Global metabolomic profiling of human synovial fluid for rheumatoid arthritis biomarkers A.K. Carlson1,2, R.A. Rawle1,3, C.W. Wallace4, E. Adams4, M.C. Greenwood5, B. Bothner1,6, R.K. June1,2,7 1Molecular Biosciences Program; 2Department of Cell Biology and Neuroscience; 3Department of Microbiology and Immunology; 4Department of Health and Human Development; 5Department of Mathematical Sciences; 6Department of Chemistry and Biochemistry; 7Department of Mechanical and Industrial Engineering, Montana State University, Bozeman, USA. Abstract Objective The objective of this study was to analyse the metabolomic profiles of rheumatoid arthritis synovial fluid to test the use of global metabolomics by liquid chromatography-mass spectrometry for clinical analysis of synovial fluid. Methods Metabolites were extracted from rheumatoid arthritis (n=3) and healthy (n=5) synovial fluid samples using 50:50 water: acetonitrile. Metabolite extracts were analysed in positive mode by normal phase liquid chromatography-mass spectrometry for global metabolomics. Statistical analyses included hierarchical clustering analysis, principal component analysis, Student’s t-test, and volcano plot analysis. Metabolites were matched with known metabolite identities using METLIN and enriched for relevant pathways using IMPaLA. Results 1018 metabolites were detected by LC-MS analysis in synovial fluid from rheumatoid arthritis and healthy patients, with 162 metabolites identified as significantly different between diseased and control. Pathways upregulated with disease included ibuprofen metabolism, glucocorticoid and mineralocorticoid metabolism, alpha-linolenic acid metabolism, and steroid hormone biosynthesis. Pathways downregulated with disease included purine and pyrimidine metabolism, biological oxidations, arginine and proline metabolism, the citrulline-nitric oxide cycle, and glutathione metabolism. Receiver operating characteristic analysis identified 30 metabolites as putative rheumatoid arthritis biomarkers including various phospholipids, diol and its derivatives, arsonoacetate, oleananoic acid acetate, docosahexaenoic acid methyl ester, and linolenic acid and eicosatrienoic acid derivatives. Conclusion This study supports the use of global metabolomic profiling by liquid chromatography-mass spectrometry for synovial fluid analysis to provide insight into the aetiology of disease. Key words rheumatoid arthritis, metabolomics, mass spectrometry, biomarkers, synovial fluid Clinical and Experimental RheumatologyClinical 2019 and Experimental Rheumatology 2019; 37: 393-399. Global metabolomics of RA synovial fluid / A.K. Carlson et al. Alyssa K. Carlson, PhD Introduction fluid (SF) (7). SF, located in the joint Rachel A. Rawle, BS Rheumatoid arthritis (RA) is one of the cavity, contains a pool of metabolites Cameron W. Wallace, BS most common autoimmune disorders produced by many types of joint cells Erik Adams, MD, PhD affecting 2% of the North American (i.e. osteoblasts, osteoclasts, osteocytes, Mark C. Greenwood, PhD Brian Bothner, PhD population (1). It is associated with sig- chondrocytes, etc.). Thus, we hypoth- Ronald K. June, PhD nificant disability and is accompanied esised that global metabolomic profiles Please address correspondence to: by a huge economic burden of $46.7 bil- of RA synovial fluid would both reveal Dr Ronald K. June II, lion (2). RA is a chronic disease charac- the local pathological changes occurring Mechanical & Industrial Engineering, terised by joint damage, synovitis, and during RA and provide metabolites as Montana State University, systemic inflammation, yet its aetiology candidate biomarkers. In this pilot study, PO Box 173800, remains unknown. Despite recent ad- we used global metabolomic profiling Bozeman, MT 59717-3800, USA. vances in treatment options for RA, ad- by LC-MS in search of biomarkers of E-mail: [email protected] equate biomarkers with high sensitivity RA in human SF. Received on April 23, 2018; accepted and specificity for early diagnosis of RA in revised form on June 18, 2018. are currently unknown (3). Biomarkers Materials and methods © Copyright CLINICAL AND of RA are imperative for early interven- Synovial fluid samples EXPERIMENTAL RHEUMATOLOGY 2019. tion to slow and/or prevent progression SF samples were obtained under IRB into the late stages of diseases. approval (RA; n=3) or purchased from Metabolomics, which measures low- Articular Engineering (Northbrook, IL) molecular weight molecules, is the most post mortem (healthy; n=5). No clinical recent of the ‘omics’ fields to be used for data were obtained for RA patients, and biomarker discovery (4). Metabolites are only partial clinical data including age, the physiological end products of gene gender, race, and cause of death were expression, and a metabolomic profile obtained for healthy patients. Samples consists of quantitative measurements of were kept at -80˚C until metabolite thousands of metabolites. By analysing extraction. the metabolite pathways, metabolomic profiling can generate phenotypes of Metabolomics disease states and insight into disease This study is an expansion of our pre- pathogenesis. Metabolomic profiling is viously published findings identifying performed either in a global or targeted potential biomarkers of OA in human manner, where a global profile seeks all SF, and metabolite extractions, HPLC- metabolites and a targeted profile exam- MS analysis, and processing of mass ines sets of pre-identified metabolites. spectra were performed as previously Previous studies have attempted to described (9). Briefly, samples were identify biomarkers of RA with targeted thawed, cells and debris were removed metabolomics using nuclear magnetic by centrifugation, polymers were pre- resonance spectroscopy (NMR), gas cipitated with acetone, and metabolites chromatography-mass spectrometry were extracted with 50:50 water: ace- (GC-MS), and liquid chromatography- tonitrile. Metabolite extracts were ana- mass spectrometry (LC-MS) (5-8). To lysed in positive mode using an Agilent our knowledge, only one prior study 1290 UPLC system with an Agilent used global metabolomic profiling for 6538 Q-TOF mass spectrometer (Agi- RA biomarkers using GC-MS (7). In lent, Santa Clara, CA) in normal phase contrast to GC-MS, LC-MS is advanta- using a Cogent Diamond Hydride geous for complex biological samples HILIC 150 x 2.1 mm column (Micro- because the samples undergo chromato- Solv, Eatontown, NJ). Funding: funding was provided by the graphic separation in addition to mass Spectra were processed for retention NSF (CMMI 1554708) and Montana State spectrometry, which provides increased time and mass-to-charge (m/z) ratio University. The Proteomics, Metabolomics, sensitivity and greater coverage of the normalisation, noise threshold (inten- and Mass Spectrometry CORE facilities metabolites present in a sample. To our sity: 1000), and peak detection using were supported by the Murdock Charitable Trust and The National Institute of knowledge, no study to date has used MZMine 2.14 (10). Metabolites with General Sciences of the National Institutes LC-MS analysis for global metabolomic median intensity values of zero for both of Health (P20GM103474). The funding profiling in search of RA biomarkers. RA and healthy datasets were eliminat- sources did not play a role in the study Despite several prior studies identifying ed from analysis. Metabolite m/z values design or execution. biomarkers of RA, only one has searched were matched to metabolite identities Competing interests: none declared. for biomarkers of RA in human synovial using METLIN with a mass tolerance 394 Clinical and Experimental Rheumatology 2019 Global metabolomics of RA synovial fluid / A.K. Carlson et al. of 15 ppm, with +1H+ or +1Na+ adducts (11). M/z values were also checked against our internal library of metabo- lite standards to confidently identify metabolite identities by both retention time and m/z value (m/z: ± 0.01, 30 ppm; retention time: ± 0.25 min) (12). Statistical analyses All statistical analyses were completed using MATLAB (Mathworks, Inc.). Significant metabolites were identified by Student’s t-test with false-discovery rate (FDR) corrections (FDR-adjusted p-value<0.05). Significantly different metabolites were visualised as a vol- cano plot by plotting the negative log10 of the p-value (y-axis) for each m/z value and the fold change (log2(RA/ healthy)) of the median m/z value in- tensity of RA and healthy SF (x-axis) against each other to illustrate both significance and magnitude of change. M/z values with zero intensity for either group were excluded from the volcano Fig. 1. Global metabolomic profiling finds distinct metabolic phenotypes of RA and healthy SF. (A) plot analysis. Fold-change normalised, HCA of RA (n=3) and healthy (H = healthy; n=5) median metabolite intensities from human SF. Of log-transformed metabolite intensities the 1018 metabolites detected in human SF, 162 were significantly different between cohorts (p<0.05). Clusters of co-regulated metabolites altered with disease are outlined in black boxes and referenced as were analysed by principal component 1.1 and 1.2. Cluster 1.1 includes 55 co-metabolites lower in RA and cluster 1.2 includes 107 metabolites analysis (PCA) to examine the variation higher in RA. (B) PCA of the metabolite intensities in RA and healthy SF. Together, PC1, PC2, and PC3 between the RA and healthy groups. were associated with 62.6% of the variation between healthy and RA SF and illustrate clear separation Metabolites