Expansion of HLA-DR⁺ Peripheral Helper T and Naïve B cells in Anti-Citrullinated Protein Antibody-Positive Individuals At-Risk for Rheumatoid Arthritis

Abstract

Objectives

To investigate immune dysregulation in the peripheral blood that contributes to the pre-rheumatoid arthritis (RA) stage of RA development in anti-citrullinated protein antibody (ACPA)+ individuals.

Methods

Using 37 markers by mass cytometry, we investigated peripheral blood mononuclear cells (PBMCs) from ACPA+ at-risk individuals (ARI), ACPA+ early untreated RA patients, and ACPA− controls in the Tokyo Women’s Medical University (TWMU) cohort (n = 17 in each group). Computational algorithms, FlowSOM and Optimized t-Distributed Stochastic Neighbor Embedding (opt-SNE), were employed to explore specific immunological differences between study groups. These findings were further evaluated, and longitudinal changes were explored, using flow cytometry and PBMCs from the USA-based Targeting Immune Responses for Prevention of RA (TIP-RA) cohort that included 11 ACPA+ individuals who later developed RA (pre-RA), of which 9 had post-RA diagnosis PBMCs (post-RA), and 11 ACPA− controls.

Results

HLA-DR⁺ peripheral helper T (Tph) cells, activated regulatory T cells, PD-1^hi CD8⁺ T cells, and CXCR5⁻CD11c⁻CD38⁺ naïve B cells were significantly expanded in PBMCs from ARI and early RA patients from the TWMU cohort. Expansion of HLA-DR⁺ Tph cells and CXCR5⁻CD11c⁻CD38⁺ naïve B cells was likewise found in both pre-RA and post-RA time points in the TIP-RA cohort.

Conclusions

The expansion of HLA-DR⁺ Tph cells and CXCR5⁻CD11c⁻CD38⁺ naïve B cells in ACPA+ individuals, including those who developed inflammatory arthritis and classified RA, supports a key role of these cells in transition from pre-RA to classified RA. These findings may identify a new mechanistic target for treatment and prevention in RA.

Graphical Abstract

Introduction

Production of highly mutated anti-citrullinated protein antibodies (ACPA) following T and B cell interactions is the hallmark of ACPA+ rheumatoid arthritis (RA). ACPA production is well established to precede the onset of inflammatory arthritis (IA) and classified RA by several years^{1 2}. The time period during which ACPA are present in the peripheral blood without detectable IA can be termed pre-RA; furthermore, one can consider individuals who are ACPA+ without IA as being in an ‘at-risk’ state³. Importantly, there are now several prevention trials implemented in ACPA+ individuals without IA, although to date no highly effective preventive intervention has been identified^4–8. As such, understanding the immunologic features associated with the initial generation of ACPA in the evolution of RA, as well as the features associated with the transition from an ACPA+ ‘at-risk’ state to classified RA, is critical to understanding the overall pathogenesis of RA, improving models to predict future RA, and ultimately to developing novel preventive approaches.

Analysis of blood B cells from RA patients identified presence of ACPA⁺ B cells in both antigen-inexperienced and affinity-matured compartments, and also demonstrated extensive somatic hypermutation and class-switching in ACPA⁺ B cells⁹. Additionally, extensive glycosylation of the IgG-ACPA variable region in RA and pre-RA subjects was identified^{10 11}. These findings imply that there is a loss of tolerance already present in naïve B cells from individuals with classified RA and that maturation of pathogenic ACPA⁺ B cells involves repeated germinal center reactions with T cell-dependent responses¹². Notably, these chronic, typically subclinical processes could potentially occur in the pre-RA stage in ectopic lymphoid structures present at sites of mucosal inflammation¹³ and potentially early joint inflammation.

Recent mass cytometry analyses have newly identified pathologically expanded peripheral helper T (Tph) cells defined as PD-1^hiCXCR5⁻ memory T cells, in the synovium and peripheral blood of seropositive (RF or ACPA) RA patients ¹⁴. The majority of Tph cells in the synovium co-expressed activation markers, HLA-DR and ICOS¹⁵, and also expressed factors enabling B cell help and antibody production. Furthermore, Tph cells had unique expression of chemokine receptors, distinct from PD-1^hiCXCR5⁺ T follicular helper (Tfh) cells, that directs migration of Tph cells to inflamed non-lymphoid sites. Additionally, Tph cells in the joint of RA patients were autoreactive¹⁶ and clonally expanded¹⁷, although specificity to citrullinated antigens was not investigated. While Tph cells are expanded in the peripheral blood and synovium of seropositive RA patients, it remains unknown if they are also expanded in individuals who are in the pre-RA stage.

In this study, we used a rigorous approach that included two independent and ethnically distinct cohorts to investigate the cellular characteristics of peripheral immune cells, including Tph cells, and expression of functional proteins to identify patterns associated with serum ACPA positivity as well as the development of classified RA. These findings add novel insights to understanding the breadth of immune dysregulations that contribute to the pre-RA stage of RA development.

Patients and methods

Study participants

Subjects for the mass cytometry analysis were identified in the Tokyo Women’s Medical University (TWMU) at-risk cohort. Serum anti-CCP+ (CCP2, Abbott) at-risk individuals (ARI) referred to the TWMU clinic with joint and musculoskeletal symptoms but no clinical arthritis, newly diagnosed anti-CCP+ RA patients with at least 1 swollen joint consistent with active arthritis and no glucocorticoids or DMARDs use meeting the 2010 ACR/EULAR RA classification criteria¹⁸, and anti-CCP− control subjects without RA were included (Supplementary Figure 1 and Supplementary Table 1). Peripheral blood mononuclear cells (PBMCs) were collected at the study visit.

We also evaluated PBMCs from the Targeting Immune Responses for Prevention of Rheumatoid Arthritis (TIP-RA) cohort conducted at the University of Colorado that longitudinally followed serum anti-CCP+ (CCP3, Werfen) individuals. Participants for this study included 11 anti-CCP+ individuals who subsequently developed IA and RA (and all met 2010 RA classification criteria¹⁸), and 11 anti-CCP− control subjects without RA (Supplementary Tables 2 and 3). Additional details of the methods of the clinical cohort studies are presented in Supplementary Methods. All study procedures were approved by the review board at TWMU and for TIP-RA by the Colorado Multiple Institutional Review Board. Written informed consent was obtained from all subjects.

Mass-cytometry and flow cytometry analysis

For the TWMU cohort, cryopreserved PBMCs were barcoded, stained, and analyzed with an anchor sample by mass cytometry as described in the previous study¹⁹. Data were normalized²⁰, batch adjusted using anchor sample²¹, and analyzed by two computational algorithms, FlowSOM²² and opt-SNE²³, followed by biaxial gating. From this result, we created T- and B-cell focused panels to explore significantly different cell populations identified in TWMU cohort, and PBMCs from TIP-RA cohort were analyzed by flow cytometry with the panels. Detailed methods and panels are described in Supplementary Methods and Supplementary Tables 4–6.

Results

Study participant demographics and immunophenotypic characterization of peripheral lymphocytes in ARI, RA, and controls

To identify immune subsets associated with ACPA+ pre-RA stage as well as classified RA, we cross-sectionally studied PBMCs from the TWMU cohort, which includes ACPA+ ARI, ACPA+ newly diagnosed RA patients, and ACPA− controls (n = 17 in each group, Supplementary Figure 1). Subject demographics are described in Supplementary Table 1. Age, sex, smoking history, and FDR status were similar between groups. Of the 17 ARI, the majority were also positive for anti-CCP3 (n = 16) and RF (n = 15)^{24 25}. During 1-year clinical follow-up, 4 ARI developed IA and were diagnosed with classified RA by 2010 ACR/EULAR RA classification criteria at 132 days in median (range, 14 to 341 days) after blood collection. Subject demographics of those who developed RA in the TWMU cohort and a comparison to pre-RA subjects from the TIP-RA cohort are described in Supplementary Table 3.

To investigate the frequency and functional states of immune cell subsets and to investigate differences between groups, we applied two computational algorithms, dimensionality reduction using t-SNE and unsupervised clustering by FlowSOM, to CD45⁺ cells including markers to identify lymphoid and myeloid cells and to evaluate their functional markers. T-SNE plot allows to visually identify major peripheral immune subsets and expression of their functional markers (Supplementary Figures 2A and B). FlowSOM builds a self-organizing map, a neural network used for clustering, creates a two-dimensional grid of nodes (clusters), and organizes the clusters into selected numbers of metaclusters (MCs). FlowSOM clustering generated 20 MCs, and significantly different populations were identified in T cells (MC6 and MC13) and B cells (MC17) but not in other populations (Supplementary Figure 2C). MC6, CD3⁺CD4⁺CD45RO⁺PD-1⁺ICOS⁺CXCR5⁻ cells in expression heatmap (Supplementary Figure 2D), was consistent with Tph cells and significantly increased in ARI and RA compared to controls. MC13 (CD3⁺TIGIT⁺) and MC17 (CD19⁺CD27⁺CD38⁺⁺ plasmablasts) were significantly increased in RA but not in ARI compared to controls. Other baseline immune subsets were comparable between the three study groups, and this was consistent with biaxial gating (Supplementary Figures 3A and 4A).

Expansion of HLA-DR⁺ Tph cells in ARI and RA

We further evaluated CD4⁺ T cells by applying FlowSOM and t-SNE to gated CD4⁺ T cells. We partitioned CD4⁺ T cells into 10 MCs containing 121 clusters with FlowSOM. MC5 and MC8 were significantly increased in ARI and RA compared to controls (Figure 1A). The heatmap of normalized expression showed very high PD-1 and ICOS expression, intermediate HLA-DR and TIGIT expression but no CXCR5 expression on MC8 (Figure 1B). MC8 was composed of two clusters (cluster A and B), which were mapped on distinct locations in the self-organizing map (Figure 1C), suggesting heterogeneous subpopulations. The abundance of cluster A was 1.6-fold and 1.8-fold higher in ARI and RA compared to controls, whereas cluster B was 3.0-fold higher in ARI and 4.1-fold higher in RA compared to controls (Figure 1D). We focused on cluster B based on its higher abundance in ARI and RA. The self-organizing map showed cluster B as PD-1^hiHLA-DR⁺ICOS⁺TIGIT⁺CXCR5⁻ cells, consistent with Tph cells¹⁴. Circulating PD-1^hiCXCR5⁺ Tfh cells were not identified as a single MC, but contained in MC3 as a cluster C, which was not different across the three study groups (Figure 1D). We found significant correlation between anti-CCP2 titer and MC8 (Spearman’s r = 0.61, p = 0.01), and cluster B (r = 0.69, p = 0.003) in ARI.

Figure 1. Identification of expanded HLA-DR⁺ Tph cells and ICOS^hiHLA-DR⁺ memory Tregs in blood of ACPA+ at-risk individuals and early RA patients.

Mass cytometry data gated on CD4⁺ T cells from Tokyo Women’s Medical University cohort were analyzed with FlowSOM and t-SNE. A, Abundance of FlowSOM metaclusters of CD4⁺ T cells. B, Heatmap of normalized expression of mass cytometry markers in each metacluster. C, FlowSOM maps illustrating the expression of indicated mass cytometry markers in individual clusters. Each node represents an individual cluster. Clusters with the same-colored ring belong in the same metacluster. D, Abundance of the FlowSOM clusters as depicted in C including 4 clusters: cluster A, cluster B, cluster C, and cluster D (% of total CD4⁺ T cells). E, Visualization of CD4⁺ T cells by t-SNE, showing HLA-DR⁻ Tph cells (green circle), HLA-DR⁺ Tph cells (red circle), Tfh cells (purple circle), and memory Treg cells (black circle) corresponding to cluster A, B, C, D in Figure D, respectively. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by Kruskal-Wallis test and Dunn’s multiple comparisons test. Error bars show medianwith interquartile range. ARI, at-risk individuals; MC, metacluster; ns, not significant; Tfh, follicular helper T; Tph, peripheral helper T; Treg, regulatory T; t-SNE, t-distributed stochastic neighbor embedding.

T-SNE clustering of CD4⁺ T cells clearly visualized PD-1 expression on memory CD4⁺ T cells (Figure 1E). PD-1^hiCXCR5⁻ Tph cells, separated from PD-1^hi CXCR5⁺ Tfh cells, were divided into two distinctive subpopulations based on HLA-DR expression. CD25^hiCD127⁻ regulatory T cells (Tregs) were aggregated in a separate region, indicating that this PD-1^hi region does not include Tregs.

Consistent with the two computational analyses, quantification by biaxial gating confirmed that the frequency and absolute number of PD-1^hiCXCR5⁻ Tph cells were significantly increased 1.6-fold in ARI and 2.5-fold in RA compared to controls (Figures 2A and B, Supplementary Figure 4B). In concordance with FlowSOM analysis, HLA-DR⁺ Tph cells were significantly increased 1.8-fold in ARI and 3.2-fold in RA compared to controls, but HLA-DR⁻ Tph cells and PD-1^hiCXCR5⁺ Tfh cells were significantly increased in RA but not in ARI compared to controls (Figures 2B and C), suggesting the primary importance of HLA-DR⁺ Tph cells during the very early development of RA.

Figure 2. Expanded HLA-DR⁺ Tph cells and ICOS^hiHLA-DR⁺ memory Tregs in blood of ACPA+ individuals and early RA patients by biaxial gating.

Mass cytometry data from Tokyo Women’s Medical University cohort were analyzed with biaxial gating. A, Representative gating for Tph cells, Tfh cells, HLA-DR⁺ Tph cells, and HLA-DR⁻ Tph cells. B, Frequencies of total Tph cells, HLA-DR⁺ Tph cells, and HLA-DR⁻ Tph cells. C, Frequency of Tfh cells. D–F, Correlation of HLA-DR+ Tph cell (red) or HLA-DR− Tph cell (blue) frequency with Anti-CCP2 titers (D) and with rheumatoid factor titers (E) in ARI and DAS28-CRP (F) in RA by Spearman’s correlation test. G, Frequencies of HLA-DR⁺ Tph cells and HLA-DR⁻ Tph cells in ARI who developed RA within one year after PBMC collection (progressor, n=4) and those who did not develop RA within that period (non-progressor, n=13). H, Representative gating for ICOS^hiHLA-DR⁺ memory Treg cells in memory CD4⁺ T cells from a control (top) and an at-risk individual (bottom). I, Frequency of ICOS^hiHLA-DR⁺ memory Treg cells. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by Kruskal-Wallis test and Dunn’s multiple comparisons test. Error bars show medianwith interquartile range. RF, rheumatoid factor. See Figure 1 for other definitions.

We next analyzed associations between the frequencies of Tph subsets and subject demographics. The frequency of HLA-DR⁺ Tph cells was not associated with age, sex, smoking history, or FDR status in ARI and RA (p > 0.05) (Supplementary Table 7). In ARI, the frequency of HLA-DR⁺ Tph cells moderately correlated with anti-CCP2 antibody titer (Spearman’s r = 0.63, p = 0.008) and non-significantly correlated with RF titer (r = 0.42, p = 0.09), whereas no correlation was found between the frequency of HLA-DR⁻ Tph cells and these two antibodies (Figures 2D and E). In RA, the frequency of HLA-DR⁺ Tph cells and HLA-DR⁻ Tph cells moderately correlated with DAS28-CRP in RA (HLA-DR⁺ Tph cells: r = 0.51, p = 0.04, HLA-DR⁻ Tph cells: r = 0.52, p = 0.04, Figure 2F). We also analyzed associations between these subsets and disease progression; progressors (n = 4), who developed classified RA within one year after sample collection, and non-progressors (n = 13), who did not develop RA within one year of sample collection. Notably, the frequency of HLA-DR⁺ Tph cells was 2.0-fold higher in progressors compared to non-progressors (p = 0.03), whereas no difference was observed in HLA-DR⁻ Tph cells (Figure 2G).

ICOS^hiHLA-DR⁺ memory Tregs were increased in ARI and RA

Besides Tph cells, FlowSOM analysis of CD4⁺ T cells identified MC5 as a significantly increased population, 1.9-fold in ARI and RA compared to controls (Figure 1A). The heatmap of normalized expression shows CD45RO⁺CD27⁺CD25^hiCD127⁻TIGIT⁺CD39⁺ cells, which is characteristic of memory Tregs (memTreg, Figure 1B). MC5 was composed of three clusters, mapped closely in the self-organizing map (Figure 1C). Within MC5, cluster D was increased 1.9-fold in ARI and RA compared to controls, although the difference between ARI and controls barely missed statistical significance (Figure 1D), and no difference was observed in the other two clusters. The self-organizing map shows HLA-DR⁺ICOS⁺TIGIT⁺CD39⁺ cells as cluster D, and T-SNE plots visualized high ICOS, HLA-DR, and TIGIT expression on CD25^hiCD127⁻ memTreg subsets (Figure 1E). We found significant correlation between anti-CCP2 titer and MC5 (Spearman’s r = 0.60, p = 0.01), and cluster D (r = 0.60, p = 0.01) in ARI. Biaxial gating also confirmed ICOS^hiHLA-DR⁺ memTregs in the frequency and absolute number was significantly increased in ARI and RA compared to controls (Figures 2H and I, Supplementary Figure 4C). The frequency of ICOS^hiHLA-DR⁺ memTregs moderately correlated with anti-CCP2 titer in ARI (r = 0.65, p = 0.005) but was not associated with disease progression. The frequency of ICOS^hiHLA-DR⁺ memTregs did not correlate with DAS28-CRP in RA (r = 0.29, p = 0.26).

Characteristics of peripherally expanded HLA-DR⁺ Tph cells

We further investigated the cellular characteristics of peripheral HLA-DR⁺ Tph cells expanded in the TWMU ARI and RA using our mass cytometry data. Representative plots for each of the markers are shown in Supplementary Figure 5A. In agreement with published data in RA¹⁴, Tph cells expressed significantly higher HLA-DR, ICOS, TIGIT, T-bet, CD38, CD57, and significantly lower CD25 and CD127 compared to the PD-1^neg memory CD4⁺ T cells in each study group (Supplementary Figure 5B). Next, we compared expression of these markers on Tph cells between the three study groups (Figure 3A). Expression of ICOS, TIGIT, and CD38 was significantly higher on Tph cells from ARI compared to controls (Figures 3B–D), whereas expression of CD57 was significantly lower (Figure 3E). HLA-DR expression on Tph cells was modestly increased in ARI compared to controls (Figure 3F). Expression of T-bet and CD27 was not different across the three study groups (Supplementary Figure 5C). The expression level of PD-1 was significantly increased on Tph and Tfh cells in ARI and RA compared to controls (Figure 3G).

Figure 3. Protein expression features of Tph cells in ACPA+ individuals.

The expression levels of indicated proteins were compared by mean metal intensity in mass cytometry data from Tokyo Women’s Medical University cohort. A, Representative gating for PD-1^negative population, PD-1^intermediate population, Tph cells, and Tfh cells in memory CD4⁺ T cells. B–F, Comparison of indicated protein expression on Tph cells (red) and on PD-1⁻ memory CD4⁺ T cells (blue) between study groups. G, Comparison of PD-1 expression on PD-1⁻ memory CD4⁺ T cells (blue), PD-1^intermediate memory CD4⁺ T cells (orange), Tph cells (red), and Tfh cells between study groups. H, Expression of indicated proteins on HLA-DR⁻ Tph cells (blue) and on HLA-DR⁺ Tph cells (red). Colored bars indicate comparison between study groups by Kruskal-Wallis with Dunn’s multiple comparisons test in B–H. Black bars indicate comparison between HLA-DR⁻ Tph cells and HLA-DR⁺ Tph cells from an indicated study group by Mann-Whitney test in H. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Error bars show medianwith interquartile range. See Figure 1 for definitions.

We further investigated cellular characteristics specifically expressed on HLA-DR⁺ Tph cells and HLA-DR⁻ Tph cells. Expression of ICOS, TIGIT, T-bet, CD38, and CD57, which were highly expressed on Tph cells, were also significantly higher on HLA-DR⁺ Tph cells compared to HLA-DR⁻ Tph cells in each study group (Figure 3H). Furthermore, within HLA-DR⁺ Tph cells, expression of PD-1 and ICOS was only significantly elevated in ARI compared to controls (Figure 3H), suggesting highly activated functional states of HLA-DR⁺ Tph cells in ARI.

PD-1^hiICOS^hi CD8⁺ T cells were increased in ARI and RA

Since the t-SNE plot of CD45⁺ lymphocytes indicated high expression of PD-1 on CD8⁺ T cells as well as CD4⁺ T cells (Supplementary Figures 2A and B), we analyzed gated CD8⁺ T cells using FlowSOM, t-SNE, and biaxial gating and found the frequency of PD-1^hiICOS^hiCD45RO⁺CD27⁺CD8⁺ T cells (PD-1^hiICOS^hi memCD8⁺ T cells) were significantly increased 2.2-fold in ARI and 2.6-fold in RA compared to controls (Supplementary Figures 4D and 6A–E). The subset of PD-1^hiICOS^hi memCD8⁺ T cells expressed HLA-DR at levels similar to Tph cells, and the frequency of HLA-DR⁺PD-1^hiICOS^hi memCD8⁺ T cells was also significantly increased in ARI and RA compared controls (Supplementary Figure 6F). Similar to Tph cells, we found significantly increased PD-1 expression on PD-1^hi CD8⁺ T cells from ARI and RA compared to controls (Supplementary Figure 6G). We did not find any association between PD-1^hiICOS^hi memCD8⁺ T cells and subject demographics or disease progression.

Expansion of CXCR5⁻CD11c⁻CD38⁺ Naïve B cells in ARI and RA

Tph cells can promote T cell-dependent B cell maturation and production of antibodies ^{14 26}; as such, we next investigated the B cell compartment. To analyze CD19⁺ B cells, we used the same strategy as T cell analysis. We partitioned CD19⁺ B cells into 15 MCs containing 121 clusters with FlowSOM. We also performed detailed quantification by biaxial gating and subsets were determined based on previous report^{27 28}. FlowSOM analysis identified MC13 (CD27⁻IgM⁺CD38⁺CXCR5⁻) as the only subset that was significantly increased in ARI and RA compared to controls (Figures 4A and B). The self-organizing map showed that MC13 consists of two clusters discriminated by IgD expression (Figure 4C). Cluster A (CD27⁻IgD⁺CXCR5⁻CD11c⁻CD38⁺ was a subpopulation of CD27⁻IgD⁺ naïve B (NAV) cells and was significantly increased in ARI and RA compared to controls (Figure 4D). Cluster B (CD27⁻IgD⁻CXCR5⁻CD11c⁻CD38⁺) was a CD38 expressing subpopulation of CD27⁻IgD⁻ double negative B (DN) cells, i.e., CD38⁺ DN3 cells, and was significantly increased in RA but not in ARI. Quantification by biaxial gating confirmed increases in these two populations in frequency and absolute number (Figures 4E–G, Supplementary Figure 4E). Visualization with t-SNE confirmed these two clusters (Supplementary Figure 7A). These findings were in accordance with high expression of CD38 and IgM and low expression of T-bet on CXCR5⁻CD11c⁻ NAV and DN3 cells (Figure 4H). The abundance of MC13 and the frequency of CXCR5⁻CD11c⁻CD38⁺ NAV cells was not associated with age, sex, smoking history, FDR status, anti-CCP2 antibody titer, or RF titer in ARI and RA (p > 0.05). There was a non-significant trend toward higher frequency of CXCR5⁻CD11c⁻CD38⁺ NAV cells in progressors compared to non-progressors (2.6-fold, p = 0.10, Figure 4I). The frequency of CXCR5⁻CD11c⁻CD38⁺ NAV cells moderately correlated with DAS28-CRP in RA (r = 0.55, p = 0.02).

Figure 4. Expansion of CXCR5⁻CD11c⁻CD38⁺ naïve B cells in blood of ACPA+ individuals.

Mass cytometry data gated on CD19⁺ B cells from Tokyo Women’s Medical University cohort were analyzed. A, Abundance of FlowSOM metaclusters of B cells. B, Heatmap of normalized expression of mass cytometry markers in each metacluster. C, FlowSOM maps illustrating the expression of indicated mass cytometry markers in individual clusters. Each node represents an individual cluster. Clusters with the same-colored ring belong in the same metacluster. D, Abundance of the FlowSOM clusters as depicted in C (cluster A and B). E, Representative gating for B cell subsets. Based on CXCR5 and CD11c expression, CD27⁻IgD⁻ B cells (double negative; DN cells) are separated into DN1, DN2, and DN3 subsets, and CD27⁻IgD⁺ B cells (naïve; NAV) are separated into resting NAV (rNAV), activated NAV (aNAV), and CXCR5⁻CD11c⁻ NAV. CXCR5⁻CD11c⁻ NAV is further subdivided based on CD38 expression. F, Frequency of CXCR5⁻CD11c⁻CD38⁺ NAV. G, Frequency of CD38⁺ DN3. H, CD38 expression of naïve B cell subsets and DN cell subsets. I, Frequency of CXCR5⁻CD11c⁻CD38⁺ NAV between in progressor (n=4) and non-progressor (n=13). *P<0.05, **P<0.01, ***P<0.001 by Kruskal-Wallis test and Dunn’s multiple comparisons test. Error bars show medianwith interquartile range.

We also identified increases in the following two naïve B cell subsets in RA compared to controls: CXCR5⁺CD11c⁻CD38⁺IgM^hi NAV (MC14, IgM^hi resting NAV) and CXCR5⁻CD11c⁻CD38⁻ NAV cells (MC11) (Supplementary Figures 7B and C). In addition, we found decreases in DN1 (CD27⁻IgD⁻CXCR5⁺CD11c⁻ B cells) and the following two memory B cell subsets in RA compared to controls: switched (IgM⁻) memory B cells (MC1) and IgM memory B cells (MC3) (Figures 4A and B, Supplementary Figure 7D).

FlowSOM analyses of CD45⁺ cells and B cells suggested plasmablasts were significantly increased in RA (Figures 4A and B, Supplementary Figures 2C and D). Biaxial gating confirmed that IgG⁺ plasmablasts but not IgA⁺ plasmablasts were increased in RA compared to ARI and controls (Supplementary Figure 7E).

Evaluation of HLA-DR⁺ Tph cells and CXCR5⁻CD11c⁻CD38⁺ NAV cells during RA development in TIP-RA cohort

To further evaluate the expansion of T cell and B cell subsets in ARI and RA identified in the cross-sectional analysis in the TWMU cohort and to explore longitudinal changes in this cell population, we performed cross-sectional and longitudinal analyses in an independent USA-based pre-RA cohort (TIP-RA). We used flow cytometry to analyze cryopreserved PBMCs from 11 ACPA+ pre-RA subjects, of which 9 also had a post-RA diagnosis sample, and 11 samples from age- and sex-matched ACPA− controls who had samples from a single time point (Supplementary Figure 1). At post-RA blood sampling, all subjects had active arthritis and met 2010 ACR/EULAR criteria for RA. Eight samples were collected on the day of RA diagnosis before starting glucocorticoids or DMARDs and 1 sample was collected at 45 days after diagnosis and on prednisone and methotrexate. The median duration between pre-RA and post-RA visits was 325 days (range; 186 to 694 days). The gating strategy is shown in Supplementary Figure 3B.

Consistent with our findings in the TWMU cohort, the frequency of HLA-DR⁺ Tph cells was significantly increased in pre-RA samples compared to control samples (1.6-fold, Figure 5A), while no significant difference was observed in the frequencies of total Tph cells and HLA-DR⁻ Tph cells between pre-RA and controls (Figures 5B and C). In addition, the frequency of HLA-DR⁺ Tph cells was consistently high at pre- and post-RA visits (0.6% vs 0.7%, Figure 5D), without a significant increase in frequency from pre- to post-RA (p = 0.57) but remaining elevated above control frequencies. Additionally, the higher expression of PD-1 on Tph and Tfh cells from the pre-RA samples was also observed (Figure 5E). A significant increase in the frequency of ICOS^hiHLA-DR⁺ memTregs was also found in pre-RA subjects compared to controls (Figure 5F), and no significant change was observed between pre- and post-RA visits (0.4% vs 0.4%. Figure 5G). PD-1^hiICOS^hiHLA-DR⁺ memCD8⁺ T cells were not elevated in pre-RA samples (Supplementary Figures 6H and I).

Figure 5. Cross-sectional and longitudinal analysis of T cell and B cell subsets in TIP-RA cohort.

A–C, Frequencies of HLA-DR⁺ Tph cells (A), total Tph cells (B), and HLA-DR⁻ Tph cells (C) in memory CD4⁺ T cells. D, Longitudinal change in frequency of HLA-DR⁺ Tph cells. E, Comparison of PD-1 expression on PD-1⁻ memory CD4⁺ T cells (blue), PD-1^intermediate memory CD4⁺ T cells (orange), Tph cells (red), and Tfh cells between Pre-RA and controls. F, Frequency of ICOS^hiHLA-DR⁺ memory Treg cells. G, Longitudinal change in frequency of ICOS^hiHLA-DR⁺ memory Treg cells. H, Frequency of CXCR5⁻CD11c⁻CD38⁺ naïve B cells. I, Longitudinal change in frequency of CXCR5⁻CD11c⁻CD38⁺ naïve B cells. J, Correlation of HLA-DR⁺ Tph cell (red) or HLA-DR⁻ Tph cell (blue) frequency with CXCR5⁻CD11c⁻CD38⁺ naïve B cell frequency in pre-RA samples by Spearman’s correlation test. K, Spearman correlation matrix of T- and B-cell subsets in pre-RA samples. Comparison between HC and Pre-RA was performed by Mann-Whitney test in A–C, E, F, and H, and comparison between Pre-RA and Post-RA was performed by Wilcoxon matched-pairs signed rank test in D, G, and I. *P<0.05, **P<0.01, ***P<0.001. Error bars show medianwith interquartile range. See Figure 1 for definitions.

In B cell analyses, the frequency of CXCR5⁻CD11c⁻CD38⁺ NAV cell was significantly increased in pre-RA subjects compared to controls (2.0-fold, Figure 5H), and the frequencies did not significantly change between pre- and post-RA samples (0.6% vs 0.4%, Figure 5I). The frequency of CXCR5⁻CD11c⁻CD38⁺ NAV cells strongly correlated with that of HLA-DR⁺ Tph cells in pre-RA samples (Spearman’s r = 0.72, p = 0.02; Figures 5J and K). The summary of findings is shown in Table 1. The correlation between the frequencies of expanded cell subsets and clinical and demographic variables is shown in Supplementary Table 7, and the correlation of T and B cell subsets is shown in Supplementary Figure 8 for the TWMU cohort and in Figure 5K for the TIP-RA cohort.

Table 1.

Summary of significantly altered cell populations in ACPA-positive at-risk individuals, pre-RA subjects and RA patients compared to controls in each cohort

Cohort	TWMU		TIP-RA
	At-Risk (n = 17)	RA (n = 17)	Pre-RA (n = 11)
CD4+ T cells	↑ HLA-DR+ Tph ↑ ICOShiHLA-DR+ memTreg	↑ HLA-DR+ Tph ↑ HLA-DR− Tph ↑ Tfh ↑ ICOShiHLA-DR+ memTreg	↑ HLA-DR+ Tph ↑ ICOShiHLA-DR+ memTreg
CD8+ T cells	↑ PD-1hiICOShi memCD8+	↑ PD-1hiICOShi memCD8+
B cells	↑ CXCR5−CD11c−CD38+ NAV	↑ CXCR5−CD11c−CD38+ NAV ↑ CXCR5−CD11c−CD38− NAV ↑ CXCR5+CD11c−CD38+IgMhi NAV ↑ CXCR5−CD11c−CD38+ DN (CD38+ DN3) ↑ IgG+ plasmablast ↓ Switched memory B ↓ IgM memory B	↑ CXCR5−CD11c−CD38+ NAV

DN, CD27⁻IgD⁻double negative B cells; memCD8⁺, memory CD8⁺ T cells; memTreg, memory regulatory T cells; NAV, CD27⁻IgD⁺ naïve B cells; Tfh, follicular helper T cells; Tph, peripheral helper T cells; TWMU, Tokyo Women’s Medical University; TIP-RA, Targeting Immune Responses for Prevention of Rheumatoid Arthritis

Discussion

Our findings demonstrate an expansion of peripheral PD1^hiCXCR5⁻HLA-DR⁺ Tph cells and CXCR5⁻CD11c⁻CD38⁺ NAV cells in ACPA+ individuals, and importantly in a subset of ACPA+ individuals who later develop RA, as well as in individuals with classified RA. While expansion of Tph cells has been reported in individuals with seropositive RA in the blood^{14 29 30}, this is the first report of these cells also being expanded in ACPA+ ARI. Our longitudinal study revealed persistent elevation of HLA-DR⁺ Tph cells and CXCR5⁻CD11c⁻CD38⁺ NAV cells in the transition from pre-RA to classified RA. Notably, our study included two independent ACPA+ pre-RA cohorts derived from two ethnically different populations, making our findings particularly robust and strongly implicating HLA-DR⁺ Tph cells and CXCR5⁻CD11c⁻CD38⁺ NAV cells as critical cell subsets involved in the transition from ACPA+ at-risk state to classified RA.

We identified expansion of HLA-DR⁺ Tph cells in ARI and pre-RA subjects but no further increase was observed in the longitudinal sampling from TIP-RA. However, there was a higher frequency of HLA-DR⁺ Tph cells in the progressors compared non-progressors in TWMU cohort. Because of this, we believe that the presence of a high frequency of HLA-DR⁺ Tph cells is an indicator of a subpopulation of ACPA+ individuals who are at high-risk for future RA. This will need to further validated but could allow for addition of HLA-DR⁺ Tph frequency to ACPA and RF testing to predict future disease.

CD4⁺ T cells have long been associated with RA pathogenesis^{31 32} and subset abnormalities were identified in pre-RA stage³³. Prior studies have identified expansion of circulating Tph cells in various autoimmune diseases related to autoantibody production, such as systemic lupus erythematosus ^{26 34}, Sjogren’s syndrome³⁵ and type 1 diabetes (T1D) ³⁶. Circulating Tph cells were shown to possess B cell help function^{26 37}. Importantly, circulating Tph cells were increased in children with newly diagnosed T1D and also in multiple autoantibody-positive at-risk children who later progressed to T1D ³⁶. This observation, with our current findings, supports a pathogenic role of Tph cells in acquisition of autoantibodies and transition to disease. Additionally, we found increased HLA-DR⁺ as well as HLA-DR⁻ Tph cells and their associations with disease activity in our RA subjects. However, it was specifically HLA-DR⁺ Tph cells that were increased in ARI and pre-RA subjects. This expression of HLA-DR in addition to ICOS on Tph cells expanded in ARI and pre-RA demonstrates a high and recent immune activation³⁸ that might reflect ongoing extra-follicular B cell development during the pre-RA state.

Tph cells have a distinct ability to migrate and promote B cell differentiation and antibody production in inflamed non-lymphoid tissues³⁹. We also found increased CXCR5⁻CD11c⁻CD38⁺ NAV cells in our ARI and pre-RA subjects, and this subset significantly correlated with HLA-DR⁺ Tph cells in our pre-RA subjects. Notably, both CXCR5⁺ and CXCR5⁻ naïve B cells with high CD38 expression were increased in RA subjects but only the CXCR5⁻ subset was increased in ARI. Although our study lacked CD10 or CD24 staining, and transitional B cells were not clearly gated, high CD38 expression on expanded CXCR5⁻ naïve B cell subsets may suggest expansion of transitional B cells, which were reported to be autoreactive⁴⁰ and immune-regulatory⁴¹, in the pre-RA stage. The lack of CXCR5 expression supports an extrafollicular association of these cells (i.e., origin of or homing to). This finding is of particular interest because multiple data support a Mucosal Origins Hypothesis that autoimmunity in RA originates at a mucosal site and prior to demonstratable joint inflammation (i.e., a pre-articular stage of RA development)¹³. While future studies are needed, perhaps accumulation of these subsets at mucosal sites, their subsequent presence in the peripheral blood, and further propagation to the joint are key features linking mucosal inflammation and RA development.

High PD-1 expression is a key feature of Tph cells, but its function is not well elucidated. Even within PD-1^hi Tph cells, we demonstrated a higher expression of PD-1 on Tph cells in ARI and RA subjects compared to controls, suggesting that in addition to an expansion of Tph cells in ARI, pre-RA, and RA, there may be unique features of Tph cells as PD-1 expression increases. Furthermore, associated with cancer immunotherapy, immune checkpoint inhibitors (ICI), including anti-PD-1 therapy, can induce IA and classifiable RA, and arthritis can persist after ICI cessation⁴², suggesting a causal relationship between PD-1 regulation and development of arthritis.

We identified expansion of ICOS^hiHLA-DR⁺ memTregs in ARI, RA and pre-RA. The frequency of ICOS^hiHLA-DR⁺ memTregs correlated with anti-CCP2 titer in ARI but did not associate with disease progression or correlated with DAS28-CRP. A previous study reported peripheral expansion of ICOS⁺ memTregs in early RA patients⁴³. In a study of pre-clinical stage of T1D, glucocorticoid-induced TNFR-related protein (GITR)⁺HLA-DR⁺ memTregs were expanded⁴⁴, and suppressive capacity was impaired in Tregs sorted from slow progressors of T1D. These studies suggest a pathogenic role of ICOS^hiHLA-DR⁺ memTregs in autoimmune diseases.

Limitations of our study include the relatively small number of individuals included in each cohort and the lack of functional assays. However, validation in an independent and ethnically distinct cohort robustly supports the relevance of our findings. In addition, plasmablasts may not be captured well in cryopreserved PBMCs, and this could be the reason that an elevation of IgA plasmablasts in ARI ⁴⁵ was not replicated in the current study. Since this is an exploratory study, multiple regression analysis was not performed. Further study is needed to determine the pathogenic roles of expanded T and B cell subsets in the pre-RA stage and significance of these subsets for prediction of imminent RA development. Finally, we do not know the antigen specificity of the expanded T and B cell subsets, which will require additional studies using appropriate tetramers and other strategies.

In summary, we found an expansion of HLA-DR⁺ Tph cells and CXCR5⁻CD11c⁻CD38⁺ NAV cells in ACPA+ ARI across two distinct study populations, including maintenance of elevations of these cells after transitions to ACPA+ RA, supporting a key role for these cells in the pathogenesis of very early RA evolution.