NMR-Based Serum Metabolomics Discriminates Takayasu Arteritis

Subscriber access provided by UNIV OF MISSISSIPPI

Article

NMR based serum metabolomics discriminates Takayasu Arteritis from Healthy Individuals: A proof of principle study Anupam Guleria, Durga Prasanna Misra, Atul Rawat, Durgesh Dubey, Chunni Lal Khetrapal, Paul Bacon, Ramnath Misra, and Dinesh Kumar J. Proteome Res., Just Accepted Manuscript • Publication Date (Web): 17 Jun 2015 Downloaded from http://pubs.acs.org on June 18, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Proteome Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

NMR based serum metabolomics discriminates Takayasu Arteritis from Healthy Individuals: A proof of principle study Anupam Guleria1*, Durga Prasanna Misra2†, Atul Rawat1†, Durgesh Dubey1, Chunni Lal Khetrapal1, Paul Bacon3, Ramnath Misra2*, and Dinesh Kumar1* 1

Centre of Biomedical Research and 2Department of Immunology, SGPGIMS Campus Lucknow-226014, India and 3Rheumatology Research Group, Division of Immunity and Infection, Birmingham University, UK *

Authors for Correspondence:

Dr. Anupam Guleria Email: [email protected] Centre of Biomedical Research (CBMR), SGPGIMS Campus, Raebareli Road, Lucknow-226014 Uttar Pradesh, India Mobile: +91-9918004592 Dr. Dinesh Kumar Email: [email protected] Centre of Biomedical Research (CBMR), SGPGIMS Campus, Raebareli Road, Lucknow-226014 Uttar Pradesh, India Mobile: +91-8953261506 Prof. R N Misra Email: [email protected] Dean and Professor and Head Clinical Immunology, SGPGIMS, Lucknow 226 014 Uttar Pradesh, India Phone +915222494284 †

Both the authors contributed equally

ABBREVIATIONS: NMR, Nuclear Magnetic Resonance; HSQC, Heteronuclear Single Quantum Correlation; TA, Takayasu Arteritis; PCA, Principal Component Analysis; PLS-DA, Partial Least Squares Discriminant Analysis; CPMG, Carr Purcell Meiboom Gill sequence; ESM, Electronic Supplementary Material

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 23

Abstract Takayasu Arteritis (TA) is a debilitating, systemic disease which involves the aorta and large arteries in a chronic inflammatory process leading to vessel stenosis. Initially, the disease remains clinically silent (or remains undetected) until the patients present with vascular occlusion. Therefore, new methods for appropriate and timely diagnosis of TA cases are needed to start proper therapy on time and also to monitor the patient’s response to the given treatment. In this context, NMR-based serum metabolomic profiling has been explored in this proof of principle study for the first time to determine characteristic metabolites that could potentially be helpful for diagnosis and prognosis of TA. Serum metabolic profiling of TA patients (n = 29) and healthy controls (n = 30) was performed using 1D 1H NMR spectroscopy and possible biomarker metabolites were identified. Using projection to least squares discriminant analysis, we could distinguish TA patients from healthy controls. Compared to healthy controls, the TA patients had (a) increased serum levels of choline metabolites, LDL cholesterol, N-acetyl glycoproteins (NAGs), and glucose and (b) decreased serum levels of lactate, lipids, HDL cholesterol, glucogenic amino acids. The results of this study are preliminary and need to be confirmed in a prospective study.

KEYWORDS: NMR based metabolomics; Takayasu Arteritis; Large Vessel Vasculitis; Serum metabolites.


Page 3 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction Takayasu Arteritis (TA) is a rare but serious form of large vessel vasculitis and adversely affects the quality of life.1-6 It involves the aorta and large arteries in a chronic inflammatory process leading to wall thickening and vessel stenosis.5,7-9 Initially the disease may be clinically silent or may remain undetected until the patients present with a complete occlusion of at least one vessel.10,11 This contrasts with the overt systemic illness seen in small vessel vasculitis such as anti-neutrophil cytoplasmic antibody (ANCA)-Associated Vasculitis (AAV)12 suggesting that TA not only involves different arteries than AAV but appears to be a very different form of vascular inflammation with different disease mechanism. However, there is a considerable lack of understanding of TA disease mechanism owing to the inability to biopsy large arteries. Basic understanding of this process is required but the immediate clinical challenge is the early appropriate diagnosis of TA as no specific auto-antibodies have been identified for the disease so far and even non-specific tests for inflammation such as acute-phase response often do not help diagnosis. Therefore, early stage molecular markers of sub-clinical processes are needed to allow initiation of therapy to control the disease, avoid further progression, and alleviate complications. The pertinent recent advance in this area is PET (Positron Emission Tomography) scanning, using fluorine-18-labeled 2-fluoro-2-deoxy-D-glucose.13-16 However, the procedure can be time-consuming and costly. The use of high radiation dose would further prohibit its routine clinical utility to scan possible cases. Therefore, cost-effective and less invasive methods for diagnosis, such as determination of clinical markers from serum, would be of significant advantage and potentially useful as an adjunct to conventional diagnostics for primary diagnosis, surveillance, and early detection of relapses. Metabolomics -a systematic approach focusing on the quantitative analysis of endogenous metabolites17-20- may significantly help addressing this issue i.e. understanding multiplex molecular events regulating the onset and progression of the disease. Quantitative analysis of metabolites -the final downstream product of the whole bio-system- helps identifying metabolic perturbations in biofluids and the metabolites significantly altered in pathological conditions can be used as possible clinical markers for diagnosis and treatment monitoring. There are several methods used for metabolomics studies and one of them is proton nuclear magnetic resonance (1H-NMR) which has extensively been used in metabolomics studies of biofluids over the past several years.19,21-28 1H NMR is especially suitable for metabolite analysis as (a) it allows quantitative profiling of a wide range of endogenous metabolites with minimal sample preparation, (b) offers the potential for high throughput (>100 samples/day is attainable) and (c) provides highly reproducible results.29,30 PET based molecular imaging with 18F fluorine-labelled deoxy-glucose revealed widespread uptake in the inner aortic wall, indicating a local metabolic abnormality.16 However, to date, there is no report studying the metabolic profiles of biofluids of TA patients to identify the disease related metabolic changes. In this preliminary study, we therefore investigated the metabolic profiles of sera derived from Takayasu Arteritis patients using NMR with an aim to assess (a) whether NMR based serum metabolomics would allow early identification of TA patients and (b) whether metabolic differences in TA patients are related to the risk of ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 23

TA progression. If this proof-of-principle study was successful, then further large-scale validation studies with larger patient cohorts would be justified. Further, if confirmed, metabolic patterns would provide valuable insights into the underlying biochemical processes and aid the understanding of the pathophysiology of this disease.

Material & Methods 1.1 Takayasu Arteritis patients and control subjects The study protocol was approved by the Institutional Research Ethics Committee, SGPGIMS, Lucknow, India. The serum samples used in this study were obtained from TA patients (fulfilling 1990 American College of Rheumatology classification criteria31) attending the Department of Clinical Immunology at SGPGIMS, Lucknow. A total of twenty nine (n=29) patients with Takayasu Arteritis (24 females, 5 males, mean age 32.9 ± 12.5 years) and thirty healthy control subjects (22 females, 8 males, mean age 35.5 ± 8.34 years) were enrolled in this study. Out of total 29 TA patients, 12 patients had active disease evaluated by the Indian Takayasu’s Arteritis Clinical Activity Score (ITAS). This is a clinical activity score devised and validated in Indian TA patients, scoring features which are new or worse in the past 3 months in the following domains: systemic, abdomen, genitourinary, renal (systolic and diastolic hypertension), nervous system and cardiovascular system. Features of diastolic hypertension, stroke, new pulse loss, bruits, pulse inequality, claudication and carotodynia are weighted to reflect a higher score. A maximum score of 51 is possible. ITAS score of 4 or more is considered clinically active.31,32 Samples used in this study were initially collected for other pathophysiological studies on TA disease with informed patient consent and stored with permission. These samples were later used for metabolomics studies. The clinical features of TA patients and use of prednisone and/or methotrexate by TA patients is tabulated in Table 1. Table 1: Patient characteristics data sheet

Age at time of diagnosis (years) Male/Female Disease activity (Active/Inactive) Disease Classification (I/IIA/IIB/III/IV/V) Methotrexate use (Yes/No) Prednisone use (Yes/No) ESR (Erythrocyte Sedimentation Rate) ITAS

Takayasu’s arteritis (n=29) 32.9±12.5

Healthy control (n=30) 35.5±8.34

5/24 12/17

8/22 --

8/1/2/-/-/18

--

10/19 13/16 58.3±28.1

----

4.55±5.72

--

Note: Values are expressed in mean ± standard deviation; ITAS is Indian Takayasu’s Arteritis Clinical Activity Score.


Page 5 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


1.2 Sample collection and Preparation Prior to blood sample collection, informed consent was obtained from all TA patients and healthy volunteers. Studies have shown intra- and inter-day variance in the levels of some amino acids, probably due to the effects of diet and/or daily activity.33,34 Therefore, venous blood samples were obtained from all subjects at early morning, after overnight fast to minimize the effect of dietary factors and inter-individual variations in metabolomics data. Blood samples were kept in vacutainer tubes for 30 minutes at room temperature for clotting. Clotted blood samples were centrifuged at 13000 rpm for 10 minutes to separate out the supernatant-serum, which was then frozen and stored at a temperature of -80 C, until the NMR measurements were performed. At the time of NMR analysis, serum samples were thawed and homogenized using a vortex mixer. To minimize the variation in pH, 250 L of saline buffer solution (in 100% D2O, NaCl 0.9%, 50 mM sodium phosphate buffer and pH 7.4) was added to 200 L of serum. After centrifugation (8000 rpm, 5 min), 450 L of each sample was transferred to 5 mm NMR tubes (Wilmad Glass, USA). A sealed capillary tube containing the known concentration of 1 mM TSP (Sodium salt of 3-trimethylsilyl(2,2,3,3-d4)-propionic acid) and dissolved in deuterium oxide (D2O) was inserted separately in all the NMR tubes both for the purpose of locking and chemical shift referencing and to aid quantitative estimation of metabolites. Deuterium oxide (D2O) and sodium salt of trimethylsilylpropionic acid-d4 (TSP) used for NMR spectroscopy were purchased from Sigma-Aldrich (Rhode Island, USA).

1.3 NMR Measurements The NMR experiments were performed at 298 K on Bruker Avance III 800 MHz NMR spectrometer (equipped with Cryoprobe) with a 5 mm broad-band inverse probe-head and a Z-shielded gradient. One‐ dimensional 1H‐NMR spectra were recorded on all the serum samples using the Carr–Purcell–Meiboom–Gill (CPMG) pulse sequence (cpmgpr1d, standard Bruker pulse program) with pre-saturation of the water peak.35 The parameters used were as follows: spectral sweep width: 12 ppm; data points: 32 K; flip angle of radiofrequnecy pulse: 90°; total relaxation delay (RD): 5 sec; T2 filtering was obtained with an echo time of 200 μs repeated 300 times, resulting in a total duration of effective echo time of 60 ms; number of scans: 128; window function: exponential and line broadening: 0.3 Hz. All the spectra or FIDs (free induction decays) were processed using Topspin-2.1 (Bruker NMR data Processing Software) using standard Fourier Transformation (FT) procedure following manual phase and baseline-correction. Prior to FT, each FID was zero-filled to 4096 data points and a sine–bell apodisation function was applied. After FT, the chemical shifts were referenced internally to the lactate methyl signal at δ 1.311 ppm. 2D homonuclear (1H_1H TOCSY and J-resolved) and heteronuclear (1H_13C HSQC) spectra were also recorded for selected samples to aid spectral assignment. The details of various NMR parameters of these 2D homonuclear and heteronuclear experiments are given in the supporting information (Appendix I).



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 23

Chemical shifts of 1D 1H NMR spectra were identified, assigned and validated (a) using the freely available software MetaboMiner36 by comparing and matching the chemical shifts for various metabolites available in the software, (b) performing spiking experiments using standard chemicals (see supplementary material, Figure S1 and S2) and (c) also using other existing databases and literature reports.37-40 Identification was achieved if there was only one candidate in the database within the specified tolerances 0.02 ppm for 1H shifts and 0.5 ppm for 13C shifts- for an observed peak and its correlated shifts. 1.4 Statistical analysis The multivariate data analysis was performed on all the 1D 1H CPMG NMR spectra obtained for the diseased and control groups. The reduction of the 1D CPMG NMR data were carried out in the chemical shift region 0.5-8.5 ppm using Bruker AMIX software (Version 3.8.7, Bruker GmbH, Germany). The spectra were then binned into 0.01 ppm integrated spectral buckets. The chemical shift region 4.7-5.2 ppm was excluded from the analysis to elminite the residual signal of water and distorted region due to water suppression. The binned data generated from the AMIX software were used for statistical analysis. The data were obtained after mean centering and normalization which was performed by dividing each data point by the sum of all data points present in the sample to compensate for the differences in concentration of metabolites among individual serum samples. The data were scaled using unit variance in which identical weight was given to all variables. The resulting data matrices were then exported into Microsoft Office Excel 2010 and used for the multivariate analysis using Unscrambler X Software (Version 10.3, CAMO USA, Norway). First, the 1D binned NMR data matrix was subjected to chemometric unsupervised principal component analysis (PCA) to check the grouping trend and outliers. After intial overveiw of data using PCA analysis, the data were subjected to supervised partial least square-discriminant analysis (PLS-DA) to identify the marker metabolites accounted for differentiation of two groups. To check the validity of the model and avoid the overfitting of the PLS model, repeated 7-fold cross validation was used. The quality of the models was assessed by R2Y(cum) and Q2(cum) values, where R2 provides the model validity in the form of explained variance result and indicates goodness of fit, and Q2 gives the proportion of variance in the data predicted by the model and indicates accuracy of prediction, respectively. The reliability of the models were further rigorously validated by the permutation tests (n = 100) and CV-ANOVA (analysis of variance testing of cross-validated predictive residuals) tests which were performed to determine significant differences between groups in the PLS-DA models. The receiver operating characteristic (ROC) analysis was also carried out to verify the robustness of the PLS-DA model in discriminating the control and diseased cohorts. The area under the ROC curve gives an accurate assessment of discriminatory ability (0.5=no discrimination; 1=perfect discrimination). The marker metabolites were identified from the loading plots (for PLS-DA) along with the help of S-plot and the scores of variable importance on projection (VIPs). A number of variables were obtained that were responsible for the difference between Takayasu arteritis and control serum samples for VIP value>1.0. ACS Paragon Plus Environment

Page 7 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Further, univariate analysis was performed by applying the independent samples T-test to several metabolites of interest (identified by the multivariate analysis) using SPSS Statistics (Version 11.2, IBM) software. For the univariate analysis, we have used NMR peak area scaled with respect to TSP resonance. A 0.05 level of probability was used as the criterion for statistical significance. The boxplot representation was used to visualize the variation in the levels of significant metabolites in control and TA samples.

Results and Discussion 1

H NMR spectra of serum samples

The representative 1D 1H CPMG NMR spectra of serum samples obtained from a healthy control and Takayasu Arteritis (TA) patient are shown in Figure 1 along with the resonances assigned to various metabolites. Assignment of metabolites was achieved using Metabominer36 and other existing databases and literature reports37-40 and confirmed by 2D 1H–1H TOCSY and 1H–13C HSQC spectra. The 1D 1H CPMG spectra of serum samples showed signals mainly from

lipids/lipoprotein fractions (e.g. HDL/LDL,

VLDL/triglycerides/UFAs etc.), glucose, amino acids, energy metabolism related molecules, ketone bodies (e.g. acetone), choline metabolites, N-acetyl-glycoproteins, and N-acetyl containing metabolites. Visual inspection of the spectra revealed lower levels of glucose and various glucogenic amino acids (i.e. which can be converted into glucose through gluconeogenesis) and higher levels of N-acetyl glycoproteins (NAGs) in the TA patients. Further extraction of metabolites associated with the disease was performed on the NMR data recorded using multivariate data analysis.

Pattern recognition analysis to access the metabolic variations between HC and TA groups: The NMR spectra recorded on 29 TA and 30 HC serum samples were subjected to multivariate data analysis to extract TA-induced changes in the serum metabolic profiles and identify the potential metabolic pathways disturbed in TA disease. PCA score plots were constructed initially for an overview of the dataset and to visualize the outliers of all samples. No significant outlier was observed in the dataset and the TA and control groups were not well separated in the PCA score plot. Therefore, to further investigate the associations of metabolites or metabolic pathways with TA disease, PLS-DA models (supervised analysis) were constructed using NMR data as the X-matrix and group information as the Y-matrix. The PLS-DA score plots showed clear separation between the healthy controls and TA patients Figure 2A. The goodness of fit and predictability for the PLS-DA models are assessed by the values of R2X (cum), R2Y (cum) and Q2 (cum) and are listed in Table 2. These values indicated that the model possessed a satisfactory fit with good predictive power. Following this, a random permutation test with 100 permutations was performed with the derived PLS-DA model to further evaluate the robustness of this method. The crossvalidation plot shown in Figure 2B reveals that the permutation testing, performed on the PLS-DA model exhibits significantly positive slopes and both the permuted R2 and Q2 values on the left were significantly ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 23

lower than the corresponding original points on the right. Also the Q2 regression line has a negative intercept (intercepts: R2 = (0.0, 0.303); Q2 = (0.0, -0.29)), which indicates that the separation model was statistically sound, and its high predictability was not due to random or over fitting of the data. Furthermore, the analysis of variance CV-ANOVA test was performed to examine the statistical significance of the differences between the two groups in the PLS-DA model (Table 2), this resulted in a score of p = 1.15 × 10-16, indicating that the differences between the groups within the model were highly significant. Further, in order to evaluate and rule out the metabolic effect of medications, PLS-DA score plot was generated from 1H NMR spectra of serum samples obtained from the TA patients with medications and without medications. The resulted PLSDA score plot as shown in the Supplementary material Figure S3 clearly revealed that there is no significant difference between groups of TA patients receiving medications and not receiving medications. Table 2: The partial least-squares-discriminant analysis (PLS-DA) model obtained from 1D CPMG NMRbased analysis of serum samples. Comparison

Control vs TA

NMR spectra

1D CPMG

R2X (cum)

R2Y (cum)

Q2 (cum)

0.562

0.864

0.763

CV-ANOVA, pvalue

Number of latent variables

1.15 × 10-16

3

Identification of metabolic markers and their quantitative analysis The present NMR-based metabolomics study aims to assess the potential of the technique as a diagnostic tool for the disease and to identify characteristic metabolites clearly differentiating the serum of TA patients from healthy controls. Figure 3 illustrates the loading plot of metabolites between healthy control and TA patients corresponding to PLS-DA score plot shown in Figure 2A. Metabolite variation could be visualized by the loading plots color-coded according to the absolute value of correlation coefficients (|r|) derived from the significant metabolites contributing to the separation in PLS-DA model, where a hot-colored signal (red) indicated more significant contribution to the class separation than a cold-colored one (blue). The upward and downward peaks in the loading plot, respectively, correspond to the elevated and depleted metabolites in the sera of TA patients. Analysis of spectra using PLS-DA is prone to over-fitting. However, permutation testing of serum PLS-DA models (Figure 2B) revealed them to be significantly different (p1), s-plots (variables with higher p and p(corr) values) along with the color-coded coefficient loading plot allowed the identification of the serum metabolites that had the greatest influence on cluster formation among the groups in the PLS-DA discriminatory model. Further independent samples t-test was performed on the above selected differential spectral bins to identify significant metabolic markers. A p value 1.0) in the PLS-DA model and the T-test (p < 0.05), a total of seventeen metabolites were identified as potential biomarkers for TA disease. These metabolites along with their chemical shifts, ACS Paragon Plus Environment

Page 9 of 23

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


VIP score and p-value are given in Table 3. The PLS-DA analysis was also carried out after removing the lipid resonances from the NMR spectra and the obtained score plot demonstrates that even the low molecular weight metabolites was able to discriminate between HC and TA groups as shown in Supplementary material, Figure S4. Further for the verification of the obtained potential biomarkers metabolites, the characteristic spectral regions of these metabolites in the 1D 1H NMR spectra were quantified based on the integration area of the identified metabolites relative to the area of the reference signal from TSP (See Supporting Information, Annexure II). The PLS-DA analysis was again repeated using the quantified metabolite data and comparable results to those mentioned in the multivariate statistical approach section were obtained. The representative box plots of potential metabolic biomarkers of TA disease (drawn from this analysis) with their relative concentrations are shown in Figure 4. To mention here is that the quantification of serum metabolites has been done using T2-edited CPMG spectra, where lipid and protein signals are attenuated. Therefore the concentrations of metabolites in serum are not absolute, but comparable between spectra. In the box plots, the boxes denote interquartile ranges, lines denote the median, and whiskers denote the 25 th and 75th percentiles. Latent and vital biomarkers were discovered and scrutinized with this procedure. The relative integral values of the significant metabolites (median and range) have also been presented in the Supplementary material, Table S1. Overall, as compared with healthy controls, a number of metabolites showed increased concentration in serum of TA patients, such as LDL, acetone, choline, glycerophosphocholine, N-acetyl glycoproteins, glucose, glycerol, while serum levels of several others metabolites such as lactate, LDL/VLDL lipids, amino acids (including valine, alanine, arginine, glutamine, proline, and 1-methylhistidine), were found to be decreased in the TA patients. Further analysis of these potential biomarkers was performed using the AUROC curve analysis. All the seventeen metabolite biomarkers are significantly different between TA and HC with AUC greater than 0.8 as enlisted in table 3. Figure 5 shows the Receiver’s operating characteristic (ROC) curves of the individual potential metabolite markers showing the diagnostic accuracy based on discriminant predicted probability. Two metabolites NAC1 and choline with an AUC of 0.98 and 0.95, respectively, has the highest potential of being the serum biomarkers for TA patients. Both metabolites were significantly up-regulated in the TA patients as compared to the healthy controls. Table 3: Chemical shift, multiplicity, resonance assignments, AUC of ROC, VIP and p-values of the individual biomarkers obtained from the PLS-DA model. The statistical significance for various metabolites was determined by independent samples t-test. Integral area of metabolites resonance was used for statistical comparison among groups. p-values less than 0.05 were considered as significant. No.

Metabolites (change in TA)

Chemical shift (multiplicity)

Assignment

AUC of ROC

VIP value

p value

1

L0 (HDL) []

0.78-0.81 (br)

CH3(CH2)n

0.86

2.20

NMR-Based Serum Metabolomics Discriminates Takayasu Arteritis

Recommend Documents