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).

Inflammatory responses to the oncogenic pathogen H. pylori involve Th17 cell activation and production of IL-17. Binding of IL-17 to its receptor IL-17RA, expressed on gastric epithelial cells, triggers intracellular signaling events leading to cytokine production. Gastric stem cells are essential for regulating injury repair that occurs within the context of inflammation; conversely, aberrant stem cell activation is linked to carcinogenesis. Leucine-rich repeats and immunoglobulin-like domains 1 (Lrig1) is a transmembrane protein that marks a distinct population of progenitor cells within the stomach and we have previously shown that H. pylori can functionally activate Lrig1⁺ cells, leading to the development of premalignant lesions. Our aim was to define the role that IL-17 exerts in regulating Lrig1 within the context of chronic H. pylori infection. Gastric tissue contains a myriad of cell types required for gastritis. Therefore, we developed an enriched gastroid system by adding gastroids harvested from mice to the upper chamber of a transwell system, and autologous splenocytes treated with anti-CD3 and anti-CD28 to induce T cell activation and macrophages to the lower chamber. H. pylori strain PMSS1 was then added to epithelial monolayers and cytokine expression was quantified using RNA from macrophage/T cell lysates. Significantly higher (p<0.05) expression levels of il-17, il-21, il-22, il-23, and il-9, all components of the Il-17 signaling pathway, were present in ex vivo infected versus uninfected samples, findings that were replicated in a human gastroid:macrophage:T cell system using an independent H. pylori strain (7.13). We extended these results by infecting wild-type (WT) or IL-17RA^-/- C57BL/6 mice with PMSS1 and sacrificed mice 8 weeks post-challenge. There were no differences in colonization; however, H. pylori significantly increased levels of chronic inflammation in IL-17RA^-/-versus WT mice (p<0.01, mean score 2.69 vs. 1.43). IL-17RA^-/- PMSS1-infected mice also harbored significantly higher numbers of proliferative Lrig1⁺ cells compared to infected WT mice in the antrum (p<0.05, 14.1% vs. 4.0%), and corpus (p<0.01, 59.9% vs. 9.8%), suggesting that IL-17 signaling may suppress injury and stem cell activation induced by H. pylori. Finally, to determine if suppression was epithelial cell autonomous or required additional cell lineages, we utilized a reductionist gastroid system generated from WT or IL-17RA^-/- mice. Infection with H. pylori significantly increased Lrig1 expression in WT but not in IL-17RA^-/- monolayers compared to controls. In conclusion, IL-17 signaling may exert a suppressive effect on pathogenesis and proliferation of Lrig1⁺ cells within the context of chronic H. pylori infection in a gastric epithelial non-autonomous manner, which likely requires immune elements present in the lamina propria.

Introduction
Increased lymphocyte infiltration at the invasive margin (IM, 500μm either side of the tumour-normal interface) or in lymph node metastases in colorectal cancers (CRCs) may lead to increased immune selection pressure. We aimed to identify whether immune escape mechanisms are selected for in IM or node metastases.

Methods
Spatially-resolved deep error-corrected targeted sequencing of genomic regions generating neoantigens binding to the ten commonest HLA alleles was performed on 82 superficial, IM and lymph node deposits from 11 stage III microsatellite-stable CRCs. Lymphocytes on H&Es were classed as intra-tumoral, adjacent or distal based on tumour proximity by a deep-learning-based classifier. The ratio of number of intra-tumoral/ adjacent lymphocytes to cancer cells was defined as intra-tumoral/ adjacent tumoral lymphocyte ratio (ITLR/ ATLR). Cyclic immunofluorescence using 22 antibodies was performed on 34 matching cancer and 10 normal regions. Cellular neighbourhoods (CNs) were defined based on cell type composition of nearest neighbours. Proportional neoantigen burden (pNAB) was defined as the proportion of subclonal non-synonymous mutations that produce strong-binding neoantigen peptides.

Results
The tumour microenvironment showed a depletion in cytotoxic T lymphocyte (CTL) fraction (p= 0.00086) and increase in CD3+CD4+CTLA4+CD45RO+FOXP3+ Treg cells (p= 6.7e-05) compared to normal mucosa.

ITLR was increased in IM (p=0.03) and ATLR was increased in IM (p=4.3e-08) and node (p=1.2e-06) compared to bulk tumour. Minimum distance between epithelial cells and closest CTLs was smaller in IM/ node compared to superficial CRC (IM/ node vs superficial: p<2e-16, Figure 1), which may facilitate immune editing and select for escape alterations.

High-frequency neoantigen mutations were depleted in IM/ node, consistent with enhanced editing (multiple regression, pNAB at purity-adjusted VAF >= 0.02 ~ sample type + patient, IM vs. superficial, missense p= 0.00679, indels p= 0.0332, node vs superficial, missense p= 3.29e-06, indels p= 1.09e-05). No depletion was observed at low-frequency, demonstrating that neoantigen mutations continue to be accrued throughout tumour evolution.

The fraction of cells assigned to CN3 (enriched for PDL1+ cancer cells) was increased in IM/ node (IM vs superficial: p=0.03, node vs superficial: p=0.03, Figure 2), suggesting that PDL1 expression in cancer cells may arise on a background of immune editing. Cellular neighbourhoods (CNs) containing PDL1+ cancer cells showed increased number of CTLs (per epithelial cell) compared to CNs with non-PDL1+ cancer cells (p<2e-16), suggesting that PDL1 expression may be driven by proximity to CTLs.

Conclusions
Increased immune infiltration in IM/ node deposits of CRCs may facilitate immune editing and selection for PDL1+ cancer cell-enriched neighbourhoods.

Figure 1: Representative regions of CRC and normal mucosa from 1 patient showing relatively increased proximity between tumour cells (red) and CTLs (green) in the invasive margin and node deposits, compared to superficial CRC. Approximate estimates of distance between example pairs of epithelial cells and proximate CTLs are shown (arrowed). Coordinates of cell centroids have been plotted on raw nuclear image used for segmentation.

Figure 1: Representative regions of CRC and normal mucosa from 1 patient showing relatively increased proximity between tumour cells (red) and CTLs (green) in the invasive margin and node deposits, compared to superficial CRC. Approximate estimates of distance between example pairs of epithelial cells and proximate CTLs are shown (arrowed). Coordinates of cell centroids have been plotted on raw nuclear image used for segmentation.

Figure 2: Representative regions of superficial, invasive margin and node deposits of CRC from 1 patient showing relative abundance of PDL1+ cancer cells (purple) within cellular neighbourhood 3 in the invasive margin and node deposits of CRC. CTLs within CN3 are shown in green. Coordinates of cell centroids have been plotted on segmentation mask.

Figure 2: Representative regions of superficial, invasive margin and node deposits of CRC from 1 patient showing relative abundance of PDL1+ cancer cells (purple) within cellular neighbourhood 3 in the invasive margin and node deposits of CRC. CTLs within CN3 are shown in green. Coordinates of cell centroids have been plotted on segmentation mask.

Andrea Sottoriva

Appearances