Society: DDW

Background: Perianal fistulae are highly morbid complications of Crohn’s disease (CD) with a cumulative incidence of 20-25%; despite demonstrated efficacy of infliximab treatment, estimates of complete resolution are only 20-30%. Ultimately, 20% of patients will require proctectomy. Across adult and pediatric populations, black CD patients are more likely to have colon-only CD, perianal fistula (OR 1.7-2.47, RR: 2.63) and progression to complex disease.
Methods: We performed single cell RNA sequencing (scRNAseq) of paired rectum and colon biopsies taken from equal African and European ancestry patients (n=14, 130,961 cells) with biopsies taken from two patients undergoing complete proctectomy, allowing for direct fistulous tract biopsies; single-nucleus RNAseq + ATAC (10X Multiome) was executed on a subset of n=6 patients (99,085 nuclei). A parallel cohort of healthy donors (n=8) and CD patients (n=11) with equal ancestry representation was recruited for peripheral blood mononuclear cell studies. We isolated CD14+ monocytes and differentiated sans microbial stimuli for 14 days; image analysis, quantitative Reverse Transcription-PCR, and Luminex immunoassays were performed.
Results: Myeloid and stromal cells were subset and re-clustered with the same cell types from the entire cohort of perianal fistula patients; n=20,704 cells were distributed among n=14 granular clusters. CHI3L1 was a top differentially expressed gene (DEG) (fold change=3.51, p-adj<1E-300) between stromal cells of fistula vs. colonic and rectal tissue origin. CHI3L1hi fibroblasts from fistulous tracts disproportionately expressed a destructive gene module (e.g. matrix metalloproteinases) and markers of macrophage activation. After two weeks’ differentiation in culture, isolated CD14+ monocytes differentially assumed a spindle-like morphology (p<0.001), expressed CHI3L1 (p=0.0145), and secreted CHI3L1 (p<0.001) in CD patients of African ancestry compared to European ancestry. Single cell gene regulatory network analyses focused on IBD-associated transcription factors confirmed cell-specific transcription factor enrichment (e.g. GATA2 in mast cells) and a broad array of predicted transcription factors in fibroblasts, including AP-1 (FOS, JUN). Multiome ATAC analysis of the CHI3L1 promoter revealed n=33 footprints bound in peaks from fibroblasts, over 60% of which were transcription factors belonging to the AP-1 family.
Conclusions: We present the first scRNAseq analysis of direct ex-vivo fistula tracts and identify CHI3L1 expressing fibroblasts that expressed common myeloid and known fibroblast destructive gene modules. Epigenetic control of inflammatory memory by AP-1 (JUN, FOS; PMID: 34320411) in a chronic inflammation context represents a promising mechanism of sustained injury in fistula by aberrantly differentiated innate immune cells.

<b>Figure 1.</b> Experimental design and sample sizes of parallel single cell direct <i>ex-vivo</i> (a) and <i>in vitro</i> (b) studies.

Figure 1. Experimental design and sample sizes of parallel single cell direct ex-vivo (a) and in vitro (b) studies.

<b>Figure 2.</b> Multimodal analyses implicate <i>CHI3L1</i> and AP-1 epigenetic control in fibroblasts.<br /> (a) Hematoxylin and eosin stain of a cross section of a fistula tract sampled for this study. (b) Uniform manifold approximation and projection (UMAP) of subclustered myeloid and stromal cells from n=14 patients. (c) Enrichment of functional fibroblast gene modules across myeloid and stromal cell types. (d) Expression of <i>CHI3L1</i> relative to housekeeping after 14 days’ culture in African and European ancestry (AA, EA) CD patients and controls. (e) CHI3L1 protein measured in day 14 CD14+ cell culture supernatants from healthy controls, CD patients, and isolated fistula single cells cultured under the same conditions. (f) Bias corrected peaks in the promoter of <i>CHI3L1</i> with top predicted transcription factor motifs bound. (g) Aggregate footprint of genome wide <i>FOS</i> occupancy across cell categories (top: stromal, myeloid, fibroblast).

Figure 2. Multimodal analyses implicate CHI3L1 and AP-1 epigenetic control in fibroblasts.
(a) Hematoxylin and eosin stain of a cross section of a fistula tract sampled for this study. (b) Uniform manifold approximation and projection (UMAP) of subclustered myeloid and stromal cells from n=14 patients. (c) Enrichment of functional fibroblast gene modules across myeloid and stromal cell types. (d) Expression of CHI3L1 relative to housekeeping after 14 days’ culture in African and European ancestry (AA, EA) CD patients and controls. (e) CHI3L1 protein measured in day 14 CD14+ cell culture supernatants from healthy controls, CD patients, and isolated fistula single cells cultured under the same conditions. (f) Bias corrected peaks in the promoter of CHI3L1 with top predicted transcription factor motifs bound. (g) Aggregate footprint of genome wide FOS occupancy across cell categories (top: stromal, myeloid, fibroblast).

Background: Single-cell analysis has transformed our ability to understand inflammatory bowel disease (IBD) but correlation to clinical treatment outcomes has yet to be done. Here, we present single-cell analysis from a cohort of patients with pediatric Crohn’s disease (pediCD) treatment naïve mucosal samples and matched follow-up mucosal samples on-treatment with anti-tumor necrosis factor (TNF) to understand single-cell intestinal shifts in clinical anti-TNF response.

Methods: Patients were enrolled from November 2017 to December 2018 and followed for 2-3 years. Biopsies were obtained from pediCD at time of diagnosis prior to initiation of therapies and from non-inflamed age-matched pediatric controls (CTRL). We performed scRNA-seq on terminal ileum cells. Patients were prospectively followed for up to 3 years including clinical response to anti-TNF. (Figure 1A)

Results: We profile diagnostic human biopsies from the terminal ileum of treatment-naïve pediatric patients with Crohn’s disease (pediCD; n=14), matched repeat biopsies (n=8), and from non-inflamed pediatric controls (CTRL; n=13). To resolve and annotate epithelial, stromal, and immune cell states among the 201,883 baseline single-cell transcriptomes, we develop a principled unbiased tiered clustering approach, ARBOL. (Figure 1B,C,D) Of the 14 pediCD patients, 4 patients with milder disease activity scores were not placed on anti-TNF blockade (NOA = not on anti-TNF, teal), 5 patients were placed onto anti-TNF therapy and sustained full clinical response (FR = full response, yellow), and 5 patients were placed onto anti-TNF but required dose escalations within first year of therapy (PR = partial response, red). We resolve a vector of T cell, innate lymphocyte, myeloid, and epithelial cell states in treatment-naïve pediCD (pediCD-TIME) samples which can distinguish patients along the trajectory of anti-TNF response. (Figure 1E,F) We then position repeat on-treatment biopsies and identify that anti-TNF treatment pushes the pediatric cellular signature towards an adult treatment-refractory state.

Conclusions: We present a treatment-naïve cohort with high-resolution principled scRNA-seq data analysis paired with clinical outcomes to understand baseline cell states to predict Crohn’s disease trajectory.

BACKGROUND & RATIONALE: Abdominal surgery induces intestinal inflammation within the muscularis externa (ME) resulting in postoperative ileus (POI), a GI motility disorder associated with prolonged hospitalizations, increased morbidity and financial burden. Enteric glia are involved but the mechanisms are poorly understood. The glial connexin-43 hemichannel (gCx43) modulates motility and is implicated in colitis models. Our aim was to investigate the role of gCx43 in intestinal inflammation using a mouse POI model or a human in vitro culture model of reactive glia (hEGC) induced by IL1β.
EXPERIMENTAL DESIGN: Cx43 signaling was studied using a Cx43 inhibitor 43Gap26 or gCx43cKO mice (Cx43flox::Sox10creERT2). In a gut surgical manipulation (IM) model of POI, we compared littermate control (LMC), sham, 3h or 24h IM mice (N=7-9). RiboTag mice (Sox10-ERT_Rosa26-YFP-RiboTag-HA:tamoxifen) were used to study glial dysregulation of connexins and IL1β in POI. IL1β induction of hEGC was used to study Cx43 signaling (N=15 specimens), since IL1β dependent enteric gliosis guides inflammation, dysmotility and immune function in POI. We analyzed glial reactivity, ME inflammation, ΔTranscriptome profiles (RNAseq, Clara-S, nanostring inflammatory panels), protein expression, 3-D LSM co-labeling, Cx43/hemichannel expression and stool formation.
RESULTS: IM induced a reactive glial phenotype with 2-5 fold upregulation of GFAP, s100β and ETBR proteins. The effect was blocked in gCx43cKO mice (p<0.01). Smooth muscle actin upregulation (ME) was further increased in gCx43cKO mice (p<0.001). Knockout of gCx43 had no effect on leukocyte infiltration. Glial Cx43cKO mice had looser stools (p <0.0001) and fewer fecal pellets (p=.0195) after POI (vs LMC IM, p>0.05). IL1β increased16-fold in glia from RiboTag mice IM (p<0.0001). IL1β upregulation was blocked in gCx43cKO mice. In a 243 gene inflammatory panel, change in expression (↑ or ↓) with IM occurred in 47% genes in gCx43cKO compared to 27% in LMC, and 8% genes in shams (p<0.001). In RiboTag mice or hEGC, Cx43 (or Cx45) mRNA levels were 30-100 fold higher than 16 other hemichannels (p<0.0001). In hEGC, 10ng/ml IL1β activated hEGC and caused a change in ΔTranscriptome Profile of 12% of genes (48.6% ↑ and 51.4% ↓, p<0.01). 43Gap26 (200µM) dysregulated 0.69% of genes and regulated 30.9% of IL1β–induced genes (p<0.01). It also decreased IL1β-induced IL6 release from hEGC (p<0.0001). IL1β induction in hEGC or IM in RiboTag Sox10-EGCs caused dysregulation of Cx43, Cx45 and Panx1 (p<0.01). LPS (1-100ng/ml) caused IL1β release in hEGC.
CONCLUSIONS & INFERENCES: Glial Cx43 hemichannels are critical determinants in mouse POI and reactive human glia, enteric gliosis and intercellular communication in immune/inflammatory cells. It’s role in GI Motility Disorders linked to inflammation warrants further investigation (DK113943, DK125809).

Fernando Ochoa Cortes

Appearances