Transcriptome network analysis implicates CX3CR1-positive type 3 dendritic cells in non-infectious uveitis

  1. Sanne Hiddingh
  2. Aridaman Pandit
  3. Fleurieke Verhagen
  4. Rianne Rijken
  5. Nila Hendrika Servaas
  6. Rina CGK Wichers
  7. Ninette H ten Dam-van Loon
  8. Saskia M Imhof
  9. Timothy RDJ Radstake
  10. Joke H de Boer
  11. Jonas JW Kuiper  Is a corresponding author
  1. Ophthalmo-Immunology, University Medical Center Utrecht, Utrecht University, Netherlands
  2. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Netherlands
  3. Department of Ophthalmology, University Medical Center Utrecht, Utrecht University, Netherlands
  4. Department of Rheumatology & Clinical Immunology, University Medical Center Utrecht, Utrecht University, Netherlands
  5. University Medical Center Utrecht and Utrecht University, Netherlands

Abstract

Background:

Type I interferons (IFNs) promote the expansion of subsets of CD1c+ conventional dendritic cells (CD1c+ DCs), but the molecular basis of CD1c+ DCs involvement in conditions not associated without elevated type I IFNs remains unclear.

Methods:

We analyzed CD1c+ DCs from two cohorts of non-infectious uveitis patients and healthy donors using RNA-sequencing followed by high-dimensional flow cytometry to characterize the CD1c+ DC populations.

Results:

We report that the CD1c+ DCs pool from patients with non-infectious uveitis is skewed toward a gene module with the chemokine receptor CX3CR1 as the key hub gene. We confirmed these results in an independent case–control cohort and show that the disease-associated gene module is not mediated by type I IFNs. An analysis of peripheral blood using flow cytometry revealed that CX3CR1+ DC3s were diminished, whereas CX3CR1− DC3s were not. Stimulated CX3CR1+ DC3s secrete high levels of inflammatory cytokines, including TNF-alpha, and CX3CR1+ DC3 like cells can be detected in inflamed eyes of patients.

Conclusions:

These results show that CX3CR1+ DC3s are implicated in non-infectious uveitis and can secrete proinflammatory mediators implicated in its pathophysiology.

Funding:

The presented work is supported by UitZicht (project number #2014-4, #2019-10, and #2021-4). The funders had no role in the design, execution, interpretation, or writing of the study.

Editor's evaluation

These findings are valuable to ocular immunologists who the study pathophysiologic mechanisms driving inflammation in human uveitis, and for future identification of novel therapeutic targets. The authors convincingly perform high dimensional multi-omic analysis of testing and replication cohorts, followed by characterization of a disease-specific cell type using comparative analysis with previously validated experimental datasets. The analysis will be of particular interest to basic and translational ocular immunologists, as well as dendritic cell biologists.

https://doi.org/10.7554/eLife.74913.sa0

Introduction

Non-infectious uveitis refers to a group of chronic inflammatory eye diseases that are among the leading causes of preventable vision loss in western countries (Thorne et al., 2016; Suttorp-Schulten and Rothova, 1996). Currently, little is known about the disease mechanism of non-infectious uveitis. A large body of mechanistic studies using experimental autoimmune uveitis (EAU) in rodents suggest that T cells play a role in non-infectious uveitis (Lee et al., 2014; Caspi, 2010). Human non-infectious uveitis is characterized by inflammation and T cells infiltrating the eyes through unknown mechanisms. The genetic association between non-infectious uveitis and MHC, and ERAP1, ERAP2 genes indicates that antigen presentation is central to the etiology (Kuiper and Venema, 2020; Huang et al., 2020; Kuiper et al., 2018; Márquez et al., 2017). A key antigen presenting cell type are dendritic cells. Despite their important role in EAU, dendritic cells have yet to be fully investigated in human non-infectious uveitis (Chen et al., 2015b; Fu et al., 2019; Wang et al., 2021). CD1c-positive conventional dendritic cells (CD1c+ DCs) have been found to be associated with disease activity (Chen et al., 2016; Chen et al., 2015a; Chen et al., 2014), and are abundant in eye fluid of patients (O’Rourke et al., 2018). In order to understand the role of CD1c+ DCs in non-infectious uveitis, it is necessary to understand their functions.

Several single-cell studies have revealed that the CD1c+ DCs (and its murine equivalent, termed ‘cDC2s’) consists of multiple subsets derived from distinct progenitors (Villani et al., 2017; Dutertre et al., 2019; Cytlak et al., 2020). The type I IFN family of cytokines promotes the expansion of a subset of CD1c+ DCs called ‘DC3’ (Dutertre et al., 2019; Girard et al., 2020; Bourdely et al., 2020). DC3s are increased in blood of type I IFN-driven systemic lupus erythematosus (SLE) patients (Dutertre et al., 2019). A significant difference between non-infectious uveitis and SLE is that active uveitis is accompanied by lower levels of type I IFN (Wang et al., 2019; Kuiper et al., 2022). It should be noted that although type I IFNs can induce lupus-like disease, they can also suppress non-infectious uveitis (Wang et al., 2019; Rönnblom et al., 1991; Rönnblom et al., 1990), pointing to an alternative disease mechanism implicating CD1c+ DCs in non-infectious uveitis. Hence, we do not fully understand the characteristics of CD1c+ DC during autoimmunity, especially in conditions not driven by type I IFNs.

For the purpose of characterizing the core transcriptional features and subset composition of CD1c+ DCs in autoimmunity of the eye, we used 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. We constructed co-expression networks that identified a robust gene module associated with non-infectious uveitis in patients that helped identify a CX3CR1-positive CD1c+ DC subset.

Materials and methods

Patients and patient material

Request a detailed protocol

This study was conducted in compliance with the Helsinki principles. Ethical approval was requested and obtained from the Medical Ethical Research Committee in Utrecht (METC protocol number #14-065/M). All patients signed written informed consent before participation. We collected blood from a discovery cohort of 23 and a replication cohort of 28 adult patients (Table 1) with HLA-B27-associated acute anterior uveitis (AU), idiopathic intermediate uveitis (IU), or HLA-A29-associated birdshot uveitis (BU). Patients were recruited at the outbound patient clinic of the Department of Ophthalmology of the University Medical Center Utrecht between July 2014 and January 2017. We recruited 27 age- and sex-matched anonymous blood donors of European Ancestry with no history of ocular inflammatory disease at the same institute to serve as unaffected controls (Table 1). Uveitis was classified and graded in accordance with the SUN classification (Jabs et al., 2005). Each patient underwent a full ophthalmological examination by an ophthalmologist experienced in uveitis, routine laboratory screening, and an X-ray of the lungs. Laboratory screening included erythrocyte sedimentation rate, renal and liver function tests, angiotensin-converting enzyme, and screening for infectious agents in the serum and an Interferon-Gamma Release Assay (IGRA) was obtained for all patients. All patients with AU and BU were HLA-B27 or HLA-A29-positive, respectively (confirmed by HLA typing). All patients had active uveitis (new onset or relapse) and there was no clinical evidence for uveitis-associated systemic inflammatory disease (e.g., rheumatic condition) till the time of sampling. None of the patients received systemic immunomodulatory treatment in the last 3 months, other than low dose (≤10 mg) oral prednisolone in one BU patient of cohort II and one AU patient of cohort I.

Table 1
Characteristics of the patients and controls from cohorts I and II.

Abbreviations: BU: birdshot uveitis, AU: HLA-B27-associated anterior uveitis, HC: healthy control, IU: idiopathic intermediate uveitis, n.a.: not applicable, *Fisher’s exact test, **ANOVA, ***Kruskal–Wallis.

Cohort IAUIUBUHCp value
N105813Total = 36
Male/female2/83/25/35/80.26*
Age in years; mean ± SD45 ± 1630 ± 942 ± 1042 ± 130.24***
Disease duration in years; median (range)8.1 (0.2–22.3)3.4 (0.4–14.1)0.9 (0.2–19.9)n.a.0.36**
Cohort IIAUIUBUHCp value
N991014Total = 42
Male/female3/62/74/66/80.8790*
Age in years; mean ± SD47 ± 1739 ± 1452 ± 1339 ± 100.06***
Disease duration in years; median (range)5.8 (0.1–39.3)3.7 (0.2–20.0)1.3 (0.2–15.1)n.a.0.14**

CD1c+ DC purification

Request a detailed protocol

Peripheral blood mononuclear cells (PBMCs) were isolated by standard ficoll density gradient centrifugation from 70 ml heparinized blood immediately after blood withdrawal (GE Healthcare, Uppsala, Sweden). For the first cohort, 10 batches (individual days) of 4–5 randomly selected patient and control samples of nitrogen stored PBMCs (mean storage time of 11 [range 0–31] months) were carefully thawed and subjected to sorting by the BD FACSAria III sorter after incubation with a panel of surface antibodies (Supplementary file 1A) and fluorescent-activated cell sorting (FACS) buffer (1% bovine serum albumin and 0.1% sodium azide in phosphate-buffered saline [PBS]). CD3−CD19−CD56−CD14−HLA-DR+CD11c+CD1c cells were sorted. The average number of collected cells by sorting was 56,881 (range 6669–243,385). For the second cohort, fresh PBMCs were immediately subjected to magnetic-activated cell sorting (MACS) for the removal (positive selection) of CD304+ cells (pDC), followed by the removal of CD19+ cells (B cell), and subsequently isolation of CD1c+ cells by using the CD1c+ (BDCA1) isolation kit (Miltenyi Biotec, Germany) according to the manufacturer’s instructions. The isolated CD1c+ fraction contained on average 147,114 cells (range 46,000–773,000) and purity was determined by flow cytometry (Supplementary file 1B) measured on the BD LSRFortessa Cell analyzer (Figure 1—figure supplement 1). Data were analyzed using FlowJo software (TreeStar Inc). MACS or FACS purified CD1c+ cells were immediately taken up in a lysis buffer (RLT plus, QIAGEN) containing 1% β-mercaptoethanol, snap frozen on dry ice, and stored at −80°C until RNA extraction was performed. Isolation of CD1c+ DC for functional experiments was done by MACS as described above. Purification of CD1c+DC subsets based on CD36 and CX3CR1 or CD14 expression from freshly isolated PBMCs was conducted by flow cytometry using the panel in Supplementary file 1C and shown in Figure 3—figure supplement 2B.

CD1c+ DC cultures and secretome analysis

Request a detailed protocol

Purified CD1c+ DCs were cultured in RPMI Glutamax (Thermo Fisher Scientific) supplemented with 10% heat-inactivated fetal bovine serum (Biowest Riverside) and 1% penicillin/streptomycin (Thermo Fisher Scientific). CD1c+ DCs were cultured at a concentration of 0.5 × 106 cells/ml in a 96-well round-bottom plate (100 μl/well). Cells were stimulated overnight (18 hr) with multiple stimuli listed in Supplementary file 1D. After stimulation, cells were lysed in an RLT plus lysis buffer (QIAGEN) and stored at −80°C until RNA extraction was performed. Cell lysates were stored at −80°C until RNA extraction was performed for qPCR. In separate cultures, CD1c+ DC subsets (sorted based on CD36 and CX3CR1 expression) were cultured in the presence of 1 µg/ml lipoteichoic acid (LTA). After 18 hr of stimulation, supernatants were harvested and IL-23 cytokine production was analyzed by ELISA (R&D Systems). The levels of IL-2, IL-5, IL-6, IL-10, IL-12p70, IL-13, IL-17, IL-22, IL-27, TNF-alpha, IFN-alpha, IFN-beta, CCL1, CXCL10, CXCL13, VEGF, CD40L, FAS, TNFR1, TNFR2, Elastase, and Granzyme B were simultaneously measured in supernatant of CD1c+ DC cultures using the in-house multiplex immunoassay based on Luminex technology, as described previously (Bakker et al., 2022). 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 file 1E).

Real-time quantitative PCR

Request a detailed protocol

First-strand cDNA was synthesized from total RNA using Superscript IV kit (Thermo Fisher Scientific), and quantitative real-time PCR (RT-qPCR) was performed on the QuantStudio 12k flex System (LifeTechnologies), following the manufacturer’s instructions. Sequences of the primers used are listed in Supplementary file 1F and the Key Resource Table. RT-qPCR data were normalized to the expression of the selected housekeeping gene GUSB (ENSG00000169919). CT values were normalized to GUSB by subtracting the CT mean of GUSB (measured in duplo) from the CT mean of the target mRNA (e.g., RUNX3) = ΔCT. The fold change (FC) of each sample was calculated compared to ΔCt of the medium control using the formula FC = 2−ΔΔCt, where ΔΔCt = ΔCt sample − ΔCt reference.

Flow cytometry of CD1c+ DC populations

Request a detailed protocol

PBMC samples from the two cohorts (HC = 11 samples; AU = 9 samples; IU = 6 samples; BU = 11 samples) were randomly selected and measured by flow cytometry in batches of 9–10 mixed samples per run, divided over 4 days. Per batch, 10 million PBMCs per sample were quickly thawed, washed with ice cold PBS and stained with the antibody panel depicted in Supplementary file 1C. PBMCs were incubated with Fixable Viability Dye eF780 (eBioscience) at room temperature for 10 min. Cells were then plated in V-bottomed plates (Greiner Bioone), washed with PBS and incubated for 30 min at 4°C in the dark with Brilliant Stain Buffer (BD) and the fluorescently conjugated antibodies. Next, the cells were washed and taken up in the FACS buffer. Flow cytometric analyses were performed on the BD FACSAria III sorter. Manual gating of data was done using FlowJo software (TreeStar inc San Carlos, CA, USA). FlowSOM v1.18.0 analysis was done as described previously (Laban et al., 2020). Lineage- (negative for CD3/CD56/CD19) HLA-DR+ data were transformed using the logicleTransform function of the flowCore v1.52.1 R package, using default parameters (Ellis et al., 2020). The SOM was trained for a 7 × 7 grid (49 clusters) with 2000 iterations. Consensus hierarchical clustering was used to annotate clusters, based on the ConsensusClusterPlus v1.50.0 R package (Wilkerson and Hayes, 2010). Principal component analysis (PCA) analysis was done on normalized expression data from flowSOM using the factoextra v 1.0.7.999 R package.

RNA isolation and RNA-sequencing

Request a detailed protocol

Total RNA from CD1c+ DC cell lysates from patients and controls was isolated using the AllPrep Universal Kit (QIAGEN) on the QIAcube (QIAGEN) according to the manufacturer’s instructions. For cohort I, RNA-seq libraries were generated by GenomeScan (Leiden, The Netherlands) with the TruSeq RNAseq RNA Library Prep Kit (Illumina Inc, Ipswich, MA, USA), and were sequenced using Illumina HiSeq 4000 generating ~20 million 150 bp paired end reads for each sample. Library preparation and Illumina sequencing was performed on samples of cohort II at BGI (Hong Kong). RNA-seq libraries were generated with the TruSeq RNAseq RNA Library Prep Kit (Illumina Inc, Ipswich, MA, USA) and were sequenced using Illumina NextSeq 500 generating approximately 20 million 100 bp paired end reads for each sample.

Power analysis

Request a detailed protocol

We conducted power analysis using the PROPER R package v 1.22.0 (Wu et al., 2015) 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) (Cheung et al., 2010). Simulation 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 log2(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

Request a detailed protocol

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 (Dobin et al., 2013) and the Python package HTSeq v0.6.1 was used to count the number of reads overlapping each annotated gene (Anders et al., 2015). We aligned the reads of the RNA-sequencing datasets to 65,217 annotated Ensemble Gene IDs. Raw count data were fed into DESeq2 v1.30.1 (Love et al., 2014) to identify differentially expressed genes between the four disease groups (AU, IU, BU, and HC). Using DESeq2, we modeled the biological variability and overdispersion in expression data following a negative binomial distribution. We used Wald’s test in each disease group versus control pairwise comparison and p values were corrected by the DESeq2 package using the Benjamini–Hochberg Procedure. We constructed co-expression gene networks with the WGCNA v 1.70-3 R package (Langfelder and Horvath, 2008) using the cumulative uveitis-associated genes from all pairwise comparisons and a soft power of 5. Module membership (MM) represents the intramodular connectivity of genes in a gene module. Gene Significance (GS >0.25) indicates a strong correlation between genes and non-infectious uveitis, whereas MM (MM >0.8) indicates a strong correlation with the EigenGene value of the modules. We calculated the intersection between the modules constructed from the two cohorts and used Fisher’s exact test to identify modules that exhibited significant overlap in genes. Gene expression data from runx3-knockout (KO) cDC2s, notch2-KO cDC2s, and ‘inflammatory’ cDC2s were obtained from the NCBI Gene Expression Omnibus (accession numbers GSE48590 [2 wild-type [WT] CD11b+ESAM+ splenic cDC2s vs. 2 CD11b+ESAM+ cDC2s from CD11c-DC-Runx3Δ mice], GSE119242 [2 untreated cDC2 vs. untreated cDC2 from CD11c-Cre notch2f/f mice], GSE149619 [5 CD172+MAR1− cDC2s in mock condition vs 3 CD172+MAR1+ cDC2 in virus condition]) using GEO2R in the GEO database, which builds on the GEOquery v2.58.0 and limma R v 3.46 packages (Davis and Meltzer, 2007; Ritchie et al., 2015). RNA-seq data from the mouse bone marrow stromal cell line OP9 expressing NOTCH ligand DLL1 (OP9-DLL1)-driven cDC2 cultures (GSE110577 [2 sorted CD11c+MHCII+B220−CD11b+ cDC2 from bone marrow cultures with FLT3L for 7 days vs. 2 sorted CD11c+MHCII+B220−CD11b +cDC2 from bone marrow cultures with FLT3L+OP9-DLL1 cells for 7 days]) were analyzed using DESeq2 and normalized count data plotted using the plotCounts function. RNA-seq count data from CD14+/− DC3 subsets from patients with SLE and systemic sclerosis were obtained via GEO (accession number: GSE136731) (Dutertre et al., 2019) and differential expression analysis was conducted using DESeq2 v1.30.1. Gene set enrichment analysis was done using the fgsea R package v1.16.0 and data plotted using the GSEA.barplot function from the PPInfer v 1.16.0 R package (Jung and Ge, 2020). Gene sets for runx3-KO, notch2-KO, inflammatory cDC2s, and cDC2s from OP9-DLL1 bone marrow cultures were generated by taking the top or bottom percentiles of ranked [−log10(p) × sign(log2(fold change))] genes from each dataset as indicated. Genes in the modules of interest that encode cell-surface proteins were identified according to surfaceome predictor SURFY (Bausch-Fluck et al., 2018).

Single-cell RNA-seq analysis of aqueous humor

Request a detailed protocol

Single-cell RNA-seq (scRNA-seq) data from a previous study of as reported by Kasper et al., 2021 of aqueous humor of four HLA-B27-positive anterior uveitis (identical to the AU group in this study) patients were obtained and downloaded via Gene Expression Omnibus (GEO) repository with the accession code GSE178833. Data were processed using the R package Seurat v4.1.0 (Stuart et al., 2019) 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 (Hafemeister and Satija, 2019). Dimensionality reduction for all cells was achieved by adapting the original UMAP coordinates for each barcode as reported by Kasper et al., 2021 (see GSE178833). Data were subjected to scGate v1.0.0 (Andreatta et al., 2022) using CLEC10A+ and C5AR1 (CD88)− cells in our gating model to purify CD1c+ DCs in the scRNAseq dataset. Dimensionality reduction for CD1c+ DCs was conducted using the R package Seurat, and cells clustered using the FindNeighbors and FindClusters functions from Seurat. After clustering and visualization with UMAP, we used the DotPlot function from the Seurat package to visualize the average expression of genes in each cluster.

Results

A CX3CR1 gene module is associated with non-infectious uveitis

We characterized the transcriptome of primary CD1c+ DCs from patients with non-infectious uveitis (Figure 1A). RNA-seq analysis (RNA-seq) was performed on lineage (CD3−CD19−CD56−CD14−)-negative, and HLA-DR-positive, CD11c and CD1c-positive DCs purified from frozen PBMCs by flow cytometry from 36 patients with anterior (AU), intermediate (IU), or posterior non-infectious uveitis (BU) and healthy controls. A co-expression network was constructed using uveitis-associated genes identified by differential expression analyses (n = 2,016 genes at P<0.05, Figure 1B), which identified six modules, of which 3 were associated with non-infectious uveitis (GS >0.25, Figure 1C). The blue module was most associated with non-infectious uveitis (Figure 1C). Based on Module Membership, CX3CR1 was the top hub gene of the blue module (Figure 1D, Supplementary file 1G). Since CX3CR1 was previously associated with a distinct subset cDC2s that may also express CD14 (Brown et al., 2019; Fujita et al., 2019), we attempted to validate and expand the gene set associated with non-infectious uveitis by MACS-isolating CD1c+ DC cells from fresh blood of 28 patients and 14 healthy controls, followed by RNA-seq analysis of the highly purified CD1c+ DCs (median [interquartile range]% = 96 [3]% pure, Figure 1—figure supplement 1). We also constructed a co-expression network for uveitis-associated genes (n = 6794, p < 0.05) in the second cohort (Figure 1E), which revealed 24 gene modules (Supplementary file 1H). Note that 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 2). The three uveitis-associated modules in cohort I shared a significant number of co-expressed genes with one module in cohort II, the black module (Figure 1F). The black module was associated with non-infectious uveitis in cohort II (GS for uveitis >0.25) and CX3CR1 was also the hub gene for this module (Black Module Membership, p = 5.9 × 10−22; Supplementary file 1H, Figure 1G). According to these findings, the overlapping disease-associated gene modules appear to represent a single gene module. In cohort I, the separation of genes into three modules was possibly due to low sensitivity to detect disease-associated genes with low expression, as replicated genes of the black module were typically higher expressed (Figure 1—figure supplement 3). In total, we replicated 147 co-expressed genes between the two cohorts (which we will refer to as the ‘black module’), of which 94% also showed consistent direction of effect (e.g., upregulated in both cohorts) (Figure 1H, Supplementary file 1I). The black module was enriched for the GO term ‘positive regulation of cytokine production’ (GO:0001819, padj = 6.9 × 10−5). In addition to CX3CR1, the black module comprised CD36, CCR2, TLR-6,-7,-8, CD180, and transcription factors RUNX3, IRF8, and NFKB1 (Figure 1I), but not CD14. In summary, these results show that a gene module characterized by CX3CR1 in blood CD1c+ DCs is associated with non-infectious uveitis.

Figure 1 with 3 supplements see all
A CX3CR1 gene module in CD1c+ dendritic cells (CD1c+ DC) is associated with non-infectious uveitis.

(A) Study design. CD1c+ DCs were purified from blood and subjected to RNA-sequencing. Co-expression network analysis was used to identify gene modules associated with uveitis. (B) Principal component analysis (PCA) of the 2016 uveitis-associated genes (p < 0.05) in 36 patients and control samples of cohort I. (C) Gene significance for uveitis for the gene modules identified by WGCNA. (D) Module membership and Gene Significance for uveitis for the blue module of cohort I. (E) PCA of the 6794 uveitis-associated genes (p < 0.05) in 42 samples of cohort II. (F) Cross-tabulation of the preservation of co-expressed genes between gene modules from cohort I and the black module from cohort II. p value is from Fisher’s exact test. (G) Same as in D, but for the black module of cohort II. (H) Heatmaps of the 147 replicated co-expressed genes (rows) for samples (columns) from cohorts I and II. The venn diagram shows the up- and downregulated genes (clusters shown in H). (I) The (log2) fold change in gene expression compared to healthy controls (x-axis) for all 147 replicated genes in patients with AU, IU, and BU. Genes encoding surface proteins are indicated in black/gray. Key transcription factors are indicated in blue. AU: anterior uveitis, IU: intermediate uveitis, BU: birdshot uveitis.

CX3CR1+ DC3 are diminished in peripheral blood of non-infectious uveitis patients

Type I IFN cytokines promote differentiation of CD1c+ DCs (Dutertre et al., 2019; Girard et al., 2020), but patients with active non-infectious uveitis have reduced blood levels of type I IFN cytokines (Wang et al., 2019; Kuiper et al., 2022). Assessment of the transcriptome of CD1c+ DCs from patients, found no enrichment for genes associated with murine type I IFN-dependent cDC2s (Bosteels et al., 2020; Figure 2A,B). Furthermore, while RUNX3 was downregulated in RNA-seq data from CD1c+ DCs from non-infectious uveitis patients (Figure 1I), stimulation of CD1c+ DCs with from healthy human donors with IFN-alpha resulted in upregulation of RUNX3 (Figure 2—figure supplement 1A). In contrast, the transcriptome of CD1c+ DCs from patients overlapped significantly with murine cDC2s knocked out for Runx3, or its upstream regulator Notch2 (Figure 2, Figure 2—figure supplement 1B ,C; Lewis et al., 2011; Fasnacht et al., 2014; Briseño et al., 2018). Given that cDC2 subsets differ by their dependence on NOTCH signaling (Lewis et al., 2011; Fasnacht et al., 2014; Briseño et al., 2018; Kirkling et al., 2018), we hypothesized that the transcriptomic signatures of the CD1c+ DC pool in patients might reflect changes in their proportions.

Figure 2 with 1 supplement see all
The CX3CR1 gene module of CD1c+ DCs is enriched for NOTCH2-RUNX3 signaling.

(A) Volcano plot for the expression of genes of the black module in cDC2s of runx3-KO mice (GSE48590), notch2-KO mice (GSE119242), and type I IFN-dependent inflammatory [inf-]cDC2s (GSE149619). Up- and downregulated genes for each condition are indicated for each condition; gray dots denote the genes with no significant change in expression. (B) Results from gene set enrichment analysis for ranked transcriptomes (using 20,668 genes with baseMean >4) for AU, IU, and BU patients. The top or bottom percentiles of the ranked [−log10(p) × sign(log2(FC))] genes from runx3-KO cDC2s, notch2-KO cDC2s, and inf-cDC2s (see a) were used as gene sets. Normalized enrichment scores (NES) and p values for each gene set are indicated. The dotted lines indicate padj = 0.05. AU: anterior uveitis, IU: intermediate uveitis, BU: birdshot uveitis.

Therefore, we used flow-cytometry analysis to identify CD1c+ DC clusters in PBMCs samples from 26 cases and 11 controls. We designed a panel based on the black module (CX3CR1, CD36, CCR2, and CD180), other CD1c+ DC markers that were not in the black module (CD1c, CD11c, CD14, CD5, and CD163) (Dutertre et al., 2019; Korenfeld et al., 2017). FlowSOM (Van Gassen et al., 2015) was used on HLA-DR+ and lineage (CD3/CD19/CD56)− PBMCs to cluster cells into a predetermined number of 49 clusters (7 × 7 grid) to facilitate detection of CD1c+ DC phenotypes in blood. The analysis with flowSOM clearly distinguished four CD1c+ DC clusters (cluster number 22, 37, 44, and 45) (Figure 3A and Figure 3—figure supplement 1A and B). We extracted the data for these four CD1c+ DC clusters and conducted principal component analysis (PCA). The PCA biplot identified CD5 and CD163 as top loadings (Figure 3—figure supplement 1C), which defines the DC2s (cluster 45), CD5−CD163− DC3s (cluster 37), and CD5−CD163+ DC3s (clusters 22 and 44) (Dutertre et al., 2019; Figure 3B, C). Among the identified clusters, we detected a significant reduction in the frequency of cluster 44 in patients compared to controls (Welch t-test, p = 0.03, Figure 3D and Figure 3—figure supplement 1D). Clusters 44 as well as cluster 22 were CD36 and CD14 positive, which indicates these clusters may represent CD14+ DC3s in human blood (Dutertre et al., 2019). However, cluster 44 had relatively higher levels of CX3CR1 than cluster 22 (Figure 3E, F). This suggests that DC3s may be phenotypically bifurcated by CX3CR1 independently of CD14. This is supported by weak correlation between CD14 and CX3CR1 in our RNA-seq data from bulk CD1c+ DCs (Pearson correlation coefficient = 0.35, Figure 3—figure supplement 2A). In addition, we sorted CD14-positive and -negative fractions from CD1c+ DCs of six healthy donors (Figure 3—figure supplement 2B) which showed no significant difference in expression levels for CX3CR1 (Figure 3—figure supplement 2C). CX3CR1 levels were also not significantly different in sorted CD5−CD163+CD14-positive and CD14-negative DC3s from patients with autoimmune diseases, further indicating that CX3CR1 expression in CD1c+ DCs may be independent from CD14 expression in CD1c+ DCs (Figure 3—figure supplement 2D).

Figure 3 with 3 supplements see all
CX3CR1+ DC3s are decreased in the blood of patients with non-infectious uveitis.

(A) Heatmap of the surface protein expression for 49 flowSOM clusters of flow-cytometry analysis of PBMC samples from 26 patients and 11 controls. The four CD1c+ (CD3−CD19−CD56−HLA-DR+CD11c+) DC clusters identified (clusters 22, 37, 44, and 45) are shown (detailed heatmap in Figure 3—figure supplement 1A). (B) Biplot of the normalized surface expression of CD5 and CD163 for the four CD1c+ DC clusters. (C) Correlation plot between manually gated CD5−CD163− DC3s and CD5−CD163+ DC3s and DC3 flowSOM clusters 22, 37, and 44. (D) The frequency of the 4 CD1c+ DC flowSOM clusters as percentage of PBMCs. p values from Welch’s t-test. (E) Principal component analysis (PCA) biplot of the DC3 clusters 22, 37, and 44. Loadings for PC1 and PC2 are shown on the right. (F) Biplots of the normalized surface expression of CD36, CD14, and CX3CR1 in the DC3 clusters 22, 37, and 44. (G) Manual gating strategy of CD1c+ DC subsets based on CD36 and CX3CR1 in PBMCs in uveitis cases and controls. p value from Welch’s t-test. Details on manual gating strategy: see Figure 3—figure supplement 3. Manual gating revealed that the CD14+CD1c+ DCs (DC3s) can be further subdivided in a CX3CR1− and a CX3CR1+ population.

We validated by manual gating that cluster 44 represents the CD36+CX3CR1+ fraction of CD1c+ DCs in peripheral blood (~25% of total CD1c+ DCs) (Figure 3—figure supplement 1E). Comparison between patients and controls corroborated that the frequency of manual gated CD36+CX3CR1+ DC3s were decreased in the blood of non-infectious uveitis patients (Welch t-test, p = 0.029, Figure 3G). In detail, we show that CD14+CD1c+ DCs double positive for CD36+ and CX3CR1 were significantly decreased (p = 0.026), while CD14+CD1c+ DCs not positive for CX3CR1 were not (p = 0.43) (Figure 3G). This supports that CX3CR1 discerns a phenotypic subpopulation of CD14+ DC3s (Figure 3—figure supplement 3), that was diminished in the blood of patients with non-infectious uveitis.

CX3CR1+ DC3s can secrete pro-inflammatory cytokines upon stimulation

We compared the cytokine-producing abilities of CX3CR1+ DC3s to their negative counterparts since the gene module associated with CX3CR1 was enriched for genes involved in cytokine regulation. To this end, we freshly sorted primary human CD1c+ DC subsets based on the surface expression of CX3CR1 and CD36, of which double-positive and -negative subsets could be sorted from the selected healthy subjects in sufficient numbers for analysis (Figure 4—figure supplement 1). Since CD36 is involved in LTA-induced cytokine production (Jimenez-Dalmaroni et al., 2009), we overnight stimulated the CD1c+ subsets with LTA. Both subsets of CD1c+ DCs secreted IL-23 equally strongly (Figure 4A). To assess the secretome of the CD1c+ DC subsets in more detail, we profiled the supernatants of LTA-stimulated CD1c+ DC subsets for additional soluble immune mediators (Supplementary file 1E): The CD1c+ DC subsets could be distinguished based on the secreted protein profile (Figure 4B), of which the levels of TNF-alpha, IL-6, VEGF-A, and TNFR1 showed significant differences between the subsets (Figure 4C). These results show that CD1c+ DC subsets defined on the basis of surface co-expression of CD36 and CX3CR1 have the capacity to secrete pro-inflammatory mediators that participate in the pathophysiology of human non-infectious uveitis.

Figure 4 with 1 supplement see all
CX3CR1+ DC3s secrete high levels of cytokines implicated in non-infectious uveitis.

(A) The CD1c+ DC cells were fluorescent-activated cell sorting (FACS) sorted into CD36+CX3CR1+ and CD36−CX3CR1−CD1c+ DCs (Figure 4—figure supplement 1). The concentration of IL-23 (ELISA) in supernatants of 18 hr cultured primary human CD1c+ DC subsets cells stimulated with lipoteichoic acid (LTA). (B) Heatmap of the levels (Z-score) of 16 detected proteins in supernatants of 18 hr cultured LTA-stimulated primary human CD1c+ DC subsets cells using an in-house multiplex Luminex assay (Supplementary file 1E). (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 (padj = p values from likelihood ratio test Bonferroni corrected for 16 detected proteins).

CX3CR1+ DC3s are detectable in the inflamed eye during non-infectious uveitis

We speculated that CX3CR1+ DC3s are important in the disease mechanisms of uveitis and may be found at increased abundance in the eye during active uveitis. We used single-cell RNA-sequencing data (scRNA-seq) of eye fluid biopsies of four noninfectious patients from Kasper et al., 2021. Cells positive for the CD1c+ DC-specific tissue-marker CLEC10A and negative for the monocyte marker C5AR1 (CD88) (Dutertre et al., 2019; Bourdely et al., 2020; Heger et al., 2018) were used to identify CD1c+ DCs among other immune cells in the scRNA-seq data (Figure 5A). Unsupervised clustering identified three clusters (1, 2, and 3) of different cells within the CD1c+ DC population (Figure 5B). We identified that cluster 1 expressed the gene profile associated with CX3CR1+ DC3s, including relatively higher levels of CX3CR1, CD36, CCR2, and lower levels of RUNX3 compared to the other two CD1c+ DC clusters (Figure 5C), which is in line with the gene profile identified by our bulk RNA-seq analysis. In summary, we conclude that CD1c+ DCs with a gene expression profile similar to CX3CR1+ DC3s can be detected in the eyes of patients during active non-infectious uveitis.

Cells with a gene profile similar to CX3CR1+ DC3s can be found in the inflamed eye during non-infectious uveitis.

(A) Single-cell RNA-sequencing (scRNAseq) analysis of eye fluid biopsies from non-infectious uveitis patients (GSE178833, reported by Kasper et al., 2021). UMAP projections of transcriptomic data from 492 cells (in red) identified by scGate analysis using CLEC10A+ and C5AR1− cells as tissue markers to identify CD1c+ DCs. (B) Unsupervised clustering of CLEC10A+C5AR1− DCs identified in a. (C) Dot plot showing average expression (color-scaled) of key marker genes of the black module and CD14 in each cluster determined in b.

Discussion

In this study of non-infectious uveitis patients and controls, we identified and replicated a CX3CR1-associated gene module in CD1c+ DCs. We were able to track back the gene module to a CX3CR1+ DC3 subset that was diminished in peripheral blood of patients with non-infectious uveitis.

Preceding studies into human CD1c+ DCs revealed functionally distinct subsets termed ‘DC2’ and ‘DC3’, with the DC3 showing both transcriptomic features reminiscent of cDC2s and monocytes – such as elevated CD36 (Villani et al., 2017; Dutertre et al., 2019). DC3s also have distinct developmental pathways and transcriptional regulators compared to DC2 (Villani et al., 2017; Dutertre et al., 2019; Cytlak et al., 2020; Bourdely et al., 2020). Recently, Cytlak et al. revealed that lower expression of IRF8 is linked to DC3 (Cytlak et al., 2020), a transcription factor that was also decreased in non-infectious uveitis. According to Brown et al., 2019, CD1c+ DCs exhibit two subsets: cDC2A and cDC2B, whereas cDC2B exhibits higher expression of CX3CR1 and produces more TNF-alpha and IL-6 than cDC2A upon stimulation. Accordingly, uveitis-associated CX3CR1+ DC3s described in this study exhibit similar phenotypical and functional features.

Dutertre et al., 2019 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. Our flow-cytometry results confirm these findings, but we also show that CD5−CD163+ DC3s that express CD14 are composed of CX3CR1-positive and CX3CR1-negative cells, of which the CX3CR1+ population is implicated in non-infectious uveitis.

Single-cell analysis supported that CD1c+ DCs in eye fluid of patients with non-infectious uveitis contain also a population that has a gene profile reminiscent of CX3CR1+ DC3s, with relatively higher levels of CX3CR1, CD36, CCR2, and lower levels of RUNX3. Patients with SLE display accumulation of CD14+ DC3s in blood (Dutertre et al., 2019), while the population of CD14+ DC3 cells was decreased in non-infectious uveitis patients. The differences between non-infectious uveitis and SLE may be related to distinct (i.e., opposite) immunopathological mechanisms; Type I IFNs drive the maturation of cDC2s into ‘inflammatory cDC2s’ (infcDC2s) (Bosteels et al., 2020) and can induce CD1c+ DCs to express a distinct set of surface receptors (Girard et al., 2020). The type I IFN-α drives immunopathology of SLE and administration of type I IFN therapy can induce lupus-like disease (Rönnblom et al., 1991; Rönnblom et al., 1990). 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 (Wang et al., 2019; Kuiper et al., 2022; Obermoser and Pascual, 2010). Despite the importance of type I IFN signaling on DC3s, our results suggest that DC3s are also dysregulated in conditions associated with decreased type I IFNs (Wang et al., 2019; Kuiper et al., 2022), supporting additional pathways involved in DC3 regulation during chronic inflammation.

We showed that the gene module of CD1c+ DCs showed overlap with the gene signature of disrupted NOTCH2 signaling in cDC2s. Notch2 signaling is mediated via the NF-κB family member Relb in murine cDC2s (Diener et al., 2021). NF-κB signaling via RelB suppresses type I IFN signaling in cDC2s (Saha et al., 2020) while selective deletion of RelB in dendritic cells protects against autoimmunity (Diener et al., 2021; Andreas et al., 2019). It is tempting to speculate that the enrichment for NOTCH gene signatures implies altered NF-κB-Relb signaling in CD1c+ DCs, with a mechanism that varies between diseases mediated by type I IFNs (e.g., SLE) and type I IFN-negative diseases (e.g., uveitis). Although this warrants further investigation, some circumstantial evidence for this is the presence of NF-κB family members in the black module, such as NFKB1 and NFKBIA, the former associated previously with a CX3CR1+ cDC2s, while the latter can regulate Relb function in dendritic cells (Shih et al., 2012; Sun, 2011). NF-κB-Relb signaling has been shown to suppress type I IFN via a histone demethylase encoded by KDM4A which was also in the black module (Jin et al., 2014). Regardless, the NF-κB pathways are regulated by the TNFR1, the main receptor for TNF-alpha (Hayden and Ghosh, 2014; Maney et al., 2014). TNFR1 is expressed at the cell surface of cDC2s and its ectodomain cleaved by the NOTCH2-pathway regulator ADAM10 (Maney et al., 2014; Yang et al., 2016; Iberg et al., 2022). We showed that both TNF-alpha and sTNFR1 were higher in the secretome of activated CX3CR1+ DC3s. This is in agreement with previous studies on CD36+ DC3s (Villani et al., 2017) or CX3CR1+ cDC2B (T-bet−) that also produced higher levels of TNF-alpha (Brown et al., 2019). Interestingly, altered NF-κB signaling specifically in cDC2 is associated with clinical response to anti-TNF-alpha therapy (Andres-Ejarque et al., 2021). Anti-TNF therapy is effective for treatment of non-infectious uveitis (Touhami et al., 2019), while anti-TNF therapy may also result in a dysregulated type I IFN response (Conrad et al., 2018) indicating potentially cross regulatory mechanisms via NF-κB signaling and type I IFN signaling affecting cDC2s. More research is needed to resolve the regulatory mechanisms driving CD1+ DC changes in type I IFN-negative inflammation, including non-infectious uveitis. It is also possible that the change in CD1c+ DCs observed in this study results from cytokine-induced precursor emigration or differentiation or that the affected peripheral blood DC3s marked by CX3CR1 are in a precursor or pre-activation state.

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 (Brown et al., 2019). This is especially interesting in light of growing evidence that microbiota-dependent signals cause autoreactive T cells to trigger uveitis (Horai et al., 2015), 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 (Ko et al., 2020). 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 (Croxford et al., 2015) and GM-CSF has been shown to stimulate the differentiation of human CD1c+ subset from progenitors (Bourdely et al., 2020). However, GM-CSF signaling in conventional dendritic cells has a minor role in the inception of EAU (Bing et al., 2020). Our data support that stimulation of CD1c+subsets with GM-CSF or TLR ligands does not induce the transcriptional features of CD1c+ DCs during non-infectious uveitis, which is in line with previous observations that support that stimulated cDC2s do not convert from one into another subset (Bourdely et al., 2020).

Note that our results of a decreased subset of CD1c+ DCs in non-infectious uveitis are in contrast with previous flow-cytometry reports in non-infectious uveitis (Chen et al., 2016; Chen et al., 2015a; Chen et al., 2014). Chen et al., 2014 reported an increase of CD1c+ myeloid dendritic cells in non-infectious uveitis. It is important to note, however, that their study did not include the DC marker CD11c, thereby including CD1c+CD11c− populations that do not cluster phenotypically with CD1c+ (CD11c+) DCs (e.g., cluster 46, se Figure 3—figure supplement 1A), which may explain the differences compared to our study.

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 toward severe visual complications or blindness. Perhaps targeting CD1c+ DCs may be achieved by dietary (microbiome) strategies and provide relatively safe preventive strategies for non-infectious uveitis.

To conclude, we have found that peripheral blood CD1c+ DCs have a gene module linked to a CX3CR1-positive CD1c+ DC subset implicated in non-infectious uveitis.

Appendix 1

Appendix 1—key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Biological sample (Homo sapiens)Aqueous humor (AqH)Department of Ophthalmology,
University Medical Center Utrecht,
Utrecht, The Netherlands
Biological sample (Homo sapiens)Blood Plasma (EDTA tubes)Department of Ophthalmology,
University Medical Center Utrecht,
Utrecht, The Netherlands
AntibodyAnti-human CD14-FITC Clone:
TÜK4 (mouse monoclonal)
MiltenyiCAT# 130-080-701(Dilution:) 1:50
AntibodyAnti-human CCR2-BV421 Clone:
K036C2 (mouse monoclonal)
BioLegendCAT# 357210(Dilution:) 1:150
AntibodyAnti-human CD11c-PerCP-Cy5.5
Clone: Bu15 (mouse monoclonal)
BioLegendCAT# 337210(Dilution:) 1:200
AntibodyAnti-human CD123-FITC Clone:
6 H6 (mouse monoclonal)
eBioscienceCAT# 11-1239-42(Dilution:) 1:20
AntibodyAnti-human CD14-PerCP-Cy5.5
Clone: HCD14 (mouse monoclonal)
BioLegendCAT# 325622(Dilution:) 1:100
AntibodyAnti-human CD14-BV785 Clone:
M5E2 (mouse monoclonal)
BioLegendCAT# 301840(Dilution:) 1:100
AntibodyAnti-human CD141 BDCA3-APC Clone:
AD5-14H12 (mouse monoclonal)
MiltenyiCAT# 130-090-907(Dilution:) 1:20
AntibodyAnti-human CD163-BV510 Clone:
GHI/61 (mouse monoclonal)
BioLegendCAT# 333628(Dilution:) 1:25
AntibodyAnti-human CD180-PE/Cy7 Clone:
MHR73-11 (mouse monoclonal)
BioLegendCAT# 312910(Dilution:) 1:100
AntibodyAnti-human CD19-eF450 Clone:
HIB19 (mouse monoclonal)
eBioscienceCAT# 48-0199-42(Dilution:) 1:25
AntibodyAnti-human CD19-BV605 Clone:
SJ25C1 (mouse monoclonal)
BDCAT# 562653(Dilution:) 1:50
AntibodyAnti-human CD19-AF700 Clone:
HIB19 (mouse monoclonal)
eBioscienceCAT# 56-0199-42(Dilution:) 1:50
AntibodyAnti-human CD1c-APC Clone:
L161 (mouse monoclonal)
eBioscienceCAT# 17-0015-42(Dilution sorting cohort 2:) 1:50
AntibodyAnti-human CD1c BDCA1-BV421 Clone:
L161 (mouse monoclonal)
BioLegendCAT# 331526(Dilution:) 1:25
AntibodyAnti-human CD1c BDCA1-APC Clone:
L161 (mouse monoclonal)
eBioscienceCAT# 17-0015-42(Dilution purity check cohort 1:) 1:20
AntibodyAnti-human CD20-PE Clone: 2 H7
(mouse monoclonal)
eBioscienceCAT# 12-0209-42(Dilution:) 1:50
AntibodyAnti-human CD3-AF700 Clone: UCHT1
(mouse monoclonal)
BioLegendCAT# 300424(Dilution:) 1:50
AntibodyAnti-human CD304 BDCA4-PE Clone:
AD5-17F6 (mouse monoclonal)
MiltenyiCAT# 130-090-533(Dilution:) 1:20
AntibodyAnti-human CD36-PE Clone: CB38
(mouse monoclonal)
BDCAT# 555455(Dilution:) 1:200
AntibodyAnti-human CD4-BV711 Clone: OKT4
(mouse monoclonal)
BioLegendCAT# 317440(Dilution:) 1:50
AntibodyAnti-human CD45-PerCP Clone: HI30
(mouse monoclonal)
BioLegendCAT# 304026(Dilution:) 1:100
AntibodyAnti-human CD5-BB515 Clone: UCHT2
(mouse monoclonal)
BDCAT# 564647(Dilution:) 1:100
AntibodyAnti-human CD56-AF700 Clone: B159
(mouse monoclonal)
BDCAT# 557919(Dilution:) 1:50
AntibodyAnti-human CD8-V500 Clone: RPA-T8
(mouse monoclonal)
BDCAT# 560774(Dilution:) 1:50
AntibodyAnti-human CX3CR1-PE/Dazzle594 Clone:
2 A9-1 (rat monoclonal)
BioLegendCAT# 341624(Dilution:) 1:100
AntibodyAnti-human HLA-DR-BV605 Clone:
G46-6 (mouse monoclonal)
BDCAT# 562845(Dilution:) 1:150
OtherViability dye and labelling reagent.
Live/Dead-APC-eF780
eBioscienceCAT# 65-0865-14(Dilution:) 1:1000
Chemical compound, drugR848InvivogenCAT# tlrl-r848(Concentration:) 1 µg/ml
Chemical compound, drugLipoteichoic acid (LTA)Sigma-AldrichCAT# L2515-5MG(Concentration:) 1 µg/ml
Chemical compound, drugLipopolysaccharide (LPS)InvivogenCAT# tlrl-3pelps(Concentration:) 10 ng/ml
Chemical compound, drugPam3CSK4InvivogenCAT# tlrl-pms(Concentration:) 5 µg/ml
Chemical compound, drugoxLDLCell BiolabsCAT# STA-214(Concentration:) 50 µg/ml
Chemical compound, drugTGFβ-b2R&D SystemsCAT# 302-B2-002/CF(Concentration:) 100 ng/ml
Chemical compound, drugFLT3LCellgenixCAT# 1415-05(Concentration:) 100 ng/ml
Chemical compound, drugTNFαR&D SystemsCAT# 210-TA-020(Concentration:) 100 ng/ml
Chemical compound, drugS100A12 (EN-RAGE)R&D SystemsCAT# 1052-ER-050(Concentration:) 1 µg/ml
Chemical compound, drugIL-4R&D SystemsCAT# 204-IL-50(Concentration:) 10 ng/ml
Chemical compound, drugIFNα-2aCell SciencesCAT# CRI003B(Concentration:) 1000 U/ml
Chemical compound, drugGM-CSFR&D SystemsCAT# 215 GM-500(Concentration:) 800 U/ml
Commercial assay or kitOlink Target 96 Immuno-OncologyOlinkCAT# 95311Olink Targeted Proteomics analysis
Commercial assay or kitTruSeq RNA Library Prep KitIlluminaCAT#RS-122-2001RNA-seq
Commercial assay or kitAllPrep DNA/RNA/miRNA Universal KitQIAGENCAT# 80224RNA isolations
Commercial assay or kitCD1c (BDCA-1)+Dendritic
Cell Isolation Kit, human
Miltenyi BiotecCAT#130-119-475isolation of CD1c+ DCs from PBMCs.
Commercial assay or kitCD19 MicroBeads, humanMiltenyi BiotecCAT#130-050-301Depletion of CD19+ B cells from PBMCs
Commercial assay or kitCD304 (BDCA-4/Neuropilin-1)
MicroBead Kit, human
Miltenyi BiotecCAT#130-090-532Depletion of plasmacytoid dendritic cells from PBMCs
Commercial assay or kitHuman IL-23 Quantikine ELISA KitR&D SystemsCAT# D2300BQuantification of IL-23 in supernatant of CD1c+ DC cultures
Recombinant DNA reagentCD36 FW (Sequence 5′–3′)
AAAGAGGTCCTTATACGTACAGAGTTCGT
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentCD36 RV (Sequence 5′–3′)
AGCCTTCTGTTCCAACTGATAGTGA
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentRUNX3 FW (Sequence 5′–3′)
CAATGACGAGAACTACTCCGC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentRUNX3 RV (Sequence 5′–3′)
GAAGCGAAGGTCGTTGAACC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentGUSB FW (Sequence 5′–3′)
CACCAGGGACCATCCAATACC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentGUSB RV (Sequence 5′–3′)
GCAGTCCAGCGTAGTTGAAAAA
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentCCR2 FW (Sequence 5′–3′)
CCACATCTCGTTCTCGGTTTATC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentCCR2 RV (Sequence 5′–3′)
CAGGGAGCACCGTAATCATAATC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentCX3CR1 FW (Sequence 5′–3′)
AGTGTCACCGACATTTACCTCC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentCX3CR1 RV (Sequence 5′–3′)
AAGGCGGTAGTGAATTTGCAC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentIRF8 FW (Sequence 5′–3′)
CGACGCGCACCATTCA
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentIRF8 RV (Sequence 5′–3′)
GCTTGCCCCCATAGTAGAAGCT
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentTLR7 FW (Sequence 5′–3′)
CAAGAAAGTTGATGCTATTGGGC
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentTLR7 RV (Sequence 5′–3′)
TGGTTGAAGAGAGCAGAGCA
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentCCR5 FW (Sequence 5′–3′)
TGCTACTCGGGAATCCTAAAAACT
Integrated DNA TechnologiesqPCR primer
Recombinant DNA reagentCCR5 RV (Sequence 5′–3′)
TTCTGAACTTCTCCCCGACAAA
Integrated DNA TechnologiesqPCR primer
Software, algorithmR Project for Statistical Computing;
R version 4.0.3 (2020-10-10)
https://www.r-project.org/RRID:SCR_001905RNA-seq,flowSOM, statistical analysis
Software, algorithmFlowJo v10.6.1BD Biosciences;
https://www.flowjo.com/solutions/flowjo
RRID:SCR_008520Flow cytometry
Software, algorithmSeurat v3.1.5Stuart et al., 2019;
http://seurat.r-forge.r-project.org/
RRID:SCR_007322scRNA-seq analysis
Software, algorithmDESeq2 v1.30.1https://bioconductor.org/packages/release/bioc/html/DESeq2.htmlRRID:SCR_015687RNA-seq analysis
Software, algorithmWGCNA v 1.70-3http://www.genetics.ucla.edu/labs/horvath/CoexpressionNetwork/RRID:SCR_003302Co-expression network analysis
Software, algorithmflowSOMhttps://github.com/SofieVG/FlowSOM, Van Gassen et al., 2023RRID:SCR_016899Flowcytometry analysis using a Self-Organizing Map.
Software, algorithmUCell v1.3.1https://github.com/carmonalab/UCell, Andreatta and Carmona, 2023scRNAseq Module score
Software, algorithmscGate v1.0.0https://github.com/carmonalab/scGate, Andreatta et al., 2023Purification of intraocular CD1c+ DCs in scRNAseq data

Data availability

All raw data and data scripts are available via dataverseNL: https://doi.org/10.34894/9Q0FVO and deposited in NCBI's Gene Expression Omnibus accessible through GEO Series accession numbers GSE195501 and GSE194060.

The following data sets were generated
    1. Kuiper JJ
    (2022) NCBI Gene Expression Omnibus
    ID GSE195501. Whole transcriptome-sequencing of CD1c+ conventional type 2 dendritic cells of human non-infectious uveitis patients [Replication cohort].
    1. Kuiper JJ
    (2022) NCBI Gene Expression Omnibus
    ID GSE194060. Whole transcriptome-sequencing of CD1c+ conventional type 2 dendritic cells of human non-infectious uveitis patients.
    1. Kuiper JW
    (2022) DataverseNL
    Data and R Scripts for: "Transcriptome network analysis implicates CX3CR1-positive type 3 dendritic cells in non-infectious uveitis".
    https://doi.org/10.34894/9Q0FVO
The following previously published data sets were used
    1. Dicken J
    2. Mildner A
    3. Leshkowitz D
    4. Touw IP
    (2014) NCBI Gene Expression Omnibus
    ID GSE48590. The affect of specific ablation of Runx3 from Esam splenic dendritic cells.
    1. Briseño CG
    2. Satpathy AT
    (2018) NCBI Gene Expression Omnibus
    ID GSE119242. Trancriptional profile of WT and Notch2 cDC2s after immunization with SRBC.
    1. Bosteels C
    2. Neyt K
    3. Vanheerswynghels M
    4. van Helden MJ
    (2020) NCBI Gene Expression Omnibus
    ID GSE149619. Inflammatory Type 2 cDCs Acquire Features of cDC1s and Macrophages to Orchestrate Immunity to Respiratory Virus Infection.
    1. Kirkling ME
    2. Cytlak U
    3. Lau CM
    4. Lewis KL
    (2018) NCBI Gene Expression Omnibus
    ID GSE110577. Notch signaling facilitates in vitro generation of cross-presenting classical dendritic cells.
    1. Kasper M
    2. Heming M
    3. Heiligenhaus A
    4. Meyer zu Hörste G
    (2021) NCBI Gene Expression Omnibus
    ID GSE178833. Intraocular dendritic cells characterize HLA-B27-associated acute anterior uveitis.

References

Decision letter

  1. Lynn M Hassman
    Reviewing Editor; Washington University School of Medicine, United States
  2. Betty Diamond
    Senior Editor; The Feinstein Institute for Medical Research, United States
  3. James Walsh
    Reviewer

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Whole transcriptome-sequencing and network analysis of CD1c+ human dendritic cells identifies cytokine-secreting subsets linked to type I IFN-negative autoimmunity to the eye" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Betty Diamond as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: James Walsh (Reviewer #1).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) Improved informatics analysis to correct for false discovery and stronger correlation with intra-ocular cDC2s, per reviewer #3

2) Revision of claims to have identified a new type of cDC2, reframed to fit this cell state into the current, more rigorously defined, classification of cDC2 subtypes (authors have already cited key papers). This analysis may also benefit from stronger analysis of the transcriptional profile of intraocular (ie tissue state) CD36/CX3CR1+ cDC2s. Conclusions must be tempered given concerns of reviewers #1 and #2

Also, please correct number of references, it was at times impossible to determine which sources were being cited.

Reviewer #1 (Recommendations for the authors):

It is unclear why the CD14 is plotted vs CD3 in Figure 1F.

Line 166: Repeated word "co-expressed expressed".

Line 247: Run on sentence (and used multiple times).

Line 384: "No minimize bias" needs to be corrected.

Reviewer #2 (Recommendations for the authors):

Specific comments:

An issue is that the cd1c+ cells used for RNAseq were isolated using microbeads which is extremely impure. What other cells contaminated the prep, can the gene expression be reliably de-convoluted? What is the cell purity prior to RNAseq?

5B and D how many patients are displayed? Patient numbers should be shown for each figure throughout the manuscript.

Phenotype of the DCs should be shown in the presence of Notch2 and ADAM10 inhibitor? Is the DC3 reduced? CX3CR1, CD36, ccr2 CD163?

What is the output of the CX3CR1, CD36, ccr2 CD163 DC3 in Notch ligand-conditioing human CD34 HPC derived DCS?

Functional analysis for understanding T cell priming by the new DC3 subsets should be shown. what is the implication of the results for autoimmune Uveitis.

Technical details that should be addressed:

– All versions of the software should be noted in the methods section, for example ConsensusClusterPlus, Deseq2, WGCNA, etc

– For logicleTransform – were the default settings used or were the parameters altered, this needs to be indicated for interpretation of the log linear transformation.

– Heatmap color palettes are not ideal for colorblind readers, would recommend using viridis palettes

– For the differential gene expression and generation of weighted gene co-expression networks, why was the raw p-value used as a cut-off and not adjusted p-value? Was there any controls for possible false positives?

– For the FlowSOM clusters – what are the approximate sizes Cluster 81, 41, 61, and 83? What percent of the peripheral blood? What is the relative proportion of these clusters in Uveitis vs control?

Author should cite Korenfeld at al JCI insight 2017 on page 9

Reviewer #3 (Recommendations for the authors):

In general, figure panels should be made larger; several -- particularly in Figure 3 -- have panels that are nearly unreadable even when printed at full-page scale.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "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" for further consideration by eLife. Your revised article has been evaluated by Betty Diamond (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

The authors utilized deep transcriptional profiling of isolated CD1c+ peripheral blood cells to identify a peripheral blood biomarker of uveitis, followed by flow cytometric analysis of protein expression to test the hypothesis that a subset of CD1c+ cells are reduced in uveitis patients. Identification of peripheral blood biomarkers of uveitis is important and their study is based on analysis of a reasonable number of study subjects with active disease, therefore the data supporting the first conclusion that a differential gene expression profile exists in uveitis is well-supported and convincing. The subsequent analysis which attempts to define and new cell type based on comparison with published expression datasets, while hypothesis-generating, is inadequate, as sufficient phenotypic and functional analyses were not performed to reach their conclusions. Similarly, the comparison with genes expressed by ocular CD1c+ cells from a separate dataset is not sufficient to conclude that the peripheral blood CD1c+ cells migrate into the eye.

Reviewer #1 (Recommendations for the authors):

The authors have addressed many of the concerns in the original reviews, and while the discovery of a subpopulation of blood-derived DCs that are changed in uveitis would be interesting the revision does not leave me convinced that this is truly a unique population of cells. Specifically the flow cytometry data in Figure 5/supplement suggests that the CD14 population is derived from a population that is bisected by the original gate, rather than from a population of CD1c+ DCs limiting functional analysis, and the scRNA seq does not show that the cells expressing black module genes are distinct from those that don't.

Reviewer #2 (Recommendations for the authors):

The authors set out to utilize advance informatics techniques to advance our understanding of a cell type previously shown to play a role in uveitis, type 2 conventional dendritic cells (previously termed mDC1). They accumulated a valuable set of samples from untreated, active disease and employed a testing and validation cohort which reproduced a core set of genes differentially expressed within the CD1c+ dendritic cells in the peripheral blood of patients with uveitis. Importantly these cells appear altered regardless of the subtype of uveitis. They then astutely question whether the transcriptional signature simply represents different proportions of cell subtypes or states and test this hypothesis using flow cytometry.

They also attempt to probe the mechanism driving this particular cell type/state by cleverly drawing on published data sets, however these hypothesis-generating experiments are not validated experimentally in a uveitis system.

They utilize in vitro analysis of similar cells isolated from human patients and show that the cells can be induced to make a specific set of cytokines which is different from a related dendritic cell type, however the lack of concordance with published activated DC2s, or the transcriptional signature in their own ex vivo activated DC2s raises more questions than it answers.

The manuscript was difficult to read, largely because it is trying to accomplish too many goals, but also because the informatics techniques and external data sets were not described sufficiently for an average reader to readily evaluate the method and conclusions, and the conclusions were overstated throughout the paper. Overall, too many hypotheses were tested, alternative hypotheses were not considered/discussed. The data should probably be divided into at least 2 papers, one which explores the peripheral blood subsets more concretely, and one which attempts to elucidate a mechanisms by which low RUNX3 expression is associated with the genes expressed more highly in DC2s in the peripheral blood of uveitis patients.

1. The rationale for depleting CD14 in the validation cohort is not clear and justified. CD14 expression on dendritic cells has been established in the literature (ex Duterte Immunity 2019). This reviewer wonders if a better flow of data would be to start with the current second cohort, CD14 depleted, ID the black module, realize that some genes discovered have been associated with CD14+DC3, then utilize a second cohort that includes CD14+CD1c+ cells to validate and expand the original gene set.

2. Experiments probing the transcriptional regulation of the uveitis-enriched gene set are not definitive, but hypothesis-generating, and left in this story, are distracting from the primary observation of a cell type differentially present in the blood of uveitis patients.

3. The authors claim to have found an inflammatory cell type, but the gene expression profile does not recapitulate inflammation-induced DC2 gene expression profile. Figure 3 attempts to shed mechanistic light but only opens more questions, like if viral infection in mice (figure 3A lower panel) and multiple inflammatory stimuli (figure 3C) induce a transcriptional program opposite what they are finding in these cells in the peripheral blood of uveitis patients, how are the cells identified in this paper likely relevant to eye inflammation?

4. Pro-inflammatory CD14+DC3 increased in blood in SLE patients, but according to authors interpretation, they find a similar cell type is decreased in peripheral blood in uveitis. The authors adhere to a possible explanation that these cells are trafficking to the eye, but by the same logic, SLE patients should have an even more significant reduction, not increase in their peripheral DC3 counts. This is not discussed.

5. Even more importantly, the Chen et al. papers which are cited generally by the authors, showed the opposite- that CD1c+ DCs are increased in the blood of patients with uveitis, and correlate with disease activity. No attempt is made by the authors to compare their data with this data derived from a larger number of patients, or discuss the difference. The authors have essentially ignored this key discrepancy with previously published data.

6. Despite changing some text around the hypothesis that this cell type is reduced in the blood because it migrates to the eye (numerically impossible as pointed out by prior reviewer #1), this notion is reiterated several other places in the paper and should be removed, and frankly, reconsidered. It is also possible that the change in peripheral blood cell type/state frequency results from cytokine-induced precursor emigration/differentiation. Perhaps there is actually a fraction of DC2-type cells that are increased in the blood in uveitis (as Chen found) but express the transcriptional program identified in the paper because they are pre-activated cells?

7. The analysis of ocular DC2s which concludes that the current cell type, CD14+DC3s with black module gene expression, are present in the eye is not convincing. The methods are not clearly explained, however it appears that rather than utilizing unbiased cluster analysis and performing differential gene expression analysis between the patients with non-infectious uveitis, and the control (endophthalmitis), or using another technique like GSEA which they used in previous figures to correlate the black module genes with the genes expressed by the ocular DC2s, the authors appear to have selected individual cells that expressed the black modules genes, then compared the expression of black module genes to the cells that did not express black module genes.

8. The authors state in their response to reviewer #1 comments "we profiled available eye fluid biopsies and paired plasma by Olink proteomics to measure immune mediators from patients and controls from this study (and several additional samples, including aqueous humor from non-inflammatory cataract controls – see revised Figure 5 panel D). This analysis shows that cytokines produced by CD36+CX3CR1+ DCs such as TNF-α and IL-6 are specifically increased in eye tissue of patients, but not in blood."-Neither this data, or discussion of it are included in the revised manuscript.

Finally, a concern is that the title is a gross overstatement of their findings:

1. They have not demonstrated that the cells in this paper induce inflammation, and especially not in the context of uveitis- only that similar cells from healthy patients produce a different set of cytokines when stimulated in vitro compared to another cell type.

2. They have not demonstrated that these very cells migrate to the eye, only that similar genes are present in a possibly similar ocular cell type in another data set.

3. They do not demonstrate type I IFN-negative autoimmunity in the eye. This was a huge stretch, presumably from prior assumptions about the mechanisms driving uveitis along with the finding that their cell type does not share a transcriptional program with murine DC2s activated in a viral infection.

In regard to addressing the Previous Editor Concerns:

1. The informatics analysis for most of the paper is likely sufficient, however the methods are not communicated succinctly and clearly such that non-informatics experts can understand the rationale and method for each analysis. The analysis of the intraocular DCs was not clear and from the details provided, did appear appropriate.

2. Revision of claim to have identified a new type of cDC2- This is still not satisfactory as:

a. authors have not ruled out the possibility that the cells are monocyte-derived by transcriptional analysis, protein expression or functional analysis. The use of CD14-deplete cells to recapitulate the gene expression profile is not sufficient to determine that the cells in this paper are not monocyte-derived, as CD14-expression is demonstrated on cells confirmed by FLT3L response to be dendritic cells in Duterte et al. Immunity 2019.

b. To be defined as a new subset of the previously defined DC3 subset, one would need to exactly replicate the marker expression and then show that the new markers subset that subset further, the current manuscript may simple be focusing on different genes/proteins expressed by one or more previously described subsets.

c. As this paper is useful for describing a cell type or state that differentiates uveitis from healthy patients, these experiments do not need to be done to publish this paper, but the naming of the cell type should be tempered to simply describe the markers that were expressed and suggest how they fit into the Duterte/Villani schema of DC2/DC3 classification. In actuality, the discrimination of cDC2 from monocyte-derived DC2-like cells has proven difficult in many papers, thus the authors are advised to stay out of the mud, so-to-speak.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Transcriptome network analysis implicates CX3CR1-positive type 3 dendritic cells in non-infectious uveitis" for further consideration by eLife. Your revised article has been evaluated by Betty Diamond (Senior Editor) and a Reviewing Editor.

Summary:

CD1c+ dendritic cells have been found in the peripheral blood and eyes of patients with active uveitis. The authors set out to characterize CD1c+ dendritic cells in uveitis. They establish that the CD1c+ population varies between uveitis and healthy controls in both gene expression and the frequency of a subset of CD1c+ DCs, recently termed DC3s. Finally, the authors utilize a previously published dataset to show that cells with similar gene expression can be found in the eye during active uveitis.

Review:

The authors compare sorted CD1c+ DCs from patients with non-infectious uveitis and healthy controls and find a gene expression signature associated with uveitis, regardless of anatomic subtype or severity, that includes expression of the chemokine receptor CX3CR1. They corroborate this finding with a second cohort of CD1c+CD14+ cells, which strengthens the uveitis-specific CD1c+ DC signature.

The authors then compare the genes enriched in these uveitis CD1c+ DCs with previously published datasets analyzing murine CD1c+ DCs. They found more overlap between uveitis patient CD1c+ DC genes and murine RUNX3/NOTCH2 KO CD1c+ DCs than with murine viral-infected cells. While the data suggests that IFN signaling may be less relevant in these cells, the authors' conclusion that the genes differentially expressed by peripheral blood CD1c+ DCs in uveitis are not mediated by type I IFNs is overstated and alternative explanations should also be considered.

Next, the authors used flow cytometry to show that blood CD36+CX3CR1+CD1c+ DCs (thus labeled DC3s) were diminished in uveitis vs healthy controls, suggesting that the difference in CD1c+ gene expression between uveitis and healthy controls may actually be due to differential presence of CD1c+ subsets. The difference is small, but statistically significant, although the observation could have been strengthened by quantifying this cell type longitudinally in the same patients during active and inactive disease.

Next the authors found that LTA-stimulated CX3CR1+ DC3s from healthy controls secrete higher levels of uveitis-relevant inflammatory cytokines, including TNF-α, compared to CX3CR1- DC3s. This experiment was performed on a small number of healthy controls and not compared with cytokine production by DC3s from uveitis patients, which could have further supported the authors conclusion that the differential gene expression identified in Figure 1 was due to reduced proportions of CX3CR1+ DC3 cells in uveitis patients vs healthy controls, rather than qualitative differences between uveitis and healthy DC3s.

Finally, the authors find expression of CD36 and CX3CR1 on CLEC10A+ (which they use as a proxy for CD1C) cells by aqueous dendritic cells from a previously published dataset, suggesting that DC3s similar to those found at reduced frequency in the peripheral blood are present in aqueous inflammation, supporting their relevance in uveitis.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Specifically, there are errors in the methods, results, and legends that must be corrected.

Reviewer #1 (Recommendations for the authors):

In the revised paper entitled "Transcriptome network analysis implicates CX3CR1-positive type 3 dendritic cells in non-infectious uveitis," Hiddingh, Vandit, Verhagen et al. explore the gene expression of PBMCs in non-infectious uveitis patients and demonstrate a CX3CR1-positive Cd1c+ gene signature that is altered in non-infectious uveitis. They show that this population is decreased in the peripheral blood, could be regulated by notch signaling, expresses proinflammatory cytokines upon stimulation with LTA, and are present in the eye during uveitis.

The hypothesis that a cD1c+ population of DCs that could be related to uveitis in humans is intriguing and deserves further study. Their use of multiple methods to explore this population and the use of multiple cohorts are a strength of the manuscript, and it raises many intriguing questions that are potentially interesting, such as if this population is expressing inflammatory cytokines upon stimulation, why are they decreased in uveitis?

There are still areas where the manuscript is hard to follow and there are some concerns with the experimental methodology. The difficulty following the author's story is partially due to errors in cross-referencing statements in the text with their figures that support the data and understanding the experiments from the figure legends. For instance, in lines 313-315, the authors state "Furthermore, in CD1c+ DCs from healthy human donors, IFN-α did not induce downregulation of RUNX3 as observed CD1c+ DCs from non-infectious uveitis patients. Figure 2 – Figure 2 supplement 1". Data to support this statement is not found in Figure 2 or Figure 2 supplements (only healthy control data). The Figure 2 Supplement 1 legend references "the notch-negative condition in d" with a d in that figure.

Methodologically, the backgating for the manual gating of CD11c/CD1c suggests that the CD36+CX3CR1+ population is really part of a larger population of CD11c+ cells, raising the question of if this population is too poorly defined in this experimental context. This concern is slightly ameliorated by the appearance of a CD36hiCX3CR1hiCD1c+ population in the unsupervised clustering.

Despite these weaknesses, there is enough strength in using multiple methods and replication with multiple patient cohorts to overcome these concerns and to utilize it as a basis to further explore the functions of this population in uveitis pathogenesis.

Reviewer #4 (Recommendations for the authors):

The authors have responded to most of the previous reviews and have generated a more clear and cohesive manuscript.

Additional recommendations:

Figure 1

The text rationale for CD14 separation is confusing, consider omitting it.

A better methodology would have been to repeat analysis with new cohort I followed by validation using new cohort II rather than simply comparing the cohorts, but this reads more clearly and logically than the prior version and the overall conclusions seem valid.

Figure 1 Sup 1 not needed, emphasized the odd methodology sorting "cohort II" for CD14- recommend omitting this from the final version, or using instead Figure 3- Supplement 2 could be moved to the supplement for Figure 1 to explain why black module (from the CD14-sorted cohort) is stronger than then enriched modules from cohort I.

CD14+ CD1c+CD11c+CD36+CXCR3+ DC3s seem to be a subset of CD1c+CD11c+CD36+CXCR3+ DC3s, which may be why there is a stronger gene expression signature black module from cohort II vs the blue and green modules from cohort 1.

The supplemental experimental data shows that sorted DC3s from healthy peripheral blood treated with a variety of inflammatory stimuli upregulate RUNX3. One alternative explanation not discussed by the authors is that peripheral blood DC3s are in a precursor or pre-activation state.

Text: in CD1c+ DCs from healthy human donors, IFN-α did not induce downregulation of RUNX3 as observed in CD1c+ DCs from non-infectious uveitis patients, however supplemental figure 2 only tests CD1c DCs from healthy patients. CD1c+ DCs from uveitis patients were never stimulated with IFN to test whether they downregulate RUNX3 after this stimuli. This textual discussion of the experimental data is misleading.

Sup figure 3 final panel should be G, not H.

Aqueous scRNA samples are listed as obtained from Utrecht in the methods section and should cite the previous dataset.

Data used from prior sources should be more clearly detailed in legends and text. As the paper reads, it appears that the authors did the murine BMDC on the OP9 culture experiment detailed in Sup Figure 2.

Figure 5 image is very misleading – "purify tissue CD1c+ DCs" suggests that cells were purified resulting in the displayed UMAP. CLEC10A and C5AR should both be shown and the label should not state CD1c+ if this expression was not assessed- this is misleading.

https://doi.org/10.7554/eLife.74913.sa1

Author response

Essential revisions:

1) Improved informatics analysis to correct for false discovery and stronger correlation with intra-ocular cDC2s, per reviewer #3

We would like to thank the editor for the time and effort to review our work. As discussed in the response to reviewer #3, we outlined a 3-step strategy to control for false positive findings in our RNA-seq analysis, which includes the use of an independent cohort and replication of gene modules over individual genes (we believe is a more challenging, but also more robust approach rarely conducted in RNA-seq analysis of dendritic cells – especially in the era of single-cell RNA seq). We have provided data analysis including adjusted P-values in Figure 1A and Figure 1- Supplement 2F, as well as in the Supplementary File 1H-1J. We also outlined in answer to reviewer #2 that we significantly expanded the analysis of intra-ocular cDC2s in Figure 6. We hope the editor agrees this has now been adequately addressed.

2) Revision of claims to have identified a new type of cDC2, reframed to fit this cell state into the current, more rigorously defined, classification of cDC2 subtypes (authors have already cited key papers). This analysis may also benefit from stronger analysis of the transcriptional profile of intraocular (ie tissue state) CD36/CX3CR1+ cDC2s. Conclusions must be tempered given concerns of reviewers #1 and #2

We have significantly reanalyzed the data presented in Figure 4 to fit the current nomenclature and show that the uveitis-associated cDC2s are a CD14+ DC3 subset most distinguished by CD36 and CX3CR1 (Figure 4, Figure 4 – Supplement 1, Figure 4 – Supplement 2). We have tempered conclusions based on the suggestions of the reviewers and adapted our discussion to include this current nomenclature. We have also substantially extended the transcriptional profile of intraocular CD36+/CX3CR1+ cDC2s by addition of Figure 6. Here, we clustered eye-infiltrating cDC2s according to the expression of the uveitis-associated gene module and show that these cells show relatively higher expression for CD36, CX3CR1, and lower RUNX3, but comparable levels of CD14 (Figure 6C) – closely corroborating our blood CD1c+ DC analyses. These DC3s were also found at higher frequency in the eye of patients (Figure 6D). We hope the editor agrees we have substantially improved the analysis of intraocular DCs.

Also, please correct number of references, it was at times impossible to determine which sources were being cited.

We have corrected the numbering of references and added references per suggestion of the reviewers.

Reviewer #1 (Recommendations for the authors):

It is unclear why the CD14 is plotted vs CD3 in Figure 1F.

We have changed this in revised Figure 1F and plotted CD14 against CD45 and added additional data on gating for all other markers (CD3, CD19, CD1c) used to assess the purity of the cell fractions used for RNA-seq in the revised Figure 1 – Supplement 1. We hope the reviewer considers this to be more appropriate.

Line 166: Repeated word "co-expressed expressed".

Line 247: Run on sentence (and used multiple times).

Line 384: "No minimize bias" needs to be corrected.

We thank the reviewer for bringing these typos to our attention and have corrected these accordingly.

Reviewer #2 (Recommendations for the authors):

Specific comments:

An issue is that the cd1c+ cells used for RNAseq were isolated using microbeads which is extremely impure. What other cells contaminated the prep, can the gene expression be reliably de-convoluted? What is the cell purity prior to RNAseq?

We agree that the reviewer that microbeads may provide a challenge in purification of cell subsets. With this in mind, we therefore went to great lengths to attribute the results to CD1c+ DCs. First of all, we used a series of MACS isolations to remove contamination from other cells. For cohort I, fresh PBMCs were immediately subjected to magnetic-activated cell sorting (MACS) for the removal (positive selection) of CD304+ cells (pDC), followed by CD19+ cells (B cell), and subsequently isolation of CD1c+ cells by using the CD1c+ (BDCA1) isolation kit (see Methods). We now have provided additional experimental data of flow cytometry analysis of the MACS-isolated cell fractions used for RNA-seq of cohort I. As shown in Figure 1—figure supplement 1, we used highly purified CD1c+ cDC2 cell fractions (median [interquartile range] % = 96[3]% pure) with no difference between groups in any of the residual cell populations. We outlined this in the Results section on page 10. As shown in revised Figure 2A-C (see also reviewer #1 comment 1), we conducted additional experiment to show that the black module genes are not from CD14+ monocytes (not dependent on CD14+ surface expression), but from CD1c+ DCs. Also, in cohort II, we used CD3-CD19-CD56-HLA-DR+CD11c+CD1c+CD14- cells for RNA-seq analysis sorted by flow cytometry. We hope that the reviewer agrees we have provided sufficient additional information to conclude that the results can be attributed to CD1c+ DCs.

5B and D how many patients are displayed? Patient numbers should be shown for each figure throughout the manuscript.

We have added the amount of patients and donors used for each experiment in Figure 1A, B, D, Figure Supplement 1, Figure 2A-D,F, Figure 3C, Figure 4A,D,G,H,I, Figure 4 – Supplement 1, Figure 5A-D, and Figure 6A, D. We hope the reviewer agrees this has now been sufficiently addressed.

Phenotype of the DCs should be shown in the presence of Notch2 and ADAM10 inhibitor? Is the DC3 reduced? CX3CR1, CD36, ccr2 CD163?

As discussed above Reviewer #1: question 3. Upon request, we have conducted experiments using the inhibitors and our CD1c+ DC flow cytometry panel, but experienced significant autofluorescence and artifacts (see gating example in Author response image 1) that hampered the correct identification of CD1c+ DCs and their subsets using CD34+ HPC derived DCs. Unfortunately, so far, our attempts have failed to overcome these artifacts. We therefore believe it is appropriate to remove the supplemental figure from the manuscript. Regardless, we show multiple complementary lines of evidence from transcriptomic analysis in transgenic and human cells that strongly link NOTCH2 signaling to the black module in cDC2s. We address this in more detail in the Discussion section. We hope the reviewer agrees that for the scope of this paper, this has been sufficiently addressed.

Author response image 1
Manual gating example of human CD34-HPC derived DCs shows substantial autofluorescence.

What is the output of the CX3CR1, CD36, ccr2 CD163 DC3 in Notch ligand-conditioing human CD34 HPC derived DCS?

Functional analysis for understanding T cell priming by the new DC3 subsets should be shown. what is the implication of the results for autoimmune Uveitis.

Technical details that should be addressed:

– All versions of the software should be noted in the methods section, for example ConsensusClusterPlus, Deseq2, WGCNA, etc

We have added the versions for software used in the method section and in the Key Resources Table form.

– For logicleTransform – were the default settings used or were the parameters altered, this needs to be indicated for interpretation of the log linear transformation.

Yes, the default settings were used for the logicleTransform() function. We have added a detailed Rmarkdown to the study DataverseNL repository (see Data availability in methods section or https://doi.org/10.34894/9Q0FVO) so readers can detail and reproduce all our analysis.

– Heatmap color palettes are not ideal for colorblind readers, would recommend using viridis palettes

We agree with the reviewer and have changed the heatmap colors to viridis palettes in Figure 4 and Figure 4 – Supplement 1.

– For the differential gene expression and generation of weighted gene co-expression networks, why was the raw p-value used as a cut-off and not adjusted p-value? Was there any controls for possible false positives?

Our aim in this study was to use network-based inference of the CD1c+ DC compartment in non-infectious uveitis with the goal to reproduce gene modules (i.e., pathways) over of individuals genes. Although WGCNA does not recommends filtering genes by differential expression, we aimed to prioritize genes linked to uveitis while controlling for possible false positives in 3 steps: (1) As a compromise between not filtering genes and using only a strict fraction of differentially expressed genes at Padj<0.05, we included genes with raw P<0.05 to construct uveitis-relevant gene modules with sufficient resolution. (2) We used an independent cohort in which we constructed modules and tested for overlap in co-expressed genes (i.e., replicate a gene module) (3) that also should show consistent direction of effect (Figure 2G). In our opinion, the strategy to replicate disease-associated gene modules is much more robust as illustrated by corroboration of these data by unbiased flow cytometry data and scRNAseq analysis. Regardless, we have now also added analysis with adjusted P-values in the result section (revised Figure 1B, Figure 1 Supplement 2F), and describe that genes with high module membership were also significantly increased in uveitis after correction for multiple testing in the result section on page 11 (Padj<0.05, Supplementary File 1H). We hope the reviewer agrees that we have used several complementary strategies to control for false positives in our analysis.

– For the FlowSOM clusters – what are the approximate sizes Cluster 81, 41, 61, and 83? What percent of the peripheral blood? What is the relative proportion of these clusters in Uveitis vs control?

We would like to thank the reviewer for raising these questions. We determined the approximate cluster sizes for cluster 81,41,61, and 83 which revealed that the clusters together represented <1% in the singlet gate (input data for flowSOM). Because identification of rare cell types is challenging for unsupervised gating algorithms, including flowSOM (reliance on cellular density which is non optimal for rare population detection), we aimed to improve detection of CD1c+ DC phenotypes by gating out lymphocytes. To this end, we reanalyzed the flowcytometry data of PBMCs of patients and controls with flowSOM using Lin-(CD3-CD19-CD56-) HLA-DR+ gated cells as input. As shown in Figure 4 – Supplement 1 this clearly identified (again) a cluster of 4 CD1c+ DC phenotypes, which we show are highly reminiscent of DC2 and DC3 cells (revised Figure 4B). We have updated Figure 4A-F with this significantly improved data and now show that the percentage of one particular cluster (now called cluster 44) is the single cluster that is significantly decreased in blood of patients (revised Figure 4B,C), and importantly, this cluster is distinguished by higher CD36 and CX3CR1 expression but not CD14, in line with our RNA-seq data and scRNAseq analysis (revised Figure 4E, 4F, 4J, Figure 4 – Supplement 2). Finally, we show that cluster 44 can be well identified by manual gating of CD36+CX3CR1+ CD1c+ DCs. We have outlined this extensive reanalysis in the result section on page 15-16.

Author should cite Korenfeld at al JCI insight 2017 on page 9.

We have added the citation at that positioning in the manuscript (which is now page 15).

Reviewer #3 (Recommendations for the authors):

In general, figure panels should be made larger; several -- particularly in Figure 3 -- have panels that are nearly unreadable even when printed at full-page scale.

We have increased the panels of the figures and show some heatmap figures in greater detail (Figure 4 – Supplement 1) in the supplemental data.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1 (Recommendations for the authors):

The authors have addressed many of the concerns in the original reviews, and while the discovery of a subpopulation of blood-derived DCs that are changed in uveitis would be interesting the revision does not leave me convinced that this is truly a unique population of cells. Specifically the flow cytometry data in Figure 5/supplement suggests that the CD14 population is derived from a population that is bisected by the original gate, rather than from a population of CD1c+ DCs limiting functional analysis.

Our sincere thanks go out to the reviewer for providing additional feedback. We think the reviewer may refer to the Figure supplement of Figure 4 (which in this revision is now Figure 3). Flow cytometry analysis in the figure supplement of Figure 3 provide more detail on CD14 expression in the CD11c+CD1c+ DC gate. We have also updated Figure 3G, which shows that CD14+CD1c+ DCs can be subdivided in CX3CR1+ and CX3CR1- cells. The reviewer will hopefully agree that we have adequately addressed this issue.

…and the scRNA seq does not show that the cells expressing black module genes are distinct from those that don't.

We agree that further improvements can be made to our intital approach. Our aim was to demonstrate that inflamed eyes of patients contain cells that express genes associated with CX3CR1+ DC3s. We reanalyzed the single-cell RNA seq data and sorted CD1c+ DCs using CLEC10A as a classical cDC2 marker (Heger et al. Front Immunol. 2018), removing monocytes and macrophages using C5AR1 (CD88- cells, according to Duterte et al. Immunity 2019, Bourdely et al. Immunity 2020). We then used unsupervised clustering to identify gene clusters in the CD1c+ DCs and compared their average gene expression pattern for genes we associated with CX3CR1+ DC3s. Now we describe that cluster 1 exhibits the pattern of expression we found in peripheral blood to be associated with CX3CR1+ DC3. We hope the reviewer agrees that we have now shown that CD1c+ DCs in eye fluid appear to cluster into distinct clusters, one of which is similar to a CX3CR1+ DC3.

Reviewer #2 (Recommendations for the authors):

The authors set out to utilize advance informatics techniques to advance our understanding of a cell type previously shown to play a role in uveitis, type 2 conventional dendritic cells (previously termed mDC1). They accumulated a valuable set of samples from untreated, active disease and employed a testing and validation cohort which reproduced a core set of genes differentially expressed within the CD1c+ dendritic cells in the peripheral blood of patients with uveitis. Importantly these cells appear altered regardless of the subtype of uveitis. They then astutely question whether the transcriptional signature simply represents different proportions of cell subtypes or states and test this hypothesis using flow cytometry.

They also attempt to probe the mechanism driving this particular cell type/state by cleverly drawing on published data sets, however these hypothesis-generating experiments are not validated experimentally in a uveitis system.

They utilize in vitro analysis of similar cells isolated from human patients and show that the cells can be induced to make a specific set of cytokines which is different from a related dendritic cell type, however the lack of concordance with published activated DC2s, or the transcriptional signature in their own ex vivo activated DC2s raises more questions than it answers.

We appreciate the reviewer's constructive comments and additional feedback. The reviewer may have been confused by the description of "in vivo activated" cDC2s, whereas this should have been “inflammatory” cDC2s, which are type I IFN-Dependent murine cDC2 subsets induced by pneumonia viruses in mice. We regret this mistake. Specifically, we wanted to demonstrate that CD1c+ DC transcriptomes of patients with non-infectious uveitis do not match those of these type I IFN dependent cDC2. The signature of patients is also not induced by type I IFN stimulation. We have clarified this in more detail in the result section.

Furthermore, we show that the CX3CR1-positive CD1c+ DCs in this study secreted similar levels of IL-23 and higher levels of TNF α and IL-6 than the CD36-CX3CR1- DC3 subsets after activation. Interestingly, this observation is in agreement with Villani et al. (Science 2017; PMID: 28428369) who observed that stimulated CD1c+CD36+ DC3s (CD1c_B population) produced more TNF α than CD36-DC3s (CD1c_A). Additionally, Brown et al. (Cell 2019) found that CX3CR1+ cDC2s produce more IL-6 and TNF-α than their CX3CR1-negative counterparts. Combined, our data on ex vivo stimulated cDC2s indicate that the secretome of activated DC3s in this study is consistent with previous findings. This has been added to the Discussion section. Hopefully, the reviewer will agree that these points do answer some questions, but further research is needed to fully understand these changes and their relation to the cause of non-infectious uveitis.

The manuscript was difficult to read, largely because it is trying to accomplish too many goals, but also because the informatics techniques and external data sets were not described sufficiently for an average reader to readily evaluate the method and conclusions, and the conclusions were overstated throughout the paper. Overall, too many hypotheses were tested, alternative hypotheses were not considered/discussed. The data should probably be divided into at least 2 papers, one which explores the peripheral blood subsets more concretely, and one which attempts to elucidate a mechanisms by which low RUNX3 expression is associated with the genes expressed more highly in DC2s in the peripheral blood of uveitis patients.

We have rewritten sections to improve readability. We have provided more details on the analysis to accommodate readers without experience in computational analyses. We have moved the bulk of the enrichment analysis to the supplement and merged figures 1 and 2. In our opinion, this should have improved the readability of the manuscript, and we hope the reviewer agrees.

1. The rationale for depleting CD14 in the validation cohort is not clear and justified. CD14 expression on dendritic cells has been established in the literature (ex Duterte Immunity 2019). This reviewer wonders if a better flow of data would be to start with the current second cohort, CD14 depleted, ID the black module, realize that some genes discovered have been associated with CD14+DC3, then utilize a second cohort that includes CD14+CD1c+ cells to validate and expand the original gene set.

Based on the reviewer's recommendation, we started with the second cohort and re-ran all analyses. We have merged figure 1 and 2 and believe this improved the flow of the work. It is our hope that the reviewer will agree that this issue has now been addressed appropriately.

2. Experiments probing the transcriptional regulation of the uveitis-enriched gene set are not definitive, but hypothesis-generating, and left in this story, are distracting from the primary observation of a cell type differentially present in the blood of uveitis patients.

Based on the reviewer's recommendation, we moved non-essential transcriptional regulation parts to the supplement and removed the separate transcriptional regulation paragraph. We believe it is important to show that there is no enrichment for type I IFN signalling, but rather for signaling pathways that distinguish discrete subsets of CD1c+ DCs, which forms the basis for flow cytometry analysis. The Results section now summarizes the potential transcriptional regulation very briefly. Our hope is that the reviewer now considers this issue better taken care of within the Results section.

3. The authors claim to have found an inflammatory cell type, but the gene expression profile does not recapitulate inflammation-induced DC2 gene expression profile. Figure 3 attempts to shed mechanistic light but only opens more questions, like if viral infection in mice (figure 3A lower panel) and multiple inflammatory stimuli (figure 3C) induce a transcriptional program opposite what they are finding in these cells in the peripheral blood of uveitis patients, how are the cells identified in this paper likely relevant to eye inflammation?

CX3CR1+ DC3s produce more TNF α and IL-6, so they were regarded as “inflammatory” by us. However, we now believe that this is somewhat confusing, particularly regarding other subsets previously called "inflammatory". The manuscript has been revised to remove this. Unlike type I IFN, this subset is associated with different transcriptional programming mediated by unknown molecular cues. CX3CR1+ cDC2s display a different gene profile than inflammatory (type I IFN) cDC2s in our study, which is consistent with previous studies in mice (PMID: 34526403). We have outlined in greater detail what we believe drives these cells in non-infectious uveitis in the Discussion section.

4. Pro-inflammatory CD14+DC3 increased in blood in SLE patients, but according to authors interpretation, they find a similar cell type is decreased in peripheral blood in uveitis. The authors adhere to a possible explanation that these cells are trafficking to the eye, but by the same logic, SLE patients should have an even more significant reduction, not increase in their peripheral DC3 counts. This is not discussed.

Our view is similar to the reviewer's that this line of reasoning is paradoxical. In the Discussion section, we have substantially revised the comparison of type I IFNs in SLE and signaling in dendritic cells of patients with non-infectious uveitis. Hopefully, the reviewer finds that the revised discussion adequately addresses this issue.

5. Even more importantly, the Chen et al. papers which are cited generally by the authors, showed the opposite- that CD1c+ DCs are increased in the blood of patients with uveitis, and correlate with disease activity. No attempt is made by the authors to compare their data with this data derived from a larger number of patients, or discuss the difference. The authors have essentially ignored this key discrepancy with previously published data.

We agree with the reviewer that a comparison with Chen et al's previous work is warranted. Chen and colleagues attempted to quantify CD1c+ myeloid dendritic cells in their work, but CD11c wasn't used in their flow cytometry panel. Based on our data, this approach would include quantification of CD1c+ CD11c-negative cells as well (Author response image 2). If we include all CD1c+ populations (regardless of CD11c), we would have an overall increase in "CD1c+ DCs" (Author response image 2) – similar to the conclusion by Chen et al. Even when CD11c is not considered in cluster analysis, however, CD1c+ CD11c- cells (cluster 46) do not cluster with our identified CD1c+ DCs, which indicates they may represent a different type of cell. In the Discussion section, we addressed this issue. We hope the reviewer agrees that this discrepancy has now been adequately discussed.

Author response image 2

6. Despite changing some text around the hypothesis that this cell type is reduced in the blood because it migrates to the eye (numerically impossible as pointed out by prior reviewer #1), this notion is reiterated several other places in the paper and should be removed, and frankly, reconsidered. It is also possible that the change in peripheral blood cell type/state frequency results from cytokine-induced precursor emigration/differentiation. Perhaps there is actually a fraction of DC2-type cells that are increased in the blood in uveitis (as Chen found) but express the transcriptional program identified in the paper because they are pre-activated cells?

Based on the reviewer's suggestion in the Discussion section, we have adopted a hypothesis that cytokine-induced precursor emigration/differentiation may also contribute to the change in peripheral blood cell type/state frequency. We have removed the hypothesis that these cells migrate to the eye.

7. The analysis of ocular DC2s which concludes that the current cell type, CD14+DC3s with black module gene expression, are present in the eye is not convincing. The methods are not clearly explained, however it appears that rather than utilizing unbiased cluster analysis and performing differential gene expression analysis between the patients with non-infectious uveitis, and the control (endophthalmitis), or using another technique like GSEA which they used in previous figures to correlate the black module genes with the genes expressed by the ocular DC2s, the authors appear to have selected individual cells that expressed the black modules genes, then compared the expression of black module genes to the cells that did not express black module genes.

We agree with the reviewer that a more unbiased cluster analysis of ocular CD1c+ dendritic cells would be more appropriate. Therefore, we reanalyzed the single cell RNA seq data and conducted cluster analysis of the CD1c+ DCs. Because we aimed to demonstrate that inflamed eyes of patients contain cells that express key genes associated with CX3CR1+ DC3s, we sorted single-cell RNA seq data for CD1c+ using CLEC10A as a classical tissue-cDC2 marker (Heger et al. Front Immunol. 2018), removing monocytes using C5AR1 (CD88- cells, according to Duterte et al. Immunity 2019, Bourdely et al. Immunity 2020). We then clustered cells and compared their gene expression for genes associated with CX3CR1+ DC3. Cluster 1 exhibits the pattern of expression we found in peripheral blood to be associated with CX3CR1+ DC3. We hope the reviewer agrees that CD1c+ DCs in eye fluid appear to cluster into distinct clusters, one of which has a gene profile of key genes that is reminiscent of CX3CR1+ DC3s. Hopefully, the reviewer will agree that we have now addressed this section more effectively.

8. The authors state in their response to reviewer #1 comments "we profiled available eye fluid biopsies and paired plasma by Olink proteomics to measure immune mediators from patients and controls from this study (and several additional samples, including aqueous humor from non-inflammatory cataract controls – see revised Figure 5 panel D). This analysis shows that cytokines produced by CD36+CX3CR1+ DCs such as TNF-α and IL-6 are specifically increased in eye tissue of patients, but not in blood."-Neither this data, or discussion of it are included in the revised manuscript.

We regret this mistake in the revision. This description should not have been included in the rebuttal.

Finally, a concern is that the title is a gross overstatement of their findings:

1. They have not demonstrated that the cells in this paper induce inflammation, and especially not in the context of uveitis- only that similar cells from healthy patients produce a different set of cytokines when stimulated in vitro compared to another cell type.

2. They have not demonstrated that these very cells migrate to the eye, only that similar genes are present in a possibly similar ocular cell type in another data set.

3. They do not demonstrate type I IFN-negative autoimmunity in the eye. This was a huge stretch, presumably from prior assumptions about the mechanisms driving uveitis along with the finding that their cell type does not share a transcriptional program with murine DC2s activated in a viral infection.

We agree with the reviewer and have changed the title to: “Transcriptome network analysis implicates CX3CR1-positive type 3 dendritic cells in non-infectious uveitis”. This seems to be more appropriate for the manuscript, and we hope the reviewer agrees.

In regard to addressing the Previous Editor Concerns:

1. The informatics analysis for most of the paper is likely sufficient, however the methods are not communicated succinctly and clearly such that non-informatics experts can understand the rationale and method for each analysis. The analysis of the intraocular DCs was not clear and from the details provided, did appear appropriate.

We would like to thank the editor for his/her time. We have amended the single cell RNA sequencing analysis in response to both reviewers to accommodate unbiased clustering and reworked the figures and manuscript as needed.

2. Revision of claim to have identified a new type of cDC2- This is still not satisfactory as:

a. authors have not ruled out the possibility that the cells are monocyte-derived by transcriptional analysis, protein expression or functional analysis. The use of CD14-deplete cells to recapitulate the gene expression profile is not sufficient to determine that the cells in this paper are not monocyte-derived, as CD14-expression is demonstrated on cells confirmed by FLT3L response to be dendritic cells in Duterte et al. Immunity 2019.

b. To be defined as a new subset of the previously defined DC3 subset, one would need to exactly replicate the marker expression and then show that the new markers subset that subset further, the current manuscript may simple be focusing on different genes/proteins expressed by one or more previously described subsets.

c. As this paper is useful for describing a cell type or state that differentiates uveitis from healthy patients, these experiments do not need to be done to publish this paper, but the naming of the cell type should be tempered to simply describe the markers that were expressed and suggest how they fit into the Duterte/Villani schema of DC2/DC3 classification. In actuality, the discrimination of cDC2 from monocyte-derived DC2-like cells has proven difficult in many papers, thus the authors are advised to stay out of the mud, so-to-speak.

Our naming of the population has been changed to CX3CR1+ DC3s, which describes the CD5- CD163+ DC3s cells that are CX3CR1-positive in flow cytometry analysis. As a result, we have removed any description related to discriminating cDC3s from monocyte-derived DC3-like cells. We hope the editors agree that this has now been dealt with more clearly and without making any unsubstantiated claims.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1 (Recommendations for the authors):

In the revised paper entitled "Transcriptome network analysis implicates CX3CR1-positive type 3 dendritic cells in non-infectious uveitis," Hiddingh, Vandit, Verhagen et al. explore the gene expression of PBMCs in non-infectious uveitis patients and demonstrate a CX3CR1-positive Cd1c+ gene signature that is altered in non-infectious uveitis. They show that this population is decreased in the peripheral blood, could be regulated by notch signaling, expresses proinflammatory cytokines upon stimulation with LTA, and are present in the eye during uveitis.

The hypothesis that a cD1c+ population of DCs that could be related to uveitis in humans is intriguing and deserves further study. Their use of multiple methods to explore this population and the use of multiple cohorts are a strength of the manuscript, and it raises many intriguing questions that are potentially interesting, such as if this population is expressing inflammatory cytokines upon stimulation, why are they decreased in uveitis?

We would like to thank the reviewer for his/hers time to review our work and pointing out the strength our our revised work.

There are still areas where the manuscript is hard to follow and there are some concerns with the experimental methodology. The difficulty following the author's story is partially due to errors in cross-referencing statements in the text with their figures that support the data and understanding the experiments from the figure legends. For instance, in lines 313-315, the authors state "Furthermore, in CD1c+ DCs from healthy human donors, IFN-α did not induce downregulation of RUNX3 as observed CD1c+ DCs from non-infectious uveitis patients. Figure 2 – Figure 2 supplement 1". Data to support this statement is not found in Figure 2 or Figure 2 supplements (only healthy control data).

Our apologies for how the sentence in line 313 was formulated. We meant to say that while RUNX3 was lower in patients with uveitis in RNA-seq data, stimulation of CD1c+ DCs from healthy controls with IFN-α did result in a decrease in RUNX3 as support for that fact that IFN-α is unlikely causing the lower expression of RUNX3 as observed by RNA-seq in uveitis patients. It has been changed to “Furthermore, while RUNX3 was downregulated in RNA-seq data from CD1c+ DCs from non-infectious uveitis patients (Figure 1I), stimulation of CD1c+ DCs with from healthy human donors with IFN-α resulted in upregulation of RUNX3”. Our hope is that the reviewer will agree that this is a more appropriate formulation.

The Figure 2 Supplement 1 legend references "the notch-negative condition in d" with a d in that figure.

This was a typo and has been changed to "the notch-negative condition in b".

Methodologically, the backgating for the manual gating of CD11c/CD1c suggests that the CD36+CX3CR1+ population is really part of a larger population of CD11c+ cells, raising the question of if this population is too poorly defined in this experimental context. This concern is slightly ameliorated by the appearance of a CD36hiCX3CR1hiCD1c+ population in the unsupervised clustering.

Despite these weaknesses, there is enough strength in using multiple methods and replication with multiple patient cohorts to overcome these concerns and to utilize it as a basis to further explore the functions of this population in uveitis pathogenesis.

We are grateful for the reviewer's time and efforts, and are pleased that he/she believes our work is strong enough to publish.

Reviewer #4 (Recommendations for the authors):

The authors have responded to most of the previous reviews and have generated a more clear and cohesive manuscript.

Thanks to the reviewer for reviewing our work and providing recommendations to improve data presentation clarity.

Additional recommendations:

Figure 1

The text rationale for CD14 separation is confusing, consider omitting it.

A better methodology would have been to repeat analysis with new cohort I followed by validation using new cohort II rather than simply comparing the cohorts, but this reads more clearly and logically than the prior version and the overall conclusions seem valid.

Figure 1 Sup 1 not needed, emphasized the odd methodology sorting "cohort II" for CD14- recommend omitting this from the final version, or using instead Figure 3- Supplement 2 could be moved to the supplement for Figure 1 to explain why black module (from the CD14-sorted cohort) is stronger than then enriched modules from cohort I.

As recommended by previous reviewers, we have included Figure 1 Supplement 1 and moved the CD14 analysis to Figure 3 – Supplement 2 (instead of near Figure 1). We believe it is important to highlight this section since it emphasises that CD14+ DC3s and CX3CR1+ DC3s are not identical populations. The current structure of the manuscript conforms to previous recommendations from other reviewers. We hope the reviewer agrees.

CD14+ CD1c+CD11c+CD36+CXCR3+ DC3s seem to be a subset of CD1c+CD11c+CD36+CXCR3+ DC3s, which may be why there is a stronger gene expression signature black module from cohort II vs the blue and green modules from cohort 1.

The possibility of such a scenario is not ruled out. However, we show that CX3CR1 and CD14 expressions are moderately correlated (Pearson's correlation coefficient = 0.35, Figure 3 Supp 2). Furthermore, Figure 3G shows that only the CD14 DC3 population that expresses CX3CR1 is significantly altered in patients when looking at CD14 DC3s.

The supplemental experimental data shows that sorted DC3s from healthy peripheral blood treated with a variety of inflammatory stimuli upregulate RUNX3. One alternative explanation not discussed by the authors is that peripheral blood DC3s are in a precursor or pre-activation state.

We agree that this may be a possibility. We have added this to the Discussion section as follows “It is also possible that the change in CD1c+ DCs observed in this study results from cytokine-induced precursor emigration or differentiation or that the affected peripheral blood DC3s marked by CX3CR1 are in a precursor or pre-activation state.”

Text: in CD1c+ DCs from healthy human donors, IFN-α did not induce downregulation of RUNX3 as observed in CD1c+ DCs from non-infectious uveitis patients, however supplemental figure 2 only tests CD1c DCs from healthy patients. CD1c+ DCs from uveitis patients were never stimulated with IFN to test whether they downregulate RUNX3 after this stimuli. This textual discussion of the experimental data is misleading.

Our apologies for how the sentence in line 313 was formulated. We meant to say that while RUNX3 was lower in patients with uveitis in RNA-seq data, stimulation of CD1c+ DCs from healthy controls with IFN-α did result in a decrease in RUNX3 as support for that fact that IFN-α is unlikely causing the lower expression of RUNX3 as observed by RNA-seq in uveitis patients. It has been changed to “Furthermore, while RUNX3 was downregulated in RNA-seq data from CD1c+ DCs from non-infectious uveitis patients (Figure 1I), stimulation of CD1c+ DCs with from healthy human donors with IFN-α resulted in upregulation of RUNX3”. Our hope is that the reviewer will agree that this is a more appropriate formulation.

Sup figure 3 final panel should be G, not H.

This was a typo. We have changed the panel figures in the legend of Figure 3 – Supplement 1.

Aqueous scRNA samples are listed as obtained from Utrecht in the methods section and should cite the previous dataset.

We regret that this was not clear for ther reviewer, the method now states that “Single cell RNA-seq (scRNA-seq) data from a previous study of as reported by Kasper et al. 2021 of aqueous humor of 4 HLA-B27-positive anterior uveitis (identical to the AU group in this study) patients were obtained and downloaded via Gene Expression Omnibus (GEO) repository with the accession code GSE178833.” We hope the reviewer agrees this is now further clarfified in sufficient details.

Data used from prior sources should be more clearly detailed in legends and text. As the paper reads, it appears that the authors did the murine BMDC on the OP9 culture experiment detailed in Sup Figure 2.

Figure legends and text now include more details on the paper references from which we obtained data.

Figure 5 image is very misleading – "purify tissue CD1c+ DCs" suggests that cells were purified resulting in the displayed UMAP. CLEC10A and C5AR should both be shown and the label should not state CD1c+ if this expression was not assessed- this is misleading.

We agree with the reviewer. We now changed this back to the description of using scGate analysis to select CLECL10A+ and C5AR1- negative cells in the figure and figure legend. We hope the reviewer agrees this better reflects the analysis conducted.

https://doi.org/10.7554/eLife.74913.sa2

Article and author information

Author details

  1. Sanne Hiddingh

    1. Ophthalmo-Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Resources, Formal analysis, Validation, Investigation, Methodology, Writing – original draft, Patient inclusions and dendritic cell purifications and dendritic cell cultures
    Contributed equally with
    Aridaman Pandit and Fleurieke Verhagen
    Competing interests
    No competing interests declared
  2. Aridaman Pandit

    1. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Department of Ophthalmology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Formal analysis, Validation, Investigation, Methodology
    Contributed equally with
    Sanne Hiddingh and Fleurieke Verhagen
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2057-9737
  3. Fleurieke Verhagen

    1. Ophthalmo-Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    3. Department of Ophthalmology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Patient inclusions and dendritic cell purifications
    Contributed equally with
    Sanne Hiddingh and Aridaman Pandit
    Competing interests
    No competing interests declared
  4. Rianne Rijken

    1. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Department of Rheumatology & Clinical Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Formal analysis, Methodology
    Competing interests
    No competing interests declared
  5. Nila Hendrika Servaas

    1. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Department of Rheumatology & Clinical Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Resources, Formal analysis, Validation, Writing – review and editing, Dendritic cell cultures
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9825-7554
  6. Rina CGK Wichers

    1. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Department of Rheumatology & Clinical Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Formal analysis, Methodology, Knock-down experiments and dendritic cell cultures
    Competing interests
    No competing interests declared
  7. Ninette H ten Dam-van Loon

    Department of Ophthalmology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Data curation, Supervision, Investigation, Methodology, Project administration
    Competing interests
    No competing interests declared
  8. Saskia M Imhof

    1. Ophthalmo-Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Department of Ophthalmology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Conceptualization, Funding acquisition, Project administration
    Competing interests
    No competing interests declared
  9. Timothy RDJ Radstake

    University Medical Center Utrecht and Utrecht University, Utrecht, Netherlands
    Contribution
    Conceptualization, Resources, Data curation, Funding acquisition, Investigation, Writing – review and editing
    Competing interests
    was a principal investigator in the immune catalyst program of GlaxoSmithKline, which was an independent research program. He did not receive any financial support. Currently, TR is an employee of Abbvie where he holds stock. TR had no part in the design and interpretation of the study results after he started at Abbvie
  10. Joke H de Boer

    1. Ophthalmo-Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Department of Ophthalmology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing – original draft, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
  11. Jonas JW Kuiper

    1. Ophthalmo-Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    2. Center for Translational Immunology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    3. Department of Ophthalmology, University Medical Center Utrecht, Utrecht University, Utrecht, Netherlands
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    For correspondence
    j.j.w.kuiper@umcutrecht.nl
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5370-6395

Funding

UitZicht (#2014-4)

  • Jonas JW Kuiper

UitZicht (#2019-10)

  • Jonas JW Kuiper

UitZicht (#2021-4)

  • Jonas JW Kuiper

The funders had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.

Ethics

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 (METC protocol number #14-065/M).

Senior Editor

  1. Betty Diamond, The Feinstein Institute for Medical Research, United States

Reviewing Editor

  1. Lynn M Hassman, Washington University School of Medicine, United States

Reviewer

  1. James Walsh

Version history

  1. Received: October 21, 2021
  2. Preprint posted: November 19, 2021 (view preprint)
  3. Accepted: April 9, 2023
  4. Accepted Manuscript published: April 12, 2023 (version 1)
  5. Version of Record published: May 15, 2023 (version 2)

Copyright

© 2023, Hiddingh, Pandit, Verhagen et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 573
    Page views
  • 68
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Sanne Hiddingh
  2. Aridaman Pandit
  3. Fleurieke Verhagen
  4. Rianne Rijken
  5. Nila Hendrika Servaas
  6. Rina CGK Wichers
  7. Ninette H ten Dam-van Loon
  8. Saskia M Imhof
  9. Timothy RDJ Radstake
  10. Joke H de Boer
  11. Jonas JW Kuiper
(2023)
Transcriptome network analysis implicates CX3CR1-positive type 3 dendritic cells in non-infectious uveitis
eLife 12:e74913.
https://doi.org/10.7554/eLife.74913

Further reading

    1. Cell Biology
    2. Immunology and Inflammation
    Yijun Zhang, Tao Wu ... Li Wu
    Research Article

    Dendritic cells (DCs), the key antigen-presenting cells, are primary regulators of immune responses. Transcriptional regulation of DC development had been one of the major research interests in DC biology, however, the epigenetic regulatory mechanisms during DC development remains unclear. Here, we report that Histone deacetylase 3 (Hdac3), an important epigenetic regulator, is highly expressed in pDCs, and its deficiency profoundly impaired the development of pDCs. Significant disturbance of homeostasis of hematopoietic progenitors was also observed in HDAC3-deficient mice, manifested by altered cell numbers of these progenitors and defective differentiation potentials for pDCs. Using the in vitro Flt3L supplemented DC culture system, we further demonstrated that HDAC3 was required for the differentiation of pDCs from progenitors at all developmental stages. Mechanistically, HDAC3 deficiency resulted in enhanced expression of cDC1-associated genes, owing to markedly elevated H3K27 acetylation (H3K27ac) at these gene sites in BM pDCs. In contrast, the expression of pDC-associated genes was significantly downregulated, leading to defective pDC differentiation.

    1. Computational and Systems Biology
    2. Immunology and Inflammation
    David J Torres, Paulus Mrass ... Judy L Cannon
    Research Article Updated

    T cells are required to clear infection, and T cell motion plays a role in how quickly a T cell finds its target, from initial naive T cell activation by a dendritic cell to interaction with target cells in infected tissue. To better understand how different tissue environments affect T cell motility, we compared multiple features of T cell motion including speed, persistence, turning angle, directionality, and confinement of T cells moving in multiple murine tissues using microscopy. We quantitatively analyzed naive T cell motility within the lymph node and compared motility parameters with activated CD8 T cells moving within the villi of small intestine and lung under different activation conditions. Our motility analysis found that while the speeds and the overall displacement of T cells vary within all tissues analyzed, T cells in all tissues tended to persist at the same speed. Interestingly, we found that T cells in the lung show a marked population of T cells turning at close to 180o, while T cells in lymph nodes and villi do not exhibit this “reversing” movement. T cells in the lung also showed significantly decreased meandering ratios and increased confinement compared to T cells in lymph nodes and villi. These differences in motility patterns led to a decrease in the total volume scanned by T cells in lung compared to T cells in lymph node and villi. These results suggest that the tissue environment in which T cells move can impact the type of motility and ultimately, the efficiency of T cell search for target cells within specialized tissues such as the lung.