöperatie in het onderzoek van gemetastaseerde nierkankerpatiënten.

Professor Kops, beste geert, dank voor het plaatsnemen in de beoordelingscommissie.

Professor Medema, beste René, je scherpte en kritische analyses hebben een constructieve bijdrage geleverd aan mijn onderzoek, waarvoor dank. livio Kleij en Rob Klompmaker bedankt voor het faciliteren van alle laboratorium gerelateerde zaken.

Alice Tondeur, dank voor de herinneringen aan deadlines en het regelwerk dat je mij uit handen hebt genomen.

182

Jeanine Roodhart, Marlies langenberg en laura Daenen, jullie toewijding aan zowel wetenschappelijk onderzoek als aan kankerpatiënten is groot. Dank voor de plezierige samenwerking, AIO-etentjes, koffiebreaks, feedback en het stoom afblazen en meedenken van de afgelopen jaren.

sander Basten, we hebben mooie koffiemomenten beleefd in onze kamer. De humor doorspekt met woordgrappen gaven veel ontspanning. Je muziekkennis is enorm en ik leef in de veronderstelling dat je bijna ieder 3FM Popduel zou hebben gewonnen. Succes met de laatste loodjes van jouw proefschrift.

frank gerritse, als master-student Biomedische Wetenschappen kwam jij je onderzoeksstage doen. Mede door jouw pionierswerk op meerdere projecten ben ik zover gekomen. Inmiddels heb jij je master behaald en ben je hard op weg SUMMA af te ronden. Succes! Mooi om straks collega’s te zijn.

Wijnand Roessingh, dank voor de vele experimenten die jij voor mij hebt gedaan. Jij brengt gezelligheid en mooie levensverhalen met je mee en dat veraangenaamde het soms wat geestdodende laboratoriumwerk. Nu je het onderzoek (tijdelijk?) hebt verlaten, ben ik benieuwd waar jij straks terecht komt.

Marco Koudijs, dank voor je positiviteit en stimulerende medewerking in de afronding van het genome sequencing onderzoek. Succes met het verdere vormgeven en voortstuwen van het grote project ‘Center for Personalized Cancer Treatment’.

Rachel giles, naast je wetenschappelijk input heb ik je ‘mentorschap’ met goede adviezen erg gewaardeerd. Dorus Mans, dank voor onze plezierige samenwerking. Ik ben benieuwd waar ons E2F1 project terecht gaat komen.

Team Urologie, stephanie Kroeze en Judith Jans, dank voor de samenwerking. Stephanie, gezellig dat je in het Antonius bent komen werken.

Medewerkers van het Laboratorium Chirurgie: professor Inne Borel Rinkes, onno Kranenburg, Benjamin emmink, Menno de Bruijn, ernst streller, frederik Hoogwater, Winan van Houdt, nikol snoeren, Danielle Raats, Klaas govaert en Maarten nijkamp, dank voor de plezierige samenwerking en ook voor de ontspanning in de vorm van een potje tennis.

Medewerkers van de onderzoeksgroepen van susanne lens en Patrick Derksen, dank voor het prettige contact.

Patrick van der Zon, dank voor het bijbrengen van de beginselen van de wereld van de complexe genetica en je hulp bij het opzetten van het sequencing project. Je Brabantse levenshouding bracht mij veel plezier.

183

lucette Teurlings en collega promovendi van Alexandre suerman / MD-PhD programma, dank voor de inspirerende bijeenkomsten! Het was opbouwend om als intervisiegroep (Annemarie schmitz, Alma van de Pol, Marie-elise nijdam en Menno de Bruijn) met elkaar op te trekken en je eigen rol binnen het promotietraject te reflecteren.

Maatschapsleden Interne Geneeskunde van het st Antonius Ziekenhuis nieuwegein, fijn om in een stimulerende omgeving en prettige sfeer te worden opgeleid. Ik kijk ernaar uit om de komende jaren met volle aandacht mij verder te ontwikkelen tot specialist. Collegae arts-assistenten, jullie betrokkenheid, (kliniek) verhalen, uitjes en warme contacten doen mij goed, met of zonder koffie van Vermaat!

Wouke de Jong, dit boekje heb jij strak en stijlvol vorm gegeven, bedankt.

lieve (schoon)ouders, familie en vrienden, jullie liefde, interesse, steun en levensverhalen blijven voor mij een stimulans en inspiratiebron. Een goed gesprek, een goed glas wijn, skiën, voetballen, zeilen, tennis, en samen optrekken in het leven zijn voor mij een verrijking en ik blijf ervan genieten! Zeilgroep, waarom denken jullie altijd dat ik de weerwolf ben?

Paranimfen, Bert en Daan, onmisbaar om vrienden zoals jullie te hebben die je door en door kennen en begrijpen! Het geeft vertrouwen om jullie ook op de promotiedag aan mijn zijde te hebben staan!

lieve Judith, allerliefste zus en beste vriendin, wat zou er van mij terecht zijn gekomen zonder jou? Jouw meeleven en onvoorwaardelijke liefde en steun, jouw reflecteren en meedenken zijn in mijn leven van grote waarde! Ik hou van je!

lieve Hanna, jij betekent zoveel voor mij! Je liefde, je altijd luisterende oor, je geduld, humor, relativeringsvermogen en wijze raad zijn voor mij onmisbaar. Jij maakt mijn leven mooier en geeft het glans en verdieping. Samen hebben wij een prachtig leven en daar hoop ik een levenlang van te genieten! Ik hou ontzettend veel van je!

Acknowledgements / Dankwoord


list of publications

185List of publications

Vermaat Js, Gerritse FL, van der Veldt AA, Roessingh WM, Niers TM, Oosting SF, Sleijfer S, Roodhart JM, Beijnen JH, Schellens JH, Gietema JA, Boven E, Richel DJ, Haanen JB, Voest EE. Validation of Serum Amyloid Alpha as an Independent Biomarker for Progression Free- and Overall Survival in Metastatic Renal Cell Cancer Patients Eur Urol. 2012 Jan 23. [Epub ahead of print]

Groenewegen G, Walraven M, Vermaat Js, de Gast B, Witteveen E, Giles R, Haanen J, Voest E. Targeting the Endothelin Axis with Atrasentan, in combination with IFN-alpha, in metastatic renal cell carcinoma.Br J Cancer 2012 Jan 17;106(2):284-9

Vermaat Js*, Nijman IJ*, Koudijs MJ, Gerritse FL, Scherer S, Mokry M, Roessingh W, Lansu N, de Bruijn E, van Hillegersberg R, Van Diest PJ, Cuppen E, Voest EE. Primary colorectal cancers and their subsequent hepatic metastases are genetically different: implications for selection of patients for targeted treatment. Clin Cancer Res. 2012 Feb 1;18(3):688-99.

Kloosterman WP, Hoogstraat M, Paling O, Tavakoli-Yaraki M, Renkens I, Vermaat Js, van Roosmalen MJ, van Lieshout S, Nijman IJ, Roessingh W, van ‘t Slot R, van de Belt J, Guryev V, Koudijs M, Voest E, Cuppen E. Chromothripsis is a common mechanism driving genomic rearrangements in primary and metastatic colorectal cancer. Genome Biology 2011 Oct 19;12(10):R103.

Roodhart JM, Daenen LG, Stigter EC, Prins HJ, Gerrits J, Houthuijzen JM, Gerritsen MG, Schipper HS, Backer MJ, van Amersfoort M, Vermaat Js, Moerer P, Ishihara K, Kalkhoven E, Beijnen JH, Derksen PW, Medema RH, Martens AC, Brenkman AB, Voest EE. Mesenchymal stem cells induce resistance to chemotherapy through the release of platinum-induced fatty acids. Cancer Cell 2011 Sep 13; 20(3):370-83.

Schackmann RC, van Amersfoort M, Haarhuis JH, Vlug EJ, Halim VA, Roodhart JM, Vermaat Js, Voest EE, van der Groep P, van Diest PJ, Jonkers J, Derksen PW. Cytosolic p120-catenin regulates growth of metastatic lobular carcinoma through Rock1-mediated anoikis resistance.J Clin Invest. 2011 Aug 1; 121(8):3176-88.

186

Kroeze SG, Vermaat Js, van Brussel A, van Melick HH, Voest EE, Jonges TG, van Diest PJ, Hinrichs J, Bosch JL, Jans JJ. Expression of nuclear FIH independently predicts overall survival of clear cell renal cell carcinoma patients.Eur J Cancer 2010 Dec;46(18):3375-82.

Langenberg MH, Nijkamp MW, Roodhart JM, Snoeren N, Tang T, Shaked Y, van Hillegersberg R, Witteveen PO, Vermaat Js, Kranenburg O, Kerbel RS, Medema RH, Borel Rinkes IH, Voest EE. Liver surgery induces an immediate mobilization of progenitor cells in liver cancer patients: A potential role for G-CSF.Cancer Biol Ther. 2010 May;9(9):743-8.

Roodhart JM, Langenberg MH, Vermaat Js, Lolkema MP, Baars A, Giles RH, Witteveen EO, Voest EE. Late release of circulating endothelial cells and endothelial progenitor cells after chemotherapy predicts response and survival in cancer patients.Neoplasia 2010 Jan;12(1):87-94.

Vermaat Js, van der Tweel I, Mehra N, Sleijfer S, Haanen JB, Roodhart JM, Engwegen JY, Korse CM, Langenberg MH, Kruit W, Groenewegen G, Giles RH, Schellens JH, Beijnen JH, Voest EE. Two-protein signature of novel serological markers apolipoprotein-A2 and serum amyloid alpha predicts prognosis in patients with metastatic renal cell cancer and improves the currently used prognostic survival models.Ann Oncol. 2010 Jul;21(7):1472-81.

Rademaker-Lakhai JM, Beerepoot LV, Mehra N, Radema SA, van Maanen R, Vermaat Js, Witteveen EO, Visseren-Grul CM, Musib L, Enas N, van Hal G, Beijnen JH, Schellens JH, Voest EE. Phase I pharmacokinetic and pharmacodynamic study of the oral protein kinase C beta-inhibitor enzastaurin in combination with gemcitabine and cisplatin in patients with advanced cancer.Clin Cancer Res. 2007 Aug 1; 13(15 Pt 1):4474-81.

Beerepoot LV, Mehra N, Vermaat Js, Zonnenberg BA, Gebbink MF, Voest EE. Increased levels of viable circulating endothelial cells are an indicator of progressive disease in cancer patients.Ann Oncol. 2004 Jan;15(1):139-45.

Crul M, Beerepoot LV, Stokvis E, Vermaat Js, Rosing H, Beijnen JH, Voest EE, Schellens JH. Clinical pharmacokinetics, pharmacodynamics and metabolism of the novel matrix metalloproteinase inhibitor ABT-518.Cancer Chemother Pharmacol. 2002 Dec; 50(6):473-8.

187

Vermaat Js*, Mans DA*, Weijts BG, van Rooijen E, van Reeuwijk J, Boldt K, Daenen LGM, van der Groep P, Rowland B, Roepman R, Voest EE, van Diest PJ, de Bruin, Giles RH Regulation of E2F1 by the von Hippel-Lindau tumor suppressor protein predicts survival in renal cell cancer patients Submitted

Vermaat Js, Roodhart JM, Gerritse FL, Gerritsen MG, Roessingh WM, van Zon PH, Daenen LGM, Mehra N, Ploos van Amstel HK, Medema RH, Voest EE Circulating cell-free mitochondrial DNA as prognostic but not predictive marker for various types of cancer Submitted

List of publications



Curriculum Vitae

189Curriculum Vitae

Joost Vermaat was born on the 11th of March 1980. He grew up in Veenendaal, Maassluis and The Hague. After graduating from secondary school (Driestar College Gouda) in 1998, he started to study Pharmaceutical Sciences at the University of Utrecht and received his Master of Science degree in Pharmaceutical Sciences in 2002. He continued his education by studying Medicine at the University of Utrecht. From November 2006 until May 2007 he did research on angiogenesis under supervision of prof. B. Anand-Apte at the Cole Eye Institute of the Cleveland Clinics in Cleveland, Ohio, USA. Shortly before graduating in 2007, he received the Alexandre Suermann / MD-PhD Stipendium, a personal scholarship for talented research students at the University Medical Center Utrecht (UMCU). With this financial support he started his PhD-studies at the laboratory of Medical Oncology at the UMCU under supervision of prof. dr. E.E. Voest. He received a Merit Award for his research on proteomics during the American Society for Clinical Oncology Annual Meeting in 2009 in Orlando, Florida, USA. In September 2010 he started his residency in Internal Medicine under supervision of dr. A.B.M. Geers and dr. W.J.M. Bos at the St Antonius Hospital in Nieuwegein and prof. dr. M.M.E. Schneider at the UMCU. Joost Vermaat is married to Hanna van Antwerpen.


Personalized Cancer Medicine

analysis of cancer mini-genome and/or proteomics

collect patient material (biopsy/blood/stool etc.)

adapt appropriateindividualized therapy

diseaseprogression

diagnosedcancer patient

Novel approaches in prognosis and personalized treatment of cancer

Documents

Transcript of Novel approaches in prognosis and personalized treatment of cancer