Dendritic cells (DCs) are professional antigen-presenting cells (APCs), which orchestrate immune responses. They provide a link between the innate and adaptive immune system by translating non-specific danger and damage signals into a targeted antipathogen response (1). DCs continuously sample their environment by micropinocytosis allowing uptake of both self and non-self proteins, which are subsequently trafficked to the cell membrane and displayed in an antigen–major histocompatibility complex (MHC) complex (2). DC receptors can sense danger, microbe, or cytokine signals, which, when triggered, drive DC maturation and activation. This promotes migration to lymph nodes, further proinflammatory cytokine release, increased stability of the antigen–MHC complex, and upregulation of co-stimulatory molecules, such as CD40 and CD86. Together, these support a targeted T cell response against the presented antigen. Thus, if activated/mature DCs present self-antigen tolerance may be breached highlighting the need for tight regulation of DC biology and their importance in autoimmunity (2, 3). Some DCs, while retaining APC capacity, are resistant to maturation-inducing signals and downregulate co-stimulatory molecule expression and pro-inflammatory cytokines while simultaneously upregulating the expression of inhibitory molecules and anti-inflammatory cytokines (4, 5). These so-called regulatory or tolerogenic DCs promote T cell anergy and regulatory T cell (Treg) generation and are being explored as a potential therapy in autoimmune disease (6–11).

Dendritic cells are a heterogeneous population and can be divided into conventional myeloid (cDCs) and plasmacytoid (pDCs) (12). cDCs express typical myeloid antigens but can be sub-divided by their expression of CD141 and CD1c, thereby generating two additional subsets; CD141+ DC (cDC1) and CD1c+ DC (cDC2), respectively. pDCs typically lack these markers, although can express low levels of CD141, and express instead CD123, CD303, and CD304. Typically both cDC and pDC subsets display APC capacity, but this function is emphasized in cDCs (13). CD141+ DC have been suggested to cross-present antigen and efficiently prime CD8+ T cells and CD1c+ DC are capable of priming CD4+ and CD8+ T cells (14). Conversely, pDCs have relatively reduced HLA-DR (MHC class II) expression (13) and are the primary type 1 interferon (IFN-I) producing cell subset, a key early cytokine in the immune defense against viral infection (15).

Dendritic cells are believed to be important in RA pathogenesis, indeed, HLA-DR variants are linked to RA susceptibility implicating DC activation of autoreactive lymphocytes in disease onset (16). Furthermore, another DC subset termed inflammatory DC, which are thought to be derived from monocytes, have been reported to be the main DC subset in the synovial fluid of RA patients and are involved in the induction and maintenance of Th17 cell responses (17–23). There is additional evidence that cDCs are involved in early breach of tolerance in animal RA models (24). However, previous studies have often not distinguished between CD141+ cDC1 and CD1c+ cDC2 subsets and referred to them collectively as myeloid cDCs or just examined CD1c+ DCs alone. Indeed, CD1c+ DCs have been found at high levels in synovial joint tissue and fluid from RA patients where they can promote pathological Th1 cytokines (25, 26). With regard to circulating CD1c+ DC in RA patients there are contrasting reports. Most studies have reported a decrease in their numbers with a comparatively immature phenotype (25, 27–31), whereas one reported an increase in number (32). Additionally, in established RA patients, CD1c+ DC levels inversely correlated with disease activity and low DC numbers were restored upon successful treatment and fall in disease burden (29). However, this was not a universal finding for all therapeutic regimens (30). Examination of CD141+ DCs specifically in RA has been limited to one published abstract where they are increased in established RA synovial fluid, displaying a relatively mature phenotype (33). Nonetheless, studies of the presumed murine equivalent demonstrate an acceleration in the onset of collagen-induced arthritis following their adoptive transfer with CD4+ T cells (34).

Conversely, while pDCs are also increased in the synovial compartment and reduced in the peripheral blood of established RA patients, the circulating pDCs are immature and numbers do not correlate with disease activity (25, 27, 29, 31, 35). Furthermore, when compared with CD1c+ DCs, synovial pDCs also have a more immature phenotype (25). Some, therefore, propose that, in RA, pDCs have an anti-inflammatory function in the context of breach of tolerance (35–37). However, this role has not been universally supported (38, 39) and the net role of pDCs in RA pathogenesis remains ambiguous. Their capacity to produce large amounts of IFN-I is likely to be important given the emerging association of the interferon gene signature (IGS) with autoimmunity. The IGS is a composite score of genes upregulated upon exposure to IFN-I [interferon response genes (IRGs)]. In both early and established RA, a subset of patients have a raised IGS, which impacts on the clinical response to certain therapies (40–45). Exposure to IFN-I is likely to be important as genetic variants that increase RA susceptibility are associated with the IFN-I pathway (46, 47) and upregulation of genes related to the IFN-I signaling pathway predicted progression to RA in seropositive arthralgia (48, 49). Furthermore, administration of IFN-α can promote an inflammatory arthritis phenotype in humans (50) and transfer of IFN-I producing DCs propagates a persistent inflammatory arthritis in mice (51). RA therapies can influence the IGS (44, 52) and, therefore, complicate study of IFN biology in RA.

Given the central role of DCs in all immune responses, and their proposed role in RA and other autoimmune inflammatory diseases, we examined CD1c+ DC, CD141+ DC, and pDCs in patients in the early stages of RA, before treatment with antirheumatic drugs.

We report for the first time an extensive examination of peripheral blood pDCs, CD1c+ DCs, and CD141+ DCs in early, drug naïve RA patients including the transcriptomic analysis of 600+ immunology-related genes in pDCs and CD1c+ DCs. Our data suggest distinct roles for DC subsets in early RA pathogenesis, an understanding of which may have important therapeutic implications.

Most investigations of the pathogenic role of pDCs in autoimmunity are focused on their marked IFN-I-producing potential and its impact on B cell function. Indeed, sustained IFN-α production from RA pDCs was sufficient to induce autoantibodies in vivo (29), and there is a high IGS in established RA synovial fluid (38, 56). We previously showed an increased IGS in early RA (44) and, when examining pDC and CD1c+ DC transcriptomes, we found increased expression of IRG, such as IFITMI or IRF4, supporting this observation. IFN-α has been shown to reduce IL6R expression in some cell lines (57), which is in keeping with our observation of reduced IL6R in both pDC and CD1c+ DC subsets. Furthermore, pDCs had reduced IFNAR1 expression, which is reduced upon ligation (58), again suggesting increased IFN-I signaling in the early RA cohort. However, when examining the relative contribution of each lymphocyte subset to the IGS as determined by five IRGs, there was no obvious subset where upregulation of these transcripts dominated. A previous study in RA examining three IRGs suggested that granulocytes may be the major contributor of the IGS although DCs were not examined in that study (59). There are thousands of potential IRGs, however (60), and their upregulation may be both ligand- and cell-dependent (61). Potentially, we did not examine the optimal IRGs to dissect a cell specific effect of IFN-I signaling or exposure.

Despite their IFN-I producing potential, we did not see association between peripheral blood pDC phenotype/frequency and the IGS in early RA. Type III interferons (IFN-III or IFN-λ) are also produced by pDCs and promote upregulation of genes normally associated with response to IFN-I (54, 62, 63). Furthermore, IFN-III are increased in RA and can associate with disease-specific antibodies (64, 65). However, pDCs had comparable IFN-I and IFN-III transcript levels to those in other circulating lymphocyte subsets, and these levels were independent of the background IGS. There are 13 IFN-α subtypes and expression of both IFN-α and IFN-λ subtypes are cell- and ligand-dependent (66), which may have contributed to these unexpected results, since we only examined two IFN-α transcripts (IFN-α2 and IFN-α1/13). Nonetheless, these data suggest that in early RA, circulating pDCs are not uniquely responsible for whole blood IGS generation by either IFN-I or IFN-III. Generation of IFN-I/III by pDCs may instead be primarily tissue-based, thereby highlighting the potential importance of migration and subsequent microenvironment on DC function. This is supported by the literature where pDCs are present in affected tissue, such as lupus nephritis or salivary tissue in Sjogren’s syndrome, where they locally produce IFN-I (67).

Unexpectedly, there was an inverse association between CD141+ DC frequency and the IGS, which was independent of CD141+ DC phenotype and circulating cytokines. This is the first time that CD141+ DCs have been examined in the circulation of early RA patients, nonetheless, CD141+ DCs are not recognized for their IFN-I production. They do, however, produce IFN-II upon stimulation (68) as well as large amounts of interferon-λ (IFN-III) following stimulation with TLR3 ligands (69–71). In SLE, there has been a suggestion that there is a large IFN-II component to the IGS (55) and, as previously discussed, IFN-III can upregulate IRGs normally associated with IFN-I exposure (54). DC activation and IFN production may also upregulate chemokine receptors resulting in trafficking of the CD141+ DCs out of the peripheral circulation. This all raises the possibility that CD141+ DCs may play an important role in RA IGS generation.

Plasmacytoid DCs and CD1c+ DCs are known to be reduced in the peripheral circulation of established RA patients, particularly in those with active disease, where they migrate to the synovial compartment (25–27, 29, 31, 35, 38, 72). CD141+ DC circulating frequency in RA has not been previously compared with healthy controls, but we have demonstrated a reduced frequency of all DC subsets in the circulation of early, drug-naïve RA patients, which was sustained up to 12 months. This emphasizes early involvement of these DC subsets in disease pathogenesis in keeping with previous observations in animal arthritis models where lymphatic CD1c+ DCs are important in breaching tolerance (24). Indeed, both CD1c+ DC and pDC transcriptomic analysis suggested increased DC proliferation despite their reduced frequency in the blood, supporting the suggestion of DC migration. Moreover, there was a further reduced pDC frequency in seropositive RA patients and these pDCs also had higher CCR7 expression, a key chemokine in promoting pDC migration (73, 74), which inversely correlated with pDC frequency. While CCR7 expression classically causes lymph node migration, its ligand, CCL19/CCL21, is significantly increased in the synovial fluid of RA patients (75) supporting the suggestion of early synovial pDC migration sustained into established RA. In keeping with our observations, seropositive established RA patients have a higher frequency of synovial fluid pDCs than seronegative patients, which co-localize with B-cells in synovial tissue (26). Synovial pDCs produce IFN-I (26) and we, therefore, propose that CCR7-driven migration of pDC into the synovial compartment in seropositive RA patients precedes localized IFN-I production and, potentially, consequent autoantibody generation. This mirrors what is known about intra-articular CD1c+ DC function where the microenvironment modifies DCs and subsequent T cell activation, in a manner that is not necessarily observed in the periphery (17, 20–22).

There were distinct differences in how the DC subsets associated with disease activity. pDCs demonstrated a comparatively immature phenotype in early RA and their frequency did not associate with disease activity, supporting what has been reported in established RA (26, 29). In contrast, both CD1c+ DCs and CD141+ DCs had an activated phenotype although only CD1c+ DCs showed an association with disease activity, a relationship again replicated in established RA (29). These observations were further corroborated by the transcriptional data, which highlighted the potential tolerizing potential of pDCs that is being increasingly appreciated in other disease states (76). Early RA pDCs were characterized by striking downregulation of TNFRSF17 with upregulation of PRDM1 and CSF1R. TNFRSF17 (BMCA) is increased upon TLR engagement (77) and is expressed at high levels on pDCs from multiple myeloma patients, where there is pathogenic expansion of plasma cells (78). Reduced expression could, therefore, be predicted to have immunoregulatory consequences. PRDM1 (BLIMP1) is increased in human pDCs in response to IFN-α and has been proposed as a mechanism to negatively control the production of effector cytokines, thereby skewing toward pDC tolerance (79). Notably, CD27 expression was minimal in the pDC cohort and markedly reduced compared with B cells, excluding plasma cell contamination. CSF1R is a tyrosine kinase receptor that causes pDC differentiation and proliferation (80–82). This pathway is also believed to be important in inflammatory macrophage differentiation. Blockers of CSF1R/CSF1 have been trialed as a potential therapy for RA (83) although the only completed phase 2 clinical trial reported little clinical benefit (84). Potentially, these therapies could also affect pDC proliferation and thus abrogate any tolerogenic function, thereby compromising clinical outcomes. In contrast, when compared with healthy controls, the most differentially increased genes in early RA CD1c+ DCs were upregulated CKDN1a and downregulated BTLA. CKDN1a is associated with an inflammatory and potentially pathogenic phenotype in mouse DCs (18, 85). BTLA is important in DC skewing of T cells toward Tregs (86) and reduced expression has been linked to an inflammatory, potentially pathogenic, monocyte-derived DC subset in RA (21, 22).

These findings are in agreement with current literature where CD1c+ DCs are likely to play a pathogenic role in RA (19, 20, 23); however, the role of pDCs in RA to date has been frequently contradictory. On the one hand, pDCs from RA synovial fluid activate T cells and trigger the production of pro-inflammatory cytokines (38), and intra-articular transfer of activated pDCs propagated an inflammatory arthritis phenotype in mice (51). On the contrary, mature peripheral blood pDCs from RA patients promoted differentiation of naive T cells into IL-10-secreting Tregs (35). Furthermore, depletion of pDCs in a RA model resulted in breach of tolerance with development of autoantibodies (36) and in various inflammatory arthritis models enhanced pDC recruitment and activation to the arthritic joints significantly eased arthritis, consistent with an anti-inflammatory or tolerogenic role (37). Our data go some way to resolving these apparently conflicting views, emphasizing the importance of anatomical microenvironment and disease stage on pDC function. Indeed, pathological processes in early RA are known to be distinct from those that dominate later (87, 88), and we observed a significant change in pDC, CD1c+, and CD141+ DC phenotype after 12 months of disease. This is in keeping with published data where established RA patients (on immunomodulatory therapy) had reduced pDC maturation compared with healthy controls (28, 31).

Finally, the understanding of the role of CD141+ DCs in RA has been mainly derived from mouse equivalent DCs (CD8+ CD103+ CD11b− DCs) (89, 90), which accelerated the onset of CIA when adoptively transferred with CD4+ T cells (34). They have been examined once in established RA where numbers were enriched in the synovial compartment and demonstrated an activated phenotype (33). This supports our observation of reduced CD141+ DC frequency in blood, with increased HLA-DR and CD80 expression. This suggests that CD141+ DCs, despite their relatively small numbers when compared with other cellular subsets, may contribute to disease pathogenesis.

In conclusion, we simultaneously examined, for the first time in early RA, pDCs, CD1c+ DCs, and CD141+ DCs. While all DC subsets are reduced, CD141 +DC, but not pDC, frequency inversely correlated with the IGS and may be a hitherto unappreciated source of IFN-I. Furthermore, pDCs have similar levels of IFN-I and IFN-III mRNA transcripts as other major leukocyte subsets. Additional marked differences exist between the DC subsets with regards to activation and relationship with disease activity. These hint that pDCs may have, in the peripheral circulation at least, a more tolerogenic role when compared with CD1c+ DCs in early disease. Future work is, therefore, justified to compare synovial pDCs and cDCs with those in the circulation to further elucidate the effect of microenvironment and disease duration on DC function and, potentially, to expose novel therapeutic avenues and targets.

This study was carried out in accordance with the recommendations of North East—Newcastle & North Tyneside 2 Research Ethics Committee with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the North East—Newcastle & North Tyneside 2 Research Ethics Committee (REC reference: 12/NE/0251).

FC performed all experiments, was involved in project development, and wrote the first draft of the manuscript. AS performed bioinformatics analysis. AA and CH assisted in experimental design. AP assisted with patient recruitment. IM and MK-S advised on manuscript structure and DC biology. JI was integral in project development and writing the manuscript. All authors reviewed manuscript content prior to submission.

There are no conflicts of interest. The submitted work was performed without of any personal, professional, or financial relationships that could potentially be construed as a conflict of interest.