Metabolic profile changes in patients with rheumatoid arthritis detected using mass spectrometry

Abstract

Background: Rheumatoid arthritis (RA) presents as pain, swelling and leads to irreversible damage in joint, and adversely affects the quality of life of patients with RA. However, the etiology of RA is still unclear, and novel biomarkers are demanded for the early prediction and diagnosis of RA and dissecting disease mechanisms.

Objective: This study aimed at profiling the disordered metabolic pathways in RA and selecting potential biomarkers to distinguish RA patients from healthy individuals, and systematically investigated the associations between metabolites and the risk of RA.

Methods: A total of 533 participants, including 382 healthy individuals and 151 RA patients, were recruited to explore altered metabolic profiles through the analysis of dried blood spot samples by mass spectrometry. Multiple algorithms were applied to identify potential biomarkers. Dose-response relationships were investigated by binary logistic regression and restricted cubic spline (RCS) analysis.

Results: There were different metabolic profiles between RA and healthy individuals. After systematic selection, a metabolic panel consisting of C20, C5, Leu, C14:1/C16, Arg/(Orn + Cit), and C2/C0 was used to differentiate the two groups. Ten-fold cross-validation and test set were employed to evaluate prediction models. The receiver operating characteristic analysis demonstrated an area under the curve of 0.920(95%CI: 0.851-0.990) in test set to distinguish the two groups. The strong correlations between the 6 metabolites and RA were observed in RCS regression model.

Conclusions: The selected biomarkers have the potential to improve the detection of RA, and may offer insights into the intervention strategies to susceptible at-risk populations of developing RA.

Keywords: Biomarker; Dried blood spot; Mass spectrometry; Metabolomics; Rheumatoid arthritis.

Fig. 1

The research design and data analysis for this study. Abbreviation: HC, healthy control, RA, rheumatoid arthritis, MS, mass spectrometry, RF, random forest, ROC, receiver operating characteristic, RCS, restricted cubic spline.

Fig. 2

Score plots of PCA and PLS-DA analyses based on the 95 metabolites for RA and HC. (A) Unsupervised PCA score plot analysis. (B) Supervised PLS-DA score plot analysis. Green circle represents for HC group, and red hexagon for RA group. (C) A 200-times permutation test for validating the performance of PLS-DA model. Abbreviation: HC, healthy control, RA, rheumatoid arthritis.

Fig. 3

Altered metabolic profile between RA patients and healthy individuals. (A) The volcano plot of VIP versus log 2 of FC. Significant points with VIP > 1 and FC > 1.2 or <-1.2 were colored in red. (B) The volcano plot of minus log 10 of adjusted p-value versus log 2 of FC. Significant points with adjusted p-value < 0.05 and FC > 1.2 or <-1.2 were colored in red. (C) Venn diagram showed 19 significantly altered metabolites between patients with RA and healthy individuals. (D) Significance analysis of microarrays to determine differentially expressed metabolites between two classes. Red points indicate the metabolites that are significantly up-regulated, and green points indicate significantly down-regulated metabolites in RA. (E) Clustering heatmap analysis of the 19 selected metabolites. (F) Pathway enrichment analysis of differential metabolites between the two groups. Abbreviation: RA, rheumatoid arthritis, HC, healthy control, VIP, variable importance in projection, FC, fold change, Ala, Alanine, Arg, Arginine, Cit, Citrulline, Leu, Leucine, Lys, Lysine, Orn, Ornithine, Val, Valine, C0, Free carnitine, C2, Acetylcarnitine, C3, Propionylcarnitine, C4, Butyrylcarnitine, C5, Isovalerylcarnitine, C8, Octanoylcarnitine, C14:1, Tetradecenoylcarnitine, C16, Palmitoylcarnitine, C16-OH, 3-Hydroxypalmitoylcarnitine, C20, Arachidic carnitine.

Fig. 4

The identification of important metabolic biomarkers to distinguish RA patients from healthy individuals. (A) Pearson’s correlation analysis. (B) Important metabolites in random forest classifier. Abbreviation: MSE, mean square error; Ala, Alanine, Arg, Arginine, Cit, Citrulline, Leu, Leucine, Lys, Lysine, Orn, Ornithine, Val, Valine, C0, Free carnitine, C2, Acetylcarnitine, C3, Propionylcarnitine, C4, Butyrylcarnitine, C5, Isovalerylcarnitine, C8, Octanoylcarnitine, C14:1, Tetradecenoylcarnitine, C16, Palmitoylcarnitine, C16-OH, 3-Hydroxypalmitoylcarnitine, C20, Arachidic carnitine.

Fig. 5

The ROC curves towards 6 selected metabolites for distinguishing RA patients from healthy individuals (A) ROC curves for binary logistic regression. (B) Adjusted ROC curves for DBP and glycaemia. The curves marked with solid blue line for training set, green dot for 10-fold cross-validation, and red star for test set. Abbreviation: ROC, Receiver-operating characteristic, DBP, diastolic blood pressure, RA, rheumatoid arthritis.

Fig. 6

The dose-response relationships between 6 selected metabolites and the risk of RA by restricted cubic spline model. The solid blue lines represent the estimated dose-response curves, and solid red lines represent the model adjusted for DBP and glycaemia. Abbreviation: Arg, Arginine, Cit, Citrulline, Leu, Leucine, Orn, Ornithine, C0, Free carnitine, C2, Acetylcarnitine, C5, Isovalerylcarnitine, C14:1, Tetradecenoylcarnitine, C16, Palmitoylcarnitine, C20, Arachidic carnitine, RA, rheumatoid arthritis, DBP, diastolic blood pressure.

Fig. 7

Logistic regression analysis on the associations between important metabolites and RA. Square represents ORs obtained from univariate logistic regression model, and circle for odd ratios obtained from the model adjusted for DBP and glycaemia. Abbreviation: OR, odd ratio, RA, rheumatoid arthritis.