Inflammatory Bowel Disease in patients with Primary Sclerosing Cholangitis: a distinct form of colitis

Sample collection and selection

A total of 47 samples were collected from the colon of subjects with either PSC-IBD, UC or without IBD during a routine colonoscopy at the University Medical Center Groningen (Groningen, the Netherlands). The diagnosis IBD was based on typical endoscopic findings confirmed by histopathological evidence. The diagnosis PSC was based on cholestatic blood levels in combination with typical radiological findings, after ruling out other possible causes for bile duct sclerosis. “Healthy control” (HC) samples were collected from patients without IBD, who underwent endoscopy for various reasons, such as rectal bleeding, check-up after polyp removal or abdominal discomfort. The absence of IBD was histologically confirmed by a trained pathologist. One “HC” sample was excluded since the patient’s diagnosis was revised to IBD at clinical re-evaluation.

Additional inclusion criteria for PSC-IBD and UC participants were no history of liver transplantation, and no isolated sigmoid involvement of the disease. Consent of all participants was obtained through Gastrointestinal Expression profiles in Intestinal Disease (GE-ID) (NL58808.042.16) and/or Parelsnoer IRB (NL24572.018.08). At the time of endoscopy, three biopsies per participant were collected in RPMI with 10% FCS at 4°C, and cryopreserved into ice cold freezing medium (FCS with 10% DMSO) within 30 minutes after harvesting. At least one biopsy from the same location as the cryopreserved biopsies was evaluated by a pathologist to examine inflammatory activity. Based on the histological evaluation, biopsies were classified as actively inflamed or non-inflamed.

Sample dissociation

Three gut mucosal biopsies per sample were thawed as described before[14,15], and subsequently dissociated using the ‘one-step collagenase’ protocol[15]. In short, biopsies were washed with ice cold PBS with 10% FCS for 15 minutes. Then they were digested with collagenase IV and DNAse II for 30 minutes in RMPI with 10% FCS at 37°C, while being pipette-strained at t=10 minutes and t=30 minutes. Finally, the biopsies were shortly incubated in TrypLE, filtered over a 70 μm filter and washed using cold PBS with 1% BSA.

Cell hashing

Cells were hash tagged using Totalseq A antibody hashtag 1 to 4 (Biolegend), according to the manufacturer’s protocol. These hashtags were later used to demultiplex pooled samples.

Viability assessment and cell counting

Viability and counting before loading were assessed using trypan blue staining and counting on a hemocytometer. At viability <60%, cells were washed once more, and when visible clumps were present, cells were filtered once more over a 70 μm filter.

Library preparation

The 10X V3 chip (B) with V3 reagents were used for library preparation. Four samples were pooled per 10X lane, aiming for a total recovery of 12,000 cells per lane. Per chip, 2 channels were loaded. cDNA library prep was done according to the 10X manual CG000185 Rev D, and hashtag libraries were prepared using the manufacturer’s protocol [16]. Libraries were indexed using 10X PN-120262 Chromium i7 Multiplex Kit.

Library quality control (QC)

Libraries were checked for concentration and peak-size distribution, using an Agilent 4200 TapeStation. All cDNA libraries featured concentration > 20 ng/ul.

Sequencing and data preparation

Sequencing was performed by BGI (Hong Kong) on an MGI2000 sequencer, featuring 100bp paired-end reads (90k reads per cell). Demultiplexing was performed at BGI, and subsequently, Cellranger v3.1.0[17] was used to perform filtering, barcode counting, and alignment, using the human GRCh38-3.0.0 genome as a reference. Detection of possible doublets was primarily performed using hashtag assignment, with additional assignment using Souporcell (v2.0)[18].

Seurat[19] v2.3.0 was used for data QC and analysis. Cells were filtered for >200 genes and <60% mitochondrial reads per cell. After integration, RNA counts were normalized with log-normalization through Seurat. Next, the first 30 principal components (based on the elbow plot) were used in the RunPCA function in Seurat for Principal Component Analysis (PCA). Finally, the FindNeighbors and FindClusters functions (with default settings) were utilized to detect cell clusters.

Cell type annotation

As a first step, all high-quality cells were divided into three compartments: epithelial, stromal, and immune. Specifically, the Seurat object was used to calculate the average expression of compartment markers for each cluster, using the function AverageExpression to identify epithelial cells (EPCAM >=4), stromal cells (THY1 >= 0.3, SOX10 >=2, MADCAM1 >=1), and immune cells (CD27 >=0.3, PTPRC >=3). The immune compartment was further filtered for <20% mitochondrial reads per cell. Next, each compartment was re-clustered and annotated separately using an automated reference-mapping algorithm (azimuth.hubmapconsortium.org) using a UC full biopsy scRNAseq dataset[20] as a reference set on the integrated (SCT) assay. This method relies on a reference single-cell dataset, where cells in the query dataset are labelled based on their closest projected reference cell (cell-to-cell approach). Lastly, biologically relevant cell types were manually annotated through marker gene clustering. Specifically, dual oxidase 2-expressing (DUOX2+) enterocytes (DUOX2>1) were annotated within the epithelial compartment; and IgA (IGHA1>115), IgG (IGHG1>90), IgM (IGHM>800), and Ig-negative plasma cells were annotated within the immune cell compartment.

Differential abundance analysis

To perform differential cell abundance analysis, the Bayesian Multinomial Logistic-Normal Linear Regression model Pibble for compositional data (R package Fido v1.0.4[21]) was used. Because the Pibble model uses Bayesian inference for parameter estimation, statistical significance is expressed as the credible interval; for example, a cell type can be deemed significantly differentially abundant if 95% of its credible interval does not contain zero. Furthermore, as the Pibble model estimates the correlation structure between all included cell types (after adjusting for covariates if any), model inference is only done once.

Differential gene expression and pathways

Differentially expressed genes were defined as genes expressed in at least 10% of either of the two comparison groups and with a minimum of 0.25-fold change in gene expression. For differential gene expression Seurat (v4.3.0.1) with the built-in ‘MAST’ method (v1.26.0) was used. All p-values were Bonferroni corrected for the number of tested genes. Pathways were then generated using the R package enrichR (v3.2) with the GO biological process 2018 database using DE genes as input[22].

A list of significantly upregulated genes (marker genes) for each cell type was generated using the above-described method, comparing the respective cell to all other cells within the cell type’s compartment (e.g. DUOX2+ enterocytes were compared to all other epithelial cells). With the list of marker genes as input, a list of marker pathways was then generated for “DUOX2+ enterocytes” and “enterocytes” accordingly.

Differentially expression of oncogenes was assessed using the above-described method, using pancancer and colorectal cancer onco(suppression)genes as input[23].

Cell-cell interaction analysis

Cell-cell interaction analysis was performed for Inflammatory Monocytes and DUOX2 Enterocytes in relation to other cell types with the R package CellChat (v.1.4.0), using a public build-in database on ligand-receptor interactions[24].

Heritability analysis

Heritability enrichment analysis was performed using the CELL-type Expression-specific integration for Complex Traits (CELLECT) v1.3.0 toolkit[25], which is designed to identify likely etiologic cell types underlying complex traits. This toolkit applies stratified LD score regression (S-LDSC)[22] to determine cell types with genetic associations significantly enriched for the disease or trait of interest. Standard settings were applied in the implementation of CELLECT, including the exclusion of genetically complex regions like the HLA locus prior to analysis.

For heritability enrichment analysis for PSC specifically expressed genes, genome-wide association study (GWAS) summary statistics of PSC from a comprehensive PSC GWAS[26]. was used as the genetic summary statistics of interest, and CELL-type EXpression-specificity (CELLEX) analysis[25] was performed to generate cell type specific gene scores.

For heritability enrichment analysis for differentially expressed genes between PSC-I and PSC-NI, we performed MAST v1.14.0, a two-part, generalized linear model with a logistic regression component for the discrete process (i.e., a gene is expressed or not) and linear regression component for the continuous process (i.e., the expression level)[27] to generate differential gene scores. Inflammation within PSC, sex, age, mitochondrial percentage, and the number of genes detected per cell[20,27] were included as fixed effect variables and individual as a random effect to control for pseudoreplication bias[28]. The Benjamini-Hochberg correction was used to control for multiple testing across all cell types. GWAS summary statistics for PSC from the previously mentioned publication[26] was used as the genetic summary statistics of interest.

For both types of heritability enrichment analysis (PSC specifically expressed genes and differentially expressed genes between PSC-I and PSC-NI), Benjamini-Hochberg correction was applied to control for the total number of tests across all cell types.

Immunohistochemistry

To confirm the existence of DUOX2+ enterocytes and the expression of HLA-DR at the protein level, we performed DUOX2 and HLA-DR immunohistochemical staining of inflamed and non-inflamed colonic mucosa. Sequential 4µm slides from resected colonic paraffin-embedded tissue were deparaffinized, followed by antigen retrieval with a Citrate buffer pH6 and blocking of endogenous peroxidase with 0.3% hydrogen peroxidase (Sigma-Aldrich). Next, the slides were incubated with PBS/ 1% BSA (Sigma-Aldrich) for 30 min, followed by incubation with a primary antibody – DUOX2 (1:100 dilution, cat. PA5-97677, Thermo Fisher) or HLA-DR (1:1000 dilution, mouse monoclonal DA6.147, Abcam) for 1 hour in the dark. Appropriate secondary and tertiary horseradish-peroxidase labelled antibodies were used for both stainings and diluted with a 1:50 ratio in staining solution (PBS with 1% BSA and 1% human serum). To complete the staining, a solution of DAB (Sigma-Aldrich) and 0.03% peroxidase was used, followed by a counterstaining of haematoxylin and mounting with Eukitt (Sigma-Aldrich). Slides were digitally scanned using a Nanozoomer (Hamamatsu). Images were taken at 5x and 40x magnification using NDP view 2 software. A pathologist specialized in gastrointestinal disease was consulted to assess the staining.

Primary results after QC

In this study we applied scRNA-seq to 47 gut biopsy samples from 24 PSC-IBD (8 inflamed (PSC-IBD-I), 16 non-inflamed samples (PSC-IBD-NI)), 18 UC (8 inflamed (UC-I), and 10 non-inflamed samples (UC-NI)), and 5 healthy control (HC) samples. The donor and treatment characteristics of these groups were comparable, except for a higher percentage of biopsies from male patients (p=0.04) and patients using of ursodeoxycholic acid in the PSC-IBD group (p=0.001) (in line with epidemiology / standard care of disease). Furthermore, UC biopsies were significantly more often taken from patients exposed to biologicals (0.0074).

After quality control and filtering, we identified 71,798 cells from 47 samples. These cells represented 54 cell types: 25 immune, 16 epithelial, 12 stromal and 1 glia cell type. Marker genes for each cell type within its compartment are available in Supplementary results 1. Figure 1A shows the non-Euclidian distribution of the cells per compartment based on their transcriptional profile in a two-dimensional visualization of multiple principal components (Uniform Manifold Approximation and Projection - UMAP).

• Download figure
• Open in new tab

Figure 1

A: UMAP view of 71,798 cells divided into the three compartments (epithelial, immune and stromal) single-cell transcriptomics coloured by annotation cell type.

B: UMAP view of the total cells divided into the three compartments (epithelial, immune, and stromal) and coloured by inflammation status. Non-inflamed state is composed of non-inflamed samples of HC, UC and PSC. Inflamed state is composed of inflamed samples of UC and PSC.

C: Cell proportion of a combination of cell types as a fraction of the total, here represented within each compartment (epithelial, stromal, and immune). Specification of cell types included in each cell group is available in supplemental results 1. The total of cells is considered within each condition (HC, UC-NI, UCI, PSC-NI, PSC-I).

D: Cell proportions of stem cells, CD4+ PD1+ T cells and inflammatory fibroblasts in different states within the entire set (epithelial, immune, and stromal compartment combined).

DUOX2+ enterocytes, a distinct epithelial cell type involved in antigen presentation in inflammation

In addition to previously reported cell types, we describe dual oxidase 2 (DUOX2)-expressing enterocytes as a new subtype of enterocyte. This enterocyte subtype is marked by the expression of DUOX2, its accessory protein dual oxidase maturation factor 2 (DUOXA2), and lipocalin-2 (LCN2). Staining of paraffin embedded inflamed colonic tissue shows DUOX2⁺ enterocytes present in patches in inflamed tissue, clearly distinguishable from goblet cells that show membrane staining of DUOX2 (Figure 2D), whereas DUOX2⁺ enterocytes show intracellular staining. This cell subtype is distinct from other epithelial cells based on both protein and gene expression. Interestingly, DUOX2⁺ enterocytes are more abundant in inflamed tissue (Figure 2A and Supplementary Results 2).

• Download figure
• Open in new tab

Figure 2:

A: Cell proportion of DUOX2+ enterocytes in different states within the entire set (epithelial, immune and stromal compartment combined). Significant differences are denoted with an asterisk.

B Alluvial plot representing top 10 most significant marker (GO) pathways of DUOX2+ enterocytes which are not marker pathways for enterocytes and marker genes of these pathways. The colours represent the odds ratio of the marker pathway.

C: Expression level of HLA-DRA and HLA-DRB1 in DUOX2+ enterocytes and enterocytes. Marker genes for the given cell types are denoted with an asterisk.

D and E: DUOX2 and HLA-DR protein stainings in UC-I colon tissue.

To define how DUOX2⁺ enterocytes differ from other enterocytes, we analyzed pathway enrichment in DUOX2 vs mature (non-DUOX2) enterocytes. Of the 158 pathways enriched in DUOX2+ enterocytes, 99 were not observed in mature (non-DUOX2) enterocytes (Supplemental results 2). The top three pathways specific to DUOX2⁺ enterocytes are those involved in cellular responses to cytokine stimuli: “cellular response to cytokine stimulus” (adjusted p-value=8.81e-06), “regulation of cytokine-mediated signalling pathway” (adjusted p-value=4.47e-04), and “negative regulation of cytokine secretion” (adjusted p-value=1.10e-03), as shown in Figure 2B. Interestingly, among DUOX2+ marker genes we find the MHC-II encoding genes HLA-DRA (adjusted p-value=2.67e-143), HLA-DRB1 (adjusted p-value=1.66e-77), HLA-DQB1 (adjusted p-value=4.69e-37), and HLA-DPA1 (adjusted p-value=7.75e-31) (Figure 2C). Moreover, protein staining shows that DUOX2⁺ enterocytes express HLA-DR on a protein level, supporting the hypothesis that these cells can present antigens to CD4⁺ T cells (Figure 2E). Notably, the HLA-DR protein distribution in DUOX2⁺ cells is specific to inflamed colon tissue and staining is restricted to subepithelial macrophage lamina propria cell infiltrate (Supplementary Figure 1F).

DUOX2⁺ enterocytes seem to behave differently upon inflammation in PSC-IBD compared to UC, as shown by differential gene expression (Supplemental results 4-7). The most significantly upregulated gene in PSC-IBD-I vs PSC-IBD-NI (not upregulated in UC-I), is HLA-DRB1 (adjusted P-value=8.71e-23), which has previously been linked to PSC-IBD pathogenesis[12]. DUOX2⁺ enterocytes in PSC-IBD-I vs PSC-IBD-NI upregulate interferon-stimulated and inducible genes (resulting in “type I interferon signaling pathway” expression) (adjusted P-value=2.67e-03) and also MHC-II encoding genes (“antigen processing and presentation of exogenous peptide antigen via MHC class II”) (adjusted P-value= 2.24e-03); whereas DUOX2⁺ enterocytes in UC-I upregulate mucin-genes involved in “O-glycan processing” (adjusted P-value= 6.47e-04). The three most significantly enriched pathways in both PSC and UC inflammation in DUOX2⁺ enterocytes involve neutrophil immunity (adjusted P-value=1.95e-11 for all). In conclusion, DUOX2⁺ enterocytes have an active role in inflammation and are marked by the PSC risk gene HLA-DRB1 in PSC-IBD-I, suggesting that a part of its role in PSC-IBD pathogenesis is through antigen presentation (Supplemental results 2).

Similar cell abundances between PSC-IBD and UC

Inflammation status predominately determines cellular composition of the biopsy. As depicted in Figure 1B and Figure 1C, the changes in cell population abundances were more pronounced in response to inflammatory conditions than variations between PSC-IBD and UC. In response to inflammation, the overall share of the epithelial compartment decreased, making room for an augmented immune compartment, while the stromal compartment remained relatively stable (Figure 1C). In general trends, both PSC-IBD and UC exhibited a decline in enterocyte subtypes during inflammation. Notably, DUOX2⁺ enterocytes showed a consistent increase in both diseases, reaching statistical significance in PSC-IBD ((Log-Ratio Value) LRV 0.8 at Credible Interval (CI) 0.95) and borderline significance in UC (LRV 0.9 at CI 0.90). Regarding immune compartment changes, both UC and PSC-IBD exhibited increases in CD8⁺IL17⁺ T cells during inflammation. B cells, Treg cells, and IgG plasma cells were significantly increased upon inflammation in both diseases. In PSC-IBD specifically, a rise in CD4⁺PD1⁺ T cells was observed (Supplemental results 3). Within the stromal compartment, inflammatory fibroblasts expanded during inflammation in both PSC-IBD (LRV 1.8 at CI 0.95) and UC (LRV 2.0 at CI 0.95). Conversely, WNT2B⁺ Fos-lo fibroblasts and microvascular cells exhibited decreases in both diseases (Supplemental results 3).

Interestingly, a significantly increased abundance of stem cells was noted in PSC-IBD compared to Healthy Controls (HC) (LRV 0.8 at CI 0.95) (Figure 1D and Supplemental Results 3).

Differential expression analysis reveals differences between PSC-IBD and UC inflammation

Next, we studied the distinctive expression patterns that delineate inflammation in PSC-IBD and UC. Our approach involved conducting differential expression (DE) analyses for PSC-IBD-I versus PSC-IBD-NI and UC-I versus UC-NI across various cell types (Supplemental Results 4 and 5).

We found substantial differences in the number of DE genes per cell type between PSC-IBD and UC, as depicted in Figure 3B. Notably, the largest number of DE genes between in PSC-IBD-I and in PSC-IBD-NI was identified in inflammatory monocytes (n=407). In contrast, a distinct profile in inflammatory fibroblasts with a large number of DE genes was observed between UC-I and UC-NI (n=1058). Furthermore, two other types of fibroblasts (WNT2B⁺ FOS-lo 1 and WNT2B⁺ FOS-hi fibroblasts) showed large differential gene expression in UC-I vs UC-NI. In both PSC-IBD and UC, the DUOX2+ enterocyte and secretory TA cells emerged as consistently prominent cells with many DE genes between inflamed and non-inflamed. These results underscore the unique cellular signatures characterizing inflammation in PSC-IBD and UC, shedding light on distinct molecular mechanisms underlying these two inflammatory conditions.

• Download figure
• Open in new tab

Figure 3:

Follicular B and CD4+ Activated Fos-hi are enriched for heritable risk of PSC.

Heatmap of partitioning heritability from GWAS of PSC into functional categories based on (A) PSC specifically expressed genes and (B) differentially expressed genes between PSC-I and PSC-NI states in each cell type. Categories reaching significance after Benjamini-Hochberg correction are denoted with an asterisk. (C) mirror barplot representing the number of differentially expressed genes on a log-scale per cell type for PSC-I compared to PSC-NI (orange) and UC-I compared to UC-NI (blue). (D) Differentially upregulated GO pathways in inflammatory monocytes in PSC-I versus PSC-NI. Pathways were manually divided into homeostasis and immune pathways. Color represents the log-normalized p-value of the pathway. Size represents the proportion of genes marking the pathways differentially upregulated in inflammatory monocytes in PSC-I as compared to PSC-NI.

PSC, Primary Sclerosing Cholangitis; GWAS, genome-wide association study; PSC-I, PSC-inflammation; PSC-NI, PSC-non inflammation. GO, Gene Ontology

Inflammatory monocytes are activated upon PSC-IBD, but not UC, inflammation

Intrigued by the vast transcriptomic changes displayed by inflammatory monocytes in PCS-IBD (as reflected by 407 DE genes), we further investigated their predicted functions within the context of PSC-IBD and UC inflammation (Supplemental results 6 and 7). Our dataset, comprising 137 PSC-IBD (69 inflamed and 68 non-inflamed) and 303 UC (257 inflamed and 46 non-inflamed) inflammatory monocytes, showed an increased abundance of this cell type upon inflammation in UC (LRV 1.23 CI 0.95), contrasting with the absence of a similar trend in PSC-IBD. Even while the absolute number of inflammatory monocytes is higher in UC samples, we find a marked difference in the number of DE genes between PSC-IBD-I versus PSC-IBD-NI (n=407) and UC-I versus UC-NI (n=3), which is unlikely to be attributable to a difference in statistical power.

The top pathways for PSC-IBD-I are housekeeping genes, however, ranking high also are many immune pathways with significant enrichment, including those associated with neutrophil immunity (adjusted p-value ∈ [5.09e-16, 7.81e-16]), proteolysis involved in cellular protein catabolic process (adjusted p-value = 1.94e-05), and antigen processing and presentation (adjusted p-value ∈ [2.65e-05, 2.69e-05]) (Figure 3C).

These findings illuminate the distinct molecular pathways associated with inflammatory monocytes in the context of PSC-IBD and UC inflammation, providing valuable insights into the nuanced immune responses characterizing these two conditions.

No differential expression oncogenes and tumour suppressor genes in PSC-IBD-I compared to UC-I

We found an increased abundance of stem cells in PSC-IBD-NI, and not UC-NI, compared to HC, which may be linked to increased CRC risk in PSC-IBD. Therefore, we compared the expression of known cancer related genes (i.e. tumour suppressor and oncogenes) in stem cells between UC and PSC-IBD, in both inflamed and non-inflamed states[29]. For both the non-inflamed state and the inflamed state, we found that similar ratios of tumour suppressor to oncogenes were upregulated in stem cells in PSC-IBD and UC. Total numbers of stem cells in both PSC-IBD and UC were relatively low, which probably rendered the current analysis underpowered (Supplemental results 9).

Cell-cell interactions show antigen presentation through inflammatory monocytes in PSC-IBD

To unravel the intricate landscape of cell-cell interactions in the context of PSC-IBD inflammation, we focused on the cell types that showed most differentially expressed genes upon inflammation —namely, DUOX2+ enterocytes and inflammatory monocytes (Figure 3C). These cell types were deemed central to the inflammatory process. Furthermore, since two previously published studies[30,31] and our data suggest a central role for antigen presentation in the pathogenesis of PSC, we focused the analysis on interactions through MHC-II encoding HLA 2 genes (Supplemental results 10).

Our analysis extended to exploring monocyte interactions with the rest of the immune cells. Remarkably, the majority of cell-cell interactions during inflammation were shared between PSC-IBD and UC. In both conditions, inflammatory monocytes and DUOX2⁺ enterocytes were predicted to communicate with innate (macrophages and dendritic cells), as well as adaptive immune cells (Tregs, cycling T cells, and CD4⁺PD1⁺ T cells) (Figure 4B).

• Download figure
• Open in new tab

Figure 4:

A: circos plot representing the unique MHC-II cell-cell interactions for inflammatory monocytes and DUOX2+ enterocytes as senders observed in PSC-I samples.

B: circos plot representing the shared MHC-II cell-cell interactions for inflammatory monocytes and DUOX2+ enterocytes as senders observed in PSC-I and UC-I samples.

C: circos plot representing the unique MHC-II cell-cell interactions for inflammatory monocytes and DUOX2+ enterocytes as senders observed in UC-I samples.

D: Expression level of CD4 in different T cell subtypes in PSC-I samples

E: Expression level of CD4 in different T cell subtypes in UC-I samples

However, a distinctive feature in PSC-IBD was the communication between both inflammatory monocytes and DUOX2⁺ enterocytes with CD8⁺IL17⁺ T cells through the CD4 coreceptor (Figure 4A). This finding is particularly noteworthy since CD8⁺IL17⁺ T cells are conventionally recognized as the central inflammatory cells in UC, with this cell-cell interaction being absent in our UC analysis (Figure 4C). Furthermore, in PSC-IBD, the communication of these cells through CD4 suggests that CD8⁺IL17⁺ T cells are CD4⁺CD8⁺ double positive T cells. Indeed, CD8⁺IL17⁺ T cells in PSC-IBD appear to express more CD4 compared to CD8⁺IL17⁺ T cells in UC (Figure 4D and Figure 4E), aligning with the findings of Shaw et al, who identified CD4+IL17+ T cells as the central inflammatory cell in PSC-IBD[31]. This observation provides a nuanced understanding of the immune cell interactions specific to PSC-IBD inflammation.

Heritability analysis and PSC risk genes exhibit distinct expression patterns in PSC-IBD inflammation

To explore the impact of established PSC risk genes on PSC-IBD, we leveraged data from the latest PSC genome-wide association study and its meta-analysis[26,32]. Our objective was to identify the cell types in PSC-IBD that carry the highest burden of known PSC risk genes and are therefore potentially crucial in PSC-IBD pathogenesis. Using a heritability analysis, correcting for the number of genes expressed by each cell and number of cells per cell types, we found that follicular B cells and CD4+PD1+ T cells express the highest number of PSC risk genes, as depicted in Figure 3A. When trying to identify which cell type carries the highest known PSC risk gene burden in PSC-IBD I vs NI, we found no significant results, potentially due to a lack of power (Figure 3A and Supplemental results 8).

When looking at differential gene expression of the known PSC risk genes in PSC-IBD-I vs PSC-IBD-NI, we found that the known PSC risk genes might play a role in cells other than adaptive immune cells during PSC-IBD inflammation (Supplementary Figure 2). In the epithelial compartment, particularly in tuft cells and enteroendocrine cells, expression of PSC risk genes was observed in PSC-IBD during inflammation. Notably, PSC risk genes such as FUT2, GPR35, PRDX5, and PSMG1 exhibited expression across multiple cell types within the epithelial compartment during inflammation. In the mesenchymal compartment, post-capillary venules, endothelial cells, and microvascular cells demonstrated expression of PSC risk genes in PSC-IBD during inflammation. Specifically, PSC risk genes including ATXN2, HDAC7, NKX2-3, RIC8B, STRN4, and TCF4 were found to be expressed across various cell types within the mesenchymal compartment upon inflammation (Supplementary Figure 2).

Intriguingly, when assessing the expression of known PSC risk genes, we observed that several of these genes—namely ATXN2, FUT2, GPR35, NKX2-3, and TCF4—exhibited minimal differential expression in the immune compartment during inflammation. This finding suggests a distinctive pattern in the regulation of these genes in immune cells compared to their more pronounced expression in epithelial and stromal cells during PSC-IBD inflammation (Supplementary Figure 2).