Dendritic cells (DC) play a pivotal role in the induction and maintenance of the immune response, linking innate and adaptive immunity (1). Together with monocytes, they form a heterogeneous group of professional antigen-presenting cells (APC) cells that display characteristic functional properties that together regulate the immune system towards both immunity or tolerance (2, 3). Although the maintenance of DCs is undergoing consistent dynamic change, monocytes, conventional DC (cDC) and plasmacytoid DC (pDC) have widely been accepted as the most dominant APCs in steady state peripheral blood (PB) (based on subtype prevalence) (4). Previously, DC classification has primarily been based on ontogeny and further demarcated by function and phenotype, resulting in three main subtypes derived from the common DC progenitor (CDP): type I conventional DCs (cDC1s), type II conventional DCs (cDC2s) and plasmacytoid DCs (pDCs) (5, 6). More recently, advances in single cell RNA sequencing (scRNA-seq) have further elaborated upon the ontogeny and heterogeneity of DCs. As recently summarized by Chen et al. (7) as well as Kvedaraite and Ginhoux (8), 4 main DC groups have now been categorized, namely the cDC1s, cDC2s and plasmacytoid DCs next to the DC3 (7, 8). Within our study, we focused on the CD141⁺ cDC1s, CD1c⁺ cDC2s and CD14+ monocytes. For simplification, we will refer to the subsets by their conventional nomenclature.

DCs are also present in tissues where they display a more mature phenotype than their blood counterparts (9, 10). However, despite this maturation difference, the main DC subsets align across tissues and species in steady state (11–14). Within these populations there is additional heterogeneity depending on, amongst others, developmental stages (11). In inflammatory settings APC phenotype and function can change (9). It remains poorly understood how the APC subtypes are functionally programmed at the site of inflammation. There is substantial support that APC may fuel the hyperactive immune response in human autoimmune disease by driving activation and differentiation of auto-reactive effector T-cell populations (15, 16). Recruitment of specific DC and/or monocytes in inflamed tissues may play a key role in the pathophysiology of the disease promoting T helper cell (Th) polarization depending on the environment. In JIA (the most common inflammatory arthropathy in children), severe chronic inflammation in the synovium leads to joint destruction and additional extra-articular manifestations (such as uveitis), as well as loss in quality of life for patients (pain and reduced functional ability in daily life) (17). Sessions with joint puncture, aspiration and intra-articular corticosteroid injection are part of early clinical management of JIA, providing access to SF exudate which can be used for quantification and analysis (17).

When compared to healthy controls, a study showed that patients with JIA have decreased PB DCs which are associated with a poor clinical outcome (18). In contrast, an enrichment of cDCs is seen in the SF from inflamed joints, with a substantial expression of maturation markers, suggesting initiation and maintenance of inflammation (18). As mentioned above, local environments and stimuli can initiate distinct modification to DC phenotype and function. When examining JIA patient SF exudate samples, a recent study found that total T cell, pDCs and cDC1s were increased with the latter being barely detectable in PB samples (6). Similarly, in the inflamed synovial joint of rheumatoid arthritis (RA) patients, cDC1s were found to be enriched and capable of activating CD4⁺ and CD8⁺ T cells, contributing to inflammation (19).

A working inflammatory arthropathy (IA) model (based on (pediatric and adult) human and murine models) suggests that autoantigen specific CD4+ T cells primed by DC enhances T cell maturation and presentation of neo-epitopes (auto-antigens). As such, autoreactive and potentially cross-reactive polarized T cells propagate autoimmune arthritis, contributing to the effector phase of the disease (20). Interestingly, at the site of inflammation in chronic arthritis, the ability of SF-derived DC from RA patients to stimulate autologous PB T cells was shown to be superior to both PB-derived DC, and PB- and SF-derived monocytes (21). Additionally, cDC2 from SF were demonstrated to be able to induce T cells to proliferate, and produce IFNγ and IL-17 (22), emphasizing their pro-inflammatory nature. Remarkably, in SF, there is a very large contribution of cDC1 to the pool of DC (~35% compared to 3% in PB) (23). cDC2 were also shown to be increased in SF (64% SF vs. 33% in PB), while pDC showed a decrease (33% PB vs. 9% SF) (23). In steady state, cDC1 have been shown to be the counterpart of the murine CD8α⁺ and CD103⁺ DC (24). However, while murine CD8α⁺ DC have been demonstrated to be the most efficient at cross-presentation of all APC, human cDC1 have superior exogenous and necrotic antigen cross presentation and produce increased amounts of type III interferons (24–26). Additional functions including IFNλ production, induction of cytotoxic T lymphocytes (CTL) (27), and potential tolerogenic features (28), suggest that cDC1 may play specific roles in immune responses.

The role of cDC1 has not been extensively described in reference to the pathogenesis of IA in general, but they were demonstrated to be distinct from PB counterparts and are capable of inducing CD4⁺ T cell responses and activating synovial fibroblasts (19).

Here, we aimed to determine whether there is distinct functional programming of human APCs in synovial fluid from inflamed joints of JIA patients. We investigated their gene expression profile and performed extensive phenotyping and functional assays to correlate the transcriptomic signatures to functional outcome. We show that while cDC2 and monocytes at the site of inflammation have a related but non-overlapping pro-inflammatory profile, the enriched cDC1 are more quiescent and less pro-inflammatory. Together the data demonstrate differential responsiveness and functional programming of cDC1s and cDC2s under inflammatory conditions.

Human cDC1 are highly enriched in SF and phenotypically distinct from other APC.

To characterize human APC under inflammatory conditions, we analyzed APC subpopulations in SF obtained from inflamed joints of juvenile arthritis patients and compared these to paired peripheral blood samples. Within HLA-DR⁺CD11c⁺lineage^- cells we distinguished 3 subpopulations for further analysis: cDC1, cDC2, and CD14⁺ monocytes ( Figure 1A ).

In SF the percentage of CD14^- DC from total APC was significantly higher than that in PB, although the majority still consisted of CD14⁺ monocytes ( Figure 1B , left panel). Furthermore, in SF we found a striking enrichment of cDC1 resulting in percentages comparable to those of cDC2 ( Figure 1B , right panel). Similar ratios of cDC subsets ( Figure 1B , right panel) were observed in SF-derived from inflamed joints of adult RA patients, indicating that our findings are not JIA-specific.

Next, we performed gene-expression profiling by RNA sequencing. Principal component analysis revealed that each APC subset plots separately, while the four replicates of each subset clustered together ( Figure 1C ). In this PC1-PC2 2D space, monocytes and cDC2 were closer to each other, indicating higher similarity. Unsupervised clustering analysis also demonstrated that cDC1 are more distinct from the other two subsets, while cDC2 and monocytes, although clustering separately, share certain features ( Supplementary Figure 1B ). We then further compared these subsets based on published markers specific for DC subsets in steady state (11, 12, 14, 35). These markers could indeed separate the three APC subsets, based on hierarchical clustering ( Figure 1D ). In concordance with the variable genes cluster analysis ( Supplementary Figure 1B ), cDC2 and monocytes partly overlapped in expression of certain genes ( Figure 1D ). When directly comparing cDC2 and monocytes without cDC1 ( Supplementary Figure 1C ), it is well possible to separate these two cell types. Since inflammation could influence marker expression, we compared expression during steady state and inflammation: 4 out of 5 genes were specific for cDC2 in inflammation, and 5 out of 13 genes remained specific for monocytes during inflammation ( Supplementary Figure 1D ; underlined). Based on our data, a signature of nine markers can optimally distinguish between monocytes and cDC2 in both inflammation and steady state (cDC2: IRF4, CD1c, FCER1A, CLEC10A; monocytes: ZEB2, FCGR2A, TLR4, MAFB, EMR1). Additional phenotyping was performed by assessing protein expression of various surface markers by flow cytometry [as reviewed by Schlitzer et al. (3)]. SF cDC1 expressed hallmark proteins CLEC9A, XCR1, and CD26 ( Figure 1E , mRNA and F, protein), while hardly expressing any of the surface markers strongly upregulated in SF monocytes and cDC2 ( Figures 1G, H ). Neither the expression of CLEC9A and CD26 nor monocyte-related markers differed between PB and SF ( Figure 1F and Supplementary Figure 1E ).

In contrast, the expression of monocyte-related markers was lower in PB cDC2 than in their SF counterparts ( Figures 1H, I ). However, our principal component analysis and clustering analysis, both taking into account thousands of genes in addition to these APC markers, demonstrated that both PB and SF cDC2 are closer to each other than they are to PB monocytes ( Supplementary Figures 1F, G ).

Until recently, monocytes and DC have mostly been characterized in steady state, however it remains unclear how these cells – especially in humans – are programmed under inflammatory conditions (2, 6, 11, 14, 18, 19). To fill this gap in knowledge, we have comprehensively profiled monocytes and cDC subsets in the exudate derived from the inflamed joint of JIA patients. Although SF from the joints of arthritis patients, may not be a complete reflection of the inflammation taking place in the tissue, it has proven to be an excellent model to study APC in a highly inflammatory environment (6, 22). Our findings show that at the site of inflammation, there is specific, functional programming of human DC subsets. In this context both monocytes and cDC2 demonstrate a clear and distinct pro-inflammatory role whereas the enriched cDC1 remain surprisingly quiescent on both a transcriptional and functional level. Despite some overlap in transcriptional profile and phenotype, we propose that inflammatory cDC2 are not monocyte-derived, but bona fide conventional DC. Based on principal component analysis and unsupervised clustering analysis, CD14, CD16, CD1c expression, but also IRF4 and IRF8 levels can be additional markers used to further distinguish cDC and monocytes (12). Recently, additional classifications have been made to DC subsets. Amongst others, Villani et al. (11) demonstrated in steady-state, sub-division of the CD1c/BDCA-1⁺ cDC2 delineate DC2 (CD5^+/-) and DC3 (cD1c⁺CD5^-CD163⁺CD14^lo-hi). The latter is of utmost interest as it strongly expresses a pro-inflammatory gene program represented by CD14, S100A9 and S100A8 and has also been implicated in the pathogenesis of systemic lupus erythematosus (SLE) (7). Although we have not yet incorporated these distinctions into our research, it is impertinent to further explore DC3s in inflammatory arthritis.

In our transcriptional data-set, both SF cDC2 and SF monocytes were very similar to their PB counterparts, indicating these SF subsets likely originate from the blood and did not differentiate from other APC subsets (monocytes) at the site of inflammation. This may seem in contrast to what Segura et al. (21), described earlier, showing a population of inflammatory DC in RA SF and inflammatory tumor ascites that they proposed to be monocyte-derived (21). This distinction may be explained by a difference in DC sorting strategy. We indeed observed CD1c expression on SF monocytes in our JIA patients, albeit at lower levels than on the CD14^-CD1c⁺ selected population and this was confirmed on the mRNA level. Supporting this hypothesis, our SF CD14⁺ monocyte population, similar to the inflammatory DC described in the paper by Segura et al. (21), produced the highest levels of the T_H17-promoting cytokines IL-6, TNF, IL-23, and IL-1β and demonstrated the strongest induction of IL-17 production by T cells. In addition, these cells were distinguished by strong expression of amongst others acute and chronic inflammatory genes, and had strong T cell-activating capacity (11). In contrast, CD14⁺ CD1c⁺ DC have been shown in blood of healthy controls and melanoma patients, albeit showing immune-inhibitory functions (41). In cancer settings, it was also demonstrated that pre-cDC1, lineage-committed progenitors that give rise to cDC1, express the T helper type 1-associated chemokine receptor CXCR3 for homing functions to the melanoma. Preservation of CXCR3-ligands CXCL9 and CXCL10 promoted cDC1 presence in tumors, highlighting a mechanism that may reflect similarly increased levels of CXCL9 and CXCL10 in SF (42).

The expression profile of cDC1-associated markers on a transcriptional and protein level confirmed that our SF-derived CD141⁺ cells consisted of bona fide cDC1 (39, 43, 44). This was further supported by the limited response to a number of TLR ligands and specifically, the production of IFNλ upon poly I:C exposure – cDC1 features that both have been described before in steady state (38, 45). Together, our findings suggest that although the APC subsets undergo specific inflammation-induced programming and differentiation/activation, the different APC subsets are still clearly distinguishable based on their steady state-defined core signature. Single-cell sequencing will shed further light on the heterogeneity and differentiation within the subsets.

We here demonstrate that SF cDC1 did not show a strong pro-inflammatory profile and induced only weak T cell proliferation and cytokine production compared to the other DC subsets. One reason for this relative unresponsiveness may be the low expression of PRR, chemokine, and cytokine receptors on the SF cDC1 that we observed. cDC1 have been reported to play a major role in cross-presentation, i.e., the presentation of acquired exogenous antigens via MHC class I which is essential for the initiation of CD8+ T cell responses (25, 46). The low MHC class I expression by cDC1 in our study may affect CD8 T cell responses, but that should be assessed with antigen-specific T cell responses and preferably also in a cross-presenting setting. However, although cross-presentation is one of the key features of cDC1 in mice, in humans it is not restricted to the cDC1 lineage (22, 24). In mice, tolerance induction has been a proposed function of cDC1, by inducing Treg (26, 47), for example via the production of retinoic acid (26). In humans, this potential has also been reported (27), describing CD141+ DC that induced T cell hypo-responsiveness and potent regulatory T cells. It is not clear however whether these are bona fide cDC1 or other cells expressing CD141, since expression of other cDC1 markers like XCR1 was not shown, and the cells were also CD14⁺. Additionally, these cells were found to produce IL-10 spontaneously, while the SF cDC1 in our study did not produce IL-10, in contrast to cDC2 and especially monocytes (data not shown). Furthermore, the aldehyde dehydrogenase genes 1-3 for the proteins needed for retinoic acid production (47) were expressed by SF cDC1, but lower than or equal to that of the other subsets (data not shown), indicating that this may not be a major mechanisms of tolerance induction by SF cDC1. Interestingly, one cytokine that was restricted to the SF cDC1 was IFNλ. IFNλ was produced by SF cDC1 upon poly I:C stimulation, but was also found in SF plasma, as measured by Luminex, about 10x as high than that in PB plasma of patients with active disease (478.1 ± 394.7 vs. 47.88 ± 79.98, respectively; data not shown). This cytokine has been shown to induce the expansion of Treg (48, 49). Mennechet et al., showed that the effect was indirect via inducing tolerogenic monocyte-derived DC, that in turn induced proliferation of Treg (49). Additionally, IFNλ has been suggested to have a dynamic in the progression of inflammation and autoimmunity specifically (50–52).

Research to the implications of APC differentiation and specialization in the pathogenesis and progression of autoimmune diseases, especially inflammatory arthritis is gaining momentum (11, 17, 19, 48). Our data contribute to the understanding of functional contribution and dynamics of APC subsets in inflammatory environments. Both cDC2 and monocytes seem instrumental in the pro-inflammatory process whereas the relative quiescence of cDC1 in combination with their selective IFNλ production suggest a potential regulatory role for this enriched DC subset in tissue inflammation.

We recognise that a limitation of this study lies in the selection and therefore possible exclusion of certain subsets based on the markers we have incorporated, as well as the diversity of patient samples (disease subtype and treatment status) included. However, we believe that the consistent finding of enrichment of cDC1s within the joint and their quiescent phenotype across all patient samples, is biologically relevant and novel. This finding is robust despite the heterogeneity in cell subtype, disease subtype and treatment status. Exploration of subsets such as DC3s is required to further elaborate upon the functional state of DCs in inflammatory arthritis and may lead to identification of potential propagators of chronic inflammation and pathogenesis.

Together, insights in the tolerance capacity of DC may be harnessed for potential immunomodulatory therapeutics in IA patients (53), allowing for long-term, efficient treatment that can be tailored to patients’ unique cases.

The data presented in the study are deposited in the GEO repository, accession number GSE220264.

The studies involving human participants were reviewed and approved by The Institutional Review Board of the University Medical Center Utrecht (UMCU) METC No.11-499/C (Pharmachild Study). Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

Conceptualization: AB, MP, SN, and FW; Selection of patients, acquisition patient data, synovial fluids and peripheral blood: JS and SV; Methodology: AB, and FW; Software: MM; Formal Analysis: AB, MM, BC, and MC; Investigation: AB; Resources: AB, AS, and FW; Writing – Original Draft: AB, AS, and MP; Writing – Review & Editing: AB, AS, MP, MM, BC, SN, JS, MC, SV, JL, and FW; Visualization: AB, AS, MP, and MM, and BC; Supervision: FW; Funding Acquisition: FW and JL. All authors contributed to the article and approved the submitted version.