Synovial fluid CD69 + CD8 + T cells with tissue‐resident phenotype mediate perforin‐dependent citrullination in rheumatoid arthritis

Abstract

Objectives

Although the importance of tissue‐resident memory T (T_RM) cells in organ‐specific chronic inflammation has been recognised, little is known about their role in rheumatoid arthritis (RA). Here, we examined the characteristics of synovial fluid CD8⁺ T cells that express canonical T_RM markers CD69 and CD103, and their role in the pathogenesis of RA.

Methods

Synovial fluid mononuclear cells (SFMCs) were obtained from patients with RA. Flow cytometric analysis of surface markers and cytotoxic molecules of CD8⁺ T cells was performed. TCR repertoire of CD8⁺ T cells was analysed by TCRVβ CDR3 sequencing. Citrullination with the formation of neutrophil extracellular trap (NET) was evaluated by immunofluorescence staining.

Results

The frequency of CD8⁺ T cells was increased in SFMCs, and these CD8⁺ T cells were primarily comprised of CD45RA^‐ memory T cells expressing CD69 and/or CD103. CD69⁺CD8⁺ T cells exhibited T_RM phenotypes, including upregulation of CXCR6, CD49a and CD101, and downregulation of S1PR1 and KLF2. TCR repertoire analysis showed that these cells were an oligoclonally expanded population with increased expression of cytotoxic molecules. The treatment of neutrophils with supernatant from IL‐15‐stimulated CD69⁺CD8⁺ T cells induced perforin‐mediated histone citrullination and NET formation irrespective of their CD103 expression. The frequency of perforin‐expressing cells among CD69⁺CD8⁺ T cells in SFMCs was significantly higher in patients with anti‐citrullinated protein antibody (ACPA) than in those without ACPA.

Conclusion

CD69⁺CD8⁺ T cells in the SFMCs of RA patients exhibit T_RM‐like features. These cells may participate in the pathogenesis of RA via perforin‐mediated citrullination.

INTRODUCTION

Rheumatoid arthritis (RA) is a chronic autoimmune disease characterised by joint swelling and destruction.¹ The hallmark manifestations of RA include hyperplastic synovial tissue, immune cell infiltration and the presence of autoantibodies, such as anti‐citrullinated protein antibody (ACPA).² Among the various types of immune cells, CD4⁺ T cells have been the focus of many research efforts because of their capacity to produce pro‐inflammatory cytokines (e.g. IFN‐γ and IL‐17A)^3‐7 and their ability to help autoantibody‐producing B‐cell maturation.⁸ In contrast, the role of CD8⁺ T cells in the pathogenesis of RA remains unclear. Previous studies have reported contradicting results related to CD8⁺ T cells in RA patients, such as the presence of anti‐inflammatory IL‐10‐producing CD8⁺ suppressor T cells,⁹ CD8⁺ T cells expressing pro‐inflammatory IL‐6 and TNF‐α,¹⁰ and resistance to regulatory T (T_reg) cell‐mediated suppression by killing T_reg cells.¹¹

Memory T cells provide rapid and effective protection against previously encountered antigens. However, when memory T cells pathologically accumulate and function in response to innocuous targets, such as self‐antigens, these cells can contribute to the pathogenesis of autoimmune diseases. Recent studies have identified the presence of a new subset of memory CD8⁺ T cells that reside in peripheral tissues without recirculation, tissue‐resident memory T (T_RM) cells, which express the characteristic markers S1P₁ antagonist CD69^12,13 and/or adhesion protein CD103 (integrin α_E).^14,15 These cells are located in various anatomical sites, including the skin, lungs and liver,^16,17 where they provide in situ immunosurveillance and protective immunity with their ability to respond rapidly.^18,19 T_RM cells have been suggested to participate in the pathogenesis of autoimmune diseases.^20‐23 The number of CD103⁺CD8⁺ T cells secreting IFN‐γ is markedly increased in the epidermis of patients with vitiligo.²⁰ In addition, a recent study demonstrated that clonally expanded CD8⁺ T cells in the synovial fluid (SF) of patients with juvenile idiopathic arthritis exhibit a feature of T_RM cells, suggesting that they may contribute to the disease pathology.²⁴

Neutrophil extracellular trap (NET), a form of neutrophil cell death, is thought to be involved in the pathogenesis of RA by providing citrullinated autoantigens. This complex contains various intracellular proteins, including citrullinated histones, which can then influence the generation of ACPA²⁵ in RA patients. Furthermore, a recent study demonstrated that cytotoxic molecules released by CD8⁺ T cells, such as perforin, can induce the formation of citrullinated proteins as well as the extracellular release of DNA.²⁶ Therefore, cytotoxic CD8⁺ T cells may contribute to the pathogenesis of RA through perforin‐mediated production of citrullinated histones with NET formation, and subsequent generation of ACPA.²⁵

In the present study, we investigated the phenotypic, transcriptional and functional characteristics of CD69⁺CD103⁺CD8⁺ and CD69⁺CD103^‐CD8⁺ T cells in the SF of RA patients. We also investigated the role of CD69⁺CD8⁺ T cells in RA, especially related to formation of citrullinated histones.

RESULTS

Synovial fluid CD8⁺ T cells from RA patients are predominantly memory cells expressing CD69 and/or CD103

First, we analysed the frequency of CD8⁺ T cells among peripheral blood mononuclear cells (PBMCs) and synovial fluid mononuclear cells (SFMCs) from patients with RA. We also analysed their expression of CD45RA. CD8⁺ T cells were identified based on CD3 and CD8 expression and the lack of CD14, CD19, CD56 and TCRγδ expression (Supplementary figure 1). The frequency of CD8⁺ T cells in the lymphocyte gate was higher among SFMCs from RA patients than PBMCs from RA patients or healthy controls (Figure 1a). Most of the CD8⁺ T cells found among SFMCs were memory cells lacking CD45RA expression (Figure 1b). To evaluate tissue‐residency marker expression in CD8⁺ memory T cells, we examined the expression of CD69 and CD103 in CD45RA^‐CD8⁺ T cells. SF CD45RA^‐CD8⁺ T cells derived from RA patients were markedly enriched with CD69⁺CD103^‐ and CD69⁺CD103⁺ cells (Figure 1c). An analysis of the memory T‐cell subset (CD45RA^‐) within SF CD8⁺ T cells showed that the proportion of effector memory T cells was higher in the CD69⁺CD103^‐ and CD69⁺CD103⁺CD8⁺ T‐cell subsets than in the CD69^‐CD103^‐CD8⁺ T‐cell subset (Figure 1d). Histopathological examination confirmed the presence of CD69⁺CD103^+/‐CD8⁺ T cells in synovial tissues from RA patients and their absence from synovial tissues from OA patients (Figure 1e). These findings indicate that RA patients have an increased frequency of CD69⁺CD103^+/‐CD8⁺ T cells in their joints.

1 Figure. CD69+CD103+/‐CD45RA‐CD8+ T cells are increased in the synovial fluid of RA patients. (a) The frequency of CD8+ T cells within the lymphocyte population in peripheral mononuclear cells (PBMCs) from healthy controls (HCs; n = 9) and rheumatoid arthritis (RA) patients (n = 13) and in synovial fluid mononuclear cells (SFMCs) from RA patients (n = 60). (b) Left: Representative histograms of CD45RA expression in CD8+ T cells among PBMCs from HCs and RA patients and SFMCs from RA patients. Right: Dot plot graph depicting the frequency of CD45RA‐CD8+T cells among PBMCs from HCs (n = 9) and RA patients (n = 13) and SFMCs from RA patients (n = 60). (c) Left: Representative flow cytometry plots for CD69 and CD103 expression in CD45RA‐CD8+ T cells among PBMCs from HCs or RA patients and SFMCs from RA patients. Right: The frequencies of CD69+CD103‐CD45RA‐ and CD69+CD103+CD45RA‐CD8+ T cells among PBMCs from HCs (n = 9) and RA patients (n = 13) and SFMCs from RA patients (n = 60). (d) The proportions of CD45RA+CCR7+ (naive), CD45RA‐CCR7+ (central memory), CD45RA‐CCR7‐ (effector memory) and CD45RA+CCR7‐ (CD45RA+ effector memory) cells among CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+CD8+ T cells from the synovial fluid in patients with RA. (e) Representative images of the immunohistochemical staining in synovial tissues from patients with RA and osteoarthritis (OA); red‐coloured cells represent CD8, white represent CD69, and green represent CD103. White arrows indicate CD69+CD103+CD8+ cells, and red arrows indicate CD69+CD103‐CD8+ cells. Scale bars = 20μm. Statistical test: one‐way ANOVA with Tukey’s multiple comparisons test; **P‐value <0.01, ***P‐value <0.001, ****P‐value <0.0001. Data were obtained from a single experiment for each donor. HC, healthy control; OA, osteoarthritis; PBMC, peripheral blood mononuclear cells; RA, rheumatoid arthritis; SFMC, synovial fluid mononuclear cells.

CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells exhibit tissue‐resident features

Next, we examined the phenotypes of CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells among SFMCs from RA patients. The expression of proteins known to be upregulated in T_RM cells (e.g. C‐X‐C chemokine receptor 6 [CXCR6], CD49a and CD101) was significantly higher in CD69⁺CD103⁺ and CD69⁺CD103^‐ than in CD69^‐CD103^‐CD45RA^‐CD8⁺ cells (Figure 2a). The CD103⁺ subset exhibited significantly higher expression of CXCR6, CD49a and CD101 than the CD103^‐ subset (Figure 2a). In addition, the mRNA expression of the tissue egress protein S1PR1 and its transcription factor, KLF2, was significantly downregulated in CD69⁺CD103⁺ and CD69⁺CD103^‐ compared to CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells (Figure 2b). Furthermore, HLA‐DR and PD‐1 expression was significantly higher in CD69⁺CD103⁺ and CD69⁺CD103^‐ than in CD69^‐CD103^‐ cells, with CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells exhibiting the highest expression (Figure 2c). The expression of activation markers CD25 and CD38 was not significantly different between the CD69^‐CD103^‐ and CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T‐cell subsets (Supplementary figure 2). The frequency of Ki‐67⁺ cells was highest in the CD69⁺CD103⁺ T‐cell subset, followed by CD69⁺CD103^‐ and CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells (Figure 2d).

2 Figure. CD69+CD103+/‐CD45RA‐CD8+ T cells from the synovial fluid of patients with RA exhibit TRM‐like signatures. (a) Upper panel: Representative histograms depicting the expression of the indicated protein on CD69‐CD103‐ (grey), CD69+CD103‐ (red) and CD69+CD103+ (blue) CD45RA‐CD8+ T cells among SFMCs from RA patients. The grey‐shaded histogram represents the isotype control. Lower panel: Dot plot graphs indicate the frequency of CXCR6, CD49a and CD101 in the indicated subsets among CD45RA‐CD8+ T cells. (b) The mRNA expression level of S1PR1 and KLF2 from isolated CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells in RA SFMCs. Data are presented as mean+SD as data were distributed normally. (c–e) The frequency of cells expressing HLA‐DR and PD‐1 (c), Ki‐67 (d), T‐bet and Eomes (e) was determined and analysed in CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells from patients with RA. Statistical test: one‐way ANOVA with Tukey’s multiple comparisons test; *P‐value <0.05, **P‐value <0.01, ***P‐value <0.001, ****P‐value <0.0001. Data were obtained from a single experiment for each patient.

To further understand the differentiation profile of SF CD8⁺ T cells, we investigated the expression of transcription factors associated with T‐cell differentiation. The frequency of T‐bet⁺ cells was similar among the three subsets, whereas the frequency of Eomes⁺ cells was higher among CD69⁺CD103⁺ and CD69⁺CD103^‐CD45RA^‐CD8⁺ T cells (Figure 2e). Collectively, CD69⁺CD103^+/‐CD8⁺ memory T cells exhibit tissue‐resident phenotypes with effector‐like features.

CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T‐cell populations exhibit increased clonal expansion with less diversity

We also investigated whether CD69⁺CD103^+/‐CD8⁺ memory T cells exhibit higher clonal expansion compared to CD69^‐CD103^‐CD8⁺ memory T cells. To analyse the TCR repertoire of SF CD8⁺ T cells, TCRVβ CDR3 sequencing was performed with sorted CD45RA^‐CD8⁺ T‐cell populations. The diversity of CDR3 calculated using Simpson’s reciprocal index tended to be lower in the CD69⁺CD103⁺CD45RA^‐CD8⁺ T‐cell population than in the CD69^‐CD103^‐CD45RA^‐CD8⁺ T‐cell population. D50 values, which represent the percentage of unique T‐cell clones that account for 50% of the total CDR3 counted, exhibited significantly less diversity in both CD69⁺CD103⁺ and CD69⁺CD103^‐CD45RA^‐CD8⁺ T cells than in CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells (Figure 3a).Figure 3b is a representative dataset describing the three subsets from a single patient with RA. The top 10 clones accounted for 60–87% of the CD69⁺CD103⁺CD45RA^‐CD8⁺ T‐cell population and 32–59% of the CD69^‐CD103^‐CD45RA^‐CD8⁺ T‐cell population, indicating a higher degree of skewed cell expansion in the CD69⁺CD103⁺CD45RA^‐CD8⁺ T‐cell population (Figure 3c). An evaluation of CDR3 usage within the total repertoire of sorted CD8⁺ T‐cell subsets showed that CD69⁺CD103^+/‐ cells exhibit reduced diversity (Supplementary figure 3a). Furthermore, V‐J combination indicated skewing of the T‐cell repertoire in CD69⁺CD103⁺ and CD69⁺CD103^‐CD8⁺ T‐cell populations (Supplementary figure 3b). The analysis of TCR overlap between the subpopulations of SF CD8⁺ memory T cells revealed greater overlap of the TCR repertoire between CD69⁺CD103^‐ and CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells than the overlap between CD69^‐CD103^‐ and CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells, with a similar proportion of unique TCRs within each subpopulation (Supplementary figure 3c and d). Collectively, these results suggest that SF CD69⁺CD103^+/‐CD8⁺ memory T cells are a distinct population characterised by oligoclonal expansion with a skewed TCR repertoire.

3 Figure. CD69+CD103+CD45RA‐CD8+ T cells are a distinct population of cells with skewed TCR repertoires against certain antigens. (a) The TCR diversity of isolated SF CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells was determined using Simpson’s reciprocal index (left) and D50 (right), which indicate the percentage of unique T‐cell clones that account for the cumulative 50% of the total CDR3 counted. (b) Representative data for detected clones with unique CDR3 regions and their frequencies. The graphs represent data from four patients. (c) The proportion of top clones with unique CDR3 regions among the total clones of CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells from the synovial fluid of patients with RA. Statistical test: unpaired Student’s t‐test; **P‐value <0.01; n.s., not significant.

IL‐15 activates synovial fluid CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells in RA patients

Next, we examined the proliferation and effector functions of SF CD8⁺ T cells. IL‐15 plays a role in the maintenance and activation of T_RM cells²⁷ and is increased in the SF of RA patients.²⁸ Therefore, we examined the effect of ex vivo IL‐15 or TCR stimulation on SF CD8⁺ T cells derived from RA patients. First, we confirmed that SF IL‐15 concentrations were significantly higher in RA patients than in OA patients, whereas IL‐2 and IL‐7 levels were not significantly different between the two groups (Supplementary figure 4a). IL‐15 and IL‐15Rα co‐localised in fibroblasts and macrophages within the synovial tissue of RA patients, but not the synovial tissue of OA patients. This indicates trans‐presentation of IL‐15 by IL‐15Rα in RA synovial tissues (Supplementary figure 4b). Furthermore, although IL‐15Rβ expression was not different between CD45RA^‐CD8⁺ T‐cell subsets, CD132 (common γ chain) was upregulated in synovial CD69⁺CD103^+/‐ compared to CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells (Supplementary figure 5). CFSE dilution assays showed that, following IL‐15 stimulation, the fold increase in the frequency of CFSE^low cells, which were undergoing proliferation, was significantly higher in CD69⁺CD103⁺ and CD69⁺CD103^‐CD45RA^‐CD8⁺ T cells compared to CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells (Figure 4a). In contrast, SF CD8⁺ T cells exhibited comparable rates of proliferation among the subpopulations in response to TCR stimulation.

4 Figure. IL‐15 stimulation induces increased proliferation and cytotoxic potential in synovial fluid CD69+CD103+/‐CD45RA‐CD8+ T cells. (a) Proliferation in response to stimulation with anti‐CD3 antibodies or IL‐15 was determined by CFSE dilution assay in sorted CD69‐CD103‐, CD69+CD103‐, and CD69+CD103+CD45RA‐CD8+ T cells from the synovial fluid of patients with RA (n = 5). The proliferation index represents the fold change (%) in CFSElow cells compared to the mock group. (b and c) Pro‐inflammatory cytokine expression (b) and the expression of cytotoxic effector molecules perforin and granzyme B (c) in CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells from the synovial fluid of RA patients (n = 29) were analysed after stimulation with or without anti‐CD3 antibodies or IL‐15. Representative flow cytometry plots are presented on the left. On the right is a comparison of the production of cytokines (b) and effector molecules (c) between subsets or stimulations. Statistical test: one‐way ANOVA with Tukey’s multiple comparisons test; *P‐value <0.05, **P‐value <0.01, ****P‐value <0.0001. Data were obtained from a single experiment for each patient. CFSE, carboxyfluorescein succinimidyl ester; Prf, perforin; GrB, granzyme B.

To determine the function of SF CD8⁺ T cells with regard to inflammation and cytotoxicity, we examined the production of pro‐inflammatory cytokines and cytotoxic molecules after stimulation with IL‐15 or anti‐CD3 antibodies. Even though IFN‐γ and TNF‐α were significantly increased by TCR stimulation in all SF CD45RA^‐CD8⁺ T‐cell subsets, IL‐15 had no significant effect on the production of these pro‐inflammatory cytokines (Figure 4b). Without stimulation, perforin and granzyme B were highly expressed in CD69⁺CD103^+/‐ CD45RA^‐CD8⁺ T cells compared to CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells (Figure 4c). The expression of perforin and granzyme B was markedly increased in CD69⁺CD103^‐ and CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells after stimulation with IL‐15 but not anti‐CD3 antibodies. These effects of IL‐15 on the expression of cytotoxic molecules were not observed in CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells. Taken together, the results indicate that the SF CD69⁺CD103^+/‐CD8⁺ memory T cells of RA patients exhibit enhanced proliferation and cytotoxic activities, particularly after IL‐15 stimulation.

IL‐15‐stimulated CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells induce perforin‐mediated citrullinated histone formation

Given the marked increase in perforin expression upon IL‐15 stimulation of CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells, we examined their involvement in the pathogenesis of RA. A recent study demonstrated that perforin can induce the citrullination of proteins through NET formation.^29,30 Indeed, high amounts of citrullinated histones were produced in perforin‐treated neutrophils and were comparable to ionomycin‐treated neutrophils; such effects were not observed following incubation with calcium‐chelating agent EGTA (Supplementary figure 6a). Next, we treated neutrophils with supernatants harvested from the sorted peripheral blood CD45RA^‐CD8⁺ T cells of healthy controls and RA patients cultured with or without IL‐15. Citrullinated histones were significantly increased when neutrophils were treated with supernatants from IL‐15‐stimulated peripheral blood CD45RA^‐CD8⁺ T cells, but without NET formation (Supplementary figure 6b). Importantly, supernatants from IL‐15‐stimulated SF CD69⁺CD103^‐ or CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells induced histone citrullination along with the formation of NET, allowing citrullinated proteins to be secreted into the extracellular space (Figure 5a and b). This effect was abolished when CD8⁺ T cells were pre‐incubated with perforin‐inhibitor concanamycin A (CMA),³¹ indicating that perforin is responsible for the citrullination with the process of NET formation. Furthermore, MPO‐dsDNA complex levels, which indicate NET release, were significantly higher in the SF of RA patients than in that of OA patients (Figure 5c). The frequency of perforin‐expressing CD69⁺CD103^‐ or CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells in SF correlated with the levels of MPO‐dsDNA complex (Figure 5d). Interestingly, perforin levels in the SF correlated with the concentrations of MPO‐dsDNA complex (Figure 5e). Furthermore, perforin‐expressing CD8⁺ T cells were located near citrullinated histones in the synovial tissue of RA patients (Supplementary figure 7). Overall, these data indicate that IL‐15‐stimulated CD69⁺CD103^+/‐CD8⁺ memory T cells induce perforin‐mediated histone citrullination with the formation of NET complex.

5 Figure. IL‐15‐stimulated synovial CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells induce perforin‐mediated citrullination in freshly isolated, healthy neutrophils. (a) CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells were sorted, pre‐incubated with or without CMA, and stimulated with IL‐15 (20ngmL‐1) for 48h. The supernatants from each subset were added to freshly isolated neutrophils from a healthy donor. Neutrophils undergoing histone citrullination were visualised by immunofluorescent confocal microscopy using antibodies against citrullinated histones with FITC (green), neutrophil elastase with Alexa Fluor 647 (red), and DNA‐binding Hoechst 33342 (blue). Scale bars=20μm. (b) The levels of histone citrullination were measured under the indicated conditions using ImageJ from at least five images. (c) Concentration of MPO‐dsDNA complex in synovial fluid of OA (n=11) and RA (n=42) patients. (d) Correlation between perforin‐expressing CD69‐CD103‐, CD69+CD103‐ and CD69+CD103+ cells among CD8+ T cells with levels of MPO‐dsDNA complex in synovial fluid of RA patients. (e) Correlation between the concentration of perforin and MPO‐dsDNA complex in synovial fluid of RA patients (n=42). Data are presented as mean+SD as data were distributed normally. Statistical tests: one‐way ANOVA with Tukey’s multiple comparisons test (b), unpaired Student’s t‐test (c) and Spearman’s rank correlation analysis (d and e); **P‐value <0.01, ****P‐value <0.0001. CMA, concanamycin A; MPO‐dsDNA, myeloperoxidase‐double‐stranded DNA; OA, osteoarthritis; RA, rheumatoid arthritis.

Perforin‐expressing CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells are associated with ACPA‐positive RA

Finally, we analysed the implication of CD69⁺CD103^+/‐CD8⁺ T cells in RA, particularly ACPA‐positive RA. The frequency of CD8⁺ T cells among lymphocyte or CD3⁺ T‐cell‐gated SFMCs was significantly higher in ACPA‐positive patients than in ACPA‐negative patients (Figure 6a). Furthermore, the frequency of perforin‐expressing cells among CD69⁺CD103^‐ or CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells was significantly higher in ACPA‐positive patients than in ACPA‐negative patients (Figure 6b). A similar result was obtained when the frequency of granzyme B‐expressing cells was analysed (Figure 6b). Importantly, the frequency of perforin‐expressing CD69⁺CD103^‐CD8⁺ or perforin‐expressing CD69⁺CD103⁺CD8⁺ T cells among CD3⁺ T cells significantly correlated with the ACPA titre in the serum of ACPA‐positive RA patients (Figure 6c). The frequency of perforin‐expressing CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells among CD3⁺ T cells did not correlate with DAS28 levels in ACPA‐positive patients (Supplementary figure 8). Taken together, these findings indicate that perforin‐expressing CD69⁺CD103^+/‐CD8⁺ memory T cells are associated with the citrullination and production of ACPA.

6 Figure. ACPA‐positive patients have an increased frequency of synovial CD8+ memory T cells with more cytotoxic effector molecules than ACPA‐negative patients. (a) Quantitation of the proportion of CD8T cells within synovial lymphocytes and CD3+ T cells in ACPA‐negative (n = 14) and ACPA‐positive (n = 45) patients. (b) The frequency of cells expressing perforin or granzyme B among synovial CD69+CD103‐ and CD69+CD103+CD45RA‐CD8+ T cells in ACPA‐negative (n = 12) and ACPA‐positive (n = 21) patients. (c) Correlation between the anti‐cyclic citrullinated protein antibody titre and the frequency of perforin‐expressing CD69+CD103‐CD45RA‐CD8+ (left) or CD69+CD103+CD45RA‐CD8+ T cells (right) in RA patients with ACPA (n = 21). Statistical tests: unpaired Student’s t‐test (a and b) and Spearman’s rank correlation analysis (c); *P‐value <0.05, **P‐value <0.01, ***P‐value <0.001. ACPA, anti‐citrullinated protein antibody; aCCP, anti‐cyclic citrullinated peptide antibodies; Prf, perforin; GrB, granzyme B.

DISCUSSION

In the present study, we demonstrated that the SF CD69⁺CD103^+/‐CD8⁺ memory T cells of RA patients exhibit the phenotype of tissue‐resident cells with effector‐like features. Moreover, a TCR repertoire analysis revealed less TCR diversity in the CD69⁺CD103^+/‐ cell population, with skewed TCR usage. In addition, IL‐15 stimulation of CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells not only increased the expression of cytotoxic molecules, but induced the citrullination of histones with formation of NETs in a perforin‐dependent manner. Furthermore, perforin‐expressing CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells were enriched in the SF of ACPA‐positive patients, and their frequency correlated with the SF MPO‐dsDNA complex level and ACPA titre.

An analysis of synovial immune cells from RA patients provided evidence for the involvement of T cells in the pathogenesis of RA. Previous studies have focused on the pathogenic role of pro‐inflammatory cytokines, such as TNF‐α and IL‐17A. TCR analysis showed that expanded circulating T cells with signatures of pro‐inflammatory cytokine production and synovium infiltration are present in autoimmune arthritis.^32,33 A few studies have highlighted the possible influence of CD8⁺ T cells in RA,^9‐11 and a strong association between RA and MHC class I polymorphisms.³⁴ However, detailed phenotypes and functions, especially within specific subsets of CD8⁺ T cells, including T_RM cells, have not been studied in RA. In the present study, we examined the SF CD8⁺ T cells of RA patients to determine whether these cells contribute to the pathogenesis of the disease.

We found that expanded SF CD8⁺ T cells contain CD69⁺CD103⁺ and CD69⁺CD103^‐ cells. CD69 and CD103 are considered canonical markers of T_RM cells, which can participate not only in protection against pathogens and tumors, but also in the pathogenesis of autoimmune disorders.^{15,20,21,35‐37} In response to IL‐15, autoreactive T_RM cells can contribute to autoimmune skin diseases, such as vitiligo.³⁸ Our present study revealed that SF CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells exhibit characteristics of T_RM cells; they downregulated the mRNA expression of S1PR1, which is critical for recirculation through the detection of S1P in blood and lymph nodes, and its transcription regulator KLF2. We also found that SF CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells express significantly higher levels of PD‐1 and CD101, which are upregulated in T_RM cells, than CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells in RA patients. The increased expression of PD‐1 and memory marker Eomes, but not activation markers CD25 and CD38, suggests that these cells are not merely activated T cells, but rather differentiated effector or memory cells. These findings indicate that SF CD69⁺CD103^+/‐CD8⁺ memory T cells in RA exhibit features of T_RM cells. However, SF is an exudate, rather than an inflamed tissue, and the tissue residency of CD69⁺CD8⁺ T cells was not assessed directly. In addition, considering that activation marker HLA‐DR was upregulated in CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells, and an increase in cytotoxic molecules in response to IL‐15 could be a feature of CD8⁺ memory T cells, we cannot rule out the possibility that SF CD69⁺CD103^+/‐CD8⁺ T cells were memory or activated T cells rather than T_RM cells. Therefore, further studies including functional analysis of tissue residency are required to confirm that these cells are bona fide T_RM cells.

Although similar expression patterns of T_RM markers were observed in both CD69⁺CD103⁺ and CD69⁺CD103^‐ subsets, the degree of expression was different between the two subsets. The expression of chemokine receptor CXCR6 and adhesion protein CD49a was significantly higher in CD69⁺CD103⁺ cells than CD69⁺CD103^‐ cells. In addition, the frequencies of PD‐1⁺ and CD101⁺ cells were higher in CD69⁺CD103⁺ cells than CD69⁺CD103^‐ cells, suggesting that CD69⁺CD103⁺ cells were a distinct population more closely resembling T_RM cells. The decreased diversity and skewing of the TCRVβ CDR3 repertoire were profound in CD69⁺CD103⁺ cells, whereas CD69⁺CD103^‐ cells exhibited a mild contraction of the repertoire compared to CD69^‐CD103^‐CD45RA^‐CD8⁺ T cells. However, the shared TCR repertoire between CD69⁺CD103^‐ and CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells suggests that two subpopulations may have originated from the same predecessors. In contrast, CD69^‐CD103^‐ cells exhibited less overlap with CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells. Further studies designed to distinguish between the two subpopulations within CD69⁺CD8⁺ T cells would contribute to understanding immune responses during joint inflammation.

IL‐15 has been considered an indispensable cytokine for the development, homeostasis and survival of T_RM cells in peripheral tissues.^14,27 The present study confirms higher levels of SF IL‐15 in RA patients than in OA patients as reported in a previous study.²⁸ In addition, we observed that IL‐15 co‐localised with IL‐15Rα in fibroblasts and macrophages from the synovial tissues of RA patients, signifying trans‐presentation of IL‐15. The proliferation and expression of cytotoxic molecules were significantly increased by IL‐15 in SF CD8⁺ T cells, particularly CD69⁺CD103^+/‐ subsets. The increased expression of CD132 may explain the augmented response to IL‐15 stimulation by SF CD69⁺CD103^‐ and CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells. Interestingly, IL‐15 stimulation also induced an increase in cytotoxic molecule expression in CD69⁺CD45RA^‐CD8⁺ T cells from the peripheral blood of HC or RA patients (data not shown); thus, the response to IL‐15 is a general feature of CD69⁺CD8⁺ memory T cells. Considering the increased IL‐15 concentration and its general effect on CD69⁺CD45RA^‐CD8⁺ T cells, especially on CD69⁺CD103^+/‐ subsets, the role of IL‐15 in the pathophysiology of RA should be studied further.

Citrullination and the presence of antibodies against to citrullinated proteins is one of the hallmarks in the pathogenesis of RA. Because NET complex contains citrullinated histones,²⁶ it can be a source of citrullinated autoantigens.²⁸ The activation of peptidylarginine deiminases, enzymes responsible for protein citrullination, is mediated by an influx of calcium ions into the cytosol of neutrophils. This leads to the formation of numerous citrullinated proteins, including histones.^28,39 Because perforin is a pore‐forming agent in the cell membrane and allows for the influx of calcium ions into target cells, it can generate citrullinated proteins in RA.²⁸ In the present study, we found that IL‐15‐stimulated CD69⁺CD103^+/‐CD8⁺ memory T cells can induce neutrophil histone citrullination and the release of neutrophil elastase into the extracellular space in the process of NET formation. The frequency of perforin‐expressing CD69⁺CD103^‐ or CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells significantly correlated with the levels of MPO‐dsDNA complex. Importantly, the expression of perforin was higher in SF CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells from ACPA‐positive RA patients than those from ACPA‐negative patients and correlated with the ACPA titre. These findings suggest that CD69⁺CD103^+/‐ subsets of SF CD8⁺ T cells may be a major trigger for citrullinated protein formation. Therefore, they likely play a role in the pathogenesis of RA, particularly in patients with ACPA‐positive RA.

In conclusion, our data indicate that expanded SF CD8⁺ memory T cells from RA patients contain CD69⁺CD103^+/‐ subsets that exhibit phenotypes of canonical T_RM cells with effector‐like features. CD69⁺CD103^+/‐ cells are more responsive to IL‐15, which was detected in the SF and tissue of patients with RA, and exhibit increased proliferation and cytotoxic potential. Furthermore, IL‐15‐stimulated CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells induce perforin‐mediated histone citrullination and NET formation. We also demonstrated that synovial perforin‐expressing CD69⁺CD103^+/‐CD45RA^‐CD8⁺ T cells are associated with ACPA‐positive RA. Taken together, our data suggest that SF CD69⁺CD103^+/‐CD8⁺ memory T cells with T_RM features play a role in the pathogenesis of ACPA‐positive RA by inducing perforin‐mediated citrullinated protein formation.

METHODS

Patients and samples

SF samples were obtained from 60 patients diagnosed with RA at Asan Medical Center (Seoul, Korea). All patients had active disease at the time of sampling, for which they required joint aspiration with or without intraarticular injection of corticosteroids. For immunofluorescence, synovial tissues were acquired from the synovium of OA and RA patients who underwent arthroplasty or synovectomy. The diagnosis of RA was based on the 2010 revised American College of Rheumatology/European League Against Rheumatism criteria for the classification of RA. The clinical and demographic data of studied patients are summarised in Supplementary table 1.

Reagents and antibodies

The following monoclonal antibodies were used for multicolour flow cytometric analysis: anti‐CD3 (UCHT1), anti‐CD8 (HIT‐8a), anti‐CD38 (HIT2), anti‐CD45RA (HI100), anti‐CD49a (SR84) and anti‐CD132 (AG184) from BD Biosciences (San Jose, CA, USA); anti‐CD8 (T8, Leu2), anti‐CD25 (BC96), anti‐CD69 (FN50), anti‐CD101 (BB27), anti‐CD103 (Ber‐ACT8), anti‐CXCR6 (K041E5), anti‐granzyme B (QA16A02), anti‐IFN‐γ (4S.B3), anti‐IL‐15Rβ (TU27), anti‐perforin (B‐D48), anti‐TCRγδ (B1) and anti‐TNF‐α (MAb11) from BioLegend (San Diego, CA, USA); anti‐CD14 (61D3), anti‐CD19 (HIB19), anti‐CD56 (CMSSB), anti‐Ki‐67 (20Raj1), anti‐T‐bet (eBio4B10) and Fixable Viability Dye eFluor^® 506 from eBioscience (San Diego, CA, USA); and anti‐Eomes (WD1928) from Invitrogen (Carlsbad, CA, USA). For immunofluorescent staining of synovial tissues and NET formation, anti‐CD8 (YTC182.20), monoclonal anti‐IL‐15, polyclonal anti‐CD68, anti‐CD103 (EPR4166(2)) and anti‐histone H3 (citrulline R2 + R8 + R17) antibodies were obtained from Abcam (Cambridge, UK). Anti‐CD69 (FN50), anti‐CD55/DAF, anti‐IL‐15Rα and anti‐neutrophil elastase (NP57) were obtained from BioLegend, Novus Biologicals (Littleton, CO, USA), R&D Systems (Minneapolis, MN, USA) and Santa Cruz (Santa Cruz, CA, USA), respectively. Recombinant human IL‐15 (PeproTech, Rocky Hill, NJ, USA) was used for the stimulation of SFMCs and sorted CD8⁺ T cells. The concentrations of IL‐15, IL‐2, IL‐7 and perforin in the SF were measured by enzyme‐linked immunosorbent assay (ELISA) using DuoSet (R&D Systems) and ELISA Set (Abcam) according to the manufacturers’ instructions.

Mononuclear cell isolation and stimulation

SF samples were collected by centrifugation at 1800 g for 10 min. The supernatants were stored at −80°C until cytokine measurement. Pellets from the SF and peripheral blood were subjected to density gradient centrifugation using Ficoll‐Paque™ PLUS (GE Healthcare, Piscataway, NJ, USA) to isolate mononuclear cells. Cells were cryopreserved using FBS (Welgene, Daejeon, South Korea) supplemented with 10% dimethyl sulfoxide (DMSO; Sigma Aldrich, St. Louis, MO, USA) and stored in liquid nitrogen until thawed for further usage. To determine the production of pro‐inflammatory cytokines and cytotoxic effector molecules, isolated mononuclear cells were incubated in RPMI 1640 medium (Welgene) supplemented with 10% FBS (Welgene) and 1% penicillin and streptomycin (P/S) (Welgene) in the presence or absence of rhIL‐15 (20 ng mL^‐1) for 24 h at 37°C in a 5% CO₂ atmosphere. When indicated, anti‐CD3 antibodies (1 μg mL^‐1; BD Biosciences), brefeldin A (GolgiStop™; BD Biosciences) and monensin (GolgiPlug™; BD Biosciences) were added in the last 5 h of the incubation.

Flow cytometric analysis

Cryopreserved SFMCs and PBMCs were thawed at 37°C and incubated with Fixable Viability Dye eFluor^® 506 (eBioscience) for 20 min at 4°C without light exposure according to the manufacturer’s instructions to exclude dead cells. The cells were then washed with DPBS containing 1% FBS (Welgene) and 0.05% sodium azide and subsequently stained with fluorochrome‐conjugated antibodies against surface proteins or fluorochrome‐matched isotype controls for 30 min at 4°C. For the staining of intracellular proteins, surface‐stained cells were fixed and permeabilised using the FoxP3 Staining Buffer Kit (eBioscience) according to the manufacturer’s instructions. The cells were stained with antibodies against intracellular proteins and matched isotype controls for 30 min at 4°C. After washing, samples were acquired on a BD FACSCanto™ II flow cytometer (BD Biosciences) and analysed using FlowJo software (TreeStar, Ashland, OR, USA).

Cell sorting

CD45RA^‐CD8⁺ T‐cell subsets were isolated according to the expression of CD69 and CD103. Cryopreserved SFMCs were thawed and washed with DPBS (Welgene) containing 2% FBS (Welgene) and 2 mm EDTA (Sigma Aldrich). CD8⁺ T cells were magnetically separated from RA SFMCs using the CD8⁺ T Cell Isolation Kit (Miltenyi Biotec, Auburn, CA, USA). CD8⁺ T cells were stained with Fixable Viability Dye eFluor^® 506 (eBioscience) and surface antibodies, filtered and further sorted into CD69^‐CD103^‐, CD69⁺CD103^‐ and CD69⁺CD103⁺CD45RA^‐CD8⁺ subsets using FACSARIA^™ III (BD Biosciences).

Quantitative RT‐PCR

Total RNA was extracted from the sorted CD69^+/‐CD103^+/‐CD45RA^‐CD8⁺ T cells from the RA SFMCs using TRIzol (Invitrogen), and extracted RNA was reverse‐transcribed into cDNA using the SuperScript™ IV First‐Strand Synthesis System (Invitrogen) according to the manufacturer’s instructions. Quantitative PCR (qPCR) was performed with TOPreal™ qPCR 2X PreMIX (SYBR Green with low ROX; Enzynomics, Daejeon, South Korea) on a LightCycler^® 480 Instrument II (Roche, Branchburg, NJ, USA). Custom‐designed primers (Cosmogenetech, Seoul, South Korea) were used (Supplementary table 2). The RNA expression levels were determined based on the normalisation of the relative expression of target genes to the control gene GAPDH.

TCR sequencing

A total of 15,000‐550,000 CD69^‐CD103^‐, CD69⁺CD103^‐ and CD69⁺CD103⁺CD45RA^‐CD8⁺ T cells from RA SFMCs were sorted and analysed (details about the cell and sequencing data are presented in Supplementary table 3). Dead cells were excluded from sorting using Fixable Viability Dye eFluor^® 506 (eBioscience). The cells were collected and stored in RNAlater™ Stabilization Solution (Thermo Fisher Scientific, Waltham, MA, USA) at −80°C. Total RNA was extracted from the samples using the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. cDNA was synthesised using the OneStep RT‐PCR Kit (Qiagen) with barcoded primers (iRepertoire, Huntsville, AL, USA). The resulting barcoded cDNA libraries were rescued with SPRIselect beads (Beckman Coulter, Brea, CA, USA) according to the manufacturer’s instructions to increase the CDR3 discovery rate and library quality. After bead clean‐up, rescued cDNA libraries were amplified exponentially using universal sequencing primers (iRepertoire) and GoTaq G2 Hot Start Master Mix (Promega, Fitchburg, WI, USA). Using NanoDrop 8000 (Thermo Fisher Scientific), the library amplification was regarded as successful if the concentration of the samples was ≥10 ng μL⁻¹. Libraries with unique barcodes were pooled and subjected to quality control by KAPA qPCR. Next‐generation sequencing was performed on the Illumina MiSeq system for 600 cycles, and the data were analysed by iRepertoire software. The output data files were used to calculate the D50, which is the percent of top dominant and unique T‐cell clones that account for the cumulative 50% of the total CDR3s counted within sample. Simpson’s reciprocal index was calculated as follows:[Image Omitted. See PDF]where D = Simpson’s reciprocal index, N = total number of all sequences and n = total number of particular sequences. Simpson’s reciprocal index approaching 1 indicates monoclonality, whereas a higher index value represents greater diversity of clonality. Circos plots were produced using the Circos software package.

Proliferation assay

Sorted cells were first labelled using the CellTrace™ CFSE Cell Proliferation Kit (Invitrogen). CFSE‐labelled cells were cultured in RPMI 1640 medium with 10% FBS and 1% P/S with or without anti‐CD3 antibodies (1 μg mL^‐1; BD Biosciences) or rhIL‐15 (20 ng mL^‐1, PeproTech) for 48 h at 37°C in a 5% CO₂ atmosphere_. After incubation, the cells were harvested and stained with fluorochrome‐conjugated antibodies against surface proteins. After washing, samples were acquired on a BD FACSCanto™ II flow cytometer (BD Biosciences) and the frequency of CFSE^low cells was determined using FlowJo software (TreeStar).

Neutrophil isolation and NET induction

Peripheral blood from healthy subjects was diluted in DPBS and centrifuged with Ficoll‐Paque™ PLUS (GE Healthcare) for density gradient separation. The supernatant including the mononuclear cell layer was removed, and the pellet at the bottom of the tube was resuspended in HBSS without phenol (Welgene). The cells were then added to an equal volume of 3% dextran (MP Biomedicals, Santa Ana, CA, USA) in HBSS and incubated for 30 min at room temperature for erythrocyte sedimentation. The upper layer containing neutrophils was transferred to a new tube and washed. The remaining erythrocytes were lysed using hypotonic lysis buffer and re‐equilibration buffer.⁴⁰ The cells were washed with HBSS before further analysis. The purity of isolated neutrophils was >90% based on CD66b and CD16 expression as assessed by flow cytometry.

To induce and visualise NET formation, 1.0 × 10⁵ neutrophils were attached to the surface of poly‐D‐lysine (Thermo Fisher Scientific) coated 12 mm cover glasses (Marienfeld, Lauda‐Königshofen, Germany) for 30 min at 37°C. The cells were then stimulated with 5 μm ionomycin (Thermo Fisher Scientific), 500 ng mL^‐1 perforin (Enzo, Farmingdale, NY, USA) or the supernatants of CD69^+/‐CD103^+/‐CD45RA^‐CD8⁺ T‐cell subsets cultured with or without IL‐15 and pre‐incubated with 30 nm CMA (Sigma Aldrich) for 2 h when required. For the inhibition of NET formation, the samples were treated with 8 mm EGTA.

Histone citrullination and NET formation

Neutrophils on coated cover glasses were fixed using paraformaldehyde at a final concentration of 4%. Cover glasses were removed from the 24‐well plate and washed with DPBS. The cells were permeabilised with PBS + Tween‐20 (Bio Basic, Markham, ON, Canada) and blocked with 5% donkey serum in 0.1% PBST for 1 h at room temperature. Citrullinated histones and NETs were stained with rabbit polyclonal anti‐histone H3 (R2 + R7 + R18) (1:250) and mouse anti‐neutrophil elastase (1:250). Coverslips were washed and stained with donkey‐raised anti‐rabbit FITC and anti‐mouse Alexa Fluor 647 antibodies. DNA was counterstained with Hoechst 33342 (Invitrogen). Citrullinated histones and NETs were visualised using the LSM 710 confocal microscope (Carl Zeiss, Göttingen, Germany). A minimum of five images per slide were taken using the 10× lens in different areas of each coverslip to ensure accurate representation. Digital images were acquired by ZEN software (Carl Zeiss) and analysed using ImageJ software (NIH, Bethesda, MD, USA).

Immunofluorescence microscopy

Synovial tissues were serially cut into sections and embedded with paraffin using microscope cover glasses. For immunofluorescent staining, tissue sections were deparaffinised and rehydrated with xylene and ethanol. The sections were immersed in antigen retrieval solution (Agilent, Santa Clara, CA, USA) citrate buffer (pH 6) and incubated in a microwave for 20 min. The slides were allowed to cool at room temperature for 30 min before washing with DPBS. Next, the slides were blocked with 1% BSA in TBS+ 0.2% Triton X‐100 (Bio Basic). The sectioned tissues were stained with the following cocktails of primary antibodies in blocking buffer overnight at 4°C: rat anti‐CD8 (1:100), mouse anti‐CD69 (1:100) and rabbit anti‐CD103 (1:100); mouse anti‐IL‐15 (1:400), goat anti‐IL‐15Rα (1:400), rabbit anti‐CD55/DAF (1:50) and/or rabbit anti‐CD68 (1:50); or rat anti‐CD8 (1:100), mouse anti‐perforin (1:50) and rabbit anti‐histone H3 (R2 + R7 + R18)(1:50). Subsequently, the sections were washed and stained for 1 h at room temperature with the following donkey‐raised secondary antibodies: anti‐mouse Alexa Fluor 647 (1:500), anti‐goat Cy3/Cy5 (1:500), anti‐rabbit FITC (1:500) and anti‐rat Cy3/Cy5 (1:500). After washing, sections were counterstained with Hoechst 33342 (Invitrogen) and mounted using ProLong Gold Antifade Mountant (Invitrogen).

MPO‐dsDNA complex measurement

96‐well ELISA microplates were coated with mouse anti‐human MPO (clone 4A4; Invitrogen) (1:2000) overnight at 4°C. After washing, the plate was blocked with 1% BSA in PBS. Samples diluted in blocking buffer were treated with DNase (Worthington, Columbus, OH, USA) for 15 min at room temperature to remove unbound DNA followed by EDTA treatment. Samples were then loaded onto plates and incubated overnight at 4°C. Plates were washed and added with diluted Picogreen reagent (1:200) in TE buffer. After incubation for 5 min, fluorescence was measured in Victor X4 (Perkin Elmer, Hopkinton, MA, USA). The concentrations of MPO‐dsDNA complex were interpolated from standard curve generated from Quant‐iT™ PicoGreen™ dsDNA Assay Kit (Invitrogen).

Statistical analysis

Statistical analyses were performed in Prism7 (GraphPad Software, San Diego, CA, USA). All data are presented as the mean±standard deviation. ACPA‐negative and ACPA‐positive patients were compared by the Mann–Whitney U‐test. Correlation studies were performed by Spearman’s rank correlation analysis. Comparisons of multiple groups, including healthy controls, RA PBMCs and RA SFMCs, and between the subsets of CD45RA^‐CD8⁺ T cells, were performed by the Student’s t‐test or ANOVA and Tukey’s multiple comparisons test. Significance is reported as P<0.05.

Acknowledgments

This research was supported by the Bio & Medical Technology Development Program of the National Research Foundation (NRF) and funded by the Korean government (MSIT) (No. 2018M3A9D3079500).

CONFLICT OF INTEREST

The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

JJ, E‐CS and SH designed the research. JJ, JSL and SH performed the experiments. JJ, Y‐GK, C‐KL, BY, E‐CS and SH analysed the data. JSL, Y‐GK, C‐KL, BY and SH provided the clinical samples. JJ, E‐CS and SH wrote the manuscript.

ETHICAL APPROVAL

All patients and healthy controls provided written informed consent. The protocols of this study were approved by the Institutional Review Board of Asan Medical Center (2017‐0866).