spiral.imperial.ac.uk · Web viewphase of the experiment (study 2, Table S1). K4IM human synovial...

of 53 /53
Targeted metabolic profiling of the Tg197 mouse model reveals itaconic acid as a marker of Rheumatoid Arthritis. Filippos Michopoulos 1,2, , Niki Karagianni 3 , Nichola Whalley 1 , Mike Firth 8 , Christoforos Nikolaou 5,6 , , Ian D Wilson 4 ,. Susan E Critchlow 1 , George Kollias 5,7* Georgios Theodoridis 2* 1 Bioscience, Oncology iMED, AstraZeneca, Alderley Park, Macclesfield, Cheshire, UK 2 Department of Chemistry, Aristotle University of Thessaloniki, 541 24 Greece. 3 Biomedcode Hellas SA, 34 Fleming Str., 16672 Vari, Greece 4 Department of Surgery and Cancer, Imperial College, London UK. 5 Biomedical Sience Research Center “Alexander Fleming”, 34 Fleming Str., 16672 Vari, Greece 6 Department of Biology, University of Crete, Heraklion, Greece 7 Department of Physiology, Faculty of Medicine, National and Kapodistrian University of Athens, Greece 8 Discovery Science, Innovative Medicine, AstraZeneca, Cambridge, UK 1

Embed Size (px)

Transcript of spiral.imperial.ac.uk · Web viewphase of the experiment (study 2, Table S1). K4IM human synovial...

_

Targeted metabolic profiling of the Tg197 mouse model reveals itaconic acid as a marker of Rheumatoid Arthritis.

Filippos Michopoulos1,2,, Niki Karagianni3, Nichola Whalley1, Mike Firth8, Christoforos Nikolaou5,6, , Ian D Wilson4,. Susan E Critchlow1, George Kollias5,7* Georgios Theodoridis2*

1Bioscience, Oncology iMED, AstraZeneca, Alderley Park, Macclesfield, Cheshire, UK

2Department of Chemistry, Aristotle University of Thessaloniki, 541 24 Greece.

3Biomedcode Hellas SA, 34 Fleming Str., 16672 Vari, Greece

4Department of Surgery and Cancer, Imperial College, London UK.

5Biomedical Sience Research Center Alexander Fleming, 34 Fleming Str., 16672 Vari, Greece

6 Department of Biology, University of Crete, Heraklion, Greece

7Department of Physiology, Faculty of Medicine, National and Kapodistrian University of Athens, Greece

8Discovery Science, Innovative Medicine, AstraZeneca, Cambridge, UK

*Author for correspondence

Email:

Georgios Theodoridis: [email protected],

George Kollias: [email protected]

Tel:+302310997718 +302109656507

Fax: :+302310997719 +302109656563

Abstract

Rheumatoid arthritis is a progressive, highly debilitating disease where early diagnosis, enabling rapid clinical intervention, would provide obvious benefits to patients, healthcare systems and society. Novel biomarkers that enable non-invasive early diagnosis of the onset and progression of the disease provide one route to achieving this goal. Here a metabolic profiling method has been applied to investigate disease development in the Tg197 arthritis mouse model. Hind limb extract profiling demonstrated clear differences in metabolic phenotypes between control (wild type), and Tg197 transgenic mice and highlighted raised concentrations of itaconic acid as a potential marker of the disease. These changes in itaconic acid concentrations were moderated or indeed reversed, when the Tg197 mice were treated with the anti-hTNF biologic infliximab (10mg/kg twice weekly for 6 weeks). Further in vitro studies on synovial fibroblasts obtained from healthy wild-type, arthritic Tg197 and infliximab-treated Tg197 transgenic mice, confirmed the association of itaconic acid with rheumatoid arthritis and disease moderating drug effects. Preliminary indications of the potential value of itaconic acid as a translational biomarker were obtained when studies on K4IM human fibroblasts treated with hTNF showed an increase in the concentrations of this metabolite.

Keywords: rheumatoid arthritis, metabolomics, intracellular metabolites, targeted analysis, biomarker, mass spectrometry

Introduction

Rheumatoid arthritis (RA) is a debilitating, progressive disease that places a considerable burden on healthcare systems. Novel biomarkers that enable non-invasive early diagnosis of the onset and progression of the disease could provide significant benefits to both patients and society by enabling early clinical intervention. They may also allow optimization of individual patient treatments especially if undertaken prior to permanent damage of the osteo-cartilage. One potential source of novel biomarkers is metabolic phenotyping (or metabotyping) of the sort practiced in metabolomics/metabonomics studies1-4. Whilst metabotyping has been employed in the investigation of the pathophysiology of RA for some time there have been relatively few publications as a result. However, such studies as have been reported demonstrate the value of metabolic phenotyping with most of this work having involved the analysis of human clinical samples such as synovial fluid5-8, plasma9-11 and urine/plasma10. For example Naughton et al,6 using 1H-NMR spectroscopy, examined the metabolic profile of synovial fluid samples and compared these profiles to those of matched serum samples and showed increased lactate concentrations and strong glucose depletion in synovial fluid compared to serum. In addition ketone bodies (3-hydroxybutyrate, acetone and acetoacetate) were enriched in the synovial fluid with significant reduction in chylomicron and VLDL-associated triacylglycerols (which also appeared to have a reduced mean chain length). This metabolic perturbation was consistent with the hypoxic conditions of the inflamed joint, restricting the diffusion of glucose and large molecules from the blood into the synovium. These effects resulted in increased lipid degradation concomitant to ketone body enrichment within the synovium to serve the energy demands of the joint tissues. Subsequently, the same group compared the metabolic phenotype of normal synovial fluid with that of RA patients7. Once again, the synovial fluid from RA patients was associated with comparatively high lactate concentrations and lower quantities of lipoprotein-associated fatty acids and glycoproteins. Williamson et al.,5 reported changes in triglycerides, glycoproteins and creatinine after the analysis of serial synovial fluid samples collected from two RA patients and the results were found to be correlated with the disease activity. Meshitsuka et al,8 reported that the lactate to alanine ratio in the synovial fluid could be used as a marker to discriminate RA from osteoarthritis12. Recent work by Giera et al.,13 used liquid chromatography coupled to mass spectrometry (LC-MS) for the analysis of synovial fluid and the characterisation of many lipid classes and lipid mediators (resolving D5, lipoxin A4, maresin 1) in RA patients.

1H NMR spectroscopic analysis of plasma from patients with RA9 revealed elevated concentrations of lactate, cholesterol, acetylated glycoprotein and unsaturated lipids and reduced amounts of high density lipoprotein (HDL). These findings were found to correlate with disease severity and support the positive association between RA and coronary artery disease14. Gas chromatography mass spectrometric (GC-MS) analysis of human serum samples11,15 reported high concentrations of L-asparagine, L-alanine, 2-oxy-butanoic acid, palmitic acid and heptanoic acid in RA patients while 2-butenoic acid, undecanoic acid, glucuronic acid and stearic acid were found in relatively larger amounts in the healthy control group. In subsequent work from the same group 10 LC-MS analysis of urine samples revealed significant changes in the acylcarnitine profile. Analysis of plasma showed increased concentrations of tryptophan, -ketoisovaleric acid, 3-methyl-2-oxovaleric acid, cholesterol sulfate, uric acid, indoxyl sulfate, 4-methyl-2-oxovaleric acid and dehydroepiandrosterone sulfate in the heat joint sample group compared to the controls.

Smolenska et al.,16 reported significant depletion in the circulating blood concentrations of hypoxanthine, uridine and uric acid in RA patients following methotrexate treatment. These investigators proposed that the observed perturbation of purine/pyrimidine metabolism affects the availability of biogenic amines for DNA/RNA synthesis which subsequently may affect immune cell proliferation and the cytokine expression profile. In addition, changes in amino acid, hypoxanthine, uric acid, lactate, uracil, trimethylamine-N-oxide and -ketoglutaric acid concentrations have recently been reported as markers of methotrexate treatment in RA patients with early disease onset17. Clearly, based on these studies, RA is associated with numerous perturbations of the metabolome.

Animal models have been (and remain) instrumental in the study of complex diseases such as RA, eliminating much of the inherent variability of human samples, providing better control and a more direct framework for the elucidation of underlying biological mechanisms driving the pathology. However, despite their undoubted value in the study of RA, to date only a limited number of metabolic profiling studies based on validated animal models have been reported. In one of these the analysis of serum samples was obtained from K/BN transgenic mice, a model of severe inflammatory arthritis, by 1H NMR spectroscopy. This study reported a range of changes in the metabolic profiles of these animals compared to healthy control mice alterations in pathways associated with nucleic acid metabolism (xanthine, hypoxanthine, uridine, uracil, and trimethylamineN-oxide), amino acid metabolism (glutamate, serine, phenylalanine, glycine, methionine, asparagine) lipid/fatty acid metabolism (glycerol, choline, 2-hydroxybutyrate, acetylcarnitine) and oxidative stress (taurine, methionine, glycine, xanthine, hypoxanthine, trimethylamine-N-oxide)18.

The human TNF transgenic (Tg197) mouse is a well-established animal model of human inflammatory polyarthritis that recapitulates human disease. In these mice, spontaneous development of pathology is due to deregulated human TNF expression produced by a 3-UTR modified human TNF transgene19. Similarly to human disease, synovial fibroblasts play a central role in pathogenesis including TNF production and gradual activation, acquisition of a, hyperplastic phenotype and release of a variety of proinflammatory cytokines, matrix metalloproteinases and others factors promoting different features of the pathology 20,21. The Tg197 arthritis model was used in the present study to obtain concise arthritis metabolic profiles of blood serum and urine collected at different stages of disease development, including samples from mice treated with anti-TNF disease-modifying drug. We applied the same methodology to analyze aqueous extracts of hind limb tissue and synovial fibroblasts to allow the comparison of modulation of metabolic pathways at organism and cellular level. More importantly, we implemented a systems biology approach that integrates the obtained metabolic profiles with gene expression data, which allows us to gain further insight on the overall modulation of metabolic pathways.

Experimental procedures

Materials

Water (18.2 M) was obtained from a Purelab Ultra System from Elga (Bucks, UK). Methanol, acetonitrile and isopropanol used for sample extraction and analysis were of HPLC grade (Sigma-Aldrich, Gillingham, UK). Tributylamine (TBA), acetic acid and all analytical standards, of the highest purity available, were purchased from Sigma-Aldrich.

In vivo sample collection

Tg197 human TNF transgenic (TG) mice were bred and maintained on CBA-C57BL/6J background in Biomedcode Hellas SA animal facility, fed a normal diet and water ad libitum. Animal study protocols were approved by the directorate of Agricultural and Veterinary Policy of the Attica Region for compliance with regulations. For sample collection a therapeutic protocol was applied, using 6 week old male and female Tg197 mice, with established disease pathology. Groups of 6 or 7 mice TG (depending upon the study) were allocated and received (TR) or not i.p. injections of the commercial anti-hTNF biologic infliximab at 10mg/kg twice weekly and for a time period of 6 weeks. A third group of 6 or 7 wild type (WT) littermates served as a control. Body weight and clinical arthritis score, based on a 3-point scale were recorded and urine sample collections were made at a standard time of the day to avoid the effects of circadian rhythm fluctuations. At the end of the study period mice were sacrificed and serum as well as hind limb tissue samples were collected. Additional serum samples were collected from groups of wild-type and transgenic mice sacrificed at 6 weeks of age, before the initiation of the treatment. Four studies (1,3 and 4), of similar study design but with modifications to the samples collected and their frequency, were performed in order to investigate particular aspects of the evolving metabolic phenotypes of TG and TR mice . All samples were stored frozen (-80C) on collection until analysis. Further information about group size and gender composition sampling time points for the in vivo sample collection is provided in Table S1.

Mouse synovial fibroblast culture

Synovial fibroblasts from healthy WT, TG and TR animals at 9 weeks of age were isolated and cultured as described previously22. Synovial fibroblasts isolated from TR animals treated in vivo with anti-hTNF were cultured in the presence of 1 g/ml infliximab to maintain treatment during the in vitro phase of the experiment (study 2, Table S1).

K4IM human synovial fibroblast cell line culture

K4IM cells were cultured in Dulbeccos Modied Eagle's Medium supplemented with 2 mM glutamine and 10% fetal calf serum. Cells were grown under normoxic conditions (5% CO2) at 37C and stimulated for 24 hours with 10 ng/ml human tumour necrosis factor (Sigma Aldrich, UK).

Sample Preparation

In the present paper a number of different specimens were analysed by LC-MS/MS (including cell lysates, hind limb extract and serum and urine). The sample preparation procedure aimed to prepare the biological sample by extracting the primary metabolites of interest, ensuring analytical stability and sensitivity whilst maintaining the integrity of the analytical system. The procedure applied was optimized for each specimen as outlined below.

Hind limb joints

Hind limb joints were ground to powder in liquid nitrogen using pestle and mortar. Powder was placed in a 2 mL Fast-Prep tubes (lysis matrix A) and extracted with 0.82 ml/100mg tissue ACN/H2O 50/50 v/v on a Fast Prep 24 module (MP Biomedicals, USA) in a sequence of two cycles (30 seconds each) of shaking at 5M/s with a 20 sec pause between cycles. The clear supernatant obtained after centrifugation at 20800 g was transferred to 2 mL glass vials and stored at -80oC. Prior to analysis 50 L of each sample was mixed with 100 L cold methanol (MeOH) and centrifuged at 20800 g. The clear supernatant was transferred to a 0.3 ml polypropylene HPLC microvial and was dried at ambient temperature in SpeedVac. Samples were resuspended in 50 L of ultrapure water and the resulting aqueous suspensions were centrifuged at 3270 g before analysis.

Urine-Serum

Urine and serum samples (10 l) were subjected to protein removal by the addition of 40 l of cold (-20C) mixture of MeOH/ACN 50:50 v/v and centrifugation at 20800g. Then 40 l of the clear supernatant were transferred to 0.3 mL polypropylene HPLC microvials and dried at ambient temperature in a Savant SPD2010 SpeedVac (Thermo Fisher Loughborough, UK). Urine extracts were resuspended in 200 l and serum extracts in 40 l of ultra pure water. Resuspended samples were centrifuged at 3270 g for 10 min at 4C before analysis.

Mouse Synovial fibroblasts

Cell pellets were extracted with 1ml ACN/H2O 50:50 v/v. The extracts were centrifuged at 20800 g and supernatants were transferred to clean tubes and stored at -20C. Prior to analysis, 100 L of the synovial fibroblast extracts were treated with 100 L cold MeOH (-20oC) to precipitate remaining soluble proteins and were centrifuged at 20800 g. The clear supernatant was transferred to a 0.3 ml polypropylene HPLC microvial and dried at ambient temperature in SpeedVac for 60 min. Samples were resuspended in 50 L of ultrapure water and aqueous suspensions were centrifuged 3270 g for 10 min at 4C before analysis.

K4IM human synovial fibroblast cell line

At the end of the incubation period the culture media was removed and cell metabolism quenched by the addition of 400 l of 40/40/20 v/v/v MeOH:ACN:H2O (-20C) to each of the wells. The cells were stored at -20C with the extraction solvent for 20 min to extract intracellular metabolites followed by cell adhesion disruption with a cell scraper (BD, Oxford, UK). The contents of each well were transferred to 1.5 ml Eppendorf tubes and centrifuged at 16,000 g (Model 5415D, Eppendorf) for 5 min to precipitate proteins and cell debris. The supernatants were transferred to clear 1.5 mL Eppendorf tubes and stored at -20C until sample analysis. For LC-MS analysis, aliquots of 100 l of each extract were transferred to polypropylene HPLC microvials (VWR Ltd, UK) and dried at ambient temperature in a Savant SPD2010 SpeedVac for approximately 1 h. Metabolite extracts were then re-suspended in 50 l of ultra pure water and centrifuged at 3270 g for 10 min at 4C, before analysis.

Quality control

For all matrices a biological quality control (QC) sample was prepared by mixing equal volumes from each of the test samples. The QC sample was treated as a test sample and was analysed periodically within each batch. To confirm metabolite retention times, a test mixture containing all of the measured metabolites at a concentration of 5 M was analysed at the beginning and the end of the analytical batch in addition to a QC sample spiked with the test mix at a final concentration of 5 M. Prior to the start of each analytical run ten injections (5 l) of the QC sample were performed to ensure adequate system conditioning. Moreover, one QC sample was analysed every five to ten samples depending on the size of the analytical batch. Samples forming the test set were run in a random order.

Sample-Data analysis.

Liquid Chromatographytandem mass spectrometry (LC-MS)

Prepared samples were analysed by LC-MS on a system consisting of an Ultimate 3000RS chromatographic system (Thermo, UK) in combination with AB4000 Q-Trap (ABSCIEX, UK) mass spectrometer operating in negative ion mode. Metabolites were resolved using gradient elution on a binary solvent system consisted of buffer A (H2O, 10 mM tributylammonium, 15 mM acetic acid) and buffer B (MeOH/Isopropanol 80/20). A full description of the methodology is given in Michopoulos et al23. To address analytical variability across sample batch a pooled sample (QC) was analysed in regular intervals24. The raw spectrometric data was analysed and peaks were integrated with MultiQuan 2.0.2 (Applied Biosystems/ MDS Sciex) and the results were exported to Excel for normalization and univariate statistical analysis by f-test and t-test. Acceptable analytical reproducibility for each peak detected was based on the determination of the coefficient of variation (CV) for each metabolite peak present in the QC samples as being lower than 30% 24,25. T-test results that gave p-values of less than 0.05 and an increase or decrease of the average metabolite peak area between groups of greater than 30% were set as the criteria for significant metabolite perturbation.

Results and Discussion

In vivo evaluation

In this mouse model signs of clinical pathology become evident by week 3 of age, while animals fail to properly gain weight, possibly due to the catabolic effect of circulating human TNF. Upon treatment with anti-hTNF antibody treatment, starting either at early disease stages (week 3 of age) or at established disease (week 6) pathology, as performed here, signs disappear or significantly ameliorate depending on treatment efficacy and dosing. The model has been instrumental in providing the proof of concept of the pathogenic role of TNF in rheumatoid arthritis and the application of anti-TNF treatment in arthritis pathology. The model has been widely used for the exploration of molecular mechanisms contributing to the development of arthritis pathology26,27 as well as for the validation of the therapeutic potential of arthritis therapeutics with a focus in anti-human TNF28,29.

Arthritis scoring, which evaluates grip strength, joint swelling and distortion, reflecting disease progress showed that in the transgenic group (TG) an increase in the arthritis score from 1.1 to 1.75 was observed over between weeks 6 and 11 whilst for the Infliximab treated group (TR) the arthritis score was reduced from 0.8 to 0.4. Significant differences were seen in weight gain between the groups ( Figure S1)

Bioanalyisis

A total of 109 metabolites ( Table S2) were profiled in all biospecimens collected in this study using a targeted ion pair LC-MS method23. Based on analysis criterion of 30 % maximum CV in QC data24,25 a number of metabolites were excluded from the dataset and the statistical analysis was based on 53 metabolites present in the hind limb extracts and 76 metabolites in the synovial fibroblast extracts.

Analysis of Hind Limbs

Data obtained for the hind limb tissue was first processed in a multivariate manner using principal component analysis (Simca P14+). Underlying non-correlated variability to group metabolic phenotype was removed with O2PLS discriminant analysis resulting a clear discrimination between WT, TG and TR phenotype (Figure 1A). Model validation was performed with a CV-Anova test resulting a p-value0.5, QC CV