An Epigenetic Basis for Sustained Inflammatory Epithelial Progenitor Cell States in Crohn’s Disease

Abstract

Background & Aims

Defining consequential differences in intestinal epithelial stem cells in healthy humans vs those with inflammatory bowel disease (Crohn’s disease and ulcerative colitis) is essential for the development of much needed therapies to restore the epithelial barrier and maintain its fidelity.

Methods

We used single-cell transcriptomic and epigenomic approaches in matched patient tissues and organoids to investigate epithelial gene expression and function in children with no pathological diagnosis in the lower gastrointestinal tract and healthy adults compared with those with Crohn’s disease.

Results

We identify an inflammatory secretory progenitor (ISP) cell state present almost exclusively in patients with Crohn’s disease compared with healthy subjects. ISPs exhibit gene expression profiles consistent with normal secretory progenitor cells but concomitantly express a suite of distinguishing pro-inflammatory genes. Mechanistically, ISPs exhibit open chromatin at ISP gene loci. Although ISP-specific genes are not expressed in intestinal stem cells, their chromatin is accessible in Crohn’s disease stem cells, suggesting that ISP genes are epigenetically poised in stem cells and subsequently transcriptionally activated in ISPs in the presence of inflammatory stimuli. Consistently, Crohn’s disease colonoids exhibit sustained ISP gene expression that can be elicited further with pro-inflammatory cytokines or via co-culture with pro-inflammatory macrophages.

Conclusions

We have defined differences in the epithelial stem and progenitor compartment of patients with Crohn’s disease that suggest aberrant stem cell differentiation and inflammatory gene expression arise and persist during disease.

Graphical abstract

Summary.

This study identifies inflammatory secretory progenitor cells in the intestinal epithelium of patients with Crohn’s disease that are largely absent in healthy individuals. Using single-cell analyses and functional studies, we found that inflammatory secretory progenitor cells express both secretory progenitor and pro-inflammatory genes, with chromatin primed for activation in Crohn’s disease stem cells. These findings suggest that aberrant stem cell differentiation and inflammatory gene expression arise and persist during disease.

The gastrointestinal tract comprises a diverse population of cells including epithelial, immune, and stromal cells, which together maintain intestinal barrier integrity. Epithelial cells are arbiters between gut luminal contents and the underlying immune cells, promoting a dynamic barrier for organismal health. Failure of barrier restoration post-damage is a hallmark of inflammatory bowel disease (IBD) (eg. Crohn’s disease [CD], ulcerative colitis [UC], and IBD undefined),1, 2, 3 the incidence of which is increasing in adults, but even more so in children.^4,5 Children with IBD often experience more severe disease compared with adults, including an increased risk for surgery in pediatric patients.⁶ Recent studies in patient tissue have advanced our understanding of cell type heterogeneity in patients with IBD7, 8, 9, 10, 11, 12; however, epithelial stem cell function is not well-characterized in the same patients. Prior studies provide evidence for both homeostatic and regenerative (ie, facultative) epithelial stem cells in human tissues, but it remains unknown how chronic inflammatory environments functionally alter these stem cells.

Epithelial stem cells driven by canonical WNT pathway activity contribute to homeostatic maintenance of the epithelial barrier. In the context of tissue injury when WNT-driven, proliferative stem cells are damaged or killed, a subset of other epithelial cells exhibit plasticity that enables a pro-regenerative state to restore the epithelial barrier.^11,13, 14, 15, 16, 17, 18 Reactivation of fetal intestinal gene expression programs is associated with regeneration from facultative stem cell populations; however, studies demonstrating these paradigms were performed in mouse models. Little is known about how epithelial stem cell populations are affected by chronic inflammation, such as in human IBD. Furthermore, mechanisms underlying changes in stem cells in human disease are not well-defined.

Epigenetic modifications contribute to stem cell fate determination. For example, prior studies in mice show increased DNA methylation of stem cell genes and decreased methylation of mature lineage genes during the transition from stem to differentiated cells.^19,20 However, recent evidence also highlights disease-driven alterations to the epigenome that are associated with chronic inflammation. Some epigenetic changes are thought to be a mechanism for inflammatory memory—an adaptive response tissues undergo after injury or inflammation. Although inflammatory memory has been typically described in immune cells, studies in mice demonstrate that mammalian epithelial cells, particularly from progenitor compartments, can acquire enhanced transcriptional responses after an initial stimulus due to persistent open chromatin in tissue-specific memory domains.21, 22, 23, 24, 25, 26, 27 Although such inflammatory memory may be beneficial for enabling more rapid and robust immune responses to repeated pathogen exposure, the same phenomenon may contribute to the pathogenesis of chronic inflammatory disorders. Patient-derived organoid systems offer a unique opportunity to study inflammatory memory because they disentangle epithelial-intrinsic phenomena from those elicited by microenvironmental cues in the inflamed in vivo environment.

Work in human organoids suggests persistent epigenetic and functional differences in the IBD epithelium. For example, aberrant transcriptional signatures have been reported in tissue from patients with IBD in remission,^28,29 and differential DNA methylation has been reported in organoids in patients with UC and CD.³⁰ Studies by our own group and others demonstrate a reduction in growth efficiency in colonoid cultures from patients with IBD that persists over numerous passages.^31,32 Finally, recent studies in mice with intestinal damage associated with graft-vs-host disease demonstrate that intestinal stem cell growth defects persist in culture, further supporting the notion that intestinal stem cells can retain memory of past damage with negative consequences.³³ Whether common epigenetic mechanisms contribute to aberrant epithelial cell behavior in contexts of chronic inflammatory diseases including CD remains unknown.

One of the most significant questions in the intestinal stem cell field is whether chronic inflammatory disease alters epithelial stem cell function to promote pathogenesis. Recent studies by Oliver et al³⁴ and Li et al³⁵ begin to address this question, identifying stem and progenitor cells as origins of metaplasia in IBD and other chronic gastrointestinal diseases. Although compelling, both studies infer but do not evaluate the function of disease-associated cell types they describe. In this study, we identify in CD a disease-specific cell state, termed inflammatory secretory progenitor (ISP), possessing chromatin and transcriptional features consistent with inflammatory memory. Epigenetic changes are evident in stem and progenitor cells and are coupled to enhanced responsiveness to myeloid-secreted cytokines and impaired stem cell fitness. Taken together, our data support the notion that CD epithelial stem cells are epigenetically rewired to permit emergence of ISPs linked to features of disease pathology.

Results

Emergence of ISP Cell States in CD Epithelium

We conducted single-cell RNA sequencing (scRNA-seq) on biopsies from endoscopically unaffected areas of ascending colon with correlating biopsies obtained within 1 to 2 cm for histologic assessment from 19 patients with CD (10 pediatric, 9 adult) and 23 control subjects (12 pediatric, 11 adult) (Figure 1A; Table 1; Supplementary Table 1). Using our optimized protocol to preserve live epithelial cells,³⁶ we acquired 106,583 unique transcriptomes (63,784 pediatric, 42,799 adult) separated by unsupervised clustering analysis into 12 clusters: epithelium (38,045), CD4⁺ T cell (7,770), CD8⁺ T cell (12,644), B cell (18,248), plasma cell (15,135), macrophage (1181), mast cell (982), endothelium (3597), fibroblast (6179), myofibroblast (745), pericyte (832), and glia (1225). The epithelium was further separated into: stem cells (3530), transit amplifying (TA) cells (2170), early progenitors (9004), OLFM4⁺REG1A⁺ secretory progenitors (2115), OLFM4^-REG1A⁺ secretory progenitors (947), OLFM4^-REG1A⁺ inflammatory secretory progenitors (905), absorptive colonocyte progenitors (5701), FABP1⁺ absorptive colonocytes (2500), AQP8⁺ absorptive colonocytes (3079), BEST4⁺ colonocytes (830), goblet cells (1010), tuft cells (286), and M cells (49) (Figure 1B; Figure 2A–D; and Supplementary Table 2).

Figure 1.

ISPs constitute a disease-specific cell state. (A) Study schematic: 3 endoscopic biopsies were collected from the same region (1–2 cm) of ascending colon. One fragment was sent for histopathology evaluation, and 2 were cryopreserved upon collection for sequencing at a later point. For generation of scRNA-seq and snATAC-seq libraries, epithelium-enriched (via crypt scraping) and lamina propria fractions were processed separately, and a part of epithelial prep was used to establish colonoids. Created in BioRender. Karakasheva, T. (2025) https://BioRender.com/sqnudqd. (B) UMAP visualization of epithelial transcriptomes (N = 12 pediatric and 11 adult controls, 10 pediatric and 9 adult CD; 32,126 transcriptomes). (C) UMAP projections of epithelial transcriptomes from control subjects or patients with CD, with ISP cluster highlighted in color (N = 18,916 and 12,233 transcriptomes for control and CD subsets, respectively). (D) Heatmap summarizing relative expression of stem cell, TA, or ISP cluster marker genes (expression levels plotted averaged per cluster). (E–F) ISP gene signature enrichment and expression of individual ISP marker genes plotted on the epithelial UMAP. In all panels, N = 19 CD (10 pediatric, 9 adult) and 23 control subjects (12 pediatric, 11 adult).

Table 1.

Patient Demographics

	CD (pediatric)	Control (pediatric)	CD (adult)	Control (adult)
Patients, n	16	15	11	11
Male, n (%)	9 (56.2)	8 (53.3)	3 (27.2)	2 (18.2)
Age at Dx, years	10.3 (±3.7)	NA	33.1 (±17.7)	NA
Age at sample collection, years	10.3 (±3.7)	12.8 (±4.0)	46.3 (±17.9)	41.8 (±15.3)
Treatment at time of sample collection
None	16 (100)	NA	8 (72.73)	NA
Immunomodulator	0	NA	3 (27.27)	NA
Physician Global Assessment at colonoscopy
Quiescent/remission	6 (37.5)	NA	8 (72.7)	NA
Mild	5 (31.25)	NA	1 (9.1)	NA
Moderate	5 (31.25)	NA	2 (18.2)	NA
Severe	0	NA	0	NA
SES-CD score (right colon)
Quiescent/remission	9 (56.25)	NA	9 (81.82)	NA
Mild	4 (25)	NA	2 (18.18)	NA
Moderate	3 (18.75)	NA	0	NA
Severe	0	NA	0	NA
sCDAI score
Remission	NA	NA	10 (90.9)	NA
Mild	NA	NA	0	NA
Moderate	NA	NA	1 (9.1)	NA
Severe	NA	NA	0	NA
PCDAI score
Quiescent	7 (43.75)	NA	NA	NA
Mild	7 (43.75)	NA	NA	NA
Moderate	2 (12.5)	NA	NA	NA
Severe	0	NA	NA	NA

NOTE. Data are presented as number (%) or mean (± standard deviation).

CD, Crohn’s disease; Dx, diagnosis; NA, not applicable; PCDAI, Pediatric Crohn’s Disease Activity Index; sCDAI, short Crohn’s Disease Activity Index; SES-CD, Simple Endoscopic Score-Crohn’s Disease.

Figure 2.

Single-cell composition of colonic biopsies from control subjects and patients with CD. (A) Expression of key marker genes for the epithelial (EPCAM), immune (PTPRC), fibroblast (PDGFRA), endothelial (PECAM1) lineages plotted on the UMAP. (B) Differentially expressed genes defining the cell clusters. The size of the dot represents the fraction of cells within cluster expressing the gene, and color represents expression level. (C) UMAP visualization of all transcriptomes in whole biopsy. (D) Differentially expressed genes defining the epithelial cell subclusters. The size of the dot represents the fraction of cells within cluster expressing the gene, and color represents expression level. (E) Epithelial subcluster abundance, per individual subject (P < .1 shown over respective comparisons). (F) UMAP projections of epithelial transcriptomes, with ISP cluster highlighted in color.

One cell cluster, which we designated ISPs, was predominantly present in CD and scarce in controls (Figure 1C, Figure 2E–F). ISPs lack markers of stem or TA cells and express genes consistent with OLFM4^-REG1A⁺ secretory progenitors from healthy tissue yet express an inflammation-related signature. The top 15 ISP cluster markers are genes encoding antimicrobial peptides, antigen presentation machinery, and enzymes involved in mucosal immunity (Figure 1D–F). Thus the disease-associated ISP cell state is characterized by enrichment for multiple genes previously linked to IBD but in a distinct population of progenitor cells.^29,37, 38, 39, 40, 41, 42, 43, 44, 45 We did not detect typical marker genes for Paneth or deep crypt secretory cells.^46,47

ISP Cell State Is Associated With Disease Activity

We observed significantly more ISPs in patients with clinically active CD (Pediatric Crohn’s Disease Activity Index [PCDAI] >10 for pediatric^48,49 or short Crohn’s Disease Activity Index (sCDAI) >150 for adults) compared with control subjects (Figure 3A–B). In addition to our cohorts, we observed ISPs in publicly available datasets using reference-based integration, including pediatric and adult CD^11,35 (Figure 3C–E), as well as UC⁸ (Figure 3F). When comparing samples from inflammation-adjacent regions with samples from actively inflamed regions in these published datasets, ISPs were more abundant in inflamed tissue¹² (Figure 3B–F). To validate the presence of ISPs in tissue, we stained ascending colon biopsies from the same patients as were sequenced (collected within 1–2 cm of the biopsies used for sequencing) for hallmark ISP proteins LCN2, CD74, and HLA-DR. No co-staining of epithelium with all 3 markers was detected in controls, whereas in CD, we detected epithelial coexpression of all markers (Figures 3A, 4A–C). Abundance of HLA-DR⁺LCN2⁺CD74⁺ cells in tissue sections correlated with abundance of ISP cells identified in the same individuals via scRNA-seq (Figure 4D).

Figure 3.

ISP state is a characteristic of active disease. (A) Schematic: experimental approach and data presentation. (B) Relative abundance of the ISP cluster in total epithelium for individual study subjects (N = 12 pediatric controls, 11 adult controls, 10 pediatric CD, 9 pediatric CD). (C) UMAP projections on a published dataset (DUOS-000145, DUOS-000146; N = 16 adult controls, 22 adult CD; 44,278 and 53,067 transcriptomes, respectively) and quantification of relative ISP abundance. (D) Relative abundance of the ISP cluster and UMAP projections on a published dataset (E-MTAB-8901; N = 11 pediatric controls, 7 pediatric CD; 8157 and 4398 transcriptomes, respectively). (E) Relative abundance of the ISP cluster and UMAP projections on a published dataset (GSE266546; N = 15 adult controls, 53 adult CD; 7444 and 46,524 transcriptomes, respectively); disease activity classification presented as reported in the original study. (F) Relative abundance of the ISP cluster and UMAP projections on a published dataset (GSE116222; N = 3 adult controls, 6 adult UC; 7288 and 11,811 transcriptomes, respectively). P values: (B–C, E–F) Kruskal-Wallis test with Dunn multiple comparisons; (D) Mann-Whitney test (unpaired, 2-tailed).

Figure 4.

Cells expressing ISP markers in colonic mucosa in CD tissue. (A–B) Representative single channel (A) and merged (B) images of immunostaining for LCN2, CD74, HLA-DR in biopsies from study subjects, quantified in (C) (N = 2 subjects per group, n = 11–16 crypts analyzed per biopsy, average per subject plotted on graph). Asterisk denotes an ISP. (D) Correlation analysis comparing abundance of ISPs in tissue (N = 4) as quantified at transcript level (scRNA-seq) to quantification at protein level (immunostaining) in matched subjects (N = 4). P values: (C) Welch’s t-test (unpaired, 2-tailed); (D) nonparametric Spearman correlation (2-tailed).

We observed reduced stem cell abundance in pediatric CD (Figure 2E), compelling us to evaluate self-renewal in Crohn’s epithelium. Clonal organoid forming efficiency (OFE) assays demonstrated reduced OFE from pediatric CD across multiple passages, consistent with stem cell self-renewal deficits. We did not observe a difference in OFE between adult Crohn’s or control colonoids (Figure 5A–C), putatively because 73% of adult patients in our cohort were in remission at the time of collection. We found significant gene expression changes in disease vs control stem cells (scRNA-seq data) in both cohorts: 87 genes in pediatric and 289 genes in adult stem cells were differentially expressed (Padj < .05) (Supplementary Table 3). Gene set enrichment analysis (GSEA) revealed upregulation of 16 and 13 pathways, mostly related to inflammation (and downregulation of 3 and 1 pathways) in stem cells from pediatric and adult Crohn’s, respectively (Figure 5D). Pathways related to cell cycle (Myc targets, E2F targets, G2M checkpoint) were most significantly upregulated in CD stem cells in adults. These data together suggest that in subjects with active CD, stem cells from noninflamed areas exhibit reduced stem cell function.

Figure 5.

Aberrant stem cell capacity in CD. (A) Representative images of colonoid cultures (N = 7 pediatric controls, 4 pediatric CD; 7 adult controls, 5 adult CD; Passage = 1–2). (B) Schematic: OFE assay. (C) Quantification of OFE for colonoids from pediatric and adult study subjects, P values: Mann-Whitney test (unpaired, 2-tailed). (D) HALLMARK pathways differentially enriched in stem cells from CD biopsies (q < 0.05).

Recent studies in mice demonstrate reactivation of a fetal intestinal gene expression program during post-injury repair in adults. This program, sometimes referred to as revival stem cell signature⁵⁰ or fetal reprogramming signature,^16,51 is driven primarily by YAP activity or interferon (IFN)-gamma.^15,52 Additionally, pediatric CD epithelial cells share transcriptional profiles with human fetal intestinal epithelium.¹¹ Interestingly, “IFN-gamma response” was among the most significantly enriched gene sets in pediatric CD stem cells (Figure 5D). We therefore interrogated our datasets for fetal gene expression programs. We found ISPs were enriched for all signatures tested, whereas there was no difference between CD and control in other cell clusters (Figure 6).

Figure 6.

ISPs are enriched for fetal gene signatures. Enrichment for the fetal gene expression signatures^11,16,50 in the epithelial clusters in CD compared with control colonic epithelium. P ≤ .1 presented on plots.

We next performed RNA velocity analysis⁵³ on our data combined with published data.¹² Results suggest ISPs originate from OLFM4^-REG1A⁺ secretory progenitor cells and have 2 putative trajectories: towards OLFM4⁺REG1A⁺ secretory or absorptive colonocyte progenitors (Figure 7A–B). These results are consistent with recent findings by Li et al,³⁵ indicating a similar cell population in adult CD termed “LND (LCN2, NOS2, DUOX2) cells” may differentiate from stem/progenitor cells into absorptive colonocytes. We analyzed raw data from Li et al using our bioinformatic pipeline, and the converse (Li et al pipeline with our data) and found ISP cells are indeed the same population as LND cells (Figure 7C–D). Furthermore, just like a subset of “late” LND cells were assigned absorptive colonocyte lineage by Li et al, we also detect a small, isolated portion of ISPs within the FABP1⁺ absorptive colonocyte cluster (Figure 1E–F, Figure 7D).

Figure 7.

ISP cells contribute to the absorptive lineage and are synonymous with “LND” cells. (A) Epithelial cluster UMAP key for plots in (B) and (D). (B) Predicted future transcriptional states for individual cells (merged data from current study and the published datasets DUOS-000145, DUOS-000146) plotted on the epithelial UMAP. (C) Epithelial transcriptomes from this study were analyzed through the pipeline from Li et al.³⁵ The cells identified as LNDs match their position on the UMAP in original paper. (D) Epithelial transcriptomes (colon) from Li et al³⁵ were analyzed through the pipeline from this study, and the LND cells are highlighted in red, demonstrating that the cluster is split between the secretory and absorptive lineages.

Epigenetic Memory Underlies Persistent Inflammatory Gene Expression in CD Epithelium

To evaluate how CD alters the epigenetic landscape of epithelial cells, including ISPs, we performed single-nucleus assay for transposase-accessible chromatin sequencing (snATAC-seq) on a subset of pediatric subjects (5 CD, 4 control) because our pediatric CD cohort had higher representation of ISPs (Figure 8A–D). We observed a modest increase in chromatin accessibility and gene activity of the ISP signature (Supplementary Table 4) in stem and progenitor cell clusters (Figure 8D). Consistent with findings from scRNA-seq, ISPs were specifically enriched in CD (P = .098) at the expense of absorptive colonocytes (P = .019) (Figure 9A–B, Figure 8E).

Figure 8.

snATAC-seq clustering of epithelium of colonic biopsies from pediatric control subjects and patients with CD. (A) UMAP representation of all epithelial cells (N = 4 control subjects (n = 3628 epithelial cells); N = 5 patients with Crohn's disease (n = 1913 epithelial cells)) sequenced from pediatric subjects color-coded by cell cluster. A total of 12 district clusters were obtained. (B) Heatmap showing the relative Gene Activity (z-score) of intestinal epithelial gene signatures (rows) across the identified epithelial cell clusters (columns). (C) UMAP representation of all epithelial cells sequenced from pediatric subjects annotated and color-coded by cell type. (D) Heatmap showing the relative Gene Activity (z-score) of intestinal epithelial gene signatures (rows) across the identified epithelial cell types (columns). (E) Density map showing UMAP embedding of chromatin accessibility features for CD and control samples. Density contours represent regions of highest cell density from each condition with color gradient reflecting the location of highest cell density. (F) UMAP of all epithelial cells from pediatric samples with ISP cells highlighted in pink (left). UMAP showing the enrichment of NFkB motifs across cell clusters (right).

Figure 9.

Chromatin accessibility is enhanced in a subset of ISP genes in CD epithelium. (A) Schematic: experimental approach and data presentation. (B) Frequency of cell types comparing controls and patients with CD. Boxplots represent the data’s interquartile range (IQR), with the median indicated. Whiskers represent the highest and lowest values within 1.5 × IQR. Points represent individual sample frequencies. (C) Schematic representation of Gene Activity score. (D) Volcano plot showing genes with different Gene Activity scores in a merged progenitor cluster (stem, secretory progenitors, early progenitors, TA) between and control and patients with CD. ISP signature genes are highlighted in red. Box: genes used for pathway analysis in (G). (E) Average chromatin accessibility for the “ISP-memory" genes and “ISP-no-memory" genes in epithelial clusters. Circles represent individual chromatin accessibility scores for each subject, dashes are mean values, solid lines are fitted curves, and shaded areas represent 95% confidence interval of the fitted curve. (F) Heatmap summarizing average normalized expression of “ISP-memory" genes in stem, progenitor, and ISP clusters (scRNA-seq data). (G) GO term analysis of genes with significantly higher Gene Activity scores (box in panel D) showing signaling pathways enriched in CD merged progenitor cluster compared with controls. (H) Volcano plots representing motifs enriched in the merged progenitor cluster in CD.

We next predicted differential gene activity using accessibility profiles of chromatin regions near gene loci in cell types presumed to give rise to ISPs—stem cells, secretory progenitors, early progenitors, and TA cells—in CD vs control (Figure 9C, D). We found a total of 393 genes (of 1858 highly variable genes) differentially accessible between conditions (Supplementary Table 5), including 8 of 23 ISP signature genes (DMBT1, HLA-DPA1, CIITA, CD74, CXCL1, CASP1, DUOX2, and DUOXA2) which we term “ISP-memory” genes (Figure 9D). Some of these genes were associated previously with enhanced chromatin accessibility in other chronic inflammatory diseases,^23,54,55 suggesting they represent common targets of epigenetic modification in response to chronic inflammation. These “ISP-memory” genes have higher chromatin accessibility in CD, perhaps a result of epigenetic memory established in vivo. We therefore calculated average accessibility for these genes in epithelial subclusters compared with remaining genes within the ISP signature (“ISP-no-memory” genes). “ISP-memory” genes have enhanced accessibility in CD stem cells (P = .0038) and persist through differentiated lineages, whereas there is no enhanced chromatin accessibility in controls. In contrast, the remaining ISP signature genes do not show enhanced chromatin accessibility in any progenitor clusters (Figure 9E). “ISP-memory” genes are expressed at lower levels in stem cells than in subsequent progenitor clusters and peak in the ISP cluster in CD but not control epithelium (Figure 9F), indicating that poised open chromatin precedes expression of these genes in CD.

In addition to ISP genes, there are 84 genes with enhanced accessibility in combined stem, secretory progenitor, early progenitor, and TA clusters in CD compared with control (Figure 9D, box). Gene Ontology (GO) enrichment revealed terms related to tumor necrosis factor (TNF)-alpha (“regulation of vascular wound healing,” “response to tumor necrosis factor”) and Signal Transducer and Activator of Transcription (STAT) (“response to type II IFN,” “response to cytokine,” “regulation of angiogenesis,” “regulation of vasculature development”) signaling, among the top enriched signaling pathways in CD (Figure 9F). Accordingly, differential transcription factor binding analysis identified nuclear factor kappa B (NFkB)/RELA motifs as enriched in open chromatin peaks in CD (Figure 9H). These motifs showed specific enrichment in stem cells and ISPs (Figure 8F).

Because NFkB/RELA motifs and interferon response factors/STAT were enriched in open chromatin of CD epithelium, we investigated the distribution of chromatin accessibility peaks surrounding these regulatory regions in ISP genes. Looking at the level of progenitor cells (merged stem cells, TA, early progenitors, secretory progenitors, ISPs) or individual subclusters, we detected peaks associated with more open chromatin in CD for HLA-DPA1, HLA-DRA, CD74, and DUOXA/DUOX2 (Figure 10A, Figure 11), but not for LCN2 (Figure 11). We observed subject-based variability in chromatin accessibility in the promoter region (Figure 10B–C). These data suggest that increased chromatin accessibility around proinflammatory transcription factor-binding motifs may contribute to expression of ISP signature genes in CD epithelium and enable reactivation of these with reintroduction of inflammation. Along with increased ISPs at the apparent expense of TA and early progenitor clusters, this suggests a possible interrelationship between ISPs and upstream cell populations.

Figure 10.

Select ISP signature genes contain open chromatin peaks enriched in CD. (A) Open chromatin peaks in a merged progenitor cluster (stem cell, TA, early progenitors, secretory progenitors, ISPs). Boxes placed to highlight the peaks with differential enrichment in CD. Corresponding gene maps positioned below chromatin traces. (B) Schematic representation of peak accessibility values. The area of peaks within promoter region (dashed outline) of genes of interest is plotted in (C) for individual subjects (N = 4 controls and 5 Crohn's).

Figure 11.

Select ISP signature genes contain open chromatin peaks enriched in CD, in multiple epithelial subclusters. Open chromatin peaks in stem cells, TAs, early progenitors, secretory progenitors, ISPs, clusters, with traces from control and subjects with CD separated by color. Corresponding gene maps positioned below the chromatin traces.

CD Colonoids Are Poised for Induction of an ISP Cell State

To evaluate the durability of CD epithelium to retain cells in an ISP state, we turned to colonoids derived from the same cell suspensions as were sequenced (Figure 12A). We found mRNA levels of ISP genes HLA-DPA, HLA-DRA, CD74, and LCN2 were elevated in colonoids derived from CD (Figure 12B), yet we did not detect equivalent staining for corresponding proteins (Figure 12C). However, when we cultured the same colonoids in differentiation medium (as opposed to expansion medium), we found a significant correlation (Pearson r = 0.87; P = .003) between HLA-DR protein on the cell surface of colonoids by flow cytometry and HLA-DRA mRNA counts in epithelial cells by scRNA-seq from the corresponding sample (Figure 12C). Because these data suggest that ISP protein expression is coupled to cell differentiation, all subsequent experiments with colonoids were conducted in differentiation medium. We reported previously the need for cytokine stimulation for reinduction of HLA-DR protein expression in colonoids from patients with very early onset IBD despite mRNA expression at baseline.³² Therefore, we hypothesized that introduction of inflammatory stimuli in vitro would result in expansion of ISPs. We employed a cocktail of 20 ng/mL interleukin (IL)-1beta, 100 ng/mL TNF-alpha, 1 μg/mL flagellin, and 10 ng/mL IL-6, previously shown to induce IBD-relevant gene expression in colonoids.^56,57 Treatment of colonoids with this cocktail resulted in clusters of cells coexpressing ISP proteins HLA-DR, LCN2, and CD74 in CD, and to a lesser extent, control colonoids (Figure 12D–E). By flow cytometry, inflammatory cocktail treatment resulted in potent upregulation of HLA-DR expression on the surface of cells in both control and CD colonoids, albeit the increase only reached statistical significance in CD colonoids (P = .006 vs P = .138 in control lines) (Figure 12F). Since snATAC-seq revealed that chromatin accessibility around the HLA-DPA gene was most potently increased in CD epithelium (Figure 9D), we used HLA-DP as a second cell surface marker to more discretely identify ISPs via flow cytometry. We confirmed that HLA-DR⁺HLA-DP⁺ cells are enriched for ISP genes by quantifying expression of ISP marker genes HLA-DPA, HLA-DRA, CD74, and LCN2, in total live cells vs HLA-DR⁺HLA-DP⁺ sorted cells from colonoids—with or without the inflammatory cocktail (Figure 13A). Cells coexpressing HLA-DP and HLA-DR were expanded significantly upon stimulation with the inflammatory cocktail in colonoids from CD but not control subjects (Figure 13B). HLA-DR⁺HLA-DP⁺ cells were absent in control colonoids at baseline, and inflammatory cocktail treatment did not elicit an expansion sufficient to sort and extract RNA (Figure 13B). However, we found that HLA-DR⁺HLA-DP⁺ cells in treated CD colonoids exhibit enrichment of hallmark ISP genes relative to total live cells (Figure 13C).

Figure 12.

ISP state can be experimentally induced in patient-derived colonoids. (A) Schematic of experimental approach and data presentation. (B) Expression of ISP marker genes in colonoids from CD or control subjects (N = 3 each, Passage = 4–6). (C) Correlation analysis of HLA-DRA gene expression in the study subjects’ biopsies (scRNA-seq data, epithelial cluster) vs HLA-DR protein expression in colonoids (N = 4 controls and 5 Crohn's) from corresponding subjects (flow cytometry), comparing colonoids grown either in stem cell-enriching expansion medium (black squares) or differentiation medium (open circles). (D) Representative images of immunostaining for LCN2, CD74, HLA-DR in colonoids from study subjects. Asterisk denotes an ISP. Quantified in (E): N = 2 lines (n = 12 high power fields) per group, Passage = 7–8. (F) Quantification (flow cytometry) of %HLA-DR+ cells in colonoids treated with the inflammatory cocktail (20 ng/mL IL-1beta, 100 ng/mL TNF-alpha, 1 μg/mL flagellin, and 10 ng/mL IL-6) for 24 hours (N = 7 or 8 lines from control subjects or patients with CD, respectively, Passage = 6–8). P values: (B) Mann-Whitney test (unpaired, 2-tailed), (E–F) ordinary 2-way analysis of variance with Sidak’s multiple comparisons.

Figure 13.

HLA-DR+HLA-DP+ cells from colonoids subjected to inflammatory cocktail express ISP marker genes. (A) Schematic: Colonoids from from control subjects or patients with CD were treated with the inflammatory cocktail for 24 hours and subjected to flow cytometry sorting of HLA-DR+ and HLA-DR+HLA-DP+ populations, as well as total live cells to use as control. (B) Quantification (flow cytometry) of %HLA-DR+HLA-DP+ cells in colonoids treated with the inflammatory cocktail for 24 hours. One line each from control subject or patient with CD, N = 3 independent experiments repeated in different passages (Passage = 4–6). (C) Expression of ISP marker genes HLA-DPA1, HLA-DRA, CD74, and LCN2, in cell populations sorted as described in (A) from control or CD colonoids. One line each from control subject or patient with CD, N = 3 independent experiments repeated in different passages (Passage = 4–6), representative data from 1 experiment are shown. P values: (B) ordinary 2-way analysis of variance with Sidak’s multiple comparisons.

Importantly, tissue-resident proinflammatory macrophages are known to secrete IL-1beta, TNF-alpha, and IL-6, which we confirmed via cytokine bead array assay⁵⁸ (Figure 14A). To test whether proinflammatory polarized macrophages can induce the ISP cell state in vitro, we obtained human monocytes from healthy donors and differentiated them ex vivo in the presence of IFN-gamma and lipopolysaccharide (LPS), followed by direct co-culture with colonoids (Figure 14B). We found that degree of macrophage differentiation varied based on the donor, with 41.5% to 85.2% of total CD14+ monocytes coexpressing macrophage marker CD64 and pro-inflammatory macrophage marker CD80 (Figure 14C). Upon co-culture, we found that colonoids from patients with CD were more responsive to ISP induction by proinflammatory macrophages than control colonoids (Figure 14D). Taken together, these data suggest that colonoids from patients with CD exhibit hyper-responsiveness to in vitro inflammatory stimuli.

Figure 14.

Proinflammatory macrophages induce expansion of cells expressing ISP marker genes in co-culture with colonoids in vitro. (A) Quantification of secretion of the cytokines TNFa, IL-1b, IL-6, IL-10, IL-8, IL-12p70 by macrophages differentiated and activated ex vivo prior to co-embedding with colonoids (N = 2 donors). (B) Schematic: monocytes from peripheral blood of healthy donors (different donor for each of the 3 independent experiments) were differentiated into proinflammatory macrophages and embedded with colonoids and 1 μg/mL flagellin (Passage = 4–7) from patients with CD or control subjects (N = 3 each), followed by flow cytometry 24 hours later. (C) Assessment of macrophage activation (CD64) and proinflammatory polarization (CD80) via flow cytometry within the CD14high monocyte population. N = 3 independent donors. (D) Quantification of %EpCAM+HLA-DR+HLA-DP+ cell induction in colonoid monocultures or co-cultures with proinflammatory macrophages (N = 3 control and 3 Crohn's colonoid lines). P values: ordinary 2-way analysis of variance with Sidak’s multiple comparisons.

Discussion

Our study supports a paradigm in which disease-associated stem cells are epigenetically poised for the emergence of cells in an ISP state in CD. Although ISP marker genes have been broadly associated with IBD, their coexpression in a distinct progenitor population supports a new model in which pathogenic cell states arise and persist in disease. Even more striking is the identification of ISPs in biopsies derived from endoscopically uninflamed tissue. This suggests that the ISP state is either maintained by exposure to systemic inflammatory molecules or it was acquired during active inflammation and that memory in the epigenome enables this state to persist in absence of active tissue inflammation.

Increased chromatin accessibility around proinflammatory transcription factor binding sites of ISP marker genes support a mechanism by which cells are poised to respond more robustly to subsequent inflammation. Data from human colonoids functionally validate that ISP marker gene expression remains elevated in homeostatic culture and that cells in this state are robustly expanded following inflammatory stimuli in CD epithelium. Although our patient-derived samples were taken from noninflamed areas, we cannot rule out systemic inflammation as a contributor to epigenetic changes. However, the persistence of the ISP transcriptional signature in CD colonoids in vitro supports an epithelial-intrinsic mechanism. Together with differential chromatin accessibility findings, our data support the notion that an altered epigenome persists in noninflamed areas that are primed to respond to subsequent inflammatory insults by producing cells in an ISP state.

Some cells in ISP state may represent a subpopulation of deep crypt secretory cells, a secretory cell population described in mice; however, we did not detect typical markers^46,47,59 for these cells in any distinct cluster within our data. Our finding that ISPs exhibit high fetal reprogramming gene expression supports the notion that these cells exhibit plasticity, and RNA velocity analyses suggest ISPs may exhibit loss of lineage fidelity. Transient lineage plasticity is a hallmark of colonic regeneration, and emerging data suggest that founder progenitor cells expressing interferon targets can clonally expand to promote wound healing after wide-spread DSS-induced epithelial cell damage in mice.⁶⁰ These findings, together with the presence of ISPs in healthy control subjects, suggest that the ISP state may represent normal wound healing and that in IBD, persistence of this state could contribute to IBD progression or even progression to colitis-associated colorectal cancer. In this vein, a recent study demonstrated that Paneth cells lacking the tumor suppressor gene Apc give rise to colonic tumors in the context of DSS-induced inflammation.⁶¹ Therefore, although ISPs may be a part of typical wound healing process, present studies indicate that in IBD there must be a mechanism by which ISPs persist.

In search of putative mechanisms that transform ISP from mediators of tissue repair into drivers of a pathologic process, we identify increased chromatin accessibility around proinflammatory transcription factor binding motifs for a subset of ISP genes. Recent studies in mice support a role of epigenetic memory in stem cells. In a mouse model of gastrointestinal acute graft-vs-host disease (GVHD), Lgr5+ stem cells exhibited changes in oxidative phosphorylation genes in vivo that persisted in enteroid culture, including a functional reduction in oxygen consumption rates. In addition, post-GVHD stem cells exhibited reduced regenerative capacity based on organoid assays.³³ Reduced regenerative capacity of intestinal stem cells following in vivo challenge has also been demonstrated in mouse models of T cell-mediated tissue injury.^62,63 Finally, additional studies of dextran sodium sulfate (DSS)-treated mice suggest that chronic inflammation promotes chromatin accessibility around AP-1 factors in stem cells contributing to tissue memory that promotes progression to adenoma.⁶⁴

Proinflammatory cell states have been described by others in adult chronic gastrointestinal diseases. As mentioned previously, Li et al³⁵ reported a cell population in adult CD termed “LND (LCN2, NOS2, DUOX2) cells that we find is synonymous with cells in the ISP state. In UC, “inflammation-activated” cell states were reported in colonoids downstream of Janus kinase (JAK)/STAT signaling,⁶⁵ whereas Oliver at al describe the presence of “inflammatory epithelial cells (INFLAREs)” in IBD, celiac disease, and colorectal cancer.³⁴ Although ISP marker genes have been associated previously with IBD, our study is the first to report an epigenetic basis for persistent inflammatory cell states in IBD and provides a foundation for expanded investigation in this realm. Future studies are needed to define the relative contribution of DNA methylation, poised vs active enhancer activity, and other histone modifications to the emergence of the ISP state. Although additional studies are needed to validate ISPs as a therapeutic target, another clinical translation of this work is to define the extent to which abundance of ISPs can serve as a prognostic marker. This possibility could be especially important as an early marker in children with IBD to predict disease progression. In conclusion, our findings suggest that epigenetic memory of intestinal stem cells not only contributes to persistence of a putatively pathogenic ISP cell state but also provides new avenues for epithelial-targeted therapies and prognostic tools for IBD.

Methods

Subject Enrollment and Demographics

This study was conducted with the approval of the Children’s Hospital of Philadelphia Institutional Review Board: #14-010826 (Principal Investigator, Dr Judith Kelsen) and the University of Pennsylvania Perelman School of Medicine Institutional Review Board: #814428 (Principal Investigator, Dr Meenakshi Bewtra). For pediatric subjects, all parents of patients provided written informed consent. For adult subjects, all patients provided written informed consent. Biopsy specimens for analyses and for histology were obtained from the same region within 1 to 2 cm from each other. These were obtained from deidentified patients collected endoscopically by Drs Kelsen and Bewtra and their clinical colleagues. In total, we processed biopsies from endoscopically unaffected areas in the ascending colon from 19 patients with CD and 23 control subjects. Histological findings ranged from moderate chronic active colitis to normal (with inflammation in other regions of the gastrointestinal tract consistent with CD). All pediatric patients were treatment naïve, with a range of disease activity as measured by the Physician Global Assessment and PDCAI.⁴⁸ In the adult group, all patients were biologic-naïve, 1 patient was treatment-naïve, and 2 patients were not receiving treatment at the time of biopsy. Disease activity was measured by the sCDAI.⁶⁶ An overview of patient demographics is listed in Table 1. Detailed patient information, including age, sex, race and ethnicity, indications for endoscopy, is available in Supplementary Table 1.

Sex as a Biological Variable

We recruited patients of both sexes with an equal distribution in control and CD cohorts. Detailed information is available in Supplementary Table 1.

Sample Processing and Preparation for Downstream Analyses

Two biopsies per site were harvested in cryopreservation media for subsequent batch sample preparations for downstream analyses as published previously.³⁶ Briefly, samples were processed to separate epithelial-enriched and nonepithelial fractions, which were processed separately for scRNA-seq, snATAC-seq, and colonoid line generation. scRNA-seq libraries were generated using Chromium Next GEM Single Cell 3' GEM, Library & Gel Bead Kit v3.1 (10X GENOMICS INC) as per manufacturer’s instructions. Illumina NovaSeq 6000 was used to sequence libraries using parameters Read1: index5: index7:Read2:: 28:8:0:87 or 28:10:10:90.

scRNA-seq Quality Control and Processing

We aligned and quantified reads using 10X Cellranger 3.1.0 and built a gene expression matrix by combining all available samples. Low-quality cells expressing less than 1000 genes and mitochondrial Unique Molecular Identifier (UMI) greater than 30% of the cell total were discarded. Only the genes that are expressed in more than 100 cells were used for analysis. The UMI counts for each gene were normalized by the cell totals followed by log transformation after addition of a pseudonumber⁶⁷ to avoid negative and undefined log values. The data are deposited to the Gene Expression Omnibus (GEO) database (accession number GSE305528). Seurat integration workflow was used to integrate the top 2000 variable genes, as anchors, across cells for all samples. After integration, principal component analysis (PCA) was used for initial dimensionality reduction and later for clustering. The components then were used as input to the Uniform Manifold Approximation and Projection (UMAP)⁶⁸ dimensionality reduction procedure. Cell types for clusters were assigned based on expression patterns of known cell markers.

Clinical Samples Included in snATAC-seq Analysis

snATAC libraries were generated from healthy and CD pediatric tissue samples of the ascending colon using Chromium Next GEM Single Cell ATAC Library & Gel Bead Kit v1.1 (10X GENOMICS INC) as per manufacturer’s instructions. Libraries were then sequenced on Illumina NovaSeq 6000 using parameters Read1: index5: index7:Read2::49:8:16:49. Samples from failed sequencing runs or with fewer than 100 high-quality sequenced cells were excluded from analysis. Supplementary Table 1 lists study subjects with snATAC-seq data available (5 CD, 4 control).

snATAC-seq Cell Type Annotation

Broad cell annotations were manually obtained by examining clusters for gene activity enrichment of canonical lineage-specific marker genes and published gene signatures⁹ for Epithelial, Immune, and Stromal compartments. Epithelial compartment cells were subsetted and reprocessed to identify finer epithelial cell types. Unsupervised cell clustering identified 12 distinct epithelial clusters from 5541 epithelial cells distributed across 4 control (3,628) and 5 CD (1,913) pediatric samples after quality control filtering (Figure 8A). Epithelial cell annotations were manually obtained by examining clusters for gene activity enrichment of canonical cell type marker genes, published signatures,⁶⁹ and a custom ISP gene signature from our scRNA-seq analysis (Supplementary Table 4). Clusters 2 and 3 showed specific enrichment for open chromatin at ISP signature genes but not for any other canonical intestinal gene programs; thus, we labeled these clusters as ISPs (Figure 8B-D).

Colonoid Line Generation, Organoid Formation Assays, and Inflammatory Cocktail Stimulation

Roughly one-third of the epithelial single-cell suspension prepared for sequencing was used to establish colonoid cultures as described.⁷⁰ Briefly, 4500 live cells from dissociated crypts were seeded in a 50-uL droplet of 80% Matrigel, allowed to solidify for 45 minutes at 37^oC. IntestiCult OGM (StemCell Tech #06010) was supplemented with ROCK inhibitor Y27632 (working concentration 10 μM) at seeding, and Primocin (InvivoGen #ant-pm-05) was used in the first passage to prevent contamination. After 14 days of culture, colonoids were dissociated via trypsinization and seeded (A) for expansion at 2,000 to 5000 live cells per 50 μL droplet of 80% Matrigel or (B) for OFE assay at 500 live cell in 10-μL droplets of 80% Matrigel. Cultures were imaged on days 7 and 14 post-seeding using Keyence BZX microscope. Organoid formation assays were performed in IntestiCult OGM, whereas all other experiments were performed in differentiation medium⁷¹ (50% L-WRN conditioned medium,⁷² 1× B27 supplement, 50 ng/mL FGF-2, 100 ng/mL IGF-1, 10 nM Gastrin, 1 mM N-Acetylcysteine, 500 nM A83-01). To induce differentiation, colonoids grown in OGM were dissociated to single cells, seeded into Matrigel droplets, and fed with differentiation medium for 14 days prior to harvest. Inflammatory cocktail treatment was performed by adding recombinant human TNF-alpha (Peprotech #300-01A, final concentration 100 ng/mL), IL-1beta (Peprotech # 200-01B, final concentration 20 ng/mL), IL-6 (Peprotech # 200-06, final concentration 10 ng/mL), and S. typhimurim flagellin (InvivoGen # FLA-ST, final concentration 1 μg/mL) for 24 hours.

Macrophage Differentiation From Monocytes

For each experiment, 10 × 10⁶ monocytes from healthy human donors were acquired from the Human Immunology Core at the University of Pennsylvania (RRID:SCR_022380) for differentiation into macrophage following the protocol established by STEMCELL Technologies. Briefly, monocytes were seeded into a 12-well plate, at 1 × 10⁶ monocytes/well, and allowed to differentiate for 6 days in ImmunoCult-SF (STEMCELL Technologies: Cat# 10961) supplemented with recombinant human macrophage colony-stimulating factor (M-CSF) (STEMCELL Technologies: Cat# 78057; 50 ng/mL). On day 4 of culture, macrophages were activated with LPS (10 ng/mL) and IFN-gamma (50 ng/mL) and on the 6th day, were harvested for co-culturing with colonoids. For quality control, 1 × 10⁶ monocytes were cultured separately to access the percentage of proinflammatory macrophages (CD14⁺CD64⁺CD80⁺) that were derived from monocytes (CD14⁺CD64^-CD80^-) using flow cytometry analysis.

Colonoid-macrophage Co-culture

Following macrophage activation, on the 6th day of culture, macrophages were harvested for co-culturing with colonoids. ImmunoCult-SF Macrophage medium was removed, and Accutase (STEMCELL Technologies: Cat# 07920) was added to macrophages, which were then incubated at 37^oC for 15 minutes. Accutase was then inactivated by 2× volume of 0.5% bovine serum albumin (BSA) in phosphate buffered saline (DPBS). The cell suspension was transferred to a conical tube, centrifuged for 5 minutes at 300 g and resuspended in medium containing a ratio 1:1 of ImmunoCult-SF to differentiation medium. Colonoids were harvested on day 12 to 13 post-seeding, washed with ice-cold DPBS, and re-embedded in 100 μL of Matrigel containing 3 × 10⁵ macrophages and ∼100 colonoids per droplet. The lines were co-cultured for another 24 hours in media at a 1:1 ratio of ImmunoCult-SF and differentiation medium containing flagellin (1 μg/mL).

Single-cell Dissociation, Fluorescence-activated Cell Sorting and Flow Cytometry Analysis of Colonoids

Colonoids (untreated, cytokine-treated, co-cultured) were harvested in DPBS and centrifuged at 700 g for 3 minutes. Samples were resuspended in TrypLE Express (Gibco: Cat# 12605-010) containing DNAse I (Sigma Aldrich: Cat# 10104159001; 0.50U/ml) and incubated in a ThermoMixer C (Eppendorf) at 37°C rotating at 800 rpm for 12 minutes to obtain a single-cell suspension. TrypLE was inactivated in sorting buffer (2% BSA in 1× Hanks Balanced Salt Solution supplemented with 25 mM HEPES, 0.5 U/mL DNAseI, and 10 μM Y-27632HCl), and samples subjected to extracellular staining (Table 2) for 30 minutes at 4°C in the dark. For fluorescence-activated cell sorting (FACS) experiments, samples were sorted into TRI-Reagent (Sigma Aldrich: Cat# T3934-100ML) using either the Cytek Aurora or the FACS Melody. For flow cytometry analysis, samples were fixed in Cytofix (BD Biosciences: Cat# 554655) for 15 to 20 minutes prior to analysis on the BD LSR Fortessa.

Table 2.

Key Reagents Used for Flow Cytometry

Antibody	Conjugation	Manufacturer	Catalog	Dilution
LIVE/DEAD fixable blue dead cell stain kit	N/A	Invitrogen	L34961A	1:1000
DAPI	N/A	Sigma Aldrich	D9542	1:10,000 (from stock of 1mg/ml)
HLA-DP	PE	BD Pharmingen	566825	1:20
HLA-DR	PerCP/Cy5.5	Biolegend	327020	1:20
EpCAM	FITC	Biolegend	324204	1:20
CD45	APC/Fire750	Biolegend	368517	1:20
CD64	APC	Biolegend	305013	1:20
CD14	PerCP/Cy5.5	Biolegend	325621	1:20
CD80	FITC	Biolegend	305205	1:20

Colonoid Formalin-fixed Paraffin-embedding and Histology

Colonoids were harvested by dislodging the Matrigel with ice-cold DPBS and fixed at 4^oC using 4% paraformaldehyde for an hour, followed by a wash with DPBS. After pelleting, fixed colonoids were resuspended in 80 μL of warmed embedding gel (2% bacto-agar + 2.5% gelatin) and transferred to chilled parafilm-wrapped block. Droplets of colonoids embedded in the gel were allowed to solidify at 4°C and then each placed into a histology cassette and stored in 70% EtOH for processing. Embedding into paraffin blocks, sectioning, and hematoxylin and eosin (H&E) staining were carried out by Center for Molecular Studies in Digestive and Liver Diseases and the Molecular Pathology and Imaging Core (RRID:SCR_022420).

Immunofluorescence

Formalin-fixed paraffin-embedded (FFPE) slides were deparaffinized and rehydrated. Antigen retrieval was done in a pressure cooker (Antigen Retriever 2100) using Tris-EDTA antigen retrieval buffer (10 mM Tris base, 1 mM EDTA solution, 0.05% Tween 20, pH 9.0). Sections were permeabilized with PBS-T (1× DPBS with 0.1% Triton-X) for 5 minutes and then washed. All washes in between steps were done with 1× DPBS. Slides were blocked at room temperature for an hour (10% donkey serum and 1% BSA in DPBS). Primary antibodies (Table 3) were diluted in 1% BSA in DPBS and incubated overnight at 4°C. After washing, slides were incubated with secondary antibodies (Table 3) and 4′,6-diamidino-2-phenylindole (DAPI) (1:5000 dilution of 1mg/mL stock) for 30 minutes in the dark at room temperature. Slides were washed and then mounted using Prolong Gold Antifade Mountant (P36930, ThermoFisher Scientific). Slides were imaged on a Leica Stellaris 5 Laser-scanning confocal microscope at the Cell and Developmental Biology Microscopy Core (RRID SCR_022373).

Table 3.

Antibodies for Immunofluorescent Staining

Primary antiboddies	Manufacturer	Catalog	Dilution
Rabbit recombinant monoclonal anti-HLA-DR [EPR3692]	Abcam	ab92511	1:100
Purified mouse anti-human CD74 Antibody	BioLegend	326802	1:100
Goat polyclonal anti-LCN2	Proteintech	AF1757	1:40
Purified mouse anti-ECAD	Cell Signaling	610182	1:50
Secondary antibodies
Cy2 AffiniPure donkey anti-rabbit IgG	Jackson IR	711-225-152	1:600
Cy3 AffiniPure donkey anti-mouse IgG	Jackson IR	715-165-151	1:600
Cy5 AffiniPure donkey anti-goat IgG	Jackson IR	705-175-147	1:600

snATAC-seq Quality Control and Processing

Raw 10X scATAC-seq data were demultiplexed and aligned to the GRCh38 genome using the mkfastq and count functions from Cell Ranger ATAC (version 2.0.0). ArchR arrow files and a 500-bp bin-by-cell matrix were created for downstream quality control (QC) and analysis with the ArchR package version 1.0.2. High-quality cell barcodes satisfying QC measures of (1) Number of fragments (2500–150,000); (2) TSS enrichment score (>12); (3) Nucleosome ratio (0.5–10); (4) Blacklist ratio (<0.5) were retained. Doublets were inferred using addDoubletScores() (k=10, knnMethod=”UMAP”) and removed. Initial dimensionality reduction on bin features was performed using iterative latent semantic indexing (LSI) implemented by addIterativeLSI() (iterations=4, varFeatures=25000, resolution=0.2). Harmony was applied to the IterativeLSI dimensionality reduction to correct for batch effects across patient samples. Cells were clustered in the harmonized space using (addClusters) (method=”Seurat”, resolution=1.7) and results were visualized using UMAP. Gene scores were computed as a function of chromatin accessibility within the gene body and putative upstream distal regulatory elements using (addImputeWeights). Accessible peaks were called using ArchR’s implementation of macs2 on pseudobulk replicates. A peak-by-cell matrix (insertions per peak and cell) was generated and used as input to ChromVAR, which calculates bias-corrected enrichment of TF motif accessibility on a per-cell basis.

Heatmap and Box Plot Analysis

Heatmap showing the relative Gene Activity (z-score) of intestinal epithelial gene signatures (rows) across the identified epithelial cell types were plotted using hierarchical clustering pheatmap v1.0.12 function in R. Published fetal and regenerative signatures in epithelial subclusters were aggregated based on individual patients. Average expression of these aggregated signatures was calculated based on each epithelial cell types and then visualized via box plots separated by different conditions using the function geom_half_boxplot implemented in gghalves v0.1.4. Peak conversion values for ISP marker genes in grouped progenitor cells (stem cells, TA, early progenitors, secretory progenitors), were visualized via box plots compared from control subjects vs patients with CD, using the function geom_boxplot implemented in ggplot2 package.

Chromatin Accessibility Analysis of ISP Genes

To measure chromatin accessibility at gene loci, addGeneScoreMatrix function in ArchR was used to compute gene activity scores. Differential accessibility in progenitor-like cells (stem cells, early progenitors, secretory progenitors, and TA cells) between patients with CD and controls was assessed using 2-sided Wilcoxon rank-sum tests. Multiple testing correction was performed using the Benjamini-Hochberg procedure, with a significance threshold of .05. The overlap between ISP genes and genes with higher accessibility in progenitor-like cells in patients with CD were defined as ISP-memory genes (HLA-DPA1, DMBT1, CXCL1, CASP1, DUOX2, DUOXA2, CD74). The rest of ISP genes were defined as ISP-no-memory genes. To evaluate chromatin accessibility for the sets of ISP-memory and ISP-no-memory genes, AddModuleScore in Seurat was used to aggregate gene activity scores, yielding a chromatin accessibility score for each gene set. Average scores were then calculated per cell type and per patient sample. Finally, a smooth curve was fitted via LOESS regression across cell types for controls and patients with CD.

RNA Velocity Analysis

Spliced, unspliced, and ambiguous expression matrices for each patient samples were generated with the tool velocyto.py v0.17.17. These matrices were then integrated and normalized into our previously generated Seurat (v4) object, which contains raw and integrated RNA expression profiles, as well as all the related meta information like annotated cell types. Only epithelial cells were extracted and included in the analysis. A UMAP was generated for this subset of cells based on 2000 highly variable genes across dataset. The integrated Seurat object was saved as H5adtable 1 format for downstream analysis. Then, the single-cell RNA velocities and transcription rates were calculated using the spliced RNA oriented model as implemented in the package UniTVelo v0.2.5.⁷³ A transition matrix of epithelial clusters based on both RNA velocity and similarities among cells was calculated and combined using the methods implemented in CellRank v2.0.2.⁷⁴ The stream of velocities was plotted onto the embedding using the function VelocityKernel.plot_projection and colored based on the pseudo-time of each cell.

Image Processing, Analysis, and Quantification

Images acquired on the Leica Stellaris 5 were processed using ImageJ Fiji v2.14.0/1.54f. Z-stacks were compressed with maximum intensity projection to generate tiff files. For patient biopsies, the following contrast/brightness settings were applied in 8-bit range (0–255) to each image: Channel 1 (DAPI, blue) min: 15 max: 125; Channel 2 (HLA-DR, cyan) min: 15 max: 150; Channel 3 (CD74, yellow) min: 15 max: 175; Channel 4 (LCN2, red) min: 15 max: 255. Scale bar was set to 100 microns. For inset images of patient biopsies, the following contrast/brightness settings were applied in 8-bit range (0–255) to each image: Channel 1 (DAPI, blue) min: 15 max: 100; Channel 2 (HLA-DR, cyan) min: 15 max: 124; Channel 3 (CD74, yellow) min: 15 max: 175; Channel 4 (LCN2, red) min: 15 max: 150. Scale bar was set to 25 microns. Where applicable, contrast/brightness settings for ECAD (Channel 3, green) were set in 8-bit range to min: 10, max: 110. For organoids, the following contrast/brightness settings were applied in 8-bit range (0–255) to each image: Channel 1 (DAPI, blue) min: 0-5 max: 25-35; Channel 2 (HLA-DR, cyan) min: 15 max: 50; Channel 3 (CD74, yellow) min: 0 max: 65; Channel 4 (LCN2, red) min: 15 max: 100.To quantify ISPs in patient biopsies, images were imported into QuPath v0.5.0 and a representative image (biopsy from patient with CD) was selected for setting quantification threshold and counts (see QuPath Script). Before running the script with newly defined “ISP” classifier, crypts were outlined as annotations in each image using E-cadherin (ECAD)-stained images for reference and to distinguish epithelial from nonepithelial cells. Therefore, the classifier was applied only to crypts across all images, and quantification and further analysis was computed by crypt (“ISPs per crypt”). To quantify ISPs in organoids, images were also imported into QuPath and single cells with all 3 markers (HLA-DR/CD74/LCN2) were annotated. These annotations were exported as counts to quantify number of ISPs per image.

RNA Isolation and Quantitative Polymerase Chain Reaction

Colonoids were harvested and resuspended in TRIzol Reagent (Catalog #15596026, Ambion). RNA was extracted using a Direct-zol RNA Microprep Kit (Catalog #R2062, Zymo Research). cDNA was generated using the High-Capacity cDNA Reverse Transcription Kit (Catalog #4368814, Invitrogen). Real-time quantitative polymerase chain reaction (RT-qPCR) was performed with TaqMan probes (TBP1: Hs00427621_m1; LCN2: Hs01008571_m1; HLA-DRA: Hs00219575_m1; HLA-DPA1: Hs00410276_m1; CD74: Hs00269961_m1) and TaqMan Fast Universal PCR Master Mix 2X (Catalog #4352042).