Serum C-X-C motif chemokine 13 is elevated in early and established rheumatoid arthritis and correlates with rheumatoid factor levels

Abstract

Introduction: We hypothesized that serum levels of C-X-C motif chemokine 13 (CXCL13), a B-cell chemokine, would delineate a subset of rheumatoid arthritis (RA) patients characterized by increased humoral immunity.

Methods: Serum from patients with established RA (the Dartmouth RA Cohort) was analyzed for CXCL13, rheumatoid factor (RF) levels, anticitrullinated peptide/protein antibody (ACPA) and total immunoglobulin G (IgG); other parameters were obtained by chart review. A confirmatory analysis was performed using samples from the Sherbrooke Early Undifferentiated PolyArthritis (EUPA) Cohort. The Wilcoxon rank-sum test, a t-test and Spearman's correlation analysis were utilized to determine relationships between variables.

Results: In both the Dartmouth and Sherbrooke cohorts, CXCL13 levels were selectively increased in seropositive relative to seronegative RA patients (P = 0.0002 and P < 0.0001 for the respective cohorts), with a strong correlation to both immunoglobulin M (IgM) and IgA RF levels (P < 0.0001). There was a weaker relationship to ACPA titers (P = 0.03 and P = 0.006, respectively) and total IgG (P = 0.02 and P = 0.14, respectively). No relationship was seen with regard to age, sex, shared epitope status or inclusion high-sensitivity C-reactive protein (hsCRP) in either cohort or regarding the presence of baseline erosions in the Sherbrooke Cohort, whereas a modest relationship with Disease Activity Score in 28 joints CRP (DAS28-CRP) was seen in the Dartmouth cohort but not the Sherbrooke cohort.

Conclusion: Using both established and early RA cohorts, marked elevations of serum CXCL13 levels resided nearly completely within the seropositive population. CXCL13 levels exhibited a strong relationship with RF, whereas the association with clinical parameters (age, sex, DAS28-CRP and erosions) or other serologic markers (ACPA and IgG) was either much weaker or absent. Elevated serum CXCL13 levels may identify a subset of seropositive RA patients whose disease is shaped by or responsive to RF production.

Figure 1

Scatterplots illustrate strong correlation of serum CXCL13 with seropositivity and immunoglobulin M rheumatoid factor in the Dartmouth RA Cohort. (A) Log-transformed C-X-C motif chemokine 13 (CXCL13) values are higher in seropositive compared to seronegative rheumatoid arthritis (RA) patients by t-test (seropositive (n = 163), geometric mean (95% CI) = 331.1 pg/ml (250.0 to 430.5); seronegative (n = 30), geometric mean (95% CI) 93.3 pg/ml (71.3 to 123.9); P = 0.0002). (B) The log-transformed CXCL13 values of seropositive patients increase with higher rheumatoid factor (RF) levels when evaluated by Spearman correlation (P < 0.0001). (C) Tertile analysis of seropositive patients comparing the highest CXCL13 tertile (third tertile) to the first and second tertiles by Wilcoxon rank-sum test identifies a strong relationship with immunoglobulin M (IgM) RF (third-tertile IgM RF: mean = 182 ± 59 IU/ml; first- and second-tertile IgM RF: mean = 113 ± 74 IU/ml; P < 0.0001). (D) CXCL13 has a weaker relationship to IgG anticitrullinated peptide/protein antibody (ACPA) (P = 0.03). *P < 0.05. Diagonal lines represent line of best fit.

Figure 2

Scatterplots show minimal or no relationship of serum CXCL13 to high-sensitivity C-reactive protein or disease activity score in 28 joints C-reactive protein in the seropositive groups of the Dartmouth RA Cohort and the Sherbrooke Early Undifferentiated PolyArthritis Cohort. (A) In the Dartmouth RA Cohort, correlation analysis of log-transformed C-X-C motif chemokine 13 (LogCXCL13) and high-sensitivity C-reactive protein (LogCRP) does not show a relationship (n = 123, P = 0.07). (B) In the Dartmouth RA Cohort, correlation analysis of CXCL13 has a modest relationship with Disease Activity Score in 28 joints CRP (DAS28-CRP) (n = 23, P = 0.01). (C) In the recent-onset rheumatoid arthritis (RA) patients from the Sherbrooke Early Undifferentiated PolyArthritis (EUPA) Cohort, no relationship is identified between LogCXCL13 and LogCRP (n = 173, P = 0.08). (D) In the recent-onset RA patients from the Sherbrooke EUPA Cohort, no relationship is identified between LogCXCL13 and DAS28-CRP (n = 170, P = 0.28). Diagonal lines represent lines of best fit.

Figure 3

Scatterplots demonstrate serum CXCL13 strongly correlates with seropositive rheumatoid arthritis and immunoglobulin M rheumatoid factor levels in the Sherbrooke Early Undifferentiated PolyArthritis Cohort, but has a weaker relationship to anticitrullinated peptide/protein antibody levels. (A) Log-transformed C-X-C motif chemokine 13 (LogCXCL13) is higher in seropositive than in seronegative RA patients as determined by t-test (seropositive: n = 173, geometric mean (95% CI) = 323.6 pg/ml (223.9 to 477.5); seronegative: n = 166, geometric mean (95% CI) = 50.1 pg/ml (35.0 to 78.0); P < 0.0001). (B) As with the Dartmouth RA Cohort, a strong relationship is seen with LogCXCL13 and immunoglobulin M rheumatoid factor (IgM RF) in seropositive samples, whether measured by Spearman correlation (P < 0.0001) (C) or by CXCL13 tertile analysis (third-tertile mean RF = 223 ± 57 IU/ml, first- and second-tertile mean RF = 141 ± 75 IU/ml; P < 0.0001). (D) A significant relationship is found between CXCL13 and anticitrullinated peptide/protein antibody (ACPA) by Spearman correlation (P = 0.006). *P < 0.05. Diagonal lines represent lines of best fit.

Figure 4

Scatterplots illustrate a strong relationship of serum CXCL13 with immunoglobulin A rheumatoid factor in seropositive rheumatoid arthritis patients in both the Dartmouth RA Cohort (established RA) and recent-onset rheumatoid arthritis patients from the Sherbrooke Early Undifferentiated PolyArthritis Cohort. (A) Pearson correlation analysis of log-transformed C-X-C motif chemokine 13 (LogCXCL13) values and log-transformed immunoglobulin A rheumatoid factor (Log IgA RF) in the Dartmouth RA Cohort shows a strong association (P < 0.0001). (B) CXCL13 tertile analysis confirms this relationship (third-tertile IgA RF: geometric mean (95% CI) = 45.0 IU/ml (28.0 to 72.3), first- and second-tertile IgA RF: geometric mean (95% CI) = 11.2 IU/ml (7.6 to 16.3); P < 0.0001). (C) In the Sherbrooke Early Undifferentiated PolyArthritis Cohort, the relationship between LogCXCL13 and Log IgA RF persists, whether measured by correlation (P < 0.0001) (D) or by CXCL13 tertile analysis (third-tertile IgA RF: geometric mean (95% CI) = 74.1 IU/ml (51.4 to 106.7), first- and second-tertile IgA RF: geometric mean (95% CI) = 20.4 IU/ml (15.1 to 27.5); P < 0.0001). *P < 0.05. Diagonal lines represent lines of best fit.