Title: Transcriptome network analysis of human CD1c+ dendritic cells identifies an inflammatory cytokine-secreting subpopulation within the CD14+ DC3s that accumulates locally in type I IFN- negative autoimmunity to the eye

Background: Inflammatory subsets of CD1c+ conventional dendritic cells (CD1c+ DCs) are promoted by type I interferons (IFN), but the molecular basis for CD1c+ DCs involvement in conditions not driven by type I IFNs is unknown. Methods: Our objective was to use RNA-sequencing of blood CD1c+ DCs and high-dimensional flow cytometry of two cohorts of non-infectious uveitis patients and healthy donors to characterize the CD1c+ DCs population of type I IFN- negative autoimmunity. Results: We report that the CD1c+ DCs pool from patients with non-infectious uveitis (n=42) is skewed towards a transcriptional network characterized by surface receptor genes CX3CR1, CCR2, and CD36, but not CD14 . We confirmed these results by RNA-sequencing in another case-control cohort (n=36) and show that this gene network is controlled by NOTCH2-RUNX3 signaling, but is not conditional on CD14 surface expression. Unbiased flow cytometry analysis based on the transcriptional network identified a CD36+CX3CR1+ subpopulation within the CD14+ DC3 subset that was diminished in peripheral blood of patients. CD36+CX3CR1+ DC3s infiltrate eyes of patients and displayed a distinctive ability to produce high levels of inflammatory cytokines that were elevated in the inflamed eye, including TNF-alpha, and IL-6, but not IL-23. Conclusion: These results show that CD36+CX3CR1+ CD1c+ DCs are a subpopulation of inflammatory CD14+ DC3s implicated in type I IFN-negative human non-infectious uveitis that secrete proinflammatory mediators that drive its pathophysiology.


Graphical Abstract:
our understanding of the CD1c+ DC characteristics during autoimmunity is incomplete, especially for conditions not driven by type I IFNs.
Our objective was to use whole transcriptome profiling by bulk RNA-sequencing of peripheral blood CD1c+ DCs and multiparameter flow cytometry of two cohorts of non-infectious uveitis patients and healthy donors to characterize the core transcriptional features and subset composition of CD1c+ DCs in autoimmunity of the eye. We constructed data-driven coexpression networks that identified a bona fide CD1c+ DC transcriptional module in patients that helped identify a cytokine-producing inflammatory CD1c+ DC subset that diminishes from peripheral blood and infiltrates the inflamed eye in non-infectious uveitis.

An CD1c+ DC transcriptome module is associated with non-infectious uveitis
We aimed to characterize the transcriptome of primary CD1c+ DCs from patients with noninfectious uveitis ( Figure 1A). We isolated CD1c+ DC cells from blood of 28 adult patients with anatomically distinct types of non-infectious uveitis ( Table 1) Table 1). Patient samples did not cluster according to clinical parameters of disease activity (e.g., cell grade in eye fluid, macular thickness) (Figure 1 - Figure Supplement 2A-E). In contrast, the expression profile of the DEGs split the samples into clear patient and control groups in twodimensional PCA subspace ( Figure 1B). To detect a biologically relevant higher order organization of the transcriptome with sufficient resolution, we constructed co-expression networks by combining the DEGs with other uveitis-associated genes detected by differential expression analyses between the groups (P<0.05, n= 6,794 unique genes, Supplemental Table 1), which discerned 24 gene modules ( Figure 1C). Inspection of the module's eigengene values across the samples revealed that the 'black' module (1,236 Table   1). Because these cell-surface receptors and other genes encoding cell-surface proteins in the black module (e.g., ESAM, CLECL12A) have been associated with distinct subsets of cDC2s (24)(25)(26)(27), we focused on the 100 genes in the black module that encode cell-surface proteins (according to surfaceome predictor SURFY) (28) and their transcriptional regulation (Supplemental Table 1); In addition to CX3CR1, the scavenger receptor CD36, and Toll like receptor (TLR) family members TLR1, TLR4, TLR6-8, and CD180, all showed significantly (Padj<0.05) higher gene expression levels in CD1c+ DC cells of patients ( Figure 1E). Eight transcription factors previously shown to be linked to distinct human CD1c+ DC subsets (29) were also detected in the black module, including RUNX3, IRF8, and NFKB1 ( Figure 1E).
The gene module encompassing CX3CR1 and CD36 has been linked to a population within the CD1c+ DC pool with a 'monocyte-like' transcriptome as well as inflammatory CD14+ monocytes, which cannot be distinguished by bulk RNA-seq analysis (16,24,30). Although we observed CD14-positive cells in our CD1c+ fractions, there was no difference between patient and control samples (Figure 1F), nor was CD14 differentially expressed by RNA-seq (P > 0.05). Furthermore, CD14 correlated weakly with the black module (Pearson correlation coefficient = 0.25, Figure 2 -Supplement 1A). To test whether the black module was directly dependent on CD14 surface expression by CD1c+ DCs, we FACS-isolated blood CD14+ and CD14-fractions from CD1c+ DCs of six healthy donors (Figure 2A). RT-qPCR analysis revealed that the expression of a panel of key genes of the black module, including CD36, CCR2, CX3CR1, and RUNX3 were not significantly different between CD14+ CD1c+ DCs and CD14-CD1c+ DCs (Figure 2B), with the exception of TLR7 (Figure 2 -Figure   Supplement 1B). In addition, RNA-sequencing data of FACS-sorted CD14+ CD1c+ DCs and CD14-CD1c+ DCs fractions purified from PBMCs of patients with type IFN-I diseases SLE and Systemic Sclerosis further verified that -unlike CD14 -the majority of genes of the black module (including CD36, CX3CR1, CCR2, and RUNX3) were not significantly affected by CD14 surface expression during chronic inflammation ( Figure 2C). We concluded that the black module was not fully dependent on CD14 expression, which would allow us to test if the association between the black module and non-infectious uveitis could also be perceived in the CD14-CD1c+ DC fraction. To ascertain that we could attribute the black module to CD1c+ DCs by bulk RNA-seq analyses, we next purified CD14-negative CD1c+ DCs by FACS from an independent cohort of 36 patients and controls and performed bulk RNA-seq ( Figure 2D and Supplemental Table 2). Co-expression network analysis of the uveitisassociated genes (n=2,016 genes at P<0.05) from the second cohort distinguished six gene modules. Three modules exhibited a significant number of overlapping co-expressed genes with the black module from the discovery cohort ( Figure 2E). The gene expression levels for these replicated co-expressed genes were nearly identical between the two cohorts (Spearman R=0.95, Figure 2 -Figure Supplement 1C), but with a clearly lower sensitivity to replicate relatively lower expressed genes (i.e., likely due to ~3-fold lower cell yield after the removal of CD14+ cells by FACS-sorting compared to cohort I, see Methods). Regardless, in total, 147 differentially co-expressed genes of the black module in the first cohort were also co-expressed in the second cohort ( Figure 2F), of which 94% of these co-expressed genes also showed consistent direction of effect (e.g., upregulated in both cohorts) (Figure 2F,   Supplementary Table 3), which substantiates this gene module as a robust and bona fide transcriptional feature of CD1c+ DCs in human non-infectious uveitis. Among the 147 genes, we replicated CX3CR1, CD36, CCR2, CCR5, TLR-6,-7,-8, CD180 (Figure 2H), but also TNFRSF1A (the main TNF-alpha receptor), CYBB, (critically involved in T cell activation by DCs in autoimmune models) (31), and transcription factors RUNX3, IRF8, and NFKBI associated with cDC2 subsets (18,29). Note, CCR2 and CX3CR1 also showed high module membership in the second cohort (Supplementary Table 3), and were also significantly associated with non-infectious uveitis after correction for multiple testing (Padj<0.05, Supplemental Table 3). Collectively, these data show a skewed transcriptional signature in blood CD1c+ DCs of non-infectious uveitis patients.

NOTCH2-RUNX3 signaling controls the CD1c+ DC gene module of non-infectious uveitis
Among the 147 genes with consistent direction of effect in both cohorts, we noticed RUNX3, IRF8, and NFKB1 as potential transcriptional regulators for the black module ( Figure 2H).
Note that RUNX3, IRF8, and NFKB1 also clustered into a single transcriptomic cluster in single cell analysis of human CD1c+ DC subsets by Brown and associates (29). Because loss of RUNX3 in mononuclear phagocytes (i.e., cDC2s) has been linked with chronic inflammation (32), we hypothesized that the signaling that involves the transcription factor RUNX3 may promote the transcriptomic signature that characterize CD1c+ DCs of patients.
Efforts to study the effects of siRNA-mediated knockdown or CRISPR-Cas9-mediated knockout (KO) of RUNX3 in CD1c+ DCs were unsuccessful (no knock-down or knock-out of RUNX3 achieved). Therefore, we analyzed gene expression data of the cDC2s (murine equivalent of human CD1c+ DCs) cells from CD11c-DC-Runx3Δ mice (33). Loss of runx3 in murine cDC2s resulted in a gene expression profile that considerably recapitulates the black module ( Figure 3A). In detail, we detected enrichment for many genes that are upregulated in runx3-KO cDCs2, including cd36, ccr2, and cx3cr1 by gene-set enrichment analysis (GSEA) ( Figure 3B). This supports our hypothesis that the 'monocyte-like' signature genes (i.e. CD36, CX3CR1, CCR2) are regulated by RUNX3. Indeed, RUNX3 directly binds the promoter of CD36 and negatively regulates its expression in myeloid cells (34). Importantly, overnight stimulation of MACS-sorted CD1c+ DCs by various key myeloid cytokines or Tolllike receptor (TLR) ligands did not result in a decrease in expression of RUNX3 nor in a concordant increase in CD36 expression, suggesting that the observed gene signature in the CD1c+ DC pool is unlikely to reflect an activation continuum ( Figure 3C). In fact, TLRstimulation (with LPS, LTA, or R848) resulted in a strong upregulation of RUNX3 and downregulation of CD36, which is the opposite of the expression pattern detected in patients.
Also, overnight stimulation with cytokines GM-CSF and IFN-alpha that are implicated in the promotion of specific cDC2 subsets (19,20) did not decrease the expression of RUNX3 ( Figure 3C). In general, the transcriptional signature of non-infectious uveitis did not resemble the gene signature of in vivo activated cDC2s (termed 'inflammatory' cDC2s' [infcDC2s]) (35). However, we did observe a significant positive enrichment score for genes down-regulated in infcDC2s in the transcriptome of IU patients (Figure 3A and 3B).
Because murine studies underscored a notch-dependent divergence of cDC2 subsets (24,36), we reasoned that the CD1c+ DC transcriptional signature driven by RUNX3 would rely on NOTCH2 signaling. To explore this, we investigated the transcriptome of dendritic cells of notch2∆-CD11c mice (37). In agreement with the expression data of the CD11c-DC-Runx3Δ mice, loss of Notch2 resulted in up regulation of ccr2, cd36, cx3cr1, and decreased expression of runx3 (Figure 3A), and we detected enrichment for genes upregulated in notch2-KO cDCs2 ( Figure 3B). This supports that NOTCH2 is upstream of RUNX3 and mediates the transcriptomic characteristics of blood CD1c+ DCs of non-infectious uveitis.
These findings were further strengthened in reanalysis of transcriptomic data of murine bone marrow progenitors cultured for 7 days with OP9 stromal cells that express the NOTCH2 ligand DLL1 or OP-9 cells without DLL1 (38,39). This analysis revealed that notch2controlled genes were enriched in the transcriptome of CD1c+ DCs of patients and that notch2-signalling governs the expression of cd36, ccr2, and cx3cr1 in cDC2s ( Figure 3D and Figure 3E). Collectively, these observations support that NOTCH2-RUNX3 signaling promotes the black module gene profile of CD1c+ DCs in human non-infectious uveitis.

CD36+CX3CR1+ DC3 are diminished in peripheral blood of non-infectious uveitis patients
We reasoned that the transcriptomic signature of the CD1c+ DC pool in patients may be an impression of changes in the proportions of CD1c+ DC subsets in blood. To allow unbiased identification of CD1c+ DC phenotypes, we first used flow cytometry analysis to identify CD1c+ DC clusters in peripheral blood mononuclear cells (PBMCs) samples from 26 cases and 11 controls. We designed a panel based on the black module (CD36, CX3CR1, CCR2, and CD180), surface markers previously linked to CD1c+ DC subsets, but that were not in the black module (CD5, and CD163) (17,40), and classical CD1c+ DC markers (CD1c and Together, these results demonstrate that CD36+CX3CR1+ DC3s were diminished in the blood of patients with non-infectious uveitis.

CD36+CX3CR1+DC3s secrete high levels of cytokines implicated in non-infectious uveitis
Next, we were interested if the CD36+CX3CR1+CD1c+ DC subset was functionally different from other CD1c+ DC subsets. To this end, we freshly sorted primary human CD1c+ DC subsets based on the surface expression of CD36 and CX3CR1, of which double-positive 1 and double-negative subsets could be sorted from the selected healthy subjects in sufficient numbers for analysis ( Figure 5 -Figure Supplement 1). Since CD36 is required for lipoteichoic acid (LTA) induced cytokine production (42), we overnight stimulated the CD1c+ subsets with LTA. Interleukin (IL)-23, a cytokine potently produced by CD1c+ DCs in general, was equally strong secreted by both subsets of CD1c+ DCs ( Figure 5A). To assess the secretome of the CD1c+ DC subsets in more detail, we profiled the supernatants of LTAstimulated CD1c+ DC subsets for additional soluble immune mediators (Supplemental Table 4): The CD1c+ DC subsets could be distinguished based on the secreted protein profile ( Figure 5B), of which the levels of TNF-alpha, IL-6, VEGF-A, and TNFR1 showed significant differences between the subsets ( Figure 5C). Proteomic analysis of eye fluid biopsies (n=20 aqueous humor samples from patients from this study available, Supplemental Table 5) and paired plasma (Supplemental Table 6) revealed eye-specific significant increase in the levels of these cytokines highly expressed by CD36+CX3CR1+ DC3s (e.g., TNF-alpha, IL-6) and ligands for chemokine receptors expressed by CD36+CX3CR1+ DC3s (e.g., CX3CL1 for CX3CR1, and MCP2 for CCR2) ( Figure 5D, Supplemental Table 7). These results show that CD1c+ DC subsets defined on the basis of surface co-expression of CD36 and CX3CR1 show a differential capacity to secrete proinflammatory mediators that participate in the pathophysiology of human non-infectious uveitis.

CD36+ CX3CR1+ DC3s infiltrate the eye during non-infectious.
We speculated that the decrease in blood CD36+CX3CR1+ CD1c+ DCs was in part the result of migration of these cells to peripheral tissues (lymph nodes) and that these cells may infiltrate the eye during active uveitis. We used single-cell RNA sequencing data of eye fluid biopsies of patients (43) and identified CD1c+ DCs by cells using the CD1c+ DC specific tissue-marker CLEC10A (20,44) ( Figure 6A). Next, we determined the relative expression of the black module in these cells by calculating the module enrichment for each cell (i.e., UCell score for genes with high module membership, Supplemental Table 8) and cluster CD1c+ DCs into black-module-negative and black-module-positive subpopulations( Figure 6B). In line with our bulk RNA-seq data, eye-infiltrating black-module-positive CD1c+ DCs expressed relatively higher levels of CD36, CX3CR1, CCR2, and lower levels of RUNX3 (but comparable levels of CD14) compared to black module-negative CD1c+DCs and therefore represent CD36+CX3CR1+DC3s ( Figure 6C). CD36+CX3CR1+DC3s were found in relatively higher abundance in eyes of non-infectious uveitis patients ( Figure 6D). In summary, we conclude that CD36+ CX3CR1+ DC3s infiltrate the eye during active noninfectious uveitis.

Discussion
In this study of non-infectious uveitis patients and controls, we identified and replicated a robust and bona fide transcriptional module in CD1c+ DCs. We were able to track back the network to a cytokine-producing CD36+CX3CR1+ CD1c+ DC subset that was diminished in peripheral blood but infiltrate the eyes of patients with non-infectious uveitis.
Using data from genetic models, we show that reciprocal expression of the gene network associated with non-infectious uveitis relies on transcription factors NOTCH2 and RUNX3. In detail, we showed that NOTCH2 signaling regulates the expression of CD36, and CX3CR1 via RUNX3 in CD1c+ DCs and that this recapitulates the (black module) gene signature of non-infectious uveitis.
Brown et al., (29) recently showed that in human blood CD1c+ DCs, differential expression of transcription factors, including RUNX3, IRF8, and NFKB1 (which were all in the transcriptional signature of CD1c+ DCs of patients) delineate CD1c+ DC subsets. Our observation that these transcription factors were differentially co-expressed support that the gene expression changes identified in uveitis patients were mediated by compositional changes in discrete subsets with reciprocal gene expression patterns (16).
Recently, Cytlak and associates revealed that lower expression of IRF8 is linked to DC3 (20), a transcription factor that was also decreased in non-infectious uveitis. Dutertre and coworkers (17) showed that the phenotype of peripheral blood CD1c+ DCs can be further segregated according to the expression of CD163 and CD5, with "DC3" cells being characterized as CD5-CD163-or CD5-CD163+cells and "DC2" as CD5+CD163 cells. We show that (CD5-CD163+) DC3 can also be further segregated based on surface expression of CD36 and CX3CR1. Importantly, we show that (CD5−CD163+) CD14+ DC3s (previously termed "inflammatory" CD14+ DC3 cells [17]), comprise of two subpopulations defined by CD36 and CX3CR1, of which the (CD14+) CD36+CX3CR1+DC3 are implicated in noninfectious uveitis, but the CD14+CD36+CX3CR1-DC3s are not ( Figure 4J). Single-cell analysis supported this diversity in CD14+ DC3s by showing that eye-infiltrating CD1c+ DCs that were enriched for the uveitis-associated (black) gene module exhibited relatively higher levels of CD36, CX3CR1, CCR2, and lower levels of RUNX3, but not CD14. Although we demonstrate that CD36+CX3CR1+ cells also express CD14, as a proof of concept, we demonstrated that the association of the black module with non-infectious uveitis can even be perceived in CD14-CD1c+ DCs, emphasizing that this gene circuit represent a previously unrecognized DC3 cell state and heterogeneity in inflammatory CD14+DC3s. Since CD36+ CX3CR1+ DC3s define phenotypically discrete populations of CD14+ DC3 in peripheral blood and inflamed tissue, it would be interesting to determine its role in other inflammatory conditions and cancer. When considering the DC3 population as a whole for comparison to previous studies, patients with Systemic lupus erythematosus (SLE) display accumulation of CD5-CD163+ DC3s in blood (17), while this population of DC3 cells (CD5-CD163+CD1c+ DCs) was decreased in non-infectious uveitis patients (P = 0.005). The differences between non-infectious uveitis and SLE may be related to distinct (i.e., opposite) immunopathological mechanisms; Type I interferons drive the maturation of cDC2s into "inflammatory cDC2s" (infcDC2s) (35) and can induce CD1c+ DCs to express a distinct set of surface-receptors (19). The type I interferon (IFN)-α drives immunopathology of SLE and administration of type I interferon therapy can induce lupus-like disease (22,23). In favor of attributing the seemingly contrasting observations in blood CD1c+ subsets between SLE and non-infectious uveitis to distinct biology is the fact that, in contrast to elevated IFN-α in patients with SLE, in non-infectious uveitis patient's disease exacerbations correlate with reduced blood type I IFN concentrations (21,45). In addition, we demonstrated that the transcriptional signature of CD1c+ DCs in non-infectious uveitis was not positively enriched for transcriptomic features of IFN-driven cDC2 subset (Figure 3).
This also indicates that unlike SLE (46), changes of the CD1c+ DC peripheral blood pool in non-infectious uveitis are not driven by increased type I interferon signaling (21), but rather by yet unspecified pathological molecular pathways. This is supported by the fact that in our stimulation experiments, IFN alpha did not induce the uveitis-associated gene signature in CD1c+ DCs (Figure 3) and that type I interferon therapy inhibits non-infectious uveitis (21).
However, an argument against this is that we detected enrichment for genes down-regulated in infcDC2s (35) in IU patients. In detail, although not all statistically significant, genes such as ccr2, cd36, cx3r1 showed a relative decrease in expression, while runx3 showed a relative increase in expression in the transcriptome of infcDC2s versus "non-inflammatory" cDC2s (based on GSE149619). In other words, this suggest that the profile of infcDC2s within the CD1c+DC pool was 'decreased' in IU patients, which is in line with the decrease in the frequency of CD5-CD163 DC3s as a whole or the inflammatory cytokine-secreting CD36+CX3CR1+ DC3 subset in patients with non-infectious uveitis. One possible explanation for the decrease in this subset may be that in the systemic condition SLE 'inflammatory' DC3s 'accumulate' in blood, while in non-infectious uveitis this population exits the circulation to infiltrate secondary lymphoid tissues and ocular structures to mediate eye inflammation. Indeed, we show by single cell analysis of ocular fluids of patients the infiltration of CD36+ CX3CR1+ DC3s in the eye during non-infectious uveitis (Figure 6). We further showed that CD36+CX3CR1+ DC3s secreted more inflammatory cytokines, such as tumor necrosis factor alpha (TNF-α) and IL-6, cytokines that we show were specifically elevated in the ocular microenvironment of patients. This is significant, because anti-TNF and anti-IL-6 therapy are both (highly) effective for treatment of non-infectious uveitis (47), which underlines the significance of the TNF-producing DC3 subset identified in this study for the pathophysiology of non-infectious uveitis. Ideally longitudinal data are used to follow the dynamics of this population in patients in relation to disease activity, which is a limitation of the current study.
Other disease modifying factors possibly affect the CD1c+ DC pool in uveitis patients. In mice, antibiotic treatment to experimentally disturb the microbiota affects a cDC2 subset phenotypically similar to CD1c+ DCs and decreases their frequency in the intestine of mice, which suggests microbiota-dependent signals involved in the maintenance of cDC2 subsets (29). This is especially interesting in light of growing evidence that microbiota dependent signals cause autoreactive T cells to trigger uveitis (48), which makes it tempting to speculate that gut-resident cDC2 subsets contribute to the activation of T cells in uveitis models. Dietary components can influence subsets of intestinal dendritic cells (49).
Regardless, most likely, an ensemble of disease modulating factors is involved. For example, myeloid cytokines, such as GM-CSF, contribute to autoimmunity of the eye (50) and GM-CSF has been shown to stimulate the differentiation of human CD1c+ subset from progenitors (20). However, GM-CSF signaling in conventional dendritic cells has a minor role in the inception of EAU (51). Our data supports that stimulation of CD1c+ subsets with GM-CSF or TLR ligands does not induce the transcriptional features of CD1c+ DCs during noninfectious uveitis, which is in line with previous observations that support that stimulated cDC2s do not convert from one into another subset (20). Note that key transcription factors (e.g. RUNX3) defining the here identified CD1c+ subsets are definitely affected by TLR stimulation, but the overall transcriptomic program of activated CD1c+ DCs is distinct (Figure   3).
Better understanding of the changes in the CD1c+ DC pool during human non-infectious uveitis will help develop strategies to pharmacologically influence putative disease pathways involved at an early disease stage, which may lay the foundation for the design of effective strategies to halt progress towards severe visual complications or blindness. Perhaps targeting CD1c+ DCs may be achieved by dietary (microbiome) strategies and provide relatively safe preventive strategies for noninfectious uveitis.
In conclusion, we discovered a CD36+CX3CR1+CD1c+ DC subset that decreased in the peripheral blood of patients with non-infectious uveitis. The fact that this population secretes high levels of TNF-alpha, is decreased in the circulation of patients, while CD1c+ DCs expressing CD36 and CX3CR1 accumulate locally in uveitis patients may not only explains the therapeutic benefit of TNF inhibition for non-infectious uveitis, it also opens new avenues for therapeutic targeting to prevent blindness due to non-infectious uveitis.

Patients and patient material
This study was conducted in compliance with the Helsinki principles. Ethical approval was requested and obtained from the Medical Ethical Research Committee in Utrecht. All patients signed written informed consent before participation. We collected blood from a discovery cohort of 29 and a replication cohort of 22 adult patients ( Table 1)  DC cultures using the in-house multiplex immunoassay based on Luminex technology, as described previously (53). Protein concentrations that were out of range were replaced with the LLOQ (lower limit of quantification) and ULOQ (upper limit of quantification) for each analyte and divided by 2 for the proteins detected below the range of detection or multiplied by 2 for values above the detection range (Supplementary Table 4).

Proteomic analysis of eye fluid biopsies and blood plasma.
Eye fluid biopsies (Aqueous humor [AqH] n =31) and plasma (n=85) were analyzed using the Immuno-Oncology panel for the simultaneous quantification of 92 protein biomarkers (Olink, Uppsala) at the Olink facility of the University Medical Center Utrecht, the Netherlands. AqH from patients from both RNA-sequencing cohorts (n=20 available) were the remainder from diagnostic anterior chamber paracentesis and compared to AqH from cataract controls (n=11) without a history of inflammatory eye disease (Supplementary Table 5). EDTA plasma samples from 74 patients and controls from this study were collected simultaneously with blood samples used for CD1c+ DC isolations (stored at −80°C). Another 10 additional plasma samples from patients with active AU, IU, and BU were collected at diagnosis (Supplementary Table 6

Real-time Quantitative PCR
First-strand cDNA was synthesized from total RNA using Superscript IV kit (Thermo Fisher

Power analysis 1
We conducted power analysis of the discovery cohort using the PROPER R package v 1.22.0 (57) with 100 simulations of the build-in RNA-seq count data from antigen presenting (B) cells from a cohort of 41 individuals (i.e., large biological variation as expected in our study) (58). Simulations parameters used the default of 20,000 genes and an estimated 10% of genes being differentially expressed. We detected 0.8 power to detect differentially expressed genes (P<0.05) at a log 2 (fold change)>1 for the smallest patient group (9 cases) and we considered the sample size reasonable for analysis.

Differential gene expression and statistical analysis
Quality check of the raw sequences was performed using the FastQC tool. Reads were aligned to the human genome (GRCh38 build 79) using STAR aligner (59)  Single cell-RNA seq analysis of aqueous humor.
Single cell RNA-seq (scRNA-seq) data from aqueous humor of 4 HLA-B27-positive anterior uveitis (identical to the AU group in this study) patients and control (Kasper et al. 2021) (43) were obtained via Gene Expression Omnibus (GEO) repository with the accession code GSE178833. Data were processed using the R package Seurat v4.1.0 (66) using R v4.0.3.
We removed low-quality cells (<200 or >2500 genes and mitochondrial percentages <5%) and normalized the data using the SCTransform() function accounting for mitochondrial percentage and cell cycle score (67). Dimensionality reduction was achieved by adapting the original UMAP coordinates for each barcode as reported by Kasper et al 2021 (43). Samples from 4 (HLA-B27-positive) AU patients and the control were subjected to scGate v1.0.0 (68) using CLEC10A in our gating model to purify CD1c+ DCs in the scRNA seq dataset. We calculated module enrichment scores for the black module genes using the UCell R package v1.3.1 (69) (using genes with high module membership [-0.85< or >0.85, n=18) in cohort I, Supplemental Table 8) and clustered CD1c+DCs into positive and negative fractions.

Data and Code Availability
The data code (R markdown), bulk RNA-Seq datasets, flow cytometry dataset, and cytokine expression dataset described in this publication are available via https://dataverse.nl/ doi: https://doi.org/10.34894/9Q0FVO and deposited in NCBI's Gene Expression Omnibus accessible through GEO Series accession numbers GSE195501 and GSE194060.

Figure Supplements
Table S1-Table S13     Luminex assay (Supplemental Table 4). C) Scatter plots with overlay boxplot with mean and interquartile range of the levels of secreted TNF-alpha, Interleukin (IL)-6, VEGF-A, and TNFR1 from the multiplex protein data in d.