Alistair Easton

Stromal cell states are altered in active Crohn’s disease (CD), but their origin and phenotypic stability are unknown. Using single-cell spatial transcriptomics, RNA sequencing and ATAC sequencing of ~2,500,000 cells, we map 18 distinct stromal cell states within their cellular and cytokine signaling environments in human full-thickness CD bowel. Inflammatory fibroblasts (IFs) reside in immune cell-rich mucosal ulcers and are induced through combinatorial cytokine exposure suppressing submucosal universal fibroblast programs. The IF state is stabilized through transcription factor (TF) activity of GLI3, TWIST1, ETV4, PRDM1 and RELB, and does not spontaneously revert; however, histone deacetylase inhibition destabilizes the IF state, preventing IF secretome-induced epithelial transmigration and activation of neutrophils. The IF open chromatin configuration is distinct from that of fibrotic contractile stroma, which populate adjoining immune-depleted submucosal fibrotic niches. These findings show that microenvironments in pathological tissue niches shape open chromatin configuration of stromal states that are amenable to modulation by epigenetic modifiers.

Chromosomal instability (CIN), a pervasive feature of esophageal adenocarcinoma (EAC), drives tumor aggressiveness and metastasis. CIN stimulates the cGAS–STING pathway, typically linked to antitumor immunity. However, despite the high CIN burden in EAC, the cGAS–STING pathway remains largely intact. To address this paradox, we interrogated multiple esophageal cancer models, finding myeloid-attracting chemokines—with CXCL8 as a prominent hit—as conserved CIN-driven targets in EAC. Using multiplexed immunofluorescence microscopy, we quantified ongoing CIN in human EAC tumors by measuring cGAS-positive micronuclei, validated by whole-genome sequencing. Coupling in situ CIN detection with single-nucleus RNA sequencing and multiplex immunophenotyping of human EAC, we link CIN to tumor-intrinsic innate immune activation, CXCL8 expression, and myeloid cell–mediated immunosuppression. In patients with EAC, CIN ^high , myeloid-dominated tumors correlate with poor outcomes and aberrant cGAS–STING signaling. These insights explain the counterintuitive maintenance of cGAS–STING and highlight the disruption of the CIN–cGAS–inflammation axis as a potential therapeutic strategy in EAC.

Supplementary Fig. S1. [A] Violin plot illustrating the distribution of OxC-EMT scores in 12 independent datasets. [B] Line graph of Hazard Ratios and associated p-values derived from fitting a loop structured cox regression model to the combined set. Y-axis on the left-hand side represents the scale for Hazard Ratio and the y-axis on the right-hand side represents the scale for the p-value. [C] Plot displaying median survival time (in days) for OxC-EMT-high (solid red line) and corresponding 95% CI (dashed red line) and median survival for OxC-EMT-low (solid blue line) and corresponding 95% CI (dashed blue line). [D] Kaplan-Meier survival plot demonstrating the probability of overall survival for OxC-EMT-high/BRCAmutated, OxC-EMT-high/BRCAwildtype, OxC-EMT-low/BRCAmutated and OxC-EMT-low/BRCAwildtype patients in the Scottish cohort. [E] Kaplan-Meier survival plot demonstrating the probability of overall survival for OxC-EMT-low risk patients in the Combined set with Oxford Classic subtypes, Cell Cycle, Differentiated, KRT17 and Mixed.

Purpose:The association between epithelial-to-mesenchymal transition (EMT) in high-grade serous ovarian cancer (HGSOC) and poor prognosis is known. However, molecularly defining a subset of tumors that reproducibly associates with poor prognosis has been an elusive goal in this disease. A molecular signature that can robustly identify patients with poor prognosis and guide treatment decisions, including surgical strategy and targeted therapies, can improve survival rates.Experimental Design:We carried out RNA sequencing of 139 tumor samples (Brescia cohort); an external validation on 362 and 126 patients from the Scottish and Garsed cohorts, respectively; and a meta-analysis of 1,023 tumors to develop clinically useful risk groups. Identification of therapeutic targets was carried out by transcriptomic analyses of fluorescence-activated cell sorted (FACS) tumor epithelial cells and multiplex immunofluorescence assessment of tissue sections.Results:In this study, we have validated the prognostic strength of the Oxford Classic–defined EMT in three independent patient cohorts: Brescia [HR = 3.6; 95% confidence interval (CI) of 1.59–7.97; P = 1.99e−03], Scottish (HR = 1.71; 95% CI of 1.08–2.70; P = 2.23e−02), and Garsed (Kruskal–Wallis P = 0.00071). OxC-based risk stratification of HGSOC could robustly identify poor-risk patients with a 5-year median survival for OxC high-risk and OxC low-risk groups of 13% and 50%, respectively (95% CI of 7.1%–23.5% vs. 36.1%–69.3%) in the Brescia cohort. Further analysis of the risk groups suggests that an alternative surgical strategy and a combination therapy involving EMT targeting drugs and immunomodulators could elicit improved clinical response in poor-risk patients.Conclusions:This study provides a clinically useful risk stratification strategy for HGSOC, as well as targeted treatment options for high-risk patients.See related commentary by Venegas et al., p. 10

Supplementary Fig. S3. [A] Horizontal bar chart representing disease enrichment analysis of 404 differentially expressed genes. [B] Bar chart illustrating a subset of 404 genes that have been previously described in other cancer types. [C] Forest plot illustrating hazard ratio from univariable cox regression analysis, 95% CI and p-value for Overall survival of OxC-EMT score in eight cancer types.

Supplementary Fig. S2. Volcano plot illustrating genes that are differentially expressed in OxC-EMT>0.5 compared to OxC-EMT = 0 in [A] Brescia cohort, [B] TCGA dataset [C] AOCS dataset [D] OVCAD dataset.

Supplementary Table S1. [Worksheet-1] List of differentially expressed genes in OxC-EMT-high compared to OxC-EMT-low bulk tumour samples from patients in the Brescia, TCGA, AOCS and OVCAD cohorts. [Worksheet-2] List of differentially expressed genes in OxC-EMT-high compared to OxC-EMT-low tumour epithelial cells.

Supplementary Fig. S4. [A] FACS images of DAPI-negative live cells from which EpCAM+ALDH+cells (EpCAM-sc) and rest of EpCAM+cells (EpCAM-bulk) were isolated from tumour tissue of patients, 11611, 11617, 11619 and 11624. All EpCAM+ cells were isolated from tumour tissue of patient 11641. Representative images of TGFβ1 and TGFβ3 expression by mIF in [B] OxC-EMT-high tumour (patient-BS14) and [C] OxC-EMT-low tumour (patient-BS141).

Supplementary Fig. S5. Representative images of an [A] OxC-EMT-high (patient BS137) and [B] OxC-EMT-low (patient BS141) tumour from the validation mIF experiments. [C] Violin plots demonstrating the difference in the abundance of M2 macrophages (CD206+ or CD206+CD163+) and CD8+TRM using immune cell counts normalised against total number of nuclei in each ROI.

Purpose: The association between epithelial-to-mesenchymal transition (EMT) in high-grade serous ovarian cancer (HGSOC) and poor prognosis is known. However, molecularly defining a subset of tumors that reproducibly associates with poor prognosis has been an elusive goal in this disease. A molecular signature that can robustly identify patients with poor prognosis and guide treatment decisions, including surgical strategy and targeted therapies, can improve survival rates. Experimental Design: We carried out RNA sequencing of 139 tumor samples (Brescia cohort); an external validation on 362 and 126 patients from the Scottish and Garsed cohorts, respectively; and a meta-analysis of 1,023 tumors to develop clinically useful risk groups. Identification of therapeutic targets was carried out by transcriptomic analyses of fluorescence-activated cell sorted (FACS) tumor epithelial cells and multiplex immunofluorescence assessment of tissue sections. Results: In this study, we have validated the prognostic strength of the Oxford Classic–defined EMT in three independent patient cohorts: Brescia [HR = 3.6; 95% confidence interval (CI) of 1.59–7.97; P = 1.99e−03], Scottish (HR = 1.71; 95% CI of 1.08–2.70; P = 2.23e−02), and Garsed (Kruskal–Wallis P = 0.00071). OxC-based risk stratification of HGSOC could robustly identify poor-risk patients with a 5-year median survival for OxC high-risk and OxC low-risk groups of 13% and 50%, respectively (95% CI of 7.1%–23.5% vs. 36.1%–69.3%) in the Brescia cohort. Further analysis of the risk groups suggests that an alternative surgical strategy and a combination therapy involving EMT targeting drugs and immunomodulators could elicit improved clinical response in poor-risk patients. Conclusions: This study provides a clinically useful risk stratification strategy for HGSOC, as well as targeted treatment options for high-risk patients. See related commentary by Venegas et al., p. 10