Clinical and biological characteristics of RA patients and controls
Study population
The characteristics of RA patients are detailed in Table 1. 211 RA patients were included (164 females, 78%, mean age 59.2±13 years). The mean disease duration of RA was 14.3±12.6 years. 80% had positive rheumatoid factor, 83% had positive anti-CCP antibodies and 57% had erosions on hand/foot x-rays. The median (IQR) DAS28-CRP was 3.8 (2.4–4.7). Among the patients, 118 patients (56%) received corticosteroid and 124 (59%) methotrexate. 92 (44%) received a targeted biologic therapy: 20 received an anti-TNF (5 Infliximab, 5 Adalimumab, 9 Etanercept, 1 Certolizumab), 11 received anti-IL6R (10 Tocilizumab and 1 Sarilumab), 14 received CTLA4-Ig (Abatacept), 14 JAK Inhibitors (6 Upadacitinib, 4 Baricitinib, 4 Tofacitinib), 33 anti-CD20 (Rituximab).
Regarding differences between the 3 RA subset, the 59 patients with refractory active RA had significantly longer disease duration (18 versus 11 years, p < 0.001) and were more likely to exhibit structural damage (73% versus 45%, p < 0.001) and use of targeted therapies (78% versus 21%, p < 0.001) (Table 1) compared to the 85 non-refractory active patients.
The CRP level was also higher in refractory active RA compared to non-refractory active RA (median: 9 mg/L vs. 5.8 mg/L), but this difference did not reach statistical significance (p = 0.09).
The mean age of the 29 healthy donors was 50.1±10 and 59.2±13 years, and they were all women.
Circulating markers in RA patients and controls
As expected, RA patients exhibited a distinct angiogenic and inflammatory profile compared to controls (Supplementary Table 2), featuring a predominance of pro-angiogenic (increase in VCAM-1, PlGF, Endostatin concentrations and decrease in IL-8 concentration, Supplementary Fig. 1a-d) and pro-inflammatory markers (increase in IL-6, IFN-γ, TNF-α, IL-4, IL610 and MMP1, concentrations, Supplementary Fig. 1e-j).
The anti-TNF treatment did not appear to affect serum TNF measurements (median concentration 10.61 pg/mL in the 20 patients treated with anti-TNF versus 14.43 pg/mL in patients without anti-TNF treatment, p = 0.24). In contrast, as previously described24, anti-IL6R treatment significantly affected serum IL-6 measurements, with increased IL-6 levels observed in anti-IL6R-treated patients compared to untreated patients (median 80 vs. 7.4 pg/mL, p < 0.001).
Circulating biomarkers concentrations according treatment resistance and disease activity
Identification of specific circulating markers of refractory active RA
In univariate analysis, no individual biomarker was able to discriminate refractory active RA from non-refractory active RA (Table 2).
Sensitivity analyses, including comparisons based on CRP levels (26 refractory active RA patients with CRP > 10 mg/L versus 28 active non-refractory RA patients with CRP > 10 mg/L), presence of Doppler ultrasound synovitis (31 refractory active RA patients with Doppler ultrasound synovitis versus 60 active non-refractory patients with Doppler ultrasound synovitis), and alternative definitions of refractory RA (failure of ≥ 2 lines of targeted therapies with different mechanisms of action or failure of ≥ 3 targeted therapies) did not lead to the identification of a specific biomarker.
Identification of specific circulating markers of active RA
We next compared the serum concentrations of 144 active RA patients to 67 patients with non-active disease. Although there was a trend for increased values of CXCL4, IL-4 and MMP-1 (Table 3; Fig. 1a–c), IL-6 was the single cytokine with a striking 3.3-fold increase in active RA compared to non-active disease (Table 3; Fig. 1d). This increase was observed both in refractory and non-refractory RA (Fig. 1e, suggesting that this elevation was not related to treatment resistance. Moreover, IL-6 levels correlated with disease activity indicators (Supplementary Fig. 2a), including the number of tender joints, the number of swollen joints, CRP, ESR, DAS28-CRP and DAS28-ESR (Supplementary Fig. 2b-g). Excluding the 11 patients receiving anti-IL6 therapy, which affects serum IL-6 measurement, did not alter these results for IL-6.
We also performed subgroup analyses according to therapeutic regimen, specifically for patients on MTX and those on targeted therapies. No cytokines showed significant differences in the MTX-treated subgroup. However, for patients on targeted therapies, those with active RA—who may be considered non-responders to these treatments—had higher IL-6 levels 25 pg/mL in non-responders vs. 5.2 pg/mL, p = 0.0003, Fig. 1f) and lower BAFF levels (862 pg/mL in non-responders vs. 1180 pg/mL, p = 0.0009, Fig. 1g) compared to patients with non-active RA, who therefore responded to the targeted therapy.
Angiogenic and inflammatory marker concentrations in rheumatoid arthritis patients subgroups. Concentrations (pg/mL) of CXCL4 (a), IL-4 (b), MMP-1 (c) and IL-6 (d) in active RA compared to non-active RA. Concentrations of IL-6 (e) in refractory active, non-refractory active RA and non-active RA. Concentration of IL-6 (f) and BAFF (g) in active RA and non-active RA under targeted therapies. Test: non-parametric two-tailed Mann-Whitney test (A–D) or Kruskall-Wallis test (E). IL: Interleukine; MMP-1: Matrix Metalloproteinase-1; RA: Rheumatoid Arthritis.
Correlations between circulating markers and disease activity indicators
While no clear individual biomarker profile of refractory RA emerged, refractory and non-refractory active RA patients exhibited marked differences regarding their correlation profile with disease activity parameters. Correlograms were conducted to examine the relationships between angiogenic and inflammatory markers with RA disease characteristics (Fig. 2).
Serum signature in rheumatoid arthritis subgroups. Correlograms of angiogenic and inflammatory markers and clinico-biological data in refractory active RA (n = 59) (a), non-refractory active RA (n = 85) (b) and non-active RA (n = 67) (c). Only significant Spearman’s correlation coefficients are represented by colour intensity and square area. Red: positive correlation. Blue: negative correlation. Larger squares indicate stronger correlations, while smaller squares indicate weaker correlations. ALAT: Alanine Aminotransferase; ASAT: Aspartate Aminotransferase; BAFF: B-cell activating factor; BMI: Body Mass Index; CRP: C-Reactive Protein; CXCL4/PF4: Platelet Factor 4; DAS28: Disease Activity Score 28; ESR: Erythrocyte Sedimentation Rate; GGT: Gamma-glutamyl transferase; HDL-C: High-density lipoprotein-cholesterol; IFN: Interferon; LDL-C: Low-density lipoprotein-cholesterol; RF: Rheumatoid factor; IL: Interleukine; MMP-1: Matrix Metalloproteinase-1; PlGF: Placenta Growth Factor; RA: Rheumatoid Arthritis; TNF: Tumor Necrosis Factor; VAS: Visual Analogue Scale; VCAM-1: Vascular Cell Adhesion Protein 1; VEGF: Vascular Endothelial Growth Factor.
Refractory active RA patients exhibited a limited correlation profile (Fig. 2a), with only few correlations with disease activity markers. Within this subgroup, we only observed positive correlations between CRP and two biomarkers (Neuropilin-1 and VEGF) (Fig. 2a and Supplementary Fig. 3a-b), as well as between DAS28 and two biomarkers (VEGF and IL-6) (Fig. 2a and Supplementary Fig. 3e-g).
In contrast, a more extensive correlation profile was detected in non-refractory active RA, with greater correlations between circulating markers and disease activity markers (Fig. 2b). We observed positive correlations between CRP levels and 10 circulating biomarkers (Angiopoietin-1, CXCL4, Neuropilin-1, PlGF, VEGF, IFN-γ, IL-4, IL-6, MMP-1, TNF-α) (Fig. 2b and Supplementary Fig. 4a-i), and between the DAS28 or the DAS28-CRP and 8 circulating biomarkers (CXCL4, IL-8, Neuropilin-1, IL-2, IL-4, IL-6, MMP-1, TNF-α) (Fig. 2b and Supplementary Fig. 4j-q).
The serum profile of non-active RA revealed correlations between CRP and 6 biomarkers (Angiopoietin-1, CXCL4, VCAM-1, BAFF, MMP-1, IL-2) (Fig. 2c and Supplementary Fig. 5a-f). There was also a negative correlation between DAS28 CRP and IL-10 (Supplementary Fig. 5).
