Proportion of T follicular helper cells in peripheral blood of rheumatoid arthritis patients: a systematic review and meta-analysis

ABSTRACT Introduction:Alterations in the levels and activity of Tfh may lead to impaired immune tolerance and autoimmune diseases. The aim of this study was to investigate the proportion and types of Tfh cells in the peripheral blood (PB) of RA patients. Areas covered:Comprehensive databases were searched for studies evaluating the proportion of Tfh cells in the PB of patients with RA compared to healthy control (HCs). The proportion of Tfh cells in RA patients was significantly higher than in HCs (SMD 0.699, [0.513, 0.884], p < 0.0001). Furthermore, Tfh cells proportion in untreated-RA and early-RA patients was markedly greater than HCs, when comparisons done without considering the definition markers, and also when Tfh cells were defined by the specified definition markers. While the proportion of Tfh cells by all definitions was higher in active-RA compared to HCs, analysis of two definitions, CD4+CXCR5+ and CD4+CXCR5+ICOS+, didn’t show significant differences. Furthermore, higher proportion of Tfh cells defined by all definitions and a specified definition (CD4+CXCR5+PD-1high) was observed when S+RA compared to S−RA patients. Expert opinion:The results demonstrate that circulating Tfh are highly elevated in RA patients highlights its potential use as a biomarker and a target for RA therapy.


Introduction
Rheumatoid arthritis (RA) is a chronic autoimmune disease with a prevalence of 0.5% to 1% of the population, and primarily affecting joint tissues [1]. Typical clinical manifestations include symmetrical inflammation and swelling in particular of the small joints of hands and feet, which can lead to joint destruction, deformity and functional disability [1,2]. RA imposes a substantial socio-economic burden on patients and societies [2,3]. Although, the exact etiology and pathogenesis of RA are not completely understood, it is well accepted that the disease is multifactorial caused by multiple genetic, environmental, and microbial factors. These risk factors together trigger uncontrolled immune responses, and development and progression of the disease [2]. Different types of innate and adaptive immune cells including dendritic cells (DCs), monocytes, macrophages, antigenspecific T and B lymphocytes along with cytokines and autoantibodies contribute to the pathogenesis of RA [3]. Among the immune cells, CD4 + T cells have critical roles in the disease induction and progression. Once activated, CD4 + T cell can differentiate into different T helper (Th) subsets including Th1, Th2, Th17, T regulatory, or T follicular helper (Tfh) cells, each play an important role in the RA [4].
Tfh cells are a subpopulation of Th cells localized in the B cell follicle of lymph nodes. These cells are essential in humoral immune response by providing help for B cell to develop germinal center, differentiate to memory B cells and long-lived antibody-secreting plasma cell, immunoglobulin class-switching and antibody affinity maturation [5]. Tfh cells can be distinguished from other differentiated CD4 + Th subsets by the expression of B-cell lymphoma 6 (BCL-6) transcription factor [6], the surface expression of CXCR5 chemokine receptor [7][8][9], CD40 ligand, programmed cell death protein 1 (PD-1), and inducible costimulatory molecule (ICOS) [5,10], and the production an release of interleukin (IL)-21 [6].
Despite the prominent role of Tfh cells in humoral immunity for protection against infectious pathogens, accumulating evidence has demonstrated that deregulated frequency and activity of Tfh may lead to impaired immune tolerance, generation of high-affinity autoantibodies and hence development of antibody-mediated autoimmune diseases [11][12][13][14]. However, anatomical localization of Tfh cells in the secondary lymphoid tissues limits their routine studies in human patients [5,7]. Recently, studies have introduced a new circulating subpopulation of CD4 + T cells that show both phenotypical and functional characteristics of classical Tfh cells residing in the lymphoid tissues [15,16]. Based on these studies, circulating Tfh (cTfh) cells, similar to the classical Tfh, could promote differentiation of naïve B cell into plasma cells and immunoglobulin secretion through IL-21 production [15]. In addition studies have shown that cTfh cells increase in proportion to their GC counterparts in secondary lymphoid tissues [16]. Thus, they represent circulating compartment of Tfh cells and evaluation of the status of cTfh would reflect the status of their counterpart in GC of secondary lymphoid tissues [15,16]. Following the introduction of cTfh, several human studies have demonstrated the increased frequency of Tfh in the peripheral blood (PB) of patients with different autoimmune diseases including multiple sclerosis (MS), systemic lupus erythematosus (SLE), autoimmune thyroiditis, myasthenia gravis (MG), Sjogren's syndrome (SjS) and RA [17].
Several research groups have explored the potential role of cTfh cells in RA pathogenesis by determining their frequency in PB of RA patients compared to the healthy controls (HCs). While some of these studies have shown increase in the frequency of Tfh in PB of patients with RA and its positive correlation with disease scores, some other studies have failed to show any differences in the frequency of these cells in patients with RA compared with HCs. Hence, in the present work, we systematically searched and reviewed the available studies documenting the proportion of Tfh among CD4 + T cells in PB of RA patients, and conducted a meta-analysis to elucidate the proportion of Tfh cells in the PB of RA patients compared to the HCs.

Search strategy
This study was carried out in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and was registered at the international prospective register of systematic reviews (PROSPERO) (CRD42020163549). We searched online databases including Embase, PubMed, Scopus and Web of Science (WOS) to identify original articles published until November 2019 reporting the frequency of Tfh among CD4 + T cells in PB of patients with RA.
The search terms used in this meta-analysis were as follows: ('Arthritis, Rheumatoid'[Mesh] OR 'Rheumatoid arthritis' OR RA) AND ('T helper follicular' OR 'T follicular helper' OR 'Follicular T helper' OR 'Follicular helper T' OR 'Tfh'). In order to reduce the risk of missing any related study, we also searched manually the reference list of review articles and the included original articles. The searches and selection of articles were conducted by two investigators independently. In the case of any disagreement, it was resolved through discussion with a third investigator.

Study selection
Inclusion criteria: 1) original human studies; 2) title or abstract containing the terms 'rheumatoid arthritis' or 'RA' and 'T helper follicular' OR 'T follicular helper' OR 'follicular T helper' OR 'follicular helper T' OR 'Tfh'; 3) documenting the frequency/ proportion of Tfh among CD4 + T cells in the PB of RA patients and HCs; 4) available as full-text article on the internet (website or PDF); 5) Available information regarding the number of patients and HCs. No restriction was applied for the subtypes of the RA disease, disease severity score, sex or race of the participant subjects in the study. Also, no time or language restrictions were imposed.
Exclusion criteria: 1) non-original studies, review articles, case reports, letter to editor or conference abstract publications; 2) Animal or in vitro studies; 3) studies without healthy control group; 3) studies without raw data in which data extraction from their graphical results were not feasible; 4) Redundancies among Embase, PubMed, Scopus, and WOS search were removed and each study was counted only once in the meta-analysis

Data extraction and quality assessment
Two independent researchers carefully reviewed the included studies and extracted data, and disagreements were resolved through discussion with the third investigator. The following data were extracted from the eligible articles: author's name, publication year, country where the authors performed the analysis, number of RA patients and HC controls, Tfh definition, mean and standard deviation (SD) of the frequency of Tfh among CD4 + T cells. When Tfh were defined by different patterns of markers in the studies, data were extracted from all of them. If instead of mean and SD, the studies reported mean and standard error (SE), medians and ranges, or median and interquartile ranges, we extracted all types of the data. In the case of some of articles presenting only graphical results without raw data in which it was not possible to extract data from their graphs, the raw data (mean and SD) were generously received from the authors after contacting the authors by e-mail. The quality of all included studies was assessed by Newcastle-Ottawa Quality Assessment Scale (NOS) (Case Control Studies) [18].

Data analysis
All statistical analyses for the Tfh frequency differences between RA and HC subjects were conducted using STATA

Article highlights
• The proportion of Tfh cells defined by all definition markers was higher in PB of RA patients than HCs. • The proportion of Tfh cells with individual definitions were higher in PB of RA patients than in HCs. • Tfh cells proportion in untreated-RA and early-RA patients was greater than HCs. • The proportion of Tfh cells defined by all definitions was higher in active -RA compared with HCs. • Seropositive RA patients had higher Tfh cells proportion compared to seronegative patients. • These results highlight potential pathogenic role of Tfh cells in RA and its potential use as a biomarker and a target for RA therapy. version 16.0 (Stata Corp., College Station, Texas). The sample size, mean, and SD of the frequency of Tfh cells were used to calculate the standardized mean difference (SMD) between RA and HC groups. For certain studies reporting only graphical results, the graphs were converted to the numerical data by using GetDat Graph Digitizer software version 2.24 (http://getdatagraph-digitizer.com/). When some studies reported SE instead of SD, the SD was calculated from sample size and SE using the following formula: SD = SE×√n. When some studies presented their results as median and range, or median and interquartile range (IQR), the mean and SD were estimated by using a standard method [19]. When there was need to combine data from two subgroups (for example combining mean and SD of 'active-RA' and those from 'remission-RA' groups), the pooled mean and SD was calculated by using a standard formula according to the Cochrane Handbook for Systematic Reviews of Interventions [20].
The meta-analysis was performed to calculate the SMD of the proportion of Tfh among CD4 + T cells in the PB of 1) RA patients versus HCs 2) untreated RA (u-RA) patients versus HCs, 3) early RA (e-RA) patients versus HCs, 4) active-RA (a-RA) patients versus in remission-RA (r-RA) patients, 5) a-RA patients versus HCs, 6) seropositive (S + ) RA patients versus seronegative (S − ) RA patients, and 7) S + RA patients versus HCs. Furthermore, since the studies used different patterns of markers to define Tfh cells, subgroup analysis was also performed to estimate the results based on each Tfh definition.
In order to evaluate the heterogeneity between the studies, the Q test and I 2 statistic was used. For the Q test, P-value < 0.1 was considered as statistically significant and I 2 values of 75, 50, and 25% were considered as evidence of high, moderate and low level of heterogeneity, respectively. SMD with 95% confidence interval (CI, Hedges' g) was calculated to assess the proportion of Tfh cells, and random effect model (REM) was used when the heterogeneity was high (I 2 > 50%), and a fixed effect model (FEM) was chosen when the heterogeneity was low or absent (I 2 ≤ 50%). In addition, the publication bias was assessed by using funnel plot in which the x -axes and y-axes of the plot were used to show SMD and SE. Egger's test was used for evaluating funnel plot asymmetry, and in the case of its presence, the 'trim and fill' method was applied to assess the effect of publication bias on the calculated results. P-value <0.05 was considered to be statistically significant except where noted.

Literature search
A total of 821 potentially eligible studies were retrieved after searching databases along with manual searching. Of these, after primary screening, 405 and 365 studies were respectively excluded due to duplication and not satisfying the inclusion criteria. Fifty one relevant full-text articles were selected and assessed for eligibility, of which 26 articles were excluded because of the following reasons: lack of healthy control group (n = 3), evaluating the Tfh cells in in vitro (n = 4), without raw data and unable to extract data from the graphical result (n = 2), evaluating changes in the proportion of Tfh in autoimmune diseases other than RA (n = 1), reporting only the absolute number of Tfh cells (n = 2), overlapping samples from other studies (n = 2), reporting only Tfh subsets (Tfh-1, −2 and −17) (n = 2), and letter to editor. Additional 10 articles were irrelevant and didn't evaluate proportion of Tfh cells, thus, finally a total of 25 articles were included in the metaanalysis ( Figure 1).

Characteristics of eligible studies
Of the included articles, 14 were performed in China, 4 in Japan, 2 in Spain, 1 in Sweden, 1 in Italy, 2 in USA, and 1 in Argentina. In total the eligible articles included 1041 RA patients and 610 HCs. According to the NOS risk of bias assessment, all the included studies had a score of 4-8, most of them had high quality (NOS score, 7 and 8), and fewer with moderate quality (NOS score, 4-6). The included studies and their basic characteristics are also listed in Table 1.

Pooled results
In initial step, we carried out meta-analysis to compare the proportion of Tfh cells between RA and HC groups regardless of the markers used for definition of Tfh cells. Some of these studies reported a significant increase in the proportion of Tfh cells in PB of RA versus HC subjects [21][22][23][24][25][26][27][28][29][30][31][32][33], some studies showed no differences [34][35][36][37][38], some studies revealed both increase in RA and no differences between the two groups depending on the definition markers [39][40][41][42], and some other reported non-significant decrease in Tfh cells proportion in RA group [43,44]. As most articles evaluated more than one Tfh definition in the same population, in order to prevent skew in the results, the number of patients and HCs was adjusted according to the number of definitions. This was also applied for other analyses in the next sections. Based on the pooled result of metaanalysis, RA patients had a significantly higher proportion of cTfh cells compared with HCs (SMD 0.699, [0.513, 0.884], p < 0.0001), although high level of heterogeneity (I 2 = 64.57, p < 0.0001) was observed between studies ( Figure 2).

Subgroup analysis
In the next step, we performed subgroup meta-analysis based of different definition markers used for Tfh cells. It should be noted that as all Tfh cells express CD3 and due to previous antigen exposure all are CD45A − and CD45RO + memory T cells. Therefore, CD3 and CD45RA/RO were not taken into consideration for subgrouping and meta-analysis. The results of the subgroup analysis based on Tfh definitions are summarized in Table 2.
We also analyzed the studies in which the 'CD4 + CXCR5 + ' Tfh cells co-expressed PD-1 and ICOS (CD4 + CXCR5 + PD-1 + ICOS + cells). From four studies, we removed two studies in which high expression of PD-1 and/or ICOS were used for Tfh cell evaluation. According to the pooled SMD result, a significant increase in the proportion of Tfh cells co-expressing ICOS and PD-1 was found in RA patients in comparison with the HCs (SMD 0.67, [0.31, 1.04], p = 0.0003, I 2 = 0%, p = 0.518) ( Supplementary Fig.7).
Finally, we assessed Tfh cells proportion from the studies in which 'CD4 + CXCR5 + CCR7 low PD-1 high ' was used to identify Tfh. Of these studies 2 reported markedly higher proportion of Tfh cells in patients with RA [33,40], and 1 reported no difference when RA group was compared with HC group [35]. Our subgroup meta-analysis revealed significant increase in the proportion of 'CD4 + CXCR5 + CCR7 low PD-1 high ' cells in PB of RA patients in comparison with HCs (SMD 0.68, [0.4, 0.96], p < 0.0001, I 2 = 4.93%, p = 0.34) (Supplementary Fig.8).

The proportion of Tfh in PB of untreated-RA patients as compared with HC
Studies in which patients were taking steroids or immunesuppressive drugs were removed from this part of the metaanalysis, since treatments were not uniform across the studies. A total of 12 articles covering 21 studies evaluated the proportion of Tfh cells in patients who did not take any medicine at the time of sample collection. Of these, some of studies included patients who didn't receive any drug for at least 1 month [21], 2 months [30,34] and 3 months [33,35,41] before blood sample collection, and 6 articles covering 12 studies included patients with lack of any treatment history [22,25,26,28,37,42]. Therefore, the proportion of Tfh cells was compared between RA patients without immunosuppressive treatments and HCs. Pooled analysis without considering the defined phenotypes for Tfh cells revealed a significant higher Tfh cells proportion in u-RA compared to the HC subjects (SMD 0.671, [0.461, 0.882], p < 0.0001, I 2 = 52.73, p = 0.003) (Figure 3). Egger's test also showed no publication bias (p = 0.32). Furthermore, we performed subgroup analysis on the studies that defined Tfh cells as 'CD4 + CXCR5 + ', 'CD4 + CXCR5 + PD-1 + ', ' CD4 + CXCR5 + PD-1 high ', ' CD4+ CXCR5+ ICOS high ', ' CD4 + CXCR5 + PD-1 + ICOS + ' and 'CD4 + CXCR5 + CCR7 low PD-1 high ' (Supplementary Fig. 9 A-F). As there was a maximum of 1 study per group for other  Supplementary Fig. 9 A-F). In addition, based on Egger's test, no evidence of publication bias was found for the studies (p > 0.1, Table 3)

The proportion of Tfh cells in PB of early RA compared with HC
Then we questioned whether the proportion of Tfh would differ between patients and HCs when the stage of the disease  (Table 1).   (Table 1).
course was taken into consideration. We included 3 articles    (Table 1).  Supplementary Fig. 10 A, B)

The proportion of Tfh in the PB of patients with active RA, patients in remission, and HCs
We next asked whether the proportion of Tfh cells of the blood CD4 + T cells, could be different between patients with active RA (a-RA) and those who were in remission (r-RA). Eight studies compared the proportion of Tfh cells between a-RA and r-RA, from which 4 studies reported higher proportion of Tfh in a-RA, 3 studies reported no differences in the Tfh cells proportion between the two groups, and 1 study reported higher Tfh cells proportion in r-RA ( Table 5). The pooled SMD result of Tfh proportion regardless of its definition revealed that there was an increasing trend toward significant in the Tfh cells proportion in the PB of a-RA in comparison with r-RA (SMD 0.275, [−0.012, 0.563], p = 0.06, I 2 = 0%, p = 0.432) ( Figure 5.A). No risk of publication bias was found based on the egger's test (p = 0.155). As evident from Table 5, different set of markers was used by the authors to evaluate the Tfh cells proportion. Thus, we also did sub-group meta-analysis on these studies to compare a-RA with r-RA. Two studies used 'CD4 + CXCR5 + ' [21,36] and 2 used 'CD4 + CXCR5 + ICOS + ' [36,39]. Four remaining studies evaluated the proportion of 'CD4 + CXCR5 + PD-1 + ' [39], 'CD4 + CXCR5 + PD-1 high ' [21], 'CD4 + CXCR5 + CCR7 low PD-1 high ' [33], and 'CD4 + CXCR5 + Foxp3 − ' [39] cells, and because meta-analysis could not be performed on only one study, they were removed . Study identifier (α, β, γ) was used to distinguish each study (Table 2).  Fig. 11 A, B) We also performed the same analysis to compare a-RA patients with HC group. Total of 8 studies compared the proportion of Tfh cells between a-RA and HC groups. Of these, 6 studies showed significant increase in the Tfh cells proportion in a-RA, and 2 remaining studies showed no differences (Table 5) Table 6, supplementary Fig. 12 A, B).

The proportion of Tfh in PB of seropositive RA patients versus seronegative RA patients
Seropositive patients were defined as anti-citrullinated protein antibodies (ACPA) + and/or rheumatoid factor (RF) + . Whether Tfh cells proportion among PB CD4 + T cells is different between S + RA and S − RA patients was evaluated. Five studies assessed the proportion of Tfh in S + RA compared with S − RA patients, of which three studies showed higher proportion of Tfh in S + RA, and 2 other showed no difference (Table 7). Pooled meta-analysis of all 5 studies, regardless of the Tfh definition revealed a significant increase in the proportion of Tfh cells in PB of patients with S + RA compared to that of those with S − RA (SMD 0.637, [0.271, 1.003], p = 0.0006, I 2 = 0%, p = 0.831) ( Figure 6A). In addition, based on the egger's test, no publication bias was found (p = 0.932). It should be noted that 4 out of the 5 studies were evaluated Tfh cells proportion in 'early RA' patients. In addition, since only two studies used the same pattern of Tfh definition (CD4 + CXCR5 + PD-1 high ) to compare S + RA and S − RA patients, we performed a sub-group meta-analysis on these studies. The sub-analysis revealed higher proportion of Tfh cells in PB of S + RA compared with S − RA patients when Tfh was defined as 'CD4 + CXCR5 + PD-1high '(SMD 0.67, [0.26, 1.07], p = 0.001, I 2 = 0%, p = 0.93) ( Table 8, supplementary Fig. 13). Table 7. Characteristics of the studies that compared Tfh cells proportion between S + RA and S − RA.

Author
Year   There was insufficient number of articles to compare Tfh cell proportion between S + RA patients and the HCs.

Geographic location of studies and Tfh cells proportion in PB of RA patients
Finally, we assessed the proportion of Tfh cells based on the geographical regions where the studies were performed. Regarding the geographic location of the included studies 14 studies were performed in China, 4 in Japan, 2 in Spain, 1 in Sweden, 1 in Italy, 1 in USA, and 1 in Argentina (Table 1) to compare RA with HC. We divided the studies into 2 groups, Asian and non-Asian studies. There were 18 articles covering 30 studies on Asian populations (Japan and China), and 6 articles covering 9 studies on the non-Asians ( Table 1). The pooled SMD revealed that there was higher Tfh cells proportion in RA patients compared to HCs in Asian populations (SMD 0.849, [0.649, 1.049], p < 0.0001, I 2 = 58.37, p < 0.0001) (Supplementary Fig. 14). There was also a higher proportion of Tfh cells in patients compared to HCs in non-Asians (SMD 0.225, [0.017, 0.432], p = 0.033, I 2 = 41.76%, p = 0.088) (Supplementary Fig. 15). There was no evidence of publication bias in the Asian as well as non-Asians analyses based on egger's test (p = 0.24, and 0.75, respectively).

Discussion
The main function of Tfh cell is to help B lymphocytes to proliferate, differentiate and promote antibody production, and regulate humoral immunity [5]. The changes in the chemokine receptors enable Tfh to localize near and interact with B cell, and promote GC formation and class-switched highaffinity antibody production [5,8]. However, Tfh deregulations can drive abnormal germinal center, B-lymphocyte differentiation and survival, autoantibody generation, and thus could be related to the development of autoimmune diseases [11,46]. Due to the difficulty in the sampling from human lymphoid tissue [5,7], and since PB Tfh cells in terms of phenotypic and B lymphocyte helper functions are similar to the GC Tfh cells [15], analysis of cTfh cells in patients has become an important clinically significant alternative strategy [17] RA is a systemic autoimmune disease characterized by the production of a large number of autoantibodies, such as ACPAs, RF and others [3], and this autoantibody production may be related to Tfh cell abnormality. In this line, while several studies have evaluated the frequency of Tfh in the PB of patients with RA compared to HCs, conflicting conclusions have been reached in some cases. The purpose of this work was to systematically evaluate the proportion of Tfh in PB of RA patients compared to HCs to clarify the proportion of cTfh in RA patients.
Meta-analysis of all included studies neglecting the definition markers, treatment status, disease activity status and serological status of the RA patients revealed a significantly higher Tfh cells proportion in the PB of patients with RA in comparison with HCs. So, in this review, we have analyzed the earned results of previous studies based on the above criteria. Our sub-analyses revealed that subgrouping the studies based on these criteria were associated with reduction in the strength of heterogeneity in most of analyses which reached to moderate, low or even no heterogeneity (Tables 2,3, 4, 6 and 8).
In the sub-group analyses, at first we performed the metaanalysis based on only definition markers used to identify Tfh cells. According to the results, the proportion of Tfh cells with the all definition markers was significantly higher in the PB of RA patients compared to HCs (Table 2). From different definitions, 'CD4 + CXCR5 + PD-1 high ' and then 'CD4 + CXCR5 + PD-1 + ' cells showed the largest SMD, and while 'CD4 + CXCR5 + ' were significantly increased in RA, it showed the lowest SMD, suggesting the importance of considering the expression level of PD-1 in RA patients.
PD-1 can promote GC B cells survival and high-affinity longlived plasma cell formation through its interaction with programmed death-ligand 1 (PD-L1) and PD-L2 on the surface of GC B cells [47,48]. Studies have shown that the expression of PD-1 is elevated on Tfh cells in patients with autoimmune diseases including RA [39,49], and PD-1 high Tfh cells have a stronger ability to activate B lymphocytes [49]. Consistently, a positive correlation has been found between the expression of PD-1 on Tfh cells, and disease activity of RA [39]. There is also a significant positive correlation between serum level of the sPD-1 and the frequency of cTfh cells, titer of auto-antibodies and DAS in RA patients [29,50]. Indeed, the inhibitory function of the increased membrane-bound PD-1 on Tfh cells is blocked in the presence of its soluble form, while its humoral assistance to antibody producing cells is intact or even hyper activated. The results of our meta-analyses revealed significantly higher proportion of Tfh cells in all the comparisons which have included PD-1 high for Tfh cells (RA versus HC, u-RA versus HC and S + RA versus S − RA). Furthermore, based on subgroup analyses, the definitions which included PD-1 ('CD4 + CXCR5 + PD-1 + ' and especially 'CD4 + CXCR5 + PD-1 high ') had the largest SMDs in comparison with the other definitions suggesting more associations between the proportion of these phenotypes and RA. Thus, in RA patients, the higher expression of PD-1 on the Tfh cells could be associated with the higher proportion of these cells.
ICOS is another surface molecule that has a pivotal role in the development of Tfh cells and also in the production of IL-21 as signature cytokine of Tfh [51]. In addition, ICOS signaling is essential for maintenance of anatomical localization of Tfh cells in B cell follicles through preserving the expression of homing receptor pattern on Tfh cells [52,53]. ICOS also promotes survival and functional maturation of GC B cells [54]. Furthermore, Tfh cells expressing the highest expression level of ICOS have the most capacity of inducing IgG production [55]. Our study demonstrated that Tfh cells expressing ICOS (CD4 + CXCR5 + ICOS + and CD4 + CXCR5 + ICOS high ) ( Table 2) were higher in RA versus HCs. In addition, 'CD4 + CXCR5 + ICOS high ' proportion had higher association with RA than 'CD4 + CXCR5 + ICOS + ' proportion as was evident from SMDs ( Table 2) which, could be justified by the above-mentioned roles defined for ICOS molecule. Considering the importance of ICOS in antibody production, unfortunately there were not at least 2 studies with CD4 + CXCR5 + ICOS +/high definitions for comparing Tfh cells proportion based on serostatus subgroup analyses.
We also analyzed the studies that only included untreated patients to exclude the effect of immuno-suppressive drugs. Higher proportions of Tfh cells regardless of Tfh cell definition, as well as when different pattern of definition markers were used observed in u-RA compared with HCs. Of comparisons from different definitions, the highest association was found for 'CD4 + CXCR5 + PD-1 high ', and 'CD4 + CXCR5 + ' had the lowest SMD, again suggesting the importance PD-1 molecules in defining the Tfh cells.
The proportion of Tfh cell was also evaluated based on the stage of the disease course. Studies have shown that the immunological aberrations during the first few months after the disease onset differ from those during later phases [56]. Due to the importance of the early diagnosis of RA and also to clarify the status of Tfh cell in the early phase of the disease, we also compared Tfh cells proportion in e-RA patients with that of HCs. All e-RA patients included in the selected articles were treatment naïve. The pooled result of the all Tfh definitions as well as subgroup analysis of Tfh cells with definitions of 'CD4 + CXCR5 +' , CD4+ CXCR5+ PD-1 + showed greater Tfh cells proportion in untreated e-RA compared to HCs, suggesting pathogenetic role of Tfh in initial stage of the RA development.
S + RA is defined as positive for RF or ACPAs and seropositivity is associated with more severe disease [3]. The pooled meta-analysis results, regardless of the Tfh definition demonstrated that S + RA patients had a significantly higher proportion of Tfh cells in their PB compared to the either S − RA patients. Also, markers-based subgroup analysis showed that the proportion of Tfh cells was significantly elevated in S + RA in comparison with S − RA patients when Tfh cells were defined as CD4 + CXCR5 + PD-1 high . This can further emphasize the previously mentioned importance of PD-1 expression level in the frequency of Tfh cells and autoantibody production in RA patients. Altogether, these results demonstrate that the production of auto-antibodies in RA patients is associated with the proportion of CD4 + CXCR5 + PD-1 high Tfh cells, which provides evidence for the connection between the seropositivity and the proportion of these cTfh cells.
Looking at different definitions used in all comparisons demonstrates that the largest SMDs for proportion of Tfh cells in comparison groups belong to the Tfh cells expressing PD-1 (PD-1 + and especially PD-1 high ), while the lowest association was related to the 'CD4 + CXCR5 + ' cells.
We also evaluated the quality and the accuracy of the data conversion by the GetDat Graph Digitizer software. Authors of some of the included articles reported numerical data in the text or table, in addition to presenting them in graph formats. In order to evaluate the accuracy and the precise of the data conversion, we converted these graphs to the numerical data. To ensure that the software cover all types of graphs, we converted different types of graphical results to the numerical data. By comparing the digitizer-extracted data from the graphs with the numerical data reported by the authors in the paper, we evaluated the quality of our conversion. Based on the comparison, there was a high degree of confidence because of the almost identical extracted data to the original reported data. Furthermore, GetDat Graph Digitizer software has been used as a reliable tool for conversion of graphical data to the numerical by other studies [57][58][59][60][61].
In the present study, high level of heterogeneity as well as high range in the percentage of Tfh cells in the PB of HC and RA patients were observed. These could be due to heterogeneity in diagnostic criteria, disease severity and activity, geographical regions of population of the studies, using different pattern/set of markers for definition of Tfh cells, and finally different gating strategies of flow cytometry that may affect the range of Tfh cell percentage for both patients and HCs.
Proliferation of auto-reactive B cells capable of generating high-affinity autoantibodies contribute to the pathology of autoimmune diseases and, thus, have led to consider Tfh cells as possible players in their pathogenesis. It should be noted that different auto-immune diseases, such as RA, SLE, MS, SjS, and MG are associated with different and specific profile of autoantibodies [17]. While the profile of autoantibodies involved in the pathogenesis of these diseases is different and specific, what is common in these disorders is the help and pathologic role of Tfh cells in the generation of disease-specific auto-reactive antibody-producing B cells. So, in different auto-immune diseases, disease-specific Tfh cells help disease-specific antibody-producing B cells. But, in the present study, the specificity of Tfh is not important, and regardless the exact specificity, its altered frequency and potential pathogenic role by providing the help for the development of these RA-specific B cell producing antibodies is important.
Studies have shown that Tfh cells have specific TCR repertoire which is determined by the cognate antigen [62]. In autoimmune diseases, an altered repertoire of T cells can contribute to the disease development and also can be used as novel diagnostic markers. For example, a recent study demonstrated that the repertoires of T cells between SLE, RA and HC groups are different and can be used as novel diagnostic marker [6364]. Consistently, animal study has demonstrated that Tfh repertoire is different in lupus-prone mice compared to HC mice, and the Tfh repertoire alteration is associated with the development of the disease [64]. Considering the importance of the presence of Tfh cells for generation of such autoantibodies and the specificity of Tfh cell repertoire, the evaluation of Tfh cell frequency or repertoire alternation potentially could be used as predictive marker, prognosis or selecting strategy of therapy.
The present work has some limitations. First, as disease duration was inconsistent across the studies we couldn't consider duration of the disease in the analysis. Second, the disease activity score of RA patients was not uniform through the studies. Third, due to heterogeneity of drugs used for treatment of patient, we couldn't evaluate the effect of treatment on the proportion of Tfh cells in RA. Fourth, insufficient number of studies with particular Tfh definitions prevented their evaluation in some of the comparisons. Fifth, among the patterns of Tfh cell definition, CD4 + CXCR5 + may contain CD4+ CXCR5+ FoxP3 + T follicular regulatory (Tfr) cells. Unfortunately, a number of study evaluated the proportion of this cell was not sufficient and similar to Tfh cells to do meta-analysis. Thus, the result of this definition should be interpreted with caution as CD4 + CXCR5+ may contain Tfr cells.

Conclusion
In conclusion, this is the first systematic review and metaanalysis that has assessed the proportion of PB Tfh cells in RA patients. Our analyses demonstrate that compared to HCs the proportions of Tfh cells in general, and in particular Tfh cells expressing PD-1 (PD-1 high and then PD-1 + ) are highly elevated in patients with RA. Therefore, this cell type likely plays a pathogenic role in RA and could potentially be used as biomarker for diagnosis, prognosis, or as a target for treatment.

Declaration of interest
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures
Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.