Induced organoids derived from patients with ulcerative colitis recapitulate colitic reactivity

The pathogenesis of ulcerative colitis (UC), a major type of inflammatory bowel disease, remains unknown. No model exists that adequately recapitulates the complexity of clinical UC. Here, we take advantage of induced pluripotent stem cells (iPSCs) to develop an induced human UC-derived organoid (iHUCO) model and compared it with the induced human normal organoid model (iHNO). Notably, iHUCOs recapitulated histological and functional features of primary colitic tissues, including the absence of acidic mucus secretion and aberrant adherens junctions in the epithelial barrier both in vitro and in vivo. We demonstrate that the CXCL8/CXCR1 axis was overexpressed in iHUCO but not in iHNO. As proof-of-principle, we show that inhibition of CXCL8 receptor by the small-molecule non-competitive inhibitor repertaxin attenuated the progression of UC phenotypes in vitro and in vivo. This patient-derived organoid model, containing both epithelial and stromal compartments, will generate new insights into the underlying pathogenesis of UC while offering opportunities to tailor interventions to the individual patient. Although ulcerative colitis (UC) is a major type of inflammatory bowel disease, attempts to model it fully have fallen short. Here the authors use patient-derived iPS cells to develop a UC organoid model that recapitulates disease histological and functional features, and confirm the role of CXCL8/CXCR1 in pathogenesis.

U lcerative colitis (UC) is one of the two principal types of inflammatory bowel disease (IBD) and is a debilitating inflammatory condition of the colon that usually begins in young adulthood 1 . Although the precise pathogenesis is unknown, the lack of mucus secretion, elevated concentrations of soluble inflammatory mediators secreted by the stroma, wounding, and epithelial barrier permeability are thought to play essential early roles in the development and progression of UC [2][3][4][5] . Clinical UC is heterogeneous, ranging from acute to chronic inflammation of the colon. As the pathogenesis is unknown, medical interventions are nonspecific, resulting in inconsistent treatment responses 6,7 . Patients with poor response to these therapies develop severe symptoms, ultimately leading to resection of the colon and rectum 8,9 . Because patient-based predictive models, which capture individual heterogeneity, are lacking, a more complete experimental model is essential.
Pluripotent stem cell (PSC)-derived human intestinal organoids have been developed for the normal mid/hind gut 10,11 and colon 12,13 and have distinct advantages in advancing precision medicine 14 . These organoids can be derived from the individual patient and exhibit recapitulation of physiological responses both in vitro and in vivo 10,12,15 . Besides the unique characteristics of PSCs, induced PSC-derived (iPSC) organoids retain the phenotype and genetic background of the individual donor as reported by recent studies modeling both normal and pathological organs [16][17][18] .
In this study, we provide evidence linking recent advances in stem cell biology 19 with intestinal development [10][11][12]15,[20][21][22] to reprogram colonic fibroblasts isolated from UC patients to iPSCs, followed by directed differentiation to induced human UC organoids (iHUCOs). Notably, iHUCOs have both epithelial and stromal compartments, preserving the colitic phenotype of their tissue of origin, such as responding to wounding by aberrant proliferation, a lack of goblet cells, and having compromised cellular tight junctions in their epithelial barrier.
High-resolution transcriptome-wide analysis of iHUCOs and UC parental fibroblasts revealed pathways and inflammatory mediators in common between both groups compared with their controls, induced human normal organoids (iHNOs) and normal fibroblasts. Specifically, the roles of the inflammatory chemokine CXCL8 and chemotactic cytokine CXCL1 were highlighted in the transcriptome and mesenchymal secretome of iHUCOs and their parental fibroblasts.
As proof-of-principle, we studied the iHUCO functional response to the CXCL8 receptor antagonist, repertaxin, and demonstrated the attenuation of colitic features in iHUCOs both in vitro and in a xenograft model in vivo. Mice have no homolog for human CXCL8. Therefore, this axis is not present in genetic or chemically induced murine models, emphasizing the essential utility of patient-derived UC models.

Results
Patterning of induced organoids recapitulates in vitro their primary tissues. Fresh surgical specimens from colitic and healthy colons were obtained (Fig. 1a), and fibroblasts were isolated 23 (Fig. 1b). We reprogrammed UC fibroblasts to iPSCs; UC fibroblasts were isolated from 6 patients with established chronic colitis along with samples from 4 participants with normal colonic fibroblasts and one from a commercially available source (Supplementary Table 1, Supplementary Fig. 1a, 1b). Pluripotency of the generated iPSCs was confirmed ( Supplementary Fig. 1c-l). Next, we applied an established protocol for intestinal organoid generation 11 to direct differentiation of the iPSCs into definitive endoderm (DE), validated by SOX17 and FOXA2 protein expression (Fig. 1a, d), followed by intestinal spheroid formation (SPH, validated by CDX2 expression, Fig. 1a, e). Normal and UC spheroids were then cultured in Matrigel for 21 days to develop induced human normal organoids (iHNOs), and iHUCOs, respectively. Representative immunofluorescent (IF) staining of iHUCOs including epithelium (stained for CK19) and mesenchyme (stained for VIM) is shown in Fig. 1f.
Both iHNO and iHUCO were characterized by comparison to their matched primary tissues. Hematoxylin and eosin (H&E) staining illustrated distinct epithelial and mesenchymal domains with an interior lumen in organoids (Fig. 1g1, g2). In addition, iHNOs frequently had a simple columnar epithelium, which is representative of the healthy colonic mucosa (Fig. 1g1) whereas iHUCOs had disorganized (pseudostratified and stratified) epithelium (Fig. 1g2). These organoid data appear to recapitulate the presence of greater pseudostratification in colonic mucosa of patients with active UC 24 caused by reparative changes in response to epithelial wounding (Fig. 1g3, g4). The percentages of simple, pseudostratified, and stratified epithelium for normal and UC organoids (Fig. 1h) and their primary tissues ( Supplementary  Fig. 2a) confirmed a significantly higher ratio of non-columnar epithelium in iHUCOs and colitic primary tissues than in the normal condition.
Immunohistochemical (IHC) staining for the proliferation marker KI67 revealed a more uniform pattern of cellular proliferation throughout the columnar epithelium of iHNOs, similar to the normal primary tissues (Fig. 1i1, i3). In contrast, regions of stratified epithelium in iHUCOs and colitic tissues had extensive and non-uniform epithelial proliferation with greater distribution (Fig. 1i2, i4, j). Epithelial expression of KI67 approached 80% in iHUCO and UC tissues, whereas it was significantly lower in normal tissue. Thus, KI67 expression in the iHUCOs suggests an accelerated rate of epithelial turnover that may be similar to the rate of epithelial cell turnover in colonic mucosa undergoing regeneration in patients with active UC 25 .
The intestinal mucus layer secreted by goblet cells in healthy mucosa includes both acid and neutral mucins to protect the epithelial barrier from luminal bacterial penetration 26 . Alcian blue and periodic acid-schiff (AB-PAS) staining revealed the secretion of both types of mucins in the iHNO lumen along with a limited numbers of goblet cells (Fig. 1k1). In contrast, iHUCOs had no or only neutral mucin staining along with a significantly lower number of goblet cells than iHNOs, suggesting a lack of acidic mucin secretory function (Fig. 1k2). The number of goblet cells in UC primary tissues (Fig. 1k4) was significantly lower than in normal tissues (Fig. 1k3, 1l, P < 0.0001). These organoid data are similar to the depletion of goblet cells and the mucus layer observed in the colonic mucosa of patients with UC 27 . Impaired function of the intestinal barrier in patients with UC leads to recurrent severe inflammation and eventual wounding. In response to tissue damage, the epithelial compartment switches to repairing the epithelium as demonstrated by an elevated number of proliferative cells (Ki67 positive) and fewer goblet cells than in homeostatic tissue 28 .
The expression of CDX2 plays a crucial role in intestinal development, including balancing proliferation with differentiation and epithelial barrier formation 11,29,30 . As expected, uniform and strong expression of CDX2 that was restricted to the epithelium was observed in iHNOs ( Supplementary Fig. 2e1) and normal colon tissues ( Supplementary Fig. 2e3). In contrast, CDX2 expression was diminished in primary UC tissues and their corresponding organoids (p < 0.0001, supplementary Fig. 2e2, e4, and f). Thus, the CDX2 expression patterns in our UC organoids were similar to inflammation-related decrease in CDX2 reported for patients with active UC 31,32 .
Recently, SATB2 has been identified as a definitive marker of distal small intestine (ileum) and colonic epithelium in humans 12 .
Thus, we conclude that iHUCOs phenocopy the features of their primary tissues, including a stratified epithelium, an aberrant rate of epithelial proliferation, and the absence of acid mucin secretion.
iHUCOs demonstrate aberrant adherens junction formation in the epithelium. In health, epithelial cells form a physical barrier within the gut lumen that protects the intestine from bacterial and inflammatory cell infiltration 33 . A dynamic combination of different apical intercellular junctions, including tight and adherens junctions, between the epithelial cells maintains this homeostasis 33,34 . Under chronic inflammatory conditions such as UC, the balance in intercellular junctions is disrupted and the integrity of the epithelial barrier is compromised 35,36 . This disruption results in an increase in the paracellular and intracellular space, bacterial invasion, dysregulation of the immune response, and ultimately a damaged, leaky epithelial barrier 2,37,38 .
The tight association between E-cadherin/β-catenin plays a main role in intestinal development and maintaining the epithelial integrity 35  cytoplasm and its eventual translocation into the nucleus is essential for canonical Wnt pathway activation and subsequent expression of tight junction proteins, the ability of membraneassociated β-catenin to bind and to co-localize with E-cadherin is a hallmark of adherens junction regulation 39 . We performed IHC to investigate the subcellular localization of β-catenin in the organoids and their matched primary tissues (Fig. 2a). IHNOs demonstrated strong membrane, cytoplasmic, and nuclear expression of β-catenin, suggesting a high degree of protein stability in these organoids (Fig. 2a1). On the other hand, the expression of β-catenin in iHUCOs was mainly limited to the plasma membrane (Fig. 2a2), and the quantification for exclusive expression on the plasma membrane was significantly higher in iHUCOs than iHNOs (p < 0.0001). However, the combined cytoplasmic and nuclear β-catenin expression was greater in iHNOs than iHUCOs (Fig. 2b). β-catenin expression was consistently and significantly lower in UC than normal primary tissues (Fig. 2a3, a4, b). E-cadherin, the main component of the adherens junction complex, had a similar expression pattern as β-catenin in both organoids and primary tissues (Fig. 2c, d). Although E-cadherin was strongly expressed in the plasma membrane, cytoplasm, and nucleus of iHNOs, cytoplasmic and nuclear expression was significantly lower in iHUCOs (P < 0.0001, Fig. 2c1, c2, d).
RhoA is one of the dominant regulators of the adherens junction complex, playing roles in cell adhesion and cytoskeleton organization 40 . To be activated, cytoplasmic RhoA (inactive) first must translocate to the plasma membrane, where it then can regulate downstream functions, including the formation of actin stress fibers (F-actin), focal adhesions, and cell motility 40 . IHC staining for RhoA showed the expression was significantly greater in both cytoplasm and plasma membrane of iHUCOs than in iHNOs, and also in UC than in normal primary tissues ( Fig. 2e1-e4, f). These data highlight the propensity for an adherens junction signature in iHUCOs.
To confirm the tight association of E-cadherin/β-catenin proteins, as well as significant differences in their sub-cellular co-localization, we performed dual IF staining in iHNOs and iHUCOs (Fig. 2g). As expected, co-localization in iHNOs extended to the cytoplasm and nucleus, indicating higher stability of both proteins (Fig. 2h) whereas in iHUCOs, co-localization was mostly limited to the plasma membrane, suggesting a propensity towards adherens junctions rather than tight junctions.
To functionally confirm these results, we used the recently described microinjection technique 20 to measure and compare the epithelial barrier permeability for both iHUCOs and iHNOs in real-time. Briefly, we microinjected the organoids with fluorescently labeled 4-kD dextran and imaged them with an inverted microscope fitted with epifluorescent filters for a total of 15 h. Real-time measurement of the barrier permeability showed significantly less dye retention in the iHUCO lumen (~50% of real-time measurement timepoints) than the lumen of iHNOs ( Supplementary Fig. 2b). Moreover, CLDN1, a major constituent of the tight junction complexes responsible for normal barrier  In summary, our in vitro findings demonstrate similar expression patterns for β-catenin and E-cadherin in the organoids and their primary tissues. We confirmed the co-expression and tight association between β-catenin and E-cadherin in organoids; both proteins co-localized predominantly in the iHUCO plasma membrane whereas expression extended to the cytoplasm and nucleus in iHNOs. Moreover, our functional study of epithelial barrier permeability in organoids with IHC for the tight junction protein CLDN1 confirmed the tight junction compromise in the iHUCO epithelium.
iHUCOs retain the colitic phenotype upon omental transplantation. To determine whether iHUCOs retain the colitic phenotype in vivo, we combined the recently reported transplantation protocol for PSC-derived organoids 42,43 using biocompatible TS-HA hydrogel to embed and transplant the organoids into the omentum of host NSG mice (Fig. 3a). At 90 days after transplantation, we collected the organoids. Further characterization by H&E demonstrated that the iHUCOs had a colon-like appearance with disorganized epithelium and short crypts whereas iHNOs recapitulated simple columnar epithelium (Fig. 3b). Summarized percentages confirmed a significantly higher ratio of stratified epithelium in UC than normal crypts (Fig. 3c). The colitic epithelium also revealed greater proliferation (Fig. 3d1, d2, 3e), a lack of acid mucin secretion, and significantly fewer goblet cells compared with normal epithelium (Fig. 3f1, f2, p < 0.0001). As expected, in normal tissue, proliferation was limited to the crypt base (Fig. 3d1), and the secreted mucus included both neutral and acid mucins (Fig. 3f1). The colonic epithelium was also characterized for the expression of CDX2 (Supplementary Fig. 2i) and SATB2 ( Supplementary Fig. 2k) proteins. Similar to the in vitro pattern, expression of both proteins was lower in UC than normal crypts ( Supplementary Fig. 2j, l).
Enteroendocrine cells, a specialized secretory cell population, in the colonic mucosa act as the functional sensors for any homeostatic dysregulation in the luminal content 44,45 . In response to alterations in the homeostasis of colonic mucosa, such as pro-inflammatory cytokines in the mucosa of the patients with active UC, greater numbers of enteroendocrine cells have been reported in several studies [46][47][48] . Among the secretory granules of enteroendocrine cells in the gut, chromogranin A, a precursor to several functional peptides, is the most abundant 46 .  IHC staining for chromogranin A showed a significantly (p < 0.0001) higher number and greater distribution of enteroendocrine cells in the epithelium of UC than normal transplanted organoids, confirming a dysregulation in iHUCO development (Fig. 3h, i). We also confirmed that the combined nuclear and cytoplasmic expression of β-catenin and E-cadherin was lower in UC than normal epithelium ( Fig. 3j-m) whereas RhoA expression was significantly higher than in normal colon (Fig. 3n, o; p < 0.0001). These expression patterns are consistent with the patterns seen in UC and normal primary tissues. Both the in vitro and in vivo data confirmed dysregulation in the developmental process of iHUCOs. We observed clinically proven colitic features including changes in crypt histology, aberrant proliferation, more enteroendocrine cells, and adherens vs. tight junction signature in UC epithelium.
Bulk RNA-sequencing reveals UC signatures in iHUCOs. To compare the transcriptional activity during differentiation and the development of organoids derived from healthy or colitic iPSCs, we conducted bulk RNA-seq on normal and UC iPSCs, definitive endoderm (DE), spheroids (SPH), and organoids (GSE117345, n = 3 each). Principal component analysis (PCA) revealed major transcriptional variations among all genes, driven by the developmental stage (Fig. 4a). To improve our understanding of the similarities and differences between UC and normal intestinal development, we conducted a PCA among DE, spheroid, and organoid stages ( Supplementary Fig. 3a). DE, as the first stage of the intestinal development, formed a distinct cluster causing a shift in the gene expression pattern between the UC and normal samples. In contrast, subsequent progression in the development to spheroids and organoids localized the normal and UC closer to each other.
Unsupervised hierarchical clustering of the global gene expression data based on the Spearman rank correlation was performed (Fig. 4b, left). Similar to PCA results, the groups segregated based on the developmental stage rather than disease status, and the organoids formed a discrete cluster separated from DEs and iPSCs but segregated closely with spheroids. Because the developmental stage was the main driver in gene expression, we compared the differentially expressed genes (DEGs) in the sequential stages of development between UC and normal. A Venn diagram of these DEGs (Fig. 4b, right) showed the greatest number of unique genes in the progression from DEs to spheroids.
To confirm the colonic identity of normal and UC organoids at the transcriptome level, we used the list of genes reported recently by Múnera et al comparing the colonic signature with small intestine signature in organoids, as well as primary tissues 12 . Heatmaps of upregulated genes in the colon were generated for the organoids at all three stages of intestinal development ( Supplementary Fig. 3c). Although both normal and UC organoids had a log 2 (RPKM) ≥ 1 for the majority of these genes, the expression pattern differed between normal and UC development ( Supplementary Fig. 3c1, c2). We listed the top 50 expressed genes in UC and normal organoids as curated heatmaps to highlight this difference ( Supplementary Fig. 3e1, e2). A Venn diagram highlights the genes exclusive to normal and UC organoids ( Supplementary Fig. 3d).
To explore the molecular states in the spheroid-to-iHUCO progression, we conducted Gene Set Enrichment Analysis (GSEA) and determined the enriched terms by applying complex network analysis 49 using Cytoscape (Fig. 4c). Significant functional terms (p < 0.01), including inflammatory and immune response, wound healing, and response to bacteria were identified in iHUCOs, which supports their UC phenotype. We generated curated heatmaps for the key genes belonging to the "inflammatory response" (Fig. 4d, top) and "wound healing" terms ( Fig. 4d, bottom) and showed that not only did iHUCOs cluster together as compared with spheroids, but they also showed a significant increase in transcriptional expression of these genes, suggesting a prominent colitic signature.
To identify dominant biological processes enriched in iHUCOs, we applied a ranked list of DEGs to Gene Ontology enrichment analysis (Gorilla, Supplementary Data File 1) and visualized the enriched GO terms using Reduce and Visualize Gene Ontology (REVIGO, Supplementary Data File 1) 50 . The terms clustered semantically along the X-axis based on the similarity: highly significant terms (p < 0.001) including "actin cytoskeleton organization" and "fiber polymerization," clustered on the left, progressed to "response to mechanical stimuli" and terminated on the right with signaling pathways, including Gprotein coupled receptors (GPCR), regulation of interleukin-8 (CXCL8) production, and Rho protein signal transduction (Fig. 4e, Supplementary Fig. 3b). The enrichment scores along with p-values for selected GO terms such as TNF, IFN-γ, CXCL1, and CXCL8 signaling are shown in Fig. 4f. Excessive production of IFN-γ and TNF-α inflammatory cytokines is associated with compromised epithelial barrier integrity 51 . TNF-α is also indirectly responsible for activating GPCRs in the epithelium and wound healing. Birkl et al. recently showed that TNF-α promotes intestinal mucosal wound repair in human colonic organoids 52 . Therefore, our data in Fig. 4f showing strong enrichment of similar GO terms confirm the inflamed nature of iHUCOs.
CXCL8, a multifunctional chemokine secreted by stromal cells in the inflammatory microenvironment, and its receptor CXCR1 have been extensively explored in wound closure 53 and tumorigenicity by us and other groups [54][55][56][57][58] . However, the effect of CXCL8-induced signaling in UC remains unclear. The highly significant roles of CXCL8 and GPCR signaling in the iHUCO transcriptome, with higher specificity of CXCR1 (binding only CXCL8 with high affinity) than CXCR2 (binding all ELR-CXC chemokines with high affinity 59 ) led us to focus on the CXCL8/ CXCR1 inflammatory mediator as a transducer of the G-proteinactivating regulatory system 60 . Dual IF staining for CXCL8 ligand and CXCR1 receptor of the organoids revealed significantly greater (p < 0.0001) expression of both proteins in the epithelium and stroma of iHUC as compared with iHN organoids (Fig. 4g-j).
To summarize, the bulk transcriptomic analyses demonstrated that the iHUCO is a feasible in vitro model for UC. Furthermore, GO analysis identified the role of GPCR, CXCL8, and Rho signaling, a CXCR1/CXCL8 signal transducer and one of the multiple downstream effects of the CXCL8/CXCR1 interaction 61 in iHUCOs. Our dual IF staining confirmed upregulation of CXCL8 ligand and CXCR1 receptor in the epithelium and stroma of iHUC vs. iHN organoids.
Single-nucleus analysis of the parental fibroblasts reveals a unique inflammatory signature. To examine the similarities between parental fibroblasts and each developmental stage, we also conducted bulk RNA-Seq on UC and normal parental fibroblasts (GSE117345).
An unsupervised hierarchical clustering based on the Canberra distance showed that parental fibroblasts shared the highest level of similarity with the organoids rather than the other stages of development (Fig. 5a). Hypothesizing that this similarity originated from the organoid mesenchymal compartment, we used single-nucleus transcriptional profiling (snRNA-Seq) to better understand the cell types and transcriptomic heterogeneity in normal and UC parental fibroblasts (n = 3 each) and compared them to their induced organoids (GSE152999). Singlenucleus profiling resulted in 12 and 9 predicted clusters in normal and UC parental fibroblasts, respectively (Fig. 5b). The list of topexpressed genes in each cluster is shown as marker plots for both normal and UC fibroblasts (Fig. 5c). Notably, IGFB5 was strongly expressed in all UC clusters shown in Fig. 5b. Overexpression of IGFBP5 has been linked to chronic inflammation via an increase in collagen synthesis, leading to mitogen-activated protein kinase (MAPK) and ultimately fibrosis 62-65 .
Next, we used the known and recently reported list of markers in stroma of normal and UC colon tissues 66  To identify the dominant biological processes enriched in UC, we applied the ranked list of DEGs in UC vs. normal fibroblasts to GOrilla, and summarized the results as a REVIGO plot (Fig. 5d). Gradual changes based on the GO term similarities along the X-axis identified highly significant terms (P < 0.001) involved in the inflammatory response and immune cell recruitment/migration, including "response to IL-6/IL-1 production" and "neutrophil degranulation," that lead to activation of biological processes such as chemokine-mediated processes and IFN-γ signaling (Fig. 5d, and Supplementary Fig. 4c). The genes playing crucial roles in enrichment of these terms were mainly Semantic Space x Semantic Space y . e DEGs in iHUCO vs. UC SPH were subjected to Gene Ontology (GO) analysis presented as REVIGO 50 scatterplots (p < 0.001). The terms extend along the X-axis based on similarity in the type of biological process (semantic space X); the difference in color highlights the variety of biological processes. Each circle represents a unique GO term. Circle size corresponds to the number of the genes associated with the GO term. f Highly enriched and significant GO terms in iHUCO shown as a bar graph for the enrichment scores. g-i Representative dual IF staining for CXCR1 (red) and CXCL8 (green) expressed in iHN and iHUC organoid epithelium and mesenchyme, along with j a violin plot for cells co-expressing CXCR1/CXCL8. All nuclei are stained for DAPI (blue). The unpaired, non-parametric two-sided Mann-Whitney U test was used to test for differences in CXCL8/CXCR1 overlap in expression between the iHNO and iHUCO epithelium vs. membrane. Scale bar = 25 μm. n = 5 iHNO and n = 6 iHUCO for IF. Source data are provided as a Source Data file.  Fig. 5e to highlight their upregulation in UC when compared with normal fibroblasts. The roles of GPCR signaling, chemokine signaling, and regulation of the GPCR downstream pathways in UC fibroblasts were also highlighted in our GSEA analysis ( Supplementary Fig. 4d). Specifically, genes such as podoplanin (PDPN), the lymphoid-tissue-like stromal marker, and oncostatin M receptor (OSMR) were expressed at an increased level 68 .
Oncostatin M and its receptor are expressed in the stroma of patients with UC, and induces the expression of PDPN. Insulinlike growth factor binding protein-5 (IGFBP5) 64 and TGFβ1 have well-known roles in fibrosis 69 . MHC Class I proteins, here noted via upregulated expression of HLA-A 70 and LIMCH1 71 , affecting actin stress fibers, have immune and wound-healing roles. These results not only confirmed the colitic signature of UC fibroblasts at the transcriptome level and the essential role of CXCL8 and CXCL1 superfamily in stromal immunoregulation, but also confirmed similar findings in a previous study showing the upregulation of the neutrophil chemoattractants CXCL1, CXCL5, CXCL8 in UC compared with non-inflamed tissues.
High-resolution transcriptomic analysis of iHUCOs highlights their parental signature. Several studies have highlighted the importance of interaction between colonic epithelium and the stromal compartment in both health and disease 72 . During UC, in addition to epithelial barrier dysfunction, mesenchymal and immune cell functions are disrupted. This disruption leads to the emergence of a population of activated mesenchymal cells that exacerbate disease severity by impairing epithelial cell maturation. Chronic inflammation also disrupts the cross-talk between the epithelium and immune cells, driving excessive local accumulation of immune cells and ultimately generating a proinflammatory mucosal state in colonic epithelium 28,73 . We used snRNA-Seq profiling to study the nuclear profile of iHUC and iHN organoids in the epithelial and stromal compartments (Figs. 6, 7).
UMAP analysis partitioned each of iHUCOs and iHNOs into 11 predicted clusters (Fig. 6a). A list of top markers in each The red and orange boxes highlight the parental fibroblasts and induced organoids-they cluster next to each other and share the highest level of the similarity compared with the other stages of development. b UMAP for the predicted clusters in normal and UC fibroblasts. Inset, pink denotes UC fibroblast clusters, while turquoise denotes normal colon fibroblast clusters. c Marker plot of the top expressed genes in the predicted clusters in normal and UC fibroblasts (Fig. 5b). Dot color and size correlate with mean expression level and fraction of cells per cluster expressing each gene. Cluster number is colored to match parental cluster. Pink number indicates UC fibroblast origin while turquoise indicates normal colon fibroblast clusters. d Highly enriched GO terms in UC vs. normal fibroblasts shown as REVIGO 50 scatterplots (P < 0.001). The terms extend along the X-axis based on similarity in the type of biological process (semantic space X); the difference in color highlights the variety of biological processes. Each circle represents a unique GO term. Circle size corresponds to the number of the genes associated with the GO term. e Violin plots show normalized gene expression levels, contrasting normal and UC fibroblasts, for each gene. A smoothed and rotated kernel density curve is plotted. The width of the violin plot represents the abundance of the number of cells at each corresponding vertical axis gene expression value. A reverse "T" shape is an indication of a universal lack of expression among all cells. Here, raw gene expression levels for each cell in normal and UC fibroblasts were extracted, scaled, and normalized. These key genes are responsible for the immune/inflammatory response and ECM remodeling in UC fibroblasts, highlighting the upregulation of these genes in UC compared with normal fibroblasts n = 30,894 cells for normal fibroblast and n = 67,654 cells for UC fibropblast.
cluster presented as a marker plot (Fig. 6b) highlights the overexpression of COL1A1, and COL3A1, in iHUCOs. The significance of these collagen genes lies in their relationship to the development of fibrosis, another hallmark of UC 74 .
To understand the differences in signaling pathways between the epithelial and stromal components of the iHUCOs and the iHNOs, we performed a Reactome pathway analysis on the DEGs between the epithelial and stromal compartments of iHUCOs and iHNOs. This analysis confirmed a tight and interactive relationship between both cell compartments in iHUCOs (Fig. 6c), in particular, the epithelial and stromal coordination in nervous system development and neural migration as highlighted by Netrin-1 signaling. Further, Reactome terms such as "anchoring fibril formation" and "laminin interactions" with very high enrichment scores in both compartments highlight the vital role of interaction between the epithelial compartment and ECM in iHUCO development. Terms such as IL-17 signaling with downstream regulation of NFKB and MAP kinases, Rho-GTPase signaling, and adherens junctions were exclusively enriched in iHUCO epithelium whereas the hallmarks of fibrosis including TGFβ signaling, excessive collagen production, and VEGF mediated permeability were unique to stroma (Fig. 6c).
Further, using established markers, annotation of iHN and IHUC organoids showed that induced organoids consist of 5 main cell types: epithelial, stromal, immune, neural and other cells (Supplementary Fig. 5a). The presence of neural cells has been recently reported in human intestinal organoids 75 . Here we show a significant increase in these cell classes, as well as immune and stromal cells in the UC compared to the normal organoids ( Supplementary Fig. 5a). Similar to parental fibroblasts, further annotation 66 of the cell subtypes in organoid epithelial and stromal compartments resulted in unique signatures in iHUCOs, correlating with the inflammatory and more immunogenic nature of UC ( Supplementary Fig. 5b, Fig. 7a, b).
In iHUCO epithelium, we identified the presence of immature signatures (vs. mature in iHNO) in enterocytes and goblet cells along with the unique secretory signature of enteroendocrine cells 45,76 (Fig. 7b). These data are consistent with the reparative nature of UC and also with our in vivo observation of an increased number of chromogranin+ (enteroendocrine cell marker) cells in the crypts formed by iHUCOs but not iHNOs (Fig. 3h, i). Moreover, it is known that in the reparative state, adult stem cell markers such as Lgr5+ cells are repressed, and distinct intestinal stem cells act to replenish lost stem cells and contribute to crypt  Fig. 5d) whereas genes such as SMOC2 78 , which promotes matrix assembly and cell adhesiveness, were overexpressed. Our in situ hybridization for Lgr5 confirmed that this stem cell marker was suppressed in iHUCOs ( Supplementary Fig. 5e, f).
Wound healing is a complex coordinated process, orchestrated by tight and dynamic interactions between different cell types and ECM components to produce successful repair. In the case of chronic wounds, this coordination is compromised and usually results in fibrosis (a hallmark of IBD) associated with excessive production of ECM components including collagen type-I 28 . Consistent with the significant signature of fibrogenic and POSTN+ subtypes in iHUCOs, IHC staining for both collagen type-I and periostin confirmed that these proteins are overexpressed in iHUCO but not iHNO ECM ( Supplementary Fig. 5g, h).
Despite the limited number of immune cells (mostly cycling B) in iHNOs, we identified extended cycling B, T, and natural killer (NK) cell sub-populations in the immune compartment of iHUCOs. We also found dendritic cells (like the UC fibroblast signature, Fig. 7b). Although the tight relationship between intestinal epithelial cells and dendritic cells is known, there is no report of direct interaction of dendritic cells with intestinal organoids in a co-culture system. However, a recent study examined the direct effect of dendritic cells on gastric organoids, confirming a steady state migration of these cells to the basolateral surface of gastric spheroids that is induced by the secretion of CXCL1 and CXCL8 chemotactic factors 79 . This study and our IHC data showing that that the immune-regulating MHC Class I, HLA-A 70 , and Limch-1are significantly upregulated in iHUCOs but not iHNOs. (Fig. 7c-f) highlight the role of epithelial stromal cross-talk in adaptive and innate immune responses where dysregulation results in a pro-inflammatory state in the GI tract.
To further explore our observations, we isolated the mesenchymal compartment of iHN and iHUC organoids, verified by IF staining for vimentin ( Supplementary Fig. 5i) and α-SMA (IF not shown) and the lack of CK19 (Supplementary Fig. 5i). We then performed a cytokine array and compared the secretome of isolated mesenchyme to parental fibroblasts (Fig. 7g)  The unpaired, non-parametric two-sided Mann-Whitney U test was used to test for the difference in percentages between iHNO and iHUCO. Scale bar = 40 μm. n = 5 iHNOs (blue) and n = 6 iHUCOs (orange). g Representative cytokine arrays of normal and UC parental fibroblasts and organoidderived mesenchyme with the quantification of three chemokines are increased in UC vs. normal: GRO-α (green), GRO − α + β + γ (blue), and CXCL8 (orange). The median value is indicated in the center of the box plot. The whiskers above and below the plot represent 1 standard deviation (SD) above and below the mean of the data, respectively. Source data are provided as a Source Data file.
In conclusion, here we show that the colitic signature of UC fibroblasts results in a unique developmental signature in iHUC vs. iHN organoids. This highlights that the microenvironment, including stromal cells and ECM proteins, is crucial in iHUCO development. Both those UC parental fibroblasts and iHUCOderived mesenchyme from the most inflamed tissues showed increased expression of CXCL8 and CXCL1 chemokine expression.
Repertaxin attenuates in vitro the progression of the colitic phenotype in iHUCOs. The upregulation of CXCL8/CXCR1 proinflammatory interaction in iHUCOs and their parental fibroblasts (Figs. 4, 5) led us to study the effect of repertaxin, a small molecule inhibitor of the CXCL8 receptor, CXCR1, on organoid development. In brief, we treated both UC and normal spheroids with repertaxin for 21 days during their development into organoids and compared the phenotypic characteristics to vehicletreated (medium lacking repertaxin) control organoids. The results confirmed significantly lower expression of CXCR1 and CXCL8 and lower co-expression in the epithelium and mesenchyme of treated iHUCOs compared to controls (Fig. 8a-d).
To functionally test the effect of repertaxin treatment on the epithelial barrier permeability, we used the microinjection technique 20 , which allows real-time analysis of changes in permeability. Repertaxin did not significantly affect the epithelial permeability of iHNOs but did decrease iHUCO epithelial permeability ( Supplementary Fig. 6h).
Therefore, CXCL8 receptor inhibition by repertaxin significantly attenuated development of the iHUCO colitic phenotype in in vitro. Repertaxin significantly affected the size and epithelial structure of iHUCOs, and it also modified the expression pattern of proteins regulating the adherens junction complex such that the pattern more closely resembled that of iHNOs. We functionally validated these observations using the microinjection technique in real-time to show that although repertaxin treatment did not significantly affect epithelial barrier permeability in iHNOs, iHUCO demonstrated diminished epithelial permeability. A cartoon illustrating the effect of repertaxin on iHUCO epithelial intercellular junctions is shown in Fig. 8u.
Repertaxin attenuates in vivo the progression of the colitic phenotype in iHUCOs. To test the significance of repertaxin treatment in vivo, we studied the effect of repertaxin on the developmental progression of spheroids to organoids implanted subcutaneously in the dorsal flank of NSG mice (Fig. 9a). In brief, we encapsulated the spheroids in TS-HA hydrogel and implanted the bead subcutaneously. Mice were then treated daily for 21 days with either 20 mg/kg repertaxin or PBS (control). The overall volume after 21 days (measured twice weekly with calipers) was significantly greater in the control than in the repertaxin-treated (20 mg/kg) groups ( Supplementary Fig. 6i). We confirmed this observation by H&E staining of the collected beads after 21 days (Fig. 9f-i). Consistent with our in vitro observation, repertaxin treatment affected iHUCO size and epithelial structure (p < 0.0001) whereas it had no significant effect on iHNO structure (Fig. 9h, i). Repertaxin also reduced the epithelial proliferation of iHUCOs as determined by Ki67 staining. However, it had no significant effect on iHNO proliferation ( Supplementary  Fig. 6j-m).
We also analyzed the beads for chemokine expression. Consistent with our in vitro data, CXCR1 and CXCL8 expression were significantly lower in both UC and normal organoids in mice treated with repertaxin than control organoids (Fig. 9b-e).
In addition, repertaxin treatment increased the membrane expression of both β-catenin and E-cadherin protein in iHNOs (Fig. 9k1, o1) whereas in iHUCOs, it resulted in greater cytoplasmic and nuclear expression of both (β-catenin, Fig. 9k2, m; and E-cadherin, Fig. 9o2, q). Further, repertaxin treatment had no significant effect on RhoA expression in iHNOs but resulted in significantly lower membrane and cytoplasmic expression of in iHUCOs (Fig. 9r-u). Claudin-1 expression in normal organoids also was not affected by repertaxin treatment, but it was significantly greater in repertaxin-treated iHUCOs, indicating greater regulation of tight junctions in iHUCO epithelium ( Supplementary Fig. 6n-q). These data demonstrate that repertaxin treatment attenuated the colitic phenotype of iHUCOs not only in vitro but also in vivo in terms of epithelial structure, size, and changes of the epithelial intercellular junction.

Discussion
The nature of UC is complex as it relies on interaction of different cellular compartments, and this has made it challenging to study the pathogenesis of colitis. Current therapeutic approaches in IBD mainly focus on the suppression of immune responses 81 ; however, these therapies often fail, suggesting that the role of the epithelial and mesenchymal components in the development and progression of UC is not appreciated.
All models have limitations. Common in vivo rodent models or epithelial organoids as an in vitro model have advantages but do not adequately recapitulate the complexity and etiology of clinical UC 82 . In particular, to date, epithelial organoids have focused solely on the epithelium and do not include the role of microenvironment in colitis [83][84][85] .
As an alternative, our induced in vitro model of human UC (iHUCO) includes both functional epithelial and stromal compartments and retains the complexity of the colitic phenotype of the tissue of origin, such as a compromised epithelial barrier, indicating a significant advance in UC modeling. Currently, in attempts to understand the mechanism by which inflammatory stroma can damage the integrity of the epithelial barrier, many studies try to develop a direct co-culture of intestinal epithelial cells, mesenchyme, and T cells. However, several layers of challenges, for example, histocompatibility issues, result in reduced mimicry of the in vivo dynamic between various cell types. Here, we introduce a human in vitro model that will allow us to identify and understand potential mechanisms behind the direct interaction of intestinal epithelial cells and stromal cells in a patientspecific manner. In particular, we identify morphologic transition, dysfunctional intercellular junctions and progression to early stages of fibrosis. High resolution transcriptional analysis of the IHUCOs and iHNOs demonstrated another level of discrimination. Increased levels of different transcripts confirmed findings of others including the presence of OSMR, PDPN, POSTN+, IGFBP5 64,65,68 which correlate with refractory disease and fibrosis. Further, key signatures demonstrating discrete cellular infiltrates were identified that distinguished the iHUCOs from the iHNOs. These signatures included epithelial enteroendocrine cells, noted to have increased prevalence in UC 66 and not surprising with respect to our histological findings demonstrating increased numbers of cells as marked by chromogranin A. Another subpopulation within the stroma included fibrogenic and WNT2/5+ fibroblasts, which suggest the presence of inflammation-associated fibroblasts 66 . Further, we identified an immune population that was entirely unique to the iHUCOs. Consistent with others 86 , B cells formed a dominant proportion of the immune subpopulation identified within the iHUCO cohort. Indeed, B cells have been the target of a small clinical trial in the past, which may not have been successful because the stromal subpopulation was not eliminated 87 . Nevertheless, these findings support the potential of our model to provide a comprehensive human model to manipulate immune and inflammatory regulators, naturally present in iHUCOs, to identify repertaxin. e, f Representative H&E of iHNO and iHUCO with and without repertaxin treatment. g, h Summarized average diameter and percentage of epithelial structure type in organoids after 21 days, with and without repertaxin treatment. Violin plots in g compare iHNO control (ctrl) vs. 20 μm repertaxin treatment. In panel h, the epithelial structure is compared between iHNO and iHUCO in the presence or absence of 20 μm repertaxin. The median value is indicated in the center of the box plot. The whiskers above and below the plot represent 1 standard deviation (SD) above and below the mean of the data, respectively. i, j, m, n Representative IHC for β-catenin and E-cadherin in control and repertaxin-treated iHNOs and iHUCOs revealing altered cellular localization of the proteins after repertaxin treatment. k, l, o, p percentage distribution of the cells positive for β-catenin and E-cadherin according to the sub-cellular compartment; plasma membrane only (Mem) or membrane extended to cytoplasm and nucleus (Cyt + Nuc) with or without repertaxin treatment of the organoids. q, r Representative IHC for RhoA in the organoids shows the changes in cellular localization based on repertaxin treatment. s, t Violin plots for the cells expressing RhoA in control and repertaxin-treated organoids according to subcellular compartment: plasma membrane only (Mem) or cytoplasm (Cyt). u Schematic representation of the mechanism underlying the changes in epithelial intercellular junction in iHUCOs with and without repertaxin treatment. IHC Scale bar = 40 μm. IF scale bar = 25 μm. n = 5 iHNOs and n = 6 iHUCOs, respectively. For all comparisons, the unpaired, non-parametric, two-sided Mann-Whitney U test was used to test the difference between percentages of enumerated observations. Source data are provided as a Source Data file. therapies to recover functioning epithelial cells in patients suffering from UC. With such a model, strategies for personalized medicine can be explored, and the mechanisms underlying the pathophysiology of human UC may be investigated in vitro.
Although the autocrine and paracrine functions of the CXCL8 chemokine and its receptor CXCR1 in the development of several types of cancer, including colorectal cancer, have been extensively studied 61,88 the downstream effect of this inflammatory interaction in UC development and progression remains unclear. Using iHUCOs, we provide evidence that shows overexpression of CXCL8/CXCR1 in UC results in a dysregulated adherens junction pattern in epithelial cells. Notably, CXCL8 lacks a murine homolog, which highlights the gap in the murine-based models and the functional importance of our model in identifying the role of CXCL8 receptor-mediated signaling in UC development and progression 89 .
We also demonstrate the functionality of the model via response to chemical perturbation by repertaxin, a CXCR1 receptor small molecule non-competitive inhibitor. Exposure of both in vitro and in vivo organoid cultures to repertaxin reduced the expression of CXCL8 ligand and CXCR1 receptor and attenuated several aspects of the colitic phenotype, including a leaky epithelial barrier, suggesting that the pro-inflammatory interaction of CXCR1-CXCL8 compromises the epithelial barrier. Our inducible organoid system provides the advantage of a patient-derived model to study the epithelial-mesenchymal and intestinal microenvironment interactions of human UC compared with the current murine models. Importantly, our model preserved the individual patient genetic background and thus, in the future, may be used to tailor IBD treatment 90 for regenerating colon and healing the damaged mucosa.
Finally, we demonstrated overexpression of CXCL8 and its receptor in UC patient tissues, validating the significance of our functional studies. Thus, using repertaxin to block this interaction may be a promising therapeutic strategy to diminish the chronic inflammatory symptoms of ulcerative colitis.   in the epithelium and mesenchyme of normal and UC organoids in the presence or absence of repertaxin. f-g Representative H&E of control and repertaxin-treated organoids collected at 21 days. h, i Average organoid diameter and percentage of each epithelial structure type in the presence or absence of repertaxin. The unpaired, non-parametric, two-sided Mann-Whitney U test was used to test whether the percentage of cells expressing the specific morphology differed. The median value is indicated in the center of the box plot. The whiskers above and below the plot represent 1 standard deviation (SD) above and below the mean of the data, respectively. j, k, n, o Representative IHC for β-catenin and E-cadherin in control and repertaxintreated organoids revealing the altered cellular localization of both proteins in iHN and iHUC organoids by repertaxin treatment. l, m, p, q Distribution of the epithelial cells positive for β-catenin and E-cadherin with or without repertaxin treatment. Results reported for the plasma membrane only (Mem) or membrane extended to cytoplasm and nucleus (Cyt + Nuc). r, s Representative IHC for RhoA in normal and UC organoids, confirming the effect of repertaxin on RhoA sub-cellular localization. t, u Percentages of cells expressing RhoA in control and repertaxin-treated organoids according to cellular compartment: plasma membrane only (Mem) or cytoplasm (Cyt). The unpaired, non-parametric, two-sided Mann-Whitney U test was used to test for the differences between all enumerated observations. IHC scale bar = 40 μm. IF scale bar = 25 μm. n = 3 each. Source data are provided as a Source Data file.

Methods
Patient population and tissue collection. Tissue specimens were obtained from the distal colon of patients with UC who underwent biopsy or surgery for resection of UC at the Cleveland Clinic. None had strictures or sclerosing cholangitis, which are known risk factors for colitis-associated cancer. Patient demographics and clinical presentations are summarized in Supplementary Table 1. Patients with UC had disease for at least 2 years, therefore classified as chronic colitis. As UC patients often have pan-colitis, manifested as inflammation along the entire tract of the large bowel, normal tissue specimens were obtained from areas at least 10 cm away from gross tumors of patients presenting for resection of sporadic colon cancer. Pathology of retrieved tissues was confirmed by the Cleveland Clinic Pathology Department and approved under IRB 13-1159.
Fibroblast isolation and characterization. To isolate patient-derived fibroblasts 23 , a segment of colonic tissue no larger than 0.5 cm 3 was retrieved from the operating room, sterilized with antifungal-antimicrobial solutions for 15 min × 3, then minced in 500-1000 μL DMEM containing collagenase Type IA (1 mg/mL) in a round bottom 2 mL microfuge tube until the tissue could pass through a 1000 μL pipette tip. The minced tissue was transferred to a 50 mL conical tube with an additional 9 mL collagenase-containing media, then placed in a water bath at 37 o C for 30 min. Collagenase activity was terminated by adding serum-containing medium to the minced tissue. The suspension was pelleted, re-suspended in 5 mL serum-containing medium and filtered through a 40-µm strainer. The filtrate was cultured on tissue culture plates in DMEM with 10% FBS and 1% pen/strep. The human colon fibroblast cell line CRL1541 was obtained from the American Type Culture Collection (ATCC). The identity of the isolated fibroblasts was confirmed by visual inspection for spindle-shaped cells, positive immunofluorescence (IF) staining for alpha-smooth muscle actin (α-SMA), and the absence of cytokeratin 19 (CK19) staining.
Short tandem repeat analysis and mycoplasma testing. A mycoplasma assay and short tandem repeat (STR) analysis (SupplementaryInformation Table 2) were conducted to verify the absence of mycoplasma and the unique origin of each fibroblast isolate, respectively. Once established, primary isolates of fibroblasts were also subjected to STR analysis to establish the unique genetic identity (Duke University DNA Analysis Facility and Genetica, Burlington, NC). Prior to experimental use, another STR analysis was performed to confirm the identity. Organoids derived from parental fibroblasts were also subjected to STR analysis, and the consistency in origin was validated (Supplementary Table 2).
Reprogramming and iPSC characterization. The CytoTune-iPS reprogramming kit (Supplementary Data File 1) was used according to the manufacturer's instructions to reprogram normal and UC fibroblasts (Supplementary Table 1) to induced pluripotent stem cells (iPSCs). The CRL1541 cell line (ATCC ® CRL1541, CCD112CoN) was included to ensure that the reprogramming protocol and colonic organoid development were reproducible by other research groups using a commercially available cell line. In addition to the visual changes in the morphology of iPSCs and the parental fibroblasts, reprogramming was confirmed by using 1) Immunofluorescent staining for the known markers Tra-1-60 and Oct-4 to delineate pluripotency in human pluripotent stem cells, and 2) the human pluripotent stem cell identification kit (R&D Systems, Supplementary Data File 1) to validate the pluripotent capacity of iPSCs by trilaminar potentiality.
Generation and patterning of induced human colonic organoids. In brief, iPSCs were differentiated to definitive endoderm (DE) by adding DE induction medium (RPMI 1640, 2 mM L-glutamine, penicillin-streptomycin and 100 ng/mL 10 Experimental design. To confirm the reproducibility of the organoid generation and their phenotype, the previously described protocol of McCracken, K. W. et al. 11 was repeated by different investigators in more than 30 events, and the colitic phenotype in iHUCOs was applicable to the majority of organoids (~80%) for UC. Briefly, following reprogramming from colitic fibroblasts to pluripotency, confluent iPSCs were then subjected to directed differentiation to definitive endoderm using Activin A, and increasing percentages of defined fetal bovine serum over 3 days (none, 0.2%, 2%, in DE induction media). On day 4, hindgut induction media was added, which includes Chiron and FGF4. On day 7, buds were pooled, harvested, and plated into 25 μL Matrigel beads. Heterogeneity between biological replicates matched observed phenotypes in the primary tissues. That is, when the chronic colitic tissue histology was less pronounced, so too was the organoid morphology. To avoid variations in phenotype, all generated organoids for the study were initiated from passage 0 originating from the iPSCs. To validate the consistency in results, all experiments were repeated on at least 3 separate occasions (independent experiments) and conducted on at least 5 independent biological replicates. A minimum number of 100 organoids per line were generated in each series of experiments.
Immunostaining and microscopy. Organoids were released from Matrigel beads through gentle washing in PBS followed by immediate fixation with 4% paraformaldehyde (PFA). The fixed organoids were concentrated into a 1.5% low melting point agarose plug prior to being embedded in paraffin. Standard conditions for antigen retrieval, including tris/borate/EDTA buffer (Discovery CC1, 950-500; Ventana) on a Discovery ULTRA automated stainer (Ventana Medical System Inc.), were implemented. Chamber slides with acetone-fixed organoids were used to perform immunohistochemistry (IHC)/immunofluorescence (IF) with the antibodies listed below. Images were captured on a Leica Microsystems microscope (version 4.3.0, Supplementary Data File 1). Cells expressing these proteins were counted manually (>500 nuclei/section/stain). IF on intact organoids was completed in suspension. Briefly, organoids were fixed for 1-3 h in 4% PFA, treated with blocking serum (5% serum in 1× PBS plus 0.5% Triton-X100) for 1 h prior to incubation with the primary antibodies overnight. The next day, the organoids were washed in PBS-0.5% Triton-X100 ×3 and exposed to the secondary antibody for 2 h. To counterstain for nuclei, organoids were suspended in a drop of VEC-TASHIELD mounting medium with DAPI and loaded on the coverwell imaging chamber for confocal microscopy. Images were captured on a Leica Microsystems confocal microscope (DM16000) using software version 2.7.3.9723, magnification ×40, and oil immersion. IHC on paraffin-embedded (4% PFA) fixed or snap-frozen sections of the primary tissues was performed with the antibodies listed in the Key Resources Table. Antibodies. The following primary antibodies were used for staining in the study: anti RNA isolation (Bulk RNA-Seq). Organoids were harvested after 21 days and washed with cold PBS. Total RNA was extracted using miRNeasy kits (Qiagen). RNA purity and integrity was assessed using a Bioanalyzer (Agilent). All samples had RNA integrity >9.5. Total RNA was processed for next-generation sequencing using TruSeq Total RNA kits (Illumina). Total RNA of each sample was depleted of ribosomal RNA using biotinylated oligomers, combined with Ribo-Zero rRNA removal beads and fragmentation of RNA using divalent cations at elevated temperatures. Cleaved RNA fragments were copied into double-stranded cDNA using (sequentially) reverse transcriptase, random primers, DNA polymerase, and RNase H. After ligation of the sequencing adapters, the cDNA products were purified.
Bulk RNA-Seq processing. RNA library construction and RNA sequencing was performed by the Case Western Reserve University School of Medicine Genomics Core, using an Illumina HiSeq2500 platform. Paired end 100-bp sequencing reads generated from this platform were assessed for quality control and trimming using FastQC, version 0.11.5 91 and multiqc, version 0.7 92 . The QC analysis identified no major issues. After removing the Illumina universal adapter sequence by Cutadapt 93 , we used the software package STAR version 2.5.3a 94 and gene annotation from Gencode 95 to align QC-ed reads to the reference transcriptome HG19. The total numbers of mapped reads to genes (read count) were counted using feature counts 96 , and log 2 (RPKM) values were calculated.
Unsupervised analyses including principal component analysis (PCA) and hierarchical clustering (HC) were performed to visualize the similarities and the distinctions among all the groups. In brief, the first three principle components were calculated using log 2 (RPKM) values and 'prcomp' function in R and were plotted into a 3-dimensional space using the 'rgl' package. The log 2 (RPKM) values were also subjected to HC analysis, using the complete agglomerative method and 'hclust' function in R. Next, the read count values and DEseq2 version 3.8 were used to identify differentially expressed genes (DEGs) and the gene signatures in developmental progression of both UC and normal groups 97 . To improve the reliability and accuracy of DEGs,-genes with minimal expression were filtered to ascertain the noise/error removal (genes less than 10 mapped reads in more than half of the total samples). Student's t-test and Benjamini-Hochberg False Discovery Rate (FDR) tests were performed for each of the DEGs in comparison groups.
The raw DEGs for iHUCO vs. spheroid were also subjected to Gene Set Enrichment Analysis (GSEA). Each gene of the DEGs was scored as 1/ (adjusted Pvalue * sgn (log 2 (fold change)) where "sgn" stands for a sign function. The adjusted P-value and fold change values were extracted from the DEG analysis results. The scored genes were sorted in descending order and converted to.rnk file as an input for GSEA preranked 98 . KEGG 99 , Reactome 100 , BioCarta 101 , and Biological Process in Gene Ontology 102 were used as gene set databases. The summary of the GSEA report for iHUCOs and spheroids was applied to create an enrichment map in Cytoscape 49 . To identify the biological processes enriched in iHUCOs vs. spheroids, we applied Analysis and Visualization (GOrilla), and Reduce and Visualize Gene Ontology (REVIGO). Specifically, the single ranked list of DEGs in iHUCO (P ≤ 0.001) was applied to GOrilla, identifying the enriched GO terms in biological process ontology. The output was shown as a 2-dimensional graph in REVIGO for the terms of interest (Supplementary Data File 1).
All heatmaps were generated using RPKM values in R, version 3.5.1 and the R packages gplots, RcolorBrewer, and pheatmap.
Nuclear isolation (snRNA-Seq). To isolate nuclei, we followed the protocol suggested by Habib et al. 103 for both fibroblasts and induced organoids. Cultured fibroblasts at lower passages (<10) or 30 days old induced organoids were collected, washed, suspended in ice-cold EZ PREP buffer (see Key Resources Table), and homogenized using a glass dounce tissue grinder (×20 for each loose and tight pestle) and incubated on ice for 5 min, with an additional 2 mL of ice-cold EZ PREP. Next, nuclei were centrifuged at 500 × g for 5 min at 4°C, washed with icecold EZ PREP and filtered through a 40-μm cell strainer. After centrifugation, the nuclei were washed in 4 mL nuclei suspension buffer (NSB; consisting of 1× PBS, 0.01% BSA and 0.1% RNase inhibitor) and filtered through a 20-μm cell strainer, and then we proceeded with the 10x Genomics pipeline for sequencing.
SnRNA-seq data processing and analysis. Using Cell Ranger 3.1.0 from 10× Genomics, raw snRNA-seq reads were processed for alignment and generating feature-barcode gene expression count matrices. A GRCh38 pre-mRNA reference genome was built and adopted to map the sequence reads. Each biological sample went through the same procedure of alignment in parallel. After individual alignment, samples were aggregated using the Cell Ranger aggr function to create the integrated data and facilitate downstream comparison. We retained genes that had expression in at least 10 nuclei and nuclei that have expression of at least 200 genes, using Seurat 3.1.2 package 104 for analyses. To ensure high quality, we filtered out the nuclei expressing high level (>10%) of genes that mapped to the mitochondrial genome. After filtering, data normalization and scaling were performed using Monocle 3 package 105,106 in R programming language, version 3.6.2.
After preprocessing and data quality control, a total of 65,963 and 67,856 nuclei were retained in normal and UC fibroblasts, respectively. Similarly, 30,819 and 44,185 nuclei were retained in iHNOs and iHUCOs. To compare UC with normal fibroblasts and organoids, we extracted the nuclei based on their type (diseased vs. normal) out of the integrated data and followed the normalization and data scaling steps before downstream analysis. In each group, we kept the first 50 principal components for data normalization using Principal Components Analysis (PCA) for Monocle 3 to for calculations. Next, the uniform Manifold Approximation and Projection (UMAP) technique was used for dimensionality reduction and visualization. Post-alignment data were read into Seurat for data normalization and scaling. Using the Wilcoxon Rank Sum test integrated in Seurat, we identified the DEGs comparing normal and UC fibroblasts. DEGs were defined as those genes that showed statistical significance and at least 0.25-fold change in log scale. Adjustment was made for multiple testing, and the FDR was controlled. Similarly, we identified DEGs comparing iHUCOs with iHNOs (see Supplemental information). Once we clustered the cells, we identified the marker genes for each cluster, using the integrated regression framework in Monocle 3 for normal and UC fibroblasts or induced organoids. Based on the list of marker genes reported by Smillie et al. 66 we annotated the cells using the SCINA algorithm 107 and each cell cluster was annotated by its dominating cell type. To identify the biological processes enriched in UC vs. normal fibroblasts, we applied the list of DEGs to GOrilla and REVIGO. Specifically, the single ranked list of DEGs in iHUCO (pvalue <= 0.001) was applied to GOrilla, identifying the enriched GO terms in biological process ontology. The output was shown as a 2-dimensional graph in REVIGO for the terms of interest. We then performed the Reactome pathway analysis on the iHUCO epithelial and stromal compartments. The list of DEGs in each compartment were applied to Gene Ontology Resource and processed for Reactome pathway analysis.
Isolation of organoid-derived mesenchyme. The 21-day normal and UC organoids were dissociated in collagenase type IA (2 mg/mL) at 37°C for 20 min. Collagenase activity was terminated by adding serum-containing medium to the mixture. The pellet was re-suspended in serum-containing medium and filtered through a 40-µm strainer. The resulting suspension was plated in DMEM with 10% FBS to isolate organoid-derived mesenchyme. Visual inspection for the spindle shape, and positive immunofluorescence for the mesenchymal marker, vimentin, and negative stain for the epithelial marker CK19 were used to confirm the mesenchymal identity of the cells. Mesenchymal isolates were propagated and used for the further analysis.
Human cytokine array C5. To perform the cytokine arrays from the conditioned media of parental fibroblasts and organoid-derived mesenchyme (N = 3 each), 10 5 cells were cultured in 1 mL of DMEM base medium for 24 h in the absence of serum. The conditioned media were harvested and subjected to quantification using the human Cytokine Array C5 (RayBiotech) according to the manufacturer's instructions.

Measurement of epithelial barrier permeability
Microinjection. Microinjections were performed using a modified protocol from Leslie et al. 108 and described in 20 . Briefly, organoids were injected using thin-wall glass capillaries and a P-30 micropipette puller (Sutter Instruments, Novato CA). Pulled microcapillaries were mounted on a Xenoworks micropipette holder with analog tubing attached to a 10-mL glass syringe filled with sterile mineral oil. Fine control of the micropipette was achieved using a micromanipulator (Narishge International Inc.) and microinjection was completed under ×1-2 magnification on an SX61 stereo dissecting scope (Olympus).
FITC-dextran permeability assay. For epithelial permeability assays, organoids were microinjected with 4 kDa FITC-dextran suspended in PBS at a concentration of 2 mg/mL as described previously 20 using the microinjection system detailed above. Images were collected at 30 min intervals at ×4 magnification on an Olympus IX71 epifluorescent microscope using a CellSens live cell imaging system (Olympus Corp). Cultures were maintained in 5% CO 2 throughout the imaging time course. The relative fluorescent signal was quantified from the collected images using ImageJ and custom macros following the protocol previously described 20 .
In vitro repertaxin experiments. Matrigel encapsulated iHUCOs vs. iHNOs were treated with 20 μM repertaxin vs. control (PBS) for 21 days. Beads were harvested, Matrigel dissolved, and placed into agarose plugs. The agarose plugs were then embedded into paraffin blocks, stained with various antibodies of interest, and enumerated. A minimum of 500 cells/condition/type were enumerated. Assays were completed in at least triplicate using N ≥ 3.
Animals. For the subcutaneous implantation and omental transplantation experiments, 4-week-old to 6-week-old and 12-week-old to 27-week-old immunedeficient NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG) mice were used, respectively. All animal experiments were performed with the approval of the Cleveland Clinic IACUC (Protocol 16-1666). All mice were housed in the animal facility at the Cleveland Clinic Lerner Research Institute in accordance with NIH Guidelines for the Care and Use of Laboratory Animals. Animals were maintained on a 12 h lightdark cycle with access to water and standard chow ad libitum. The animals were identified by numbered cages and by ear punches. Ambient temperature was maintained at 71°F, with 45-50% humidity.
Omental transplantation of organoids. The 21-day normal and UC organoids (3 independent biological replicates) were removed from Matrigel and encapsulated in tyramine-based hyaluronan (TS-HA) hydrogel (a kind gift from Dr. A. Calabro) for a total of 10 beads (20 organoids per bead) per group (technical replicates). The hydrogel was prepared at 4 mg/mL in PBS and was activated by horseradish peroxidase at a ratio of 500:1. Equal numbers of spheroids (between control and treatment groups) were suspended in 25 µL activated hydrogel and cross-linked by exposure to 0.03% hydrogen peroxide (1 µL). Cross-linked hydrogel beads were transferred to 24-well plates and incubated in organoid medium for 2 h, then the medium was replaced with fresh organoid medium to assure removal of residual chemicals. We adapted a transplantation protocol 43 to xenograft each seeded bead surgically within the omentum of a host NSG mouse. The omentum was exposed from the mouse peritoneum, wrapped around the seeded scaffold, and tied with a 6-0 prolene purse string suture to securely contain the bead. The wrapped unit was then reduced back into the peritoneal cavity and the surgical incision was sutured closed. Post-operative treatment included buprenex, 0.1 mg.kg twice daily for 48 h. After 90 days, host mice were euthanized, beads were harvested and fixed in 4% PFA prior to embedding in paraffin blocks for histologic evaluation.
Subcutaneous implantation of spheroids. TS-HA hydrogel (a kind gift from Dr. A. Calabro) was used to encapsulate the intestinal spheroids preceding the subcutaneous implantation 109 . The hydrogel beads were prepared as above for the omental encapsulation experiments. The day after, the beads were subcutaneously implanted in the right dorsal flank of NSG mice 23 . The growth was measured twice per week using calipers. On day 2 post-implantation, repertaxin, 20 mg/kg vs. control diluent (PBS) was injected in rotating sites near the site of implantation for the duration of the experiments. Tumor volumes were calculated using the formula: volume = width 2 × length (mm, length as the longer measurement). The beads were harvested 21 days after implantation for further characterization.
For these experiments, blinding occurred at two levels: the investigator implanting the beads was blinded as to the contents of the beads. The investigator doing the treatments and the measurements was also blinded. The investigator performing the enumeration was likewise blinded.
Quantification and statistical analysis. The data presented as individual points or violin plots represent the percentage of positive cells for the target protein in n = 5 for normal and n = 6 for UC (biological replicates). A minimum of 500 nuclei per group in epithelia and 100 nuclei per group in mesenchyme (if applicable) were evaluated. Each independent biological sample was comprised 50-70 organoids pooled together per experiment for further analyses. Because the data were not normally distributed, an unpaired, non-parametric Mann-Whitney t test in GraphPad Prism 8.0 was performed for comparison of two groups. For the microinjection assay, unless otherwise indicated in the figure legends, differences between experimental groups or conditions were evaluated using a non-parametric Wilcoxon rank-sum test. All statistical analyses in the microinjection assay were conducted using R, version 3.5.3 (2019-03-11), and the plots were generated using the R package ggplot2.
Statistical power, single nuclear analyses. Based on the sequencing depth and the abundance of cells in this snRNA-seq, sufficient statistical power was achieved to detect differentially expressed genes and cell clusters. We had 5 biological samples for the normal condition and 6 biological samples for the colitis condition. These yielded a total of~174,000 cells for all organoids combined. On average, we observed 23,871 number of reads per cell before normalization. We conducted simulation studies using the state-of-the art power assessment tool POWSC 110 , which modeled expression level using a mixture of zero-inflated Poisson (ZIP) and log-normal Poisson (LNP) distributions. For small group comparisons with only 2k cells per sample, the power ranges from 0.788 to 0.985, with a marginal power of 0.922. The range of power comprehensively reflected the different strata of average reads. In all our single-cell analyses, we have achieved higher power, as the cell number is already much higher than in the simulation setting.