A sample article title

A systems oncology approach identifies NT5E as a key metabolic regulator in tumour cells and modulator of platinum sensitivity

Ekaterina Nevedomskaya1$*, Richard Perryman2,3, Shyam Solanki2, Nelofer Syed3, Oleg A. Mayboroda1, Hector C. Keun2*

1 Center for Proteomics and Metabolomics, Leiden University Medical Center (LUMC), L4-Q, PO Box 9600, 2300RC Leiden Leiden, The Netherlands

2 Division of Cancer, Department of Surgery and Cancer, Imperial College London, Hammersmith Hospital, London W12 0NN, UK

3 Division of Brain Sciences, Department of Medicine, Imperial College London, London, UK

* Corresponding authors

E-mail: e.nevedomskaya@nki.nl (EN), h.keun@imperial.ac.uk (HCK)

Tel: +31 (0)20 512 7920 (EN), +44-(0) 20-7594-3161 (HCK)

§ Current address: Department of Molecular Pathology, The Netherlands Cancer Institute, Amsterdam, The Netherlands

Abstract

Altered metabolism in tumor cells is required for rapid proliferation but also can influence other phenotypes which affect clinical outcomes such as metastasis and sensitivity to chemotherapy. GWAS-guided integration of NCI-60 transcriptome and metabolome data identified ecto-5’-nucleotidase (NT5E or CD73) as a major determinant of metabolic phenotypes in cancer cells. NT5E expression and associated metabolome variations were also correlated with sensitivity to several chemotherapeutics including platinum-based treatment. NT5E mRNA levels were observed to be elevated in cells upon in vitro and in vivo acquisition of platinum resistance in ovarian cancer cells and specific targeting of NT5E increased tumor cell sensitivity to platinum. We observed that tumor NT5E levels were prognostic for outcomes in ovarian cancer and were elevated after treatment with platinum, supporting the translational relevance of our findings. In this work we integrated and analyzed a plethora of public data, demonstating the merit of such ‘systems oncology’ approach for discovery of novel players in cancer biology and therapy. We experimentally validated the main findings of NT5E gene being involved in both intrinsic and acquired resistance to platinum-based drugs. We propose that the efficacy of conventional chemotherapy could be improved by NT5E inhibition and that NT5E expression may be a useful prognostic and predictive clinical biomarker.

Keywords: O2PLS; Chemometrics; Cancer; Metabonomics; Transcriptomics; Data Integration; chemotherapy; cisplatin resistance

Introduction

Cancer is characterized by a high degree of genetic heterogeneity not only between different tumor types but also even between different cell populations within one tumor1. There are multiple mutations that can lead to a malignant cell with multiple downstream pathways being affected2. However, one common characteristic of cancer, irrespective of the causal mutations, is altered metabolism that supports tumor growth and persistence3. Altered metabolism allows malignant cells to survive in the abnormal tumor microenvironment and cope with hypoxia, low pH and low-nutrient content. Such adaptations are absolutely necessary for the survival of malignant cells and have been linked to clinical outcomes, such as metastasis and sensitivity to therapy4,5. Expanding the knowledge of metabolic regulation would not only help understanding the origins of malignant transformations, but also unravel possibilities for specifically disrupting cancerous metabolism and thus new therapeutic options6.

Metabolic reprogramming of a cancer cell is a complex process, the origin and regulation of which is poorly understood. The precise metabolic alterations in a given tumor depend on a variety of factors, including driver mutations, overall genomic landscape and tumor microenvironment. Integrating system-level omics data can shed light on the inter-connectivity of those components and regulation of metabolism in cancer7. We approached such integration using systematic joint analysis of metabolic and gene expression data. For this we used prior knowledge of genetically influenced metabotypes 8 from genome-wide association studies (GWASs) and a statistical integration method, O2PLS, for combining metabolic and gene expression data from the well-characterized NCI-60 cell line panel. O2PLS is a method of bidirectional statistical modeling between sets of multidimensional data that allows focusing on the correlations of interest and improved interpretability of those correlations9. O2PLS has been successfully used for combining transcriptomic and metabolomic10, proteomic and metabolic profiling data11 and data on different classes of lipid metabolites12. Here our initial hypothesis was that using O2PLS to conduct a joint analysis of gene expression and metabolic profiles in a panel of diverse cancer cell lines could identify a gene or geneset with significant influence over cancer cell metabolism. Since tumour metabolism has been previously described to be associated with therapy resistance, tumour subtypes and outcome, we also hypothesised that the gene(s) thus defined through O2PLS analysis were related to such traits.

Using this methodology we identified the extracellular nucleotidase, NT5E – which encodes for an ecto-5’-nucleotidase as a major determinant of cancer cellular metabolism. NT5E (CD73) has been previously associated with impaired anti-tumor immune response via production of extracellular adenosine that suppresses T-cells 13 14. Aberrant expression of NT5E has been shown in a number of cancers: melanoma15, leukemia16, glioma17, colorectal18 and breast, notably triple negative,19 cancers. Subsequent bioinformatics investigation confirmed that NT5E was differentially expressed in cancer cell lines depending on oncogene, epithelial-mesenchymal transition (EMT) or hormonal receptor status and that both expression and NT5E activity was correlated with sensitivity to a number of cytotoxic drugs.

Importantly, we observed in several independent datasets that NT5E expression was elevated upon in vitro and in vivo acquisition of resistance to cisplatin in ovarian cancer cell lines. We confirmed experimentally that cisplatin resistance could be reversed in vitro by targeting NT5E. The translational significance of our findings was supported by the association of NT5E expression to survival in ovarian cancer patients and its increased expression upon cisplatin therapy. Collectively our data indicate a role for NT5E in mediating sensitivity to platinum chemotherapy in tumors via a direct, cell autonomous alteration in metabolism, and suggest that NT5E inhibition may be a potential therapeutic strategy in platinum resistant disease or for potentiating response in combination with platinum agents.

Experimental Procedures Data

List and details of all the publicly available datasets used for the analysis is available in Supplementary Table S1.

Data integration

To limit the number of genes used in the integrative analysis, we focused on the genes that have been previously described to have an influence on metabolic profiles in humans. For this we selected 7 published GWAS studies (Supplementary Table S2) that explored the influence of genetic loci on metabolites levels in blood and urine8. Collectively, 75 genes mentioned in these studies as associated to metabolism were selected for further analysis. The list of the selected genes can be found in Supplementary Table S3.

Orthogonal Projections to Latent Structures, OPLS20, is an extension of a supervised multivariate analysis method - Partial Least Squares (PLS)21 – with an integrated Orthogonal Signal Correction filter (OSC)22. The special feature of OPLS method is that it separates the variation of the data matrix X into the following parts: correlated to the response Y, systemically non-related (orthogonal) to Y and the residual variance. Such a separation allows examining the sources of variation separately and improves the interpretability of the model and makes the method widely applicable in the field of chemometrics and ‘omics’ data analysis23. O2PLS9,24 is a generalization of OPLS that allows prediction in both directions between multivariate matrices X and Y. The model is separated into the joint variation between the matrices and the orthogonal variation present in each of the matrices. O2PLS has been proposed as a method for integration of different ‘omics’ data and successfully used in a number of applications 10-12.

Gene expression and metabolite data were integrated using O2PLS modeling. X and Y data matrices were represented by mRNA and metabolite levels respectively. 7-fold cross-validation was performed to assess the predictive ability of the model, Q2 values were calculated for the model as a whole and for each of the variables in both matrices. The number of predictive O2PLS components was determined based on the scree plot of the PCA analysis of the matrix of covariance between X and Y. The number of orthogonal components was determined by maximizing joint Q2 values for prediction of both X and Y. Sample permutation was also performed in order to estimate the significance of the obtained Q2 levels. Predictive X and Y loadings represent co-variance between gene expression and metabolomics data.

Survival analysis

Cox proportional hazard model was used to obtain hazard ratios in each of the ovarian cancer datasets separately. Meta-analysis was done with R package survcomp25, the set of Cox regression coefficients was combined using a random effect model.

All the analyses were performed using in-house scripts and publicly available packages in R statistical software (R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/). Scripts are available upon request.

Gene Set Enrichment Analysis was performed using GSEA software (http://www.broad.mit.edu/gsea/)26 using the collection of gene ontologies as gene sets.

Reagents

Unless otherwise stated, reagents were obtained from Sigma (Sigma-Aldrich Ltd, Dorset, UK).

Cell Culture

PEO1 and PEO4 ovarian cancer cells were obtained from Euan Stronach, Imperial College London. BT549 and MCF7 breast cancer cells were obtained from the Developmental Therapeutics Program, National Cancer Institute. Cells were maintained in RPMI 1640 (phenol red; no glutamine; 10% FBS), supplemented with penicillin/streptomycin, L-glutamine and NEAA (hereafter referred to as growth media), incubated at 37°C and 5% CO2.

Transient NT5E knockdown and cisplatin sensitivity

NT5E expression was knocked down using the siPORT protocol. The siRNA’s used were either Silencer Select validated siRNA targeting NT5E ((Life Technologies Ltd, Paisley, UK; sense: 5’- GUAUCCAUGUGCAUUUUAAtt-3’, antisense: 5’-UUAAAAUGCACAUGGAUACgt-3’) or AllStars Negative Control siRNA (QIAGEN Ltd, Manchester, UK). Briefly, siPORT was diluted 1:20 in Opti-MEM and incubated at room temperature for 10 minutes. AllStars Negative Control siRNA and NT5E siRNA were diluted in Opti-MEM to the final concentration of 10nM. The siPORT (Life Technologies Ltd, Paisley, UK) was then added to each siRNA and incubated at room temperature for 10 minutes. 20μL of either AllStars Negative Control siRNA or siRNA targeting NT5E transfection solution was then added in to each well of a 96 well plate. 6 x 103 cells were added in to each well of the 96 well plates, and media was added to a total volume of 100μL. Cells were incubated at 37°C and 5% CO2 for 24 hours, at which point the media was replaced with growth media supplemented with cisplatin (0µM, 2.5µM). The cells were incubated at 37°C and 5% CO2 for 72 hours, at which point 7.5μL of CCK8 solution was added and they were placed back in to the incubator. After 4 hours the colour change was analysed on a Synergy H1 hybrid plate reader (BioTek UK, Bedfordshire, UK) at 450nm wavelength.

Western blot for NT5E protein

2.5 x 105 cells were added in to each well of a 6 well plate and transfected using the siPORT protocol mentioned above, using 50nM AllStars Negative Control siRNA or siRNA targeting NT5E in a total of 2.5mL of media, and left for 24 hours. The media was then removed and replaced with 2.5mL of normal growth media. After 72 hours, the media was removed and the cells were lysed using RIPA solution (RIPA buffer with 1X protease cocktail inhibitor), centrifuged at 15,000rpm for 15 minutes and the protein supernatant removed. The amount of protein was determined by BCA assay (Thermo Fisher Scientific UK, Hemel Hempstead, UK) with standard protein controls, and analysed on a Synergy H1 hybrid plate reader (BioTek UK, Bedfordshire, UK) at 562nm wavelength.

Western blot was carried out using a standard SDS polyacrylamide gel electrophoresis (PAGE) protocol. 15µg of protein was separated by gel electrophoresis using a 10% poly-acrylamide/bis-acrylamide gel (Tris 600mM, SDS 6mM, 0.75% weight/volume APS, 0.05% TEMED) in SDS PAGE running buffer (Tris 25mM, SDS 17mM, glycine 200mM), and then transferred on to a nitrocellulose membrane in transfer buffer (Tris 25mM, glycine 187mM, 20% methanol). Membranes were stained with ponceau to ensure protein was present, and then washed using wash buffer (1X PBS in dH2O with 0.5% TWEEN). The blots were exposed to blocking solution (wash buffer with 5% powdered milk), and then probed with the following antibodies in blocking solution: rabbit anti-CD73 mAb (1:800, Abcam PLC, Cambridge, UK), donkey anti-rabbit mAb linked to a horseradish peroxidase (HRP) reporter (1:5000, Abcam PLC, Cambridge, UK), mouse anti-β-actin mAb (1:10,000, Sigma Aldrich Ltd, Dorset, UK) and goat anti-mouse mAb linked to a HRP reporter (1:20,000, Abcam PLC, Cambridge, UK). The blots were exposed to enhanced chemiluminescent solution (Supersignal West Pico Chemiluminescent substrate by PIERCE, Thermo Fisher Scientific UK, Hemel Hempstead, UK) for 5 minutes, and immunoreactive bands were detected on a Kodak 4000MM Image Station (Kodak Ltd, Herts, UK).

Results NT5E is a major determinant of metabolism in cancer cells

Transcriptomics data for 60 of the NCI60 cell lines was obtained using the Agilent-012391 Whole Human Genome Oligo Microarray G4112A chip (41000 probes); metabolite levels were available for 154 uniquely identified metabolites (Supplementary Table S4) across 58 cell lines. To reduce dimensionality in the analysis and to maximize the likelihood of identifying biologically meaningful associations, when integrating mRNA and metabolic data we focused on genes previously shown to be related to metabolic traits on a systemic level (genetically influenced metabotypes). Genetic variation in a number of genes (Table S2) was previously associated to metabolites concentrations in biofluids in genome-wide association studies (GWAS). 8. We limited our analysis to these GWAS-based genes, which corresponded to 75 unique genes in NCI60 gene expression data, which in turn equated to 119 mRNA probes. The matched gene expression and metabolite levels data for 58 cell lines were used as X and Y inputs for O2PLS analysis, respectively. The optimal O2PLS model comprised 2 predictive, 3 Y- and 3 X-orthogonal components. The amount of variation in mRNA data that correlated to metabolites was found to be 12% and the amount of variation in metabolite data correlated to gene expression 27%. The overall proportion of variance predicted in cross-validation (Q2 values) was 4 and 12%, for mRNA probes and metabolites respectively. To confirm the significance of this level of predictivity the null distribution of the model Q2 was estimated using 1000 datasets with random sample permutations: none of the permuted models produced Q2 values exceeding those of the initial model. The cross-validation procedure in combination with the generation of random null models indicated that the statistical model was not over-fitted. Subsets of variables in each matrix with the highest Q2 values, which contributed the most to the to model predictivity, were identified and are displayed in Figure 1. As can be seen from the scores plot the three probes of NT5E gene have the highest predictive Q2 value. This suggested that of the panel selected the expression of this gene was the most strongly associated to variation in the levels of metabolites across cancer cell lines in the NCI60 panel.

Using OPLS regression to create a model linking intracellular metabolite variation specifically to NT5E expression we then identified metabolites that were correlated explicitly to NT5E, namely: cholesterol, glycerol, guanine, guanosine, inosine, inositol 1-phosphate, N-acetylneuraminic acid, phosphate, uracil (Supplementary Figure S1). Significantly, inosine concentration has been previously shown to be associated with NT5E nucleotide polymorphism in a GWAS study27, in support of our observations.

Validation of the relationship between NT5E expression and metabolite levels was performed using an independent dataset which characterized consumption and release (CORE) profiles of metabolites from media across the NCI-60 cancer cell lines28. As anticipated, NT5E expression showed significant positive correlation to inosine (among other metabolites) and negative correlation to a number of nucleoside-monophosphates (Supplementary Figure S2).

In conclusion, NT5E is not only associated with certain metabolic traits on a systemic level, but also determines metabolic phenotype among cancer cell lines.

NT5E expression and related cellular metabolic phenotypes are associated to chemosensitivity in cell lines

Using the data on the sensitivity of NCI-60 cell lines to a panel of 118 ‘mechanism of action’ drugs29, we found that NT5E expression was negatively correlated with sensitivity to a number of the drugs (namely, high expression of NT5E is associated with low potency of the drug; Table 1).

Sensitivity to two nucleoside analogues, azacytidine and inosine-glycodialdehyde, exhibited the most significant correlations to NT5E expression. As the sensitivity to these chemotherapeutic agents is related to the expression of nucleoside transporters30, it is possible that elevated levels of extracellular nucleosides associated with increased NT5E expression, leads to competition for cellular uptake, thus resulting in less of the chemotherapeutics entering the cell and decreased potency. Interestingly, we also observed that sensitivity to a number of platinum-based drugs was significantly (4/5 compounds significant in univariate test and tetraplatin significant after stringent Bonferroni correction) negatively correlated with the expression of NT5E gene (Table 1). Using the CORE dataset we were further able to confirm that the metabolites associated with NT5E expression were also associated with the sensitivity to platinum agents (Supplementary Figure S3). We have previously reported associations between a number of the metabolites (inosine, guanine, guanosine, uracil, cholesterol) found to be associated with NT5E expression in our current analysis and platinum sensitivity using a completely different bioinformatics approach of data integration based on over-representation31. These complimentary results obtained using different bioinformatics approaches suggest the existence of a specific metabolic phenotype that is associated with the sensitivity to cytotoxic agents, such as for instance platinum-based drugs.

We also investigated the degree to which the link between NT5E and sensitivity to cytotoxic agents could be explained by differences in growth rates of the cell lines. Although NT5E expression correlated with the doubling times of the cell lines, and negatively with platinum sensitivity (Supplementary Figure S4 a-b), this correlation could not be completely explained by growth rates. Within a homogeneous group of fast-growing cell lines there was a considerable range of NT5E expression, which did not correlate to cell growth, but correlated significantly to drug potency (Supplementary Figure S4 c-d).

NT5E is upregulated in the models of intrinsic and acquired platinum resistance in ovarian cancer

The observed correlation between the NT5E expression and sensitivity to platinum-based drugs is of particular interest in the light of ovarian cancer, for which treatment options are currently limited and typically confined to chemotherapy32.

To assess whether NT5E expression is associated with resistance to cisplatin treatment, we evaluated expression in the publicly available panel of ovarian cancer cell lines for which sensitivity to cisplatin was determined33. Consistent with our previous observations, we found NT5E expression elevated in the subgroup of ovarian cancer cell lines resistant to cisplatin (p = 0.006; Figure 2a).

However, the observed correlations are indicative of intrinsic sensitivity of the cell lines to the drugs, whereas the major challenge in ovarian cancer is the resistance to chemotherapy that develops after the initial response to the therapy in the majority of patients34. We analyzed several publicly available expression datasets of the ovarian cancer cell lines that are sensitive to platinum and resistant derivatives (in vitro acquired resistance), as well as sensitive and resistant cells derived from the same patient (in vivo acquired resistance). NT5E expression was elevated in 5/6 cell lines with acquired platinum-resistance compared to their sensitive counterparts across independent studies (Figure 2b).

NT5E silencing increases sensitivity to platinum

Given our findings, we sought experimental validation and functional confirmation of the association between NT5E and platinum sensitivity. In an initial experiment we characterized two breast cancer cell lines, BT-549 and MDA-MB-231, with relatively high expression of NT5E (as determined in NCI-60 cell line panel data). As we would predict from this, both cell lines showed relatively limited sensitivity to cisplatin (BT-549 being more sensitive then MDA-MB-231). Silencing of NT5E expression led to increased sensitivity to cisplatin in both these cell lines (Figure 3 a-b), supporting our hypothesis that high NT5E expression is causally involved in intrinsic cisplatin insensitivity.

Next, we examined two ovarian cancer cell lines that originated from the same patient tumor before and after the clinical acquisition of resistance to cisplatin: PEO1 (sensitive) and PEO4 (resistant)35. Western blot analysis confirmed that the resistant cell line PEO4 had a higher NT5E protein expression than the sensitive PEO1 cell line (Figure 3e), consistent with our observations on the transcriptional level (Figure 2a). Silencing of NT5E in PEO1 cell line slightly, though not significantly, increased sensitivity to cisplatin (Figure 3c). While in resistant PEO4 cell line it significantly increased sensitivity to the drug (Figure 3d). These results support our hypothesis that NT5E is involved not only in intrinsic, but also acquired resistance to cisplatin. Furthermore, they suggest that interfering with NT5E expression can re-sensitize resistant cancer cells to cisplatin.

NT5E expression predicts disease outcome in ovarian cancer patients and is elevated upon treatment with platinum-based agents

Sensitivity to chemotherapy is a crucial factor influencing patient survival in ovarian cancer particularly since debulking surgery is often suboptimal, especially in advanced cancers36. Association of NT5E expression and drug sensitivity is of potential clinical significance, since platinum- and taxane-based drugs are the major chemotherapeutic agents used for treatment of ovarian cancer 32. NT5E expression is strongly negatively associated to sensitivity of cancer cell lines to platinum-based drugs (Table 1). To evaluate the translational relevance of NT5E we conducted a meta analysis of survival outcomes in 6 cohorts of ovarian cancer patients. NT5E was significantly associated to overall survival of ovarian cancer patients (overall hazard ratio 1.21, p=0.005, Figure 4).

Using publicly available gene expression data, we also examined changes in NT5E expression in patient biopsies obtained before and after treatment with either carboplatin or paclitaxel37. We found that NT5E levels were significantly elevated in the post-treatment compared to the pre-treatment samples in patients that had been treated with carboplatin or a combination of paclitaxel/carboplatin (p = 0.0003 in paired t-test; Figure S5). NT5E levels remained unchanged upon treatment with paclitaxel alone (Figure S5). This suggests, in line with our hypotheses, that NT5E upregulation is part of the tumor cell response to stress induced by exposure to platinum-based chemotherapy.

NT5E and cancer subtypes

NT5E showed differential expression throughout the NCI60 panel (Figure S6), with one of the highest variation being in breast cancer cell lines (Figure 5a). This variation corresponded to key clinical subtypes of breast cancer – estrogen receptor positive (ER+) and negative (ER-) disease. NT5E was also found to exhibit higher expression in androgen-insensitive prostate cancer cell lines compared to androgen-sensitive38 (Figure 5b). Ovarian tumors can also be characterized by hormone receptor status39. Using the gene expression data from the cancer genome atlas (TCGA) we stratified ovarian cancer patients into estrogen-positive or -negative based on the median expression of estrogen-receptor alpha gene (ESR1). We found that the expression of NT5E was higher in tumors with ESR1 expression below the median, compared to the tumors with ESR1 expression above the median, i.e. an inverse association between NT5E and ESR1 expression (p < 10-5; Figure 5c). Examination of other datasets showed that silencing of estrogen-receptor (ER) in breast cancer cells induces upregulation of NT5E expression (Supplementary Figure S7) which is consistent with previous reports40.

NT5E expression has been previously shown to correlate positively to a known component of cell migration (PLAU) in breast cancer tumors41, and implicated in cell adhesion in glioma17. To investigate whether NT5E is involved in epithelial-mesenchymal transition (EMT), we explored correlation of expression to known markers of EMT42 in NCI60 panel. There was a strong positive correlation of NT5E expression and expression of genes associated with EMT (Supplementary Figure S8a), such as the well known mesenchymal markers, SLUG and Fibronectin (Supplementary Figure S8b-c). Similarly, there was a moderate negative correlation to the expression of the genes decreasing in EMT. This suggests that high NT5E expression might be associated with a more mesenchymal, and thus more aggressive, metastatic phenotype.

In order to get a global view on the potential pathways associated with NT5E expression we performed correlation and geneset enrichment analysis of gene expression across the cancer cell line encyclopedia (CCLE) panel. Gene Ontology terms related to extracellular matrix, regulation of cell migration and cell junctions were enriched among the genes positively correlated with NT5E expression (Supplementary Table S5). Nucleic acid-related metabolic processes, mRNA processing and splicing, as well as DNA repair and replication processes were enriched among the genes negatively correlated to NT5E expression (Supplementary Table S5).

There is a continuous effort to stratify different cancers into subtypes based on their genomic architecture which can be differentially managed in the clinic43. We investigated whether NT5E exhibits various levels of expression in cell lines depending on their mutational status. A panel of 474 cancer cell lines from CCLE44 was used, for which both mutational and gene expression data are available. We found that NT5E expression was significantly elevated in the cell lines that have an oncogenic mutation in either KRAS and/or BRAF, compared to wild type, while expression was lower in cells with mutations in the tumor suppressors PTEN and/or TP53 compared to wild type (Supplementary Figure S9). This result is in accordance to the results of in vivo transcriptomics analysis, in which NT5E expression was reported to be up-regulated in BRAF-mutant serous ovarian tumors compared to wild type tumors45. These findings suggest that NT5E elevation may contribute to any differential response to chemotherapy associated with genomic subtypes of disease.

Discussion

In the present study we identified NT5E (CD73) as a major determinant of metabolism in cancer cell lines and demonstrated its importance for survival of ovarian cancer patients and resistance to chemotherapy, in particular platinum-based treatment. While the role of NT5E in immunosuppression has been widely reported 13,14,46, our work suggests an additional, cell-autonomous role for NT5E in intrinsic sensitivity to chemotherapy and acquired resistance. We also show that for ovarian cancer cell lines the in vivo acquired resistance to cisplatin can be reversed by blocking the activity of NT5E. Ovarian cancer is generally characterized by very poor prognosis with the treatment options being limited and consisting of cytoreductive surgery followed by platinum- or taxane-based treatment. Despite the initial response to chemotherapy in the majority of cases the resistance to therapy develops and disease recurs. Identification of new targets for overcoming resistance could therefore have a major impact on patient survival.

NT5E expression has been demonstrated to be implicated in a number of cancers: melanoma15, colorectal cancer18, chronic lymphocytic leukemia16, glioma17 and breast cancer19,41,47,48. The mRNA expression and methylation status of NT5E have been correlated to the clinical survival outcome19,41 in breast cancer patients. Antibody therapy against NT5E was suggested as a possible treatment in breast cancer14 and has been shown to suppress metastasis49 and tumor angiogenesis50. Recently NT5E expression has been shown to be negatively associated with ovarian cancer outcome in serous high grade ovarian tumors 51. This confirms our observation of NT5E as a favorable prognostic marker in a broader selection of ovarian cancer patients. Importantly, we show that NT5E expression is also upregulated in ovarian tumors upon chemotherapy. We investigated if upregulation of NT5E expression was a characteristic feature of acquired resistance to cisplatin in ovarian cancer cell lines and if inhibiting its activity could reverse the resistance. We show, using publicly available datasets, that in isogenic pairs of cell lines originated from in vivo and in vitro acquired resistance, NT5E expression is increased upon the acquisition of resistance in the majority (with one cell line as an exception) of cell lines and is consistent through multiple datasets available. The reverse association of NT5E expression in cell lines and sensitivity to chemotherapy was confirmed experimentally. We also demonstrated in vitro that silencing NT5E can reverse the acquired resistance to cisplatin. To the best of our knowledge there are no previous reports regarding the role of the enzyme in sensitivity to chemotherapy in ovarian cancer.

Despite the suggested importance of metabolism in cancer cells, its relationship to drug sensitivity is rarely discussed52. Our results, as well as previous observations by Cavill et al.31, suggest that metabolism, and nucleotide salvage pathway in particular, can play a pivotal role.

Previously the role of NT5E in promoting survival of cancer cells and metastasis was attributed to its immunosuppressive role due to extracellular accumulation of adenosine in various cancers 13,53, including ovarian 54,55. Recently it has been shown that the inhibition of NT5E does not influence the sensitivity to a chemotherapeutic drug, namely docetaxel, in vitro in breast cancer cell lines, however, is efficient in vivo, indicating the importance of the immune response41. By contrast we show that sensitivity and, more importantly, acquired resistance to cisplatin can be modulated in vitro, which suggests that there are mechanisms other than immunosuppression by which NT5E can influence chemo-sensitivity and resistance. Analysis of the clinical data for ovarian cancer patients provides translational validity of the findings. Surprisingly, a recent paper Virtanen et al. has shown that extracellular adenosine, generated by NT5E, can be rapidly taken up by cells, inhibiting cell invasion and motility, as well as phospho-AMPK1α 56. The latter, if confirmed in our cell lines system, might explain the NT5E involvement in cisplatin resistance: phosphor-AMPK1 inhibits mTOR, inhibition of which has been shown to sensitize resistant ovarian cancer cells to cisplatin 57. It is also possible that NT5E could regulate the availability of nucleotide triphosphates required for DNA synthesis and repair processes58. Our gene set enrichment analysis also shows correlation between NT5E expression and processes related to mRNA and DNA metabolism, processing and replication. However, the exact mechanism that leads to the effect we observe is still to be elucidated.

Furthermore, NT5E is differentially expressed in cell lines and tumors that correspond to different subtypes of cancer and can be potentially used for identifying more invasive, metastasis-prone subtypes of various cancers. NT5E is highly expressed in cell lines corresponding to the more aggressive ER-negative breast tumors compared to ER-positive cells. In a large panel of cancer cell lines it is differentially expressed between cell lines baring KRAS/BRAF or TP53/PTEN mutations. It has been shown for a number of cancer subtypes that KRAS and TP53 mutations are mainly mutually exclusive and represent two distinct tumor origins59,60. Recently, ovarian cancers have been classified into two distinct groups: Type I with frequent KRAS/BRAF mutations and Type II characterized by TP53 aberrations61. Such classification has therapeutic implications with Type I being less responsive to platinum-based chemotherapy62,63.

Additionally, Ras oncogenes have been associated to clinical and experimental resistance to platinum-based drugs in ovarian cancer64,65. Our analysis suggests that high NT5E expression is associated with a more mesenchymal phenotype of cancer cells due to the correlation of expression to known markers of EMT, such as TWIST, SNAIL, SLUG and Vimentin. Association of NT5E expression and EMT has also recently been shown for ovarian cancer patients 51. EMT has been implicated in resistance to cisplatin 66,67. This is in contrast to the recent paper by Miow et al. 33, in which epithelial ovarian cancer cell lines are shown to be more resistant to cisplatin than the mesenchymal cell lines. While NT5E expression in these ovarian cell lines is consistent with our hypothesis of high NT5E expression in cisplatin-resistant cell lines, the association and sole definition of the epithelial/mesenchymal status might need further refinement.

Collectively, this suggests NT5E expression being an important marker relevant for subtyping and assessing difference in chemosensitivity in various types of cancer.

Limitations of the study. The variables entering the initial O2PLS model were limited by the number of genes described to be associated with metabolic profiles in GWAS studies as well as by the metabolomics data publicly available to date. O2PLS analysis could be redone with the larger panel of genes and metabolites, possibly bringing new knowledge on metabolic control in cancer. For selective interference experiments, work in other cell lines and ultimately in vivo experiments would add further confidence in our findings.

Conclusions

Using a systems oncology approach that integrates knowledge of metabolic alterations on a systemic level (through GWAS studies) with omics- data on a cellular level, we identified NT5E as a major determinant of metabolism in cancer cell lines. Although a role for NT5E in cancer is known, our data demonstrates its involvement in sensitivity and acquired resistance to chemotherapy, which is independent of immune system. The strong influence of NT5E on cellular metabolism suggests that metabolism and regulation of extra- and intra- cellular metabolic homeostasis is pivotal in NT5E-promoted metastasis and treatment-resistance. Taken all our data together, we propose that NT5E could represent a novel target for treating platinum-resistance in ovarian cancer.

Competing interests

The authors declare that they have no competing interests.

Acknowledgements

The authors thank the developers of NCBI GEO database and all the contributors of the public datasets used, as well as CCLE and TCGA investigators for making their data accessible. We also thank Dr Euan Stronach for providing the PEO1/PEO4 cell lines.

SS was supported by a Biotechnology and Biological Sciences Research Council(BBSRC)/AstraZeneca CASE award (BB/I532588/1). RP was supported by an Medical Research Council (MRC) Advanced Masters Studentship (MR/J015938/1).

Supporting Information

Figure S1 – OPLS loadings

Figure S2 – Correlation of NT5E and CORE metabolites

Figure S3 – Correlation of NT5E and sensitivity too tetraplatin

Figure S4 – Correlation of NT5E, platinum sensitivity and cell growth rates

Figure S5 – Change of NT5E expression upon platinum therapy

Figure S6 – NT5E expression across NCI-60 cell line panel

Figure S7 – NT5E expression upon ESR1 silencing

Figure S8 - NT5E expression and epithelial–mesenchymal transition

Figure S9 – NT5E expression and cancer subtypes

Table S1 - List of publicly available datasets used in the analysis

Table S2 - List of GWAS studies used for the selection of genes for the analysis

Table S3 - List of the genes used for data integration

Table S4 - List of metabolites used in the analysis

Table S5 – Results of Gene Set Enrichment Analysis

References

(1)Fisher, R.; Pusztai, L.; Swanton, C. Cancer heterogeneity: implications for targeted therapeutics. Br. J. Cancer 2013, 108, 479–485.

(2)Kandoth, C.; McLellan, M. D.; Vandin, F.; Ye, K.; Niu, B.; Lu, C.; Xie, M.; Zhang, Q.; McMichael, J. F.; Wyczalkowski, M. A.; et al. Mutational landscape and significance across 12 major cancer types. Nature 2013, 502, 333–339.

(3)Cairns, R. A.; Harris, I. S.; Mak, T. W. Regulation of cancer cell metabolism. Nat. Rev. Cancer 2011, 11, 85–95.

(4)Zhao, Y.; Butler, E. B.; Tan, M. Targeting cellular metabolism to improve cancer therapeutics. Cell Death Dis 2013, 4, e532.

(5)Buchakjian, M. R.; Kornbluth, S. The engine driving the ship: metabolic steering of cell proliferation and death. Nat. Rev. Mol. Cell Biol. 2010, 11, 715–727.

(6)Galluzzi, L.; Kepp, O.; Heiden, M. G. V.; Kroemer, G. Metabolic targets for cancer therapy. Nat Rev Drug Discov 2013.

(7)Joyce, A. R.; Palsson, B. Ø. The model organism as a system: integrating “omics” data sets. Nat. Rev. Mol. Cell Biol. 2006, 7, 198–210.

(8)Suhre, K.; Gieger, C. Genetic variation in metabolic phenotypes: study designs and applications. Nat. Rev. Genet. 2012, 13, 759–769.

(9)Trygg, J. O2-PLS for qualitative and quantitative analysis in multivariate calibration. J. Chemometrics 2002, 16, 283–293.

(10)Bylesjö, M.; Eriksson, D.; Kusano, M.; Moritz, T.; Trygg, J. Data integration in plant biology: the O2PLS method for combined modeling of transcript and metabolite data. Plant J. 2007, 52, 1181–1191.

(11)Rantalainen, M.; Cloarec, O.; Beckonert, O.; Wilson, I. D.; Jackson, D.; Tonge, R.; Rowlinson, R.; Rayner, S.; Nickson, J.; Wilkinson, R. W.; et al. Statistically integrated metabonomic-proteomic studies on a human prostate cancer xenograft model in mice. J. Proteome Res. 2006, 5, 2642–2655.

(12)Kirwan, G. M.; Johansson, E.; Kleemann, R.; Verheij, E. R.; Wheelock, Å. M.; Goto, S.; Trygg, J.; Wheelock, C. E. Building multivariate systems biology models. Anal. Chem. 2012, 84, 7064–7071.

(13)Jin, D.; Fan, J.; Wang, L.; Thompson, L. F.; Liu, A.; Daniel, B. J.; Shin, T.; Curiel, T. J.; Zhang, B. CD73 on tumor cells impairs antitumor T-cell responses: a novel mechanism of tumor-induced immune suppression. Cancer Res. 2010, 70, 2245–2255.

(14)Stagg, J.; Divisekera, U.; McLaughlin, N.; Sharkey, J.; Pommey, S.; Denoyer, D.; Dwyer, K. M.; Smyth, M. J. Anti-CD73 antibody therapy inhibits breast tumor growth and metastasis. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 1547–1552.

(15)Wang, H.; Lee, S.; Nigro, C. L.; Lattanzio, L.; Merlano, M.; Monteverde, M.; Matin, R.; Purdie, K.; Mladkova, N.; Bergamaschi, D.; et al. NT5E (CD73) is epigenetically regulated in malignant melanoma and associated with metastatic site specificity. Br. J. Cancer 2012, 106, 1446–1452.

(16)Serra, S.; Horenstein, A. L.; Vaisitti, T.; Brusa, D.; Rossi, D.; Laurenti, L.; D'Arena, G.; Coscia, M.; Tripodo, C.; Inghirami, G.; et al. CD73-generated extracellular adenosine in chronic lymphocytic leukemia creates local conditions counteracting drug-induced cell death. Blood 2011, 118, 6141–6152.

(17)Cappellari, A. R.; Vasques, G. J.; Bavaresco, L.; Braganhol, E.; Battastini, A. M. O. Involvement of ecto-5'-nucleotidase/CD73 in U138MG glioma cell adhesion. Mol. Cell. Biochem. 2012, 359, 315–322.

(18)Wu, X.-R.; He, X.-S.; Chen, Y.-F.; Yuan, R.-X.; Zeng, Y.; Lian, L.; Zou, Y.-F.; Lan, N.; Wu, X.-J.; Lan, P. High expression of CD73 as a poor prognostic biomarker in human colorectal cancer. J Surg Oncol 2012, 106, 130–137.

(19)Nigro, Lo, C.; Monteverde, M.; Lee, S.; Lattanzio, L.; Vivenza, D.; Comino, A.; Syed, N.; McHugh, A.; Wang, H.; Proby, C.; et al. NT5E CpG island methylation is a favourable breast cancer biomarker. Br. J. Cancer 2012, 107, 75–83.

(20)Trygg, J.; Wold, S. Orthogonal projections to latent structures (O-PLS). J. Chemometrics 2002, 16, 119–128.

(21)Wold, S.; Sjöström, M.; Eriksson, L. PLS-regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems 2001, 58, 109–130.

(22)Wold, S.; Antti, H.; Lindgren, F.; Öhman, J. Orthogonal signal correction of near-infrared spectra. Chemometrics and Intelligent Laboratory Systems 1998, 44, 175–185.

(23)Pinto, R. C.; Trygg, J.; Gottfries, J. Advantages of orthogonal inspection in chemometrics. J. Chemometrics 2012, 26, 231–235.

(24)Trygg, J.; Wold, S. O2-PLS, a two-block (X-Y) latent variable regression (LVR) method with an integral OSC filter. J. Chemometrics 2003, 17, 53–64.

(25)Schröder, M. S.; Culhane, A. C.; Quackenbush, J.; Haibe-Kains, B. survcomp: an R/Bioconductor package for performance assessment and comparison of survival models. Bioinformatics 2011, 27, 3206–3208.

(26)Subramanian, A.; Tamayo, P.; Mootha, V. K.; Mukherjee, S.; Ebert, B. L.; Gillette, M. A.; Paulovich, A.; Pomeroy, S. L.; Golub, T. R.; Lander, E. S.; et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 15545–15550.

(27)Suhre, K.; Shin, S.-Y.; Petersen, A.-K.; Mohney, R. P.; Meredith, D.; Wägele, B.; Altmaier, E.; CARDIoGRAM; Deloukas, P.; Erdmann, J.; et al. Human metabolic individuality in biomedical and pharmaceutical research. Nature 2011, 477, 54–60.

(28)Jain, M.; Nilsson, R.; Sharma, S.; Madhusudhan, N.; Kitami, T.; Souza, A. L.; Kafri, R.; Kirschner, M. W.; Clish, C. B.; Mootha, V. K. Metabolite profiling identifies a key role for glycine in rapid cancer cell proliferation. Science 2012, 336, 1040–1044.

(29)Bussey, K. J.; Chin, K.; Lababidi, S.; Reimers, M.; Reinhold, W. C.; Kuo, W.-L.; Gwadry, F.; Ajay; Kouros-Mehr, H.; Fridlyand, J.; et al. Integrating data on DNA copy number with gene expression levels and drug sensitivities in the NCI-60 cell line panel. Mol. Cancer Ther. 2006, 5, 853–867.

(30)Huang, Y.; Anderle, P.; Bussey, K. J.; Barbacioru, C.; Shankavaram, U.; Dai, Z.; Reinhold, W. C.; Papp, A.; Weinstein, J. N.; Sadée, W. Membrane transporters and channels: role of the transportome in cancer chemosensitivity and chemoresistance. Cancer Res. 2004, 64, 4294–4301.

(31)Cavill, R.; Kamburov, A.; Ellis, J. K.; Athersuch, T. J.; Blagrove, M. S. C.; Herwig, R.; Ebbels, T. M. D.; Keun, H. C. Consensus-phenotype integration of transcriptomic and metabolomic data implies a role for metabolism in the chemosensitivity of tumour cells. PLoS Comput. Biol. 2011, 7, e1001113.

(32)Luvero, D.; Milani, A.; Ledermann, J. A. Treatment options in recurrent ovarian cancer: latest evidence and clinical potential. Ther Adv Med Oncol 2014, 6, 229–239.

(33)Miow, Q. H.; Tan, T. Z.; Ye, J.; Lau, J. A.; Yokomizo, T.; Thiery, J.-P.; Mori, S. Epithelial-mesenchymal status renders differential responses to cisplatin in ovarian cancer. Oncogene 2014.

(34)Cannistra, S. A. Cancer of the ovary. N. Engl. J. Med. 2004, 351, 2519–2529.

(35)Langdon, S. P.; Lawrie, S. S.; Hay, F. G.; Hawkes, M. M.; McDonald, A.; Hayward, I. P.; Schol, D. J.; Hilgers, J.; Leonard, R. C.; Smyth, J. F. Characterization and properties of nine human ovarian adenocarcinoma cell lines. Cancer Res. 1988, 48, 6166–6172.

(36)Bristow, R. E.; Tomacruz, R. S.; Armstrong, D. K.; Trimble, E. L.; Montz, F. J. Survival effect of maximal cytoreductive surgery for advanced ovarian carcinoma during the platinum era: a meta-analysis. J. Clin. Oncol. 2002, 20, 1248–1259.

(37)Ahmed, A. A.; Mills, A. D.; Ibrahim, A. E. K.; Temple, J.; Blenkiron, C.; Vias, M.; Massie, C. E.; Iyer, N. G.; McGeoch, A.; Crawford, R.; et al. The extracellular matrix protein TGFBI induces microtubule stabilization and sensitizes ovarian cancers to paclitaxel. Cancer Cell 2007, 12, 514–527.

(38)Zhao, H.; Kim, Y.; Wang, P.; Lapointe, J.; Tibshirani, R.; Pollack, J. R.; Brooks, J. D. Genome-wide characterization of gene expression variations and DNA copy number changes in prostate cancer cell lines. Prostate 2005, 63, 187–197.

(39)Sieh, W.; Köbel, M.; Longacre, T. A.; Bowtell, D. D.; deFazio, A.; Goodman, M. T.; Høgdall, E.; Deen, S.; Wentzensen, N.; Moysich, K. B.; et al. Hormone-receptor expression and ovarian cancer survival: an Ovarian Tumor Tissue Analysis consortium study. Lancet Oncol. 2013, 14, 853–862.

(40)Spychala, J.; Lazarowski, E.; Ostapkowicz, A.; Ayscue, L. H.; Jin, A.; Mitchell, B. S. Role of estrogen receptor in the regulation of ecto-5'-nucleotidase and adenosine in breast cancer. Clin. Cancer Res. 2004, 10, 708–717.

(41)Loi, S.; Pommey, S.; Haibe-Kains, B.; Beavis, P. A.; Darcy, P. K.; Smyth, M. J.; Stagg, J. CD73 promotes anthracycline resistance and poor prognosis in triple negative breast cancer. Proc. Natl. Acad. Sci. U.S.A. 2013, 110, 11091–11096.

(42)Lee, J. M.; Dedhar, S.; Kalluri, R.; Thompson, E. W. The epithelial-mesenchymal transition: new insights in signaling, development, and disease. J. Cell Biol. 2006, 172, 973–981.

(43)Banerji, S.; Cibulskis, K.; Rangel-Escareno, C.; Brown, K. K.; Carter, S. L.; Frederick, A. M.; Lawrence, M. S.; Sivachenko, A. Y.; Sougnez, C.; Zou, L.; et al. Sequence analysis of mutations and translocations across breast cancer subtypes. Nature 2012, 486, 405–409.

(44)Barretina, J.; Caponigro, G.; Stransky, N.; Venkatesan, K.; Margolin, A. A.; Kim, S.; Wilson, C. J.; Lehár, J.; Kryukov, G. V.; Sonkin, D.; et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 2012, 483, 603–607.

(45)Wong, K.-K.; Tsang, Y. T. M.; Deavers, M. T.; Mok, S. C.; Zu, Z.; Sun, C.; Malpica, A.; Wolf, J. K.; Lu, K. H.; Gershenson, D. M. BRAF mutation is rare in advanced-stage low-grade ovarian serous carcinomas. Am. J. Pathol. 2010, 177, 1611–1617.

(46)Yegutkin, G. G.; Marttila-Ichihara, F.; Karikoski, M.; Niemelä, J.; Laurila, J. P.; Elima, K.; Jalkanen, S.; Salmi, M. Altered purinergic signaling in CD73-deficient mice inhibits tumor progression. Eur. J. Immunol. 2011, 41, 1231–1241.

(47)Wang, L.; Zhou, X.; Zhou, T.; Ma, D.; Chen, S.; Zhi, X.; Yin, L.; Shao, Z.; Ou, Z.; Zhou, P. Ecto-5'-nucleotidase promotes invasion, migration and adhesion of human breast cancer cells. J. Cancer Res. Clin. Oncol. 2008, 134, 365–372.

(48)Zhi, X.; Wang, Y.; Zhou, X.; Yu, J.; Jian, R.; Tang, S.; Yin, L.; Zhou, P. RNAi-mediated CD73 suppression induces apoptosis and cell-cycle arrest in human breast cancer cells. Cancer Sci. 2010, 101, 2561–2569.

(49)Terp, M. G.; Olesen, K. A.; Arnspang, E. C.; Lund, R. R.; Lagerholm, B. C.; Ditzel, H. J.; Leth-Larsen, R. Anti-human CD73 monoclonal antibody inhibits metastasis formation in human breast cancer by inducing clustering and internalization of CD73 expressed on the surface of cancer cells. J. Immunol. 2013, 191, 4165–4173.

(50)Allard, B.; Turcotte, M.; Spring, K.; Pommey, S.; Royal, I.; Stagg, J. Anti-CD73 therapy impairs tumor angiogenesis. Int. J. Cancer 2013.

(51)Turcotte, M.; Spring, K.; Pommey, S.; Chouinard, G.; Cousineau, I.; George, J.; Chen, G. M.; Gendoo, D. M. A.; Haibe-Kains, B.; Karn, T.; et al. CD73 Is Associated with Poor Prognosis in High-Grade Serous Ovarian Cancer. Cancer Res. 2015, 75, 4494–4503.

(52)Ferreira, L. M. R.; Hebrant, A.; Dumont, J. E. Metabolic reprogramming of the tumor. Oncogene 2012, 31, 3999–4011.

(53)Wang, L.; Fan, J.; Thompson, L. F.; Zhang, Y.; Shin, T.; Curiel, T. J.; Zhang, B. CD73 has distinct roles in nonhematopoietic and hematopoietic cells to promote tumor growth in mice. J. Clin. Invest. 2011, 121, 2371–2382.

(54)Govindaraj, C.; Scalzo-Inguanti, K.; Madondo, M.; Hallo, J.; Flanagan, K.; Quinn, M.; Plebanski, M. Impaired Th1 immunity in ovarian cancer patients is mediated by TNFR2+ Tregs within the tumor microenvironment. Clin. Immunol. 2013, 149, 97–110.

(55)Häusler, S. F. M.; Montalbán del Barrio, I.; Strohschein, J.; Chandran, P. A.; Engel, J. B.; Hönig, A.; Ossadnik, M.; Horn, E.; Fischer, B.; Krockenberger, M.; et al. Ectonucleotidases CD39 and CD73 on OvCA cells are potent adenosine-generating enzymes responsible for adenosine receptor 2A-dependent suppression of T cell function and NK cell cytotoxicity. Cancer Immunol. Immunother. 2011, 60, 1405–1418.

(56)Virtanen, S. S.; Kukkonen-Macchi, A.; Vainio, M.; Elima, K.; Härkönen, P. L.; Jalkanen, S.; Yegutkin, G. G. Adenosine inhibits tumor cell invasion via receptor-independent mechanisms. Mol. Cancer Res. 2014, 12, 1863–1874.

(57)Peng, D.-J.; Wang, J.; Zhou, J.-Y.; Wu, G. S. Role of the Akt/mTOR survival pathway in cisplatin resistance in ovarian cancer cells. Biochem. Biophys. Res. Commun. 2010, 394, 600–605.

(58)Bogan, K. L.; Brenner, C. 5′-Nucleotidases and their new roles in NAD+ and phosphate metabolism. New J. Chem. 2010, 34, 845–853.

(59)Ciriello, G.; Cerami, E.; Sander, C.; Schultz, N. Mutual exclusivity analysis identifies oncogenic network modules. Genome Res. 2012, 22, 398–406.

(60)Leslie, A.; Pratt, N. R.; Gillespie, K.; Sales, M.; Kernohan, N. M.; Smith, G.; Wolf, C. R.; Carey, F. A.; Steele, R. J. C. Mutations of APC, K-ras, and p53 are associated with specific chromosomal aberrations in colorectal adenocarcinomas. Cancer Res. 2003, 63, 4656–4661.

(61)Shih, I.-M.; Kurman, R. J. Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. Am. J. Pathol. 2004, 164, 1511–1518.

(62)Schmeler, K. M.; Sun, C. C.; Bodurka, D. C.; Deavers, M. T.; Malpica, A.; Coleman, R. L.; Ramirez, P. T.; Gershenson, D. M. Neoadjuvant chemotherapy for low-grade serous carcinoma of the ovary or peritoneum. Gynecol. Oncol. 2008, 108, 510–514.

(63)Gershenson, D. M.; Sun, C. C.; Bodurka, D.; Coleman, R. L.; Lu, K. H.; Sood, A. K.; Deavers, M.; Malpica, A. L.; Kavanagh, J. J. Recurrent low-grade serous ovarian carcinoma is relatively chemoresistant. Gynecol. Oncol. 2009, 114, 48–52.

(64)Ratner, E. S.; Keane, F. K.; Lindner, R.; Tassi, R. A.; Paranjape, T.; Glasgow, M.; Nallur, S.; Deng, Y.; Lu, L.; Steele, L.; et al. A KRAS variant is a biomarker of poor outcome, platinum chemotherapy resistance and a potential target for therapy in ovarian cancer. Oncogene 2012, 31, 4559–4566.

(65)Youn, C.-K.; Kim, M.-H.; Cho, H.-J.; Kim, H.-B.; Chang, I.-Y.; Chung, M.-H.; You, H. J. Oncogenic H-Ras up-regulates expression of ERCC1 to protect cells from platinum-based anticancer agents. Cancer Res. 2004, 64, 4849–4857.

(66)Rosanò, L.; Cianfrocca, R.; Spinella, F.; Di Castro, V.; Nicotra, M. R.; Lucidi, A.; Ferrandina, G.; Natali, P. G.; Bagnato, A. Acquisition of chemoresistance and EMT phenotype is linked with activation of the endothelin A receptor pathway in ovarian carcinoma cells. Clin. Cancer Res. 2011, 17, 2350–2360.

(67)Marchini, S.; Fruscio, R.; Clivio, L.; Beltrame, L.; Porcu, L.; Fuso Nerini, I.; Cavalieri, D.; Chiorino, G.; Cattoretti, G.; Mangioni, C.; et al. Resistance to platinum-based chemotherapy is associated with epithelial to mesenchymal transition in epithelial ovarian cancer. Eur. J. Cancer 2013, 49, 520–530.

TablesTable 1 - Analysis of correlation of NT5E expression and sensitivity to chemotherapeutic drugs in NCI60 panel

Name

Pearson correlation coefficient

p

p adjusted*

Inosine-glycodialdehyde

-0.59

1.00E-06

0.0001

Tetraplatin

-0.52

3.00E-05

0.004

5-6-Dihydro-5-azacytidine

-0.49

0.0001

0.01

Nitrogen mustard hydrochloride

-0.48

0.0001

0.01

Dichloroallyl-lawsone

-0.48

0.00015

0.02

Fluorodopan

-0.47

0.0002

0.02

Daunorubicin hydrochloride

-0.47

0.0002

0.02

Melphalan

-0.45

0.0004

0.05

…

…

…

…

Iproplatin

-0.43

0.0006

0.08

Diaminocyclohexyl-Pt-II

-0.43

0.0007

0.08

cis-Diamminedichloroplatinum(II) (cisplatin)

-0.26

0.05

1

Carboplatin

-0.21

0.1

1

* Bonferroni corrected

Figures LegendsFigure 1 - O2PLS analysis

Loadings plots of O2PLS model integrating gene expression (a) and metabolite data (b). In this display the relative position of each point in the two loadings plots indicates if regression coefficients for a given pair of mRNA probe and metabolite correlated or anti-correlated to each other, while distance from the origin and color indicate the magnitude of the association and magnitude of contribution to Q2 respectively. Variables with Q2 values above 0.2 are labeled.

Figure 2 - NT5E expression in ovarian cancer cell lines with acquired and intrinsic resistance to cisplatin

(a) Fold change of NT5E expression in resistant cell lines (PEO4, A2780CP, CHICisR, M41CisR, TYKnuCisR, OV90C) relative to their parental sensitive counterparts (PEO1, A2780, CHI, M41, TYKnu, OV90). ±SD of biological replicates is shown where applicable. (b) Boxplot of the NT5E expression in two groups of ovarian cell lines: intrinsically sensitive and resistant to cisplatin.

Figure 3 - Effect of NT5E silencing on cisplatin sensitivity in breast and ovarian cancer cell lines.

Barplots show cell number relative to control in MDA-MB-231 (a), BT-549 (b), PEO1 (c) or PEO4 (d) cells transfected with NT5E siRNA (white bar) or with scrambled control (black bar) upon treatment with 0 or 2.5 uM of cisplatin. Error bars show standard deviation over biological replicates (n = 5 for breast cancer cell lines, n = 3 for ovarian cancer cell lines). Below: Western blot analysis of the efficiency of knockdown. (e) Western blot analysis of NT5E protein level in PEO1 and PEO4 cell lines (representative of 3). n.s. not significant, * p < 0.05, ** p < 0.005 (independent t-test)

Figure 4 - Meta-analysis of survival data

Forest plot of NT5E expression association with survival in 6 ovarian cancer studies. Solid blue lines denote HRs 95% confidence intervals; boxes denote the relative influence of each study over the results; diamond marks the overall HR and its 95% confidence interval.

Figure 5 - NT5E expression and cancer subtypes

NT5E expression according to sex-hormone sensitivity in (a) breast cancer ER+ (white) and ER- (grey) cell lines, error bars indicate standard deviation across multiple probes; (b) prostate cancer cell lines that are androgen-sensitive (AS, white) or insensitive (AI, grey); (c) ovarian cancer tumors stratified by estrogen-receptor level.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

For TOC only

- 1 -

Dataset N HR (95% CI)

GSE14764 80 0.98 (0.59-1.65)GSE15622 35 0.77 (0.38-1.56)GSE26712 185 1.40 (1.02-1.93)GSE30161 54 0.91 (0.50-1.64)GSE9891 278 1.2 (0.90-1.60)TCGA 395 1.29 (1.02-1.63)

Overall 1.21 (1.05-1.39)

0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4

MC

F7

NT

5E e

xpre

ssio

n

_

_

_

_

_

_

_

_

_

_

ER+ER−

Breast cancer cell lines (GSE22821)

T47

D

HS

578T

MD

A-M

B-2

31

BT-

549

NT

5E e

xpre

ssio

n

ASAI

22R

v1

LAP

C4

LNC

aP

MD

A2B

MD

A2A

DU

145

PC

3

PP

C ERS1 above median

NT

5E e

xpre

ssio

n

a b c-6

-4-2

02

-4-3

-2-1

0

Prostate cancer cell lines (GSE4016) Ovarian tumors (TCGA)

-2-1

01

23

4

ESR1 below median

p < 10-5

-0.2 -0.1 0.0 0.1 0.2

-0.1

0.0

0.1

0.2

X lo

adin

gs 2

Gene Expression Data

PDXDC1

SLC16A10

ELOVL2

PLEKHH1

ETFDH

SLC22A4

SPARC

NT5E

NT5ENT5E

-0.1

30.

41

Q2

-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15

-0.0

50.

000.

050.

100.

15Y

load

ings

2

Metabolites

4-Guanidinobutanoic acid

adenine

aspartate

cholesterolcytidine 5’-monophosphateglycerol

guanosine

inosine

inositol 1-phosphate

L-beta-imidazolelactic acid

malic acid

methionine

N-acetylneuraminic acid

n-hexadecanoic acid

palmitoleic acid

pantothenic acid

phosphate

tetradecanoic acid

trans-4-hydroxyproline

tyramine

uracil

valine

-0.0

80.

44

Q2

X loadings 1 Y loadings 1

S-(5’-adenosyl)-L-methionine

N-formyl-L-methionine

-0.2 -0.1 0.0 0.1 0.2

-0.1

0.0

0.1

0.2

X lo

adin

gs 2

mRNAs

PDXDC1

SLC16A10

ELOVL2

PLEKHH1

ETFDH

SLC22A4

SPARC

NT5E

NT5E

NT5E

-0.1

30.

41

Q2

-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15

-0.0

50.

000.

050.

100.

15Y

load

ings

2

Metabolites

4-Guanidinobutanoic acid

adenine

aspartate

cholesterolcytidine 5’-monophosphateglycerol

guanosine

inosine

inositol 1-phosphate

L-beta-imidazolelactic acid

malic acid

methionine

N-acetylneuraminic acid

n-hexadecanoic acid

palmitoleic acid

pantothenic acid

phosphate

tetradecanoic acid

trans-4-hydroxyproline

tyramine

uracil

valine

-0.0

80.

44

Q2

X loadings 1 Y loadings 1

a b

S-(5’-adenosyl)-L-methionine

N-formyl-L-methionine

PE

O1

CH

I

CH

ICis

R

M41

M41

Cis

R

TY

Knu

TY

Knu

Cis

R

OV

90

OV

90C

A27

80

A27

80C

P

A27

80

A27

80C

P

A27

80

A27

80C

isR

Exp

ress

ion

fold

cha

nge

01

23

45

6

PE

O4

PE

O1

PE

O4

In vivo acquired resistance In vitro acquired resistance

Sensitive

Resistant

GSE41499 GSE34615 GSE26465 GSE34615 GSE28646 GSE47856

2.5

5.0

7.5

10.0

Cisplatinresistant

Cisplatinsensitive

RM

A n

orm

aliz

ed g

ene

expr

essi

on

p = 0.006

a

b

GSE47856

PEO4

0.0

0.2

0.4

0.6

0.8

1.0PEO1

MDAMB231

0 2.50.0

0.2

0.4

0.6

0.8

1.0

1.2SCRsiNT5E

*

cisplatin (uM)

Rel

ativ

e ce

ll n

um

ber

(A

.U.)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

**

BT549

Rel

ativ

e ce

ll nu

mbe

r (A

.U.)

NT5E (64 kDa)

PEO1 PEO4

0 2.5

cisplatin (uM)

SCRsiNT5E

SCR siNT5ENT5E (64 kDa)

B-Tubulin

Rel

ativ

e ce

ll n

um

ber

(A

.U.)

*

0 2.5

cisplatin (uM)

NT5E (64 kDa)

B-Actin

SCR siNT5E

SCR siNT5E

B-Actin

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Rel

ativ

e ce

ll n

um

ber

(A

.U.)

n.s.

SCRsiNT5E

SCRsiNT5E

0 2.5

cisplatin (uM)

NT5E (64 kDa)

B-Actin

a b

c d

e