The impact of inflammation on metabolomic profiles in patients with arthritis

Abstract

Objective: Inflammatory arthritis is associated with systemic manifestations including alterations in metabolism. We used nuclear magnetic resonance (NMR) spectroscopy-based metabolomics to assess metabolic fingerprints in serum from patients with established rheumatoid arthritis (RA) and those with early arthritis.

Methods: Serum samples were collected from newly presenting patients with established RA who were naive for disease-modifying antirheumatic drugs, matched healthy controls, and 2 groups of patients with synovitis of ≤3 months' duration whose outcomes were determined at clinical followup. Serum metabolomic profiles were assessed using 1-dimensional (1) H-NMR spectroscopy. Discriminating metabolites were identified, and the relationships between metabolomic profiles and clinical variables including outcomes were examined.

Results: The serum metabolic fingerprint in established RA was clearly distinct from that of healthy controls. In early arthritis, we were able to stratify the patients according to the level of current inflammation, with C-reactive protein correlating with metabolic differences in 2 separate groups (P < 0.001). Lactate and lipids were important discriminators of inflammatory burden in both early arthritis patient groups. The sensitivities and specificities of models to predict the development of either RA or persistent arthritis in patients with early arthritis were low.

Conclusion: The metabolic fingerprint reflects inflammatory disease activity in patients with synovitis, demonstrating that underlying inflammatory processes drive significant changes in metabolism that can be measured in the peripheral blood. The identification of metabolic alterations may provide insights into disease mechanisms operating in patients with inflammatory arthritis.

Figure 1

Metabolic fingerprinting distinguishes between sera from patients with established rheumatoid arthritis (RA) and matched healthy controls and illustrates that metabolomic profiles in patients with early arthritis are altered by control of or resolution of inflammation. A and B, One-dimensional ¹H–nuclear magnetic resonance (NMR) spectra of serum obtained from disease-modifying antirheumatic drug–naive patients with RA (solid circles) and healthy controls (open circles) were subjected to principal components analysis (PCA) (A) and to supervised analysis (partial least-squares discriminant analysis) (B). C and D One-dimensional ¹H-NMR spectra of serum obtained from a subset of patients with early arthritis before (solid circles) and after (open circles) a decrease in the C-reactive protein level following therapy or spontaneous resolution were subjected to PCA (C) and to supervised analysis (partial least-squares discriminant analysis with orthogonal signal correction) (D). The values on the axis labels indicate the proportion of the variance captured by each principal component (PC1 and PC2) or latent variable (LV1 and LV2).

Figure 2

The metabolic fingerprints of sera from patients with early arthritis prior to treatment with disease-modifying antirheumatic drugs are strongly influenced by the level of inflammation. A, One-dimensional ¹H-NMR spectra of serum obtained from patients with very early arthritis (in group 1) were subjected to PCA. Solid circles indicate C-reactive protein (CRP) levels in the highest tertile, shaded circles indicate CRP levels in the middle tertile, and open circles indicate CRP levels in the lowest tertile. B and C, Strong correlations between measured CRP and predicted CRP values were found for patients with early arthritis in group 1 (B) and those in group 2 (C) (P < 0.001 for both groups). The predicted values were calculated from the concentrations of a series of metabolites that were discovered using partial least-squares regression analysis. Insets show the optimization of the multivariate regression, with the highest correlation between measured and predicted CRP occurring with 154 NMR bins (maximum R² of 0.671) for group 1 and with 1,136 NMR bins (maximum R² of 0.4157) for group 2. See Figure 1 for other definitions.

Figure 3

Metabolic fingerprints of sera from patients with early arthritis from 2 different patient groups. A and B, Serum samples from patients with early arthritis from group 1 (A) and group 2 (B) at first presentation were assessed using partial least-squares discriminant analysis to distinguish patients whose disease was resolving (solid circles) from those whose disease was persistent (open circles). Sensitivity and specificity were 59.4% and 58.9%, respectively, for group 1 and 59.5% and 56.4%, respectively, for group 2. C and D, Serum samples from patients with early arthritis from group 1 (C) and group 2 (D) at first presentation were assessed using partial least-squares discriminant analysis with orthogonal signal correction, to distinguish patients whose disease was resolving (solid circles) from those who developed persistent RA (open circles). Sensitivity and specificity were 50% and 69.6%, respectively, for group 1 and 73.1% and 67.6%, respectively, for group 2. See Figure 1 for definitions.