TL1A-activated T cells remodel the rectal mucosa in patients with Crohn’s disease with perianal fistulising disease Authors: Victoria Gudiño, Jae Won Cho, Berta Caballol, Ángela Sanzo-Machuca, Ana Corraliza, Marisol Veny, Isabella Dotti, Livia Moreira Genaro, Elisa Melón-Ardanaz, Maria Carme Masamunt, Miriam Esteller, Iris Teubel, Lisseth Robbins, Àngel Giner, Cristina Prieto, Elena Ferrer, Raquel Franco Leal, Albert Martin-Cardona, Carme Loras, Maria Esteve, Jordi Rimola, Agnès Fernández-Clotet, Ingrid Ordás, Elena Ricart, Julian Panés, Martin Hemberg, Azucena Salas Published: 2025 ABSTRACT AbstractBackgroundPerianal fistulising disease (PFD) is a complication that affects about 20% of patients with Crohn’s disease (CD) whose aetiology remains unknown.ObjectivesTo identify predisposing events driving fistula formation.DesignRectal biopsies from patients with CD with or without PFD (CD+PFD and CD, respectively; n=31) were collected and subjected to single-cell RNA sequencing. Functional analyses were conducted using peripheral CD3+ T cells, intestinal tissue explants, primary fibroblasts and two-dimensional epithelial monolayer cell cultures.ResultsThe rectal mucosa of patients with CD+PFD is imprinted with cellular and transcriptomic alterations specific to PFD and independent of luminal inflammation, potentially driven by tumour necrosis factor-like ligand 1A (TL1A) activation in CD4+ T cells. We identified lymphotoxin beta (LTB or its functional heterotrimer LTα1β2) as a novel mediator downstream of TL1A that, along with interleukin (IL)-22, induces a PFD-associated signature in rectal fibroblast and epithelial cells, respectively. This signature includes an increased abundance of fibroblasts, an induction of matrix-degrading enzymes, transcriptomic rewiring of the lamina propria S1 fibroblasts and an anti-bacterial and immune responses in epithelial cells. Notably, the induction of LTα1β2 and IL-22 occurs independently of tumour necrosis factor (TNF) signalling, revealing a new TL1A-LTα1β2/IL-22 axis that remains active under anti-TNF therapy.ConclusionOur findings revealed unique cellular alterations in the rectum of patients with CD+PFD, highlighting the previously unrecognised involvement of TL1A in mediating this signature and supporting the need for exploring the role of TL1A inhibition as a therapeutic approach for PFD.Keywords: SINGLE-CELL ANALYSIS, CROHN'S DISEASE, T LYMPHOCYTES, GASTROINTESTINAL FISTULAE, CYTOKINES WHAT IS ALREADY KNOWN ON THIS TOPIC One-fifth of patients with Crohn’s disease suffers from perianal fistulising disease (PFD). One-fifth of patients with Crohn’s disease suffers from perianal fistulising disease (PFD). Current biologics or surgical treatment is not always efficacious in the long term. Current biologics or surgical treatment is not always efficacious in the long term. Epithelial transformation by pro-inflammatory cytokines is suggested to drive fistula formation, but there is still a poor characterisation of PFD pathogenesis. Epithelial transformation by pro-inflammatory cytokines is suggested to drive fistula formation, but there is still a poor characterisation of PFD pathogenesis. WHAT THIS STUDY ADDS This study provides a cell-specific rectal transcriptional signature associated with fistula formation, involving lymphocyte activation by tumour necrosis factor-like ligand 1A (TL1A). This study provides a cell-specific rectal transcriptional signature associated with fistula formation, involving lymphocyte activation by tumour necrosis factor-like ligand 1A (TL1A). Together with interleukin (IL)-22, we identified lymphotoxin beta (LTα1β2) as a new mediator downstream TL1A activation, whose activation is independent of tumour necrosis factor (TNF) signalling. Together with interleukin (IL)-22, we identified lymphotoxin beta (LTα1β2) as a new mediator downstream TL1A activation, whose activation is independent of tumour necrosis factor (TNF) signalling. Using primary intestinal cultures, we characterised the stromal and epithelial alterations associated with PFD and driven by LTα1β2 and IL-22, respectively. Using primary intestinal cultures, we characterised the stromal and epithelial alterations associated with PFD and driven by LTα1β2 and IL-22, respectively. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY We provide a new approach to study predisposing events in PFD by investigating the transcriptomic alterations in the rectal mucosa of these patients. We provide a new approach to study predisposing events in PFD by investigating the transcriptomic alterations in the rectal mucosa of these patients. Our results provide molecular data to support the involvement of TNF in PFD, but uncover a new signalling axis (TL1A-LTα1β2/IL-22) with potential therapeutic value. Our results provide molecular data to support the involvement of TNF in PFD, but uncover a new signalling axis (TL1A-LTα1β2/IL-22) with potential therapeutic value. Introduction Crohn’s disease (CD) is a chronic remitting and relapsing inflammatory condition of the gastrointestinal tract often complicated by fibro-stenotic, fistulising and/or perianal disease. Perianal lesions are present in approximately 12% of patients at diagnosis,1 with their prevalence rising to nearly 20% over the course of the disease.2 Several factors, such as rectal luminal disease, have been associated with an increased risk of developing perianal fistulising disease (PFD).3 However, a large proportion of patients with PFD present no rectal involvement, suggesting that the mechanisms driving fistula formation may differ from those supporting luminal inflammation. The current model used to explain the pathogenesis of PFD involves the transformation of epithelial cells through epithelial-to-mesenchymal transition (EMT) as a response to cytokines such as transforming growth factor beta (TGFβ), tumour necrosis factor (TNF), or interleukin (IL)-13.4–6 This is accompanied by the remodelling of the extracellular matrix (ECM), which facilitates the migration of transformed epithelial cells to the site of fistula formation, as seen by the enhanced expression of matrix-degrading enzymes or metalloproteinases (MMPs) in the fistula compared with intestinal or peri-fistula tissues.7 8 Importantly, experiments using cells from the fistula tract revealed an expansion of CD4+ over CD8+ T cells,6 as well as a heightened fibroblast response to IL-13 or TGFβ compared with fibroblasts from non-fistulising tissue.4 However, these observations are largely based on analyses from fistula-derived tissues, which are more likely to reflect the end-stage of the disease rather than the early mechanisms driving fistula initiation or progression. To address this limitation, we analysed the rectal mucosa of patients with CD with (CD+PFD) and without (CD) perianal disease, generating a comprehensive single-cell atlas of the rectal cellular populations and states. Considering the increased risk of fistula development in patients with rectal involvement and the possibility that fistulae may originate from both the anal canal and the rectum, we hypothesised that alterations in the rectum may precede and contribute to the development of PFD, through mechanisms potentially independent from those driving luminal inflammation. Our data reveals that the transcriptomic rewiring of the rectal mucosa can occur independently of luminal inflammation and is, at least in part, mediated by the activation of CD4+ T cells in response to tumour necrosis factor-like ligand 1A (TL1A). Herein, we provide novel insights into the pathophysiology of perianal disease, defining a distinct stromal and epithelial signature associated with PFD and identifying TL1A along with lymphotoxin β and IL-22 as key mediators of these alterations with potential therapeutic value. Methods Patient and public involvement Patients or members of the public were not involved in the design, conduct, reporting or dissemination plans of this research. Further detailed methods are outlined in the online supplemental information. Results Defining a PFD signature in the rectal mucosa To investigate whether changes in the rectum could contribute to the development of fistulae, we analysed a cohort of patients with CD with (CD+PFD; n=19) or without (CD; n=12) PFD (figure 1A, online supplemental table 1). Patients were further stratified according to the presence or absence of active rectal inflammation. Baseline demographics showed a broad age range (24–72 years) and disease duration (5–47 years). Disease location and phenotype varied, with ileocolonic location and inflammatory behaviour being the most frequent presentations. Importantly, patients with perianal fistulae had increased exposure to anti-TNF and prior perianal surgery. Within the CD+PFD group, most patients presented complex fistulae according to the AGA classification and high MAGNIFI and Van Assche indices regardless of the presence of rectal inflammation, highlighting the structural burden of perianal disease independent of luminal inflammation. Most fistulae were inter-sphincteric, with a few trans-sphincteric and supra-sphincteric. Biopsies were freshly processed for single-cell RNA sequencing (scRNA-seq) as described.9 We obtained the transcriptomes from 81 370 high quality cells, which we annotated into five subsets: epithelium, stroma, T cells, B and plasma cells and myeloid cells (figure 1B). These were then subclustered to identify cell types at a higher resolution (online supplemental figure 1A, online supplemental table 2). A comparison of the cell abundances between the two groups revealed a significant increase in the number of lamina propria S1 (ABCA8), bottom-crypt S3 (OGN) and immediate-early response (IERs) fibroblasts (FOS) and glia (S100B) in CD+PFD compared with patients with CD (figure 1C). In contrast, CD samples showed an increase in the abundance of pericytes (RSG5), IgG plasma cells, plasmablasts (DERL3) and neutrophils (PROK2). Importantly, changes in all the latter populations could be attributed to differences in the severity of inflammation within samples from both groups, rather than changes associated with the presence of PFD. Indeed, each group contained samples from endoscopically inflamed rectum (6/19 in the CD+PFD group -CD+PFDinf- and 5/12 in the CD group -CDinf-; figure 1D, online supplemental figure 1B, C). In addition, some of the patients in the CD+PFD group without rectal inflammation at the time of sampling had previous rectal involvement (CD+PFD with rectal mucosa in remission, CD+PFDre, n=6). Given the impact that current and even previous disease has on the transcriptome of the intestinal mucosa,10 we focused our analysis on samples from uninvolved rectum (patients with no current or previous rectal inflammation) from CD+PFD (CD+PFDun, n=7) and CD (CDun, n=7) patients. Of note, none of the patients in the CDun group developed PFD even after 5 years since inclusion, confirming their stable non-fistulising phenotype. Interestingly, when comparing CD+PFDun and CDun groups, we confirmed the increase in different fibroblast subsets and glia in the fistulising group, along with myeloid, IgG plasmablasts and some T-cell types (figure 1E), suggesting that not all of them are driven by inflammation. Beyond changes in cell abundances, differentially expressed gene (DEG) analysis between these groups revealed significant changes in gene expression across lymphoid, epithelial and stromal cells (figure 1F, online supplemental figure 1D). In sum, these results indicate that the rectal mucosa of patients with PFD, even in the absence of ongoing or previous inflammation, is imprinted with a unique transcriptional profile that involves lymphocytes, various epithelial cell types and fibroblasts. Patients with PFD present cellular and transcriptomic alterations in the rectal mucosa. (A) Schematic representation of the study. (B) UMAP representation of the scRNA-seq dataset coloured by patient group. (C) Cell abundance analysis comparing patients with CD and CD+PFD. (D) Chart indicating the categorisation of patients from the cohort based on the presence of PFD and luminal inflammation of the rectum. (E) Cell abundance analysis comparing CDun and CD+PFDun patients. For C and E, cell frequency differences are represented as the Log2FC, and coloured bars indicate significant differences (|FC| ≥1.5 and adjusted p value <0.05. Fisher’s exact test, Benjamini-Hochberg correction. (F) Heatmap representing the number of genes upregulated in each group per cell type. Only cell types with cell counts above 10 are shown. CD, Crohn’s disease; Log2FC, logarithm of fold change in base 2; PFD, perianal fistulising disease; scRNA-seq, single-cell RNA sequencing; UMAP, Uniform Manifold Approximation and Projection. Changes in lamina propria fibroblasts from patients with PFD occur independently of TGFβ Cell abundance analysis pointed to fibroblasts as significantly altered in patients with PFD regardless of their inflammatory state. Among them, lamina propria S1 fibroblasts showed the highest number of DEGs when comparing CDun with CD+PFDun patients (figures 1F and 2A, online supplemental table 3). Pathway analysis suggests a metabolic reprogramming of CD+PFDun S1 fibroblasts, with pathways related to cellular respiration and lipid metabolism being significantly deregulated between groups (figure 2B). Intriguingly, response to TGFβ signalling, which has previously been suggested to play a role in the pathophysiology of PFD4 5 was found to be downregulated in the CD-PFDun group (figure 2B). To confirm this, we stimulated primary intestinal fibroblasts with TGFβ and measured the expression of several DEGs between groups. Indeed, TGFβ significantly downregulated the expression of some of those genes, while it induced the expression of CTGF, the second-most downregulated gene in the CD+PFDun group (figure 2C). Interestingly, TGFβ also decreased the expression of MMP1, MMP3 and CHI3L1, which previous studies found upregulated in perianal fistula tracts.7 8 11 In agreement with the scRNA-seq data, bulk RNA from the rectal biopsies showed not only the significant upregulation of TFPI2, though not EMILIN1 and PLAC9, but also the downregulation of CTGF in CD+PFDun biopsies (figure 2D, online supplemental figure 2A). Interestingly, we also observed a significant upregulation of MMP1, MMP3 and CHI3L1 in bulk RNA from CD+PFDun samples, validating their previously described association with PFD (figure 2D). TGFβ signalling is downregulated in the S1 fibroblasts of CD+PFDun patients. (A) Volcano plot representing the DEGs in the lamina propria S1 fibroblasts between CDun and CD+PFDun patients. A two-sided Wilcoxon rank-sum test was applied. Genes with an adjusted p value <0.05, and an avgLog2FC ˃ log2 (1.2) or avgLog2FC ˂ −log2 (1.2) were considered upregulated (UPP) or downregulated (DWW), respectively. Genes with a nominal p value˂0.05 are also indicated in light colours if avgLog2FC ˃ log2 (1.2) (UP) or avgLog2FC ˂−log2 (1.2) (DW). (B) Bar plot representing significantly enriched pathways between CDun and CD+PFDun patients obtained with SCPA. Bar length indicates the logarithm of the fold change, while the bar colour indicates the q value, a score generated by SCPA that represents the size of the distribution change. Arrow highlights the ‘Response to transforming growth factor beta’ pathway. (C) qRT-PCR of primary intestinal fibroblasts stimulated with TGFβ (5 ng/mL) for 24 hours. Data is shown as a Log2FC comparing TGFβ-stimulated versus unstimulated fibroblasts. One sample t-test, parametric, *q<0.05, **q<0.01, n=5. (D) Whole-tissue qRT-PCR analysis of rectal biopsies from CDun (n=8) and CD+PFDun (n=12) patients; Mann-Whitney test, *q≤0.05, **q≤0.01, ****q≤0.0001. (E) UMAP of Kong et al12 dataset coloured by cell subset (epithelium, immune and stroma). (F) Jaccard analysis comparing the similarity between the DEGs and the S1 fibroblasts of Kong et al12 with the DEGs from the S1 fibroblasts of the current study. (G) Violin plot comparing the TGFβ signature scores of S1 fibroblasts from Kong et al12 between patients without (CD) and with PFD (CD+PFD). The signature is generated from bulk RNA sequencing data of TGFβ-stimulated primary fibroblasts. The Wilcoxon rank-sum test was applied. AU, arbitrary units; CD, Crohn’s disease; DEGs, differentially expressed genes; Log2FC, logarithm of the fold change in base 2; PFD, perianal fistulising disease; qRT-PCR, quantitative reverse transcription polymerase chain reaction; SCPA, single-cell pathway analysis; TGFβ, transforming growth factor beta; UMAP, uniform manifold approximation and projection. We then explored whether this signature was also observed in the context of mucosal inflammation. While there were no DEGs between CDinf and CD+PFDinf patients in the S1 fibroblasts (online supplemental figure 2B), a fibroblast-specific TGFβ response signature was also downregulated in CD+PFDinf compared with CDinf, suggesting that this might be a specific PFD trait (online supplemental figure 2C). Finally, we used an independent scRNA-seq dataset by Kong et al12 to validate our observations. In brief, we selected non-inflamed samples from patients with CD with or without PFD (n=24, online supplemental table 1, figure 2E). Interestingly, we observed that the DEGs in S1 fibroblasts from this new cohort were highly similar to their counterparts in our cohort (figure 2F). This is in line with the upregulation of TFPI2 in patients with CD+PFD but the downregulation of CTGF in the same group (online supplemental figure 2D). Moreover, the TGFβ response signature was also downregulated in the S1 fibroblasts of patients with PFD in Kong et al, thus pointing to a stromal remodelling in patients with PFD similar to that seen in our cohort. In summary, we describe a transcriptional signature expressed by fibroblasts and associated with the PFD phenotype. This signature was not driven by TGFβ signalling, suggesting alternative mechanisms may be at work. CD4+ T cell-expressed lymphotoxin β imprints a PFD signature on intestinal fibroblasts We next hypothesised that immune-derived signals, independent of TGFβ, may be acting on the stromal compartment of patients with PFD. Within immune cells, T cells, particularly CD4+ T cells, were the cell type with the highest number of DEGs when comparing CDun and CD+PFDun patients (figure 1F, online supplemental table 3). Among DEGs, LTB, encoding for the lymphotoxin β chain, presented the highest fold change increase across all T-cell subsets in patients with perianal disease (figure 3A and online supplemental figure 3A). This lymphotoxin subunit forms a membrane-bound heterotrimer with LTα (LTα1β2) and signals through the LTβ receptor (LTBR).13 LTBR is expressed by stromal and epithelial cells, in addition to some myeloid subsets (figure 3B). We then investigated LTB-LTBR interactions across all cell types using CellPhoneDB.14 In agreement with the expression pattern of LTBR, cells expressing LTB interacted with epithelial, mesenchymal and myeloid cells, being the strength and interaction of LTB-LTBR higher in CD+PFDun compared with CDun patients (figure 3C). Based on this, we hypothesised that LTB in its heterotrimer form LTα1β2 could be driving the transcriptional signature we observed in fibroblasts from CD+PFDun patients. LTα1β2 derived from TL1A-activated T cells induces the PFD-associated signature in the stroma. (A) Volcano plot representing the DEGs in the CD4+ T cells between CDun and CD+PFDun patients. A two-sided Wilcoxon rank-sum test was applied. Genes with an adjusted p value < 0.05, and an avgLog2FC ˃ log2 (1.2) or avgLog2FC ˂ −log2 (1.2) were considered upregulated (UPP) or downregulated (DWW), respectively. Genes with a nominal p value ˂0.05 are also indicated in light colours if avgLog2FC ˃ log2 (1.2) (UP) or avgLog2FC ˂ −log2 (1.2) (DW). (B) UMAP plot showing the expression of LTB and LTBR in the whole object. (C) Heatmaps representing the interactions predicted for LTB (source cells) and LTBR (target cells) for CDun and CD+PFDun by CellPhoneDB analysis. Only significant interactions are represented among clusters with cell counts above 10, and colour scale indicates the predicted interaction strength. (D) qRT-PCR of primary intestinal fibroblasts stimulated with LTα1β2 (100 ng/mL) for 24 hours. Data is shown as a Log2FC comparing LTα1β2-stimulated versus unstimulated fibroblasts. One sample t-test, parametric, *q<0.05, **q<0.01, n=7. (E) Cytokine enrichment plot based on the immune response enrichment analysis15 conducted with the top 50 genes overexpressed in the CD4+ T cells of CD+PFDun patients. Bar length represents the enrichment score, and the colour scale indicates the FDR-adjusted p value (two-sided Wilcoxon rank-sum test). (F) Violin plot of TNFRSF25 expression within the T-cell compartment. The adjusted p value was generated by DEG analysis. (G) Flow cytometry plots of peripheral T cells stimulated with anti-CD3 or anti-CD3+TL1A for 48 hours and stained with a-LTα and LTβR-Fc. Double-positive cells correspond to LTα1β2+ T cells. Plots are from a representative replicate, and percentages of LTα1β2+ T cells comparing the two conditions are shown on the right. Paired t-test, parametric, **p<0.01, n=7. CD, Crohn’s disease; DEGs, differentially expressed genes; FDR, false discovery rate, Log2FC, logarithm of the fold change in base 2; PFD, perianal fistulising disease; qRT-PCR, quantitative reverse transcription polymerase chain reaction; TL1A, tumour necrosis factor-like ligand 1A; UMAP, uniform manifold approximation and projection. Indeed, primary intestinal fibroblasts stimulated with LTα1β2 significantly upregulated the expression of several DEGs, in addition to MMP1, MMP3 and CHI3L1, while downregulating CTGF and CARMN, partially reproducing the signature we observed in S1 fibroblasts from CD+PFDun patients (figure 3D). Modulation of TFPI2 by LTα1β2 was also validated at the protein level (online supplemental figure 3B). Altogether, these results suggest that increased LTα1β2 expression by CD4+ T cells could be driving the fibroblast signature we observed in CD+PFDun rectal samples. A TL1A-LTα1β2 axis is enhanced in CD+PFDun mucosal T cells To identify the potential upstream regulators of the CD4+ T cell transcriptional signature in PFD, including the overexpression of LTB, we submitted the 50 most significantly regulated genes to the immune dictionary (www.immune-dictionary.org)15 (online supplemental table 3). This tool uses single-cell transcriptomic profiles of response to cytokines in murine lymph nodes to perform an immune response enrichment analysis, which infers those potential cytokines that induce the input gene set from a given cell type. This analysis identified TL1A as the most likely regulator of the CD4+ T-cell signature in CD+PFDun samples (figure 3E). Furthermore, TNFRSF25, encoding for the TL1A receptor (DR3), was upregulated in CD+PFDun rectal CD4+ T cells (figure 3F), further supporting a role for TL1A in inducing the changes observed in the CD4+ T cells of these patients. Conversely, in the context of rectal inflammation (CDun vs CDinf), both LTB and TNFRSF25 were significantly upregulated in CD4+ T cells (online supplemental figure 3C), with no differences observed in the active mucosa of patients with PFD (CDinf vs CD+PFDinf, online supplemental figure 3D). Remarkably, significantly higher LTB and a trend towards increased expression in TNFRSF25 were also confirmed in the independent public dataset used by Kong et al (online supplemental figure 4E). TL1A (TNFSF15) is a T-cell activator cytokine from the TNF superfamily that plays a prominent role in IBD pathophysiology.16 The literature indicates that macrophages, T cells and dendritic cells are the main producers of TL1A.17–20 In contrast, our scRNA-seq data suggests that TNFSF15 is mainly expressed by epithelial cells followed by macrophages and fibroblasts, particularly under inflammatory conditions (online supplemental figure 4A). Membrane-bound and secreted epithelial expression of TL1A was confirmed in mucosal cells and primary epithelial cultures, respectively (online supplemental figure 4B,C). However, no significant differences in TNFSF15 expression were observed between CDun and CD+PFDun patients (online supplemental figure 4A). Despite the link between TL1A and LTα1β2 suggested by the immune dictionary, to our knowledge, this has not been functionally described. To investigate this, we activated sorted T cells with an anti-CD3 antibody followed by TL1A, or vehicle, for 48 hours. Given the lack of commercial anti-LTβ antibodies, cell-surface detection of the functional heterotrimer (LTα1β2) was performed using a combination of an anti-LTα and an LTβR-Fc chimaera (online supplemental figure 5A). In agreement with our predictions, stimulation with TL1A induced a significant increase in LTα1β2 surface expression (figure 3G). TL1A-induced LTα1β2-producing cells were enriched for CD4+ effector memory T cells (CD62L−CD45RA−), and showed a marked increase in interferon-γ, IL-17 and IL-22-producing cells compared with LTα1β2- cells (online supplemental figure 5B). These results demonstrate a previously unrecognised TL1A/LTα1β2 axis and suggest its potential activation in the rectal mucosa of patients with PFD in the absence of rectal involvement. TL1A-LTα1β2 signalling as an alternative pathway in the context of TNF inhibition TNF inhibitors are currently the only recommended biologic to treat active PFD in patients with CD,21 suggesting the involvement of TNF in the development or maintenance of fistula tracts.22 Hence, we investigated whether TNF could also induce the PFD-associated signature. We first tested the effect of TNF on CD3+ peripheral T cells and observed that TNF increases LTB transcription (figure 4A), in agreement with previous observations.23 However, in contrast to TL1A, TNF had no effect on the protein expression of the functional heterotrimer LTα1β2 (figure 4B). Conversely, TNF, as observed with LTα1β2, significantly modulated the expression of the S1 CD+PFDun deregulated genes TFPI2 and CTGF, as well as of the PFD-associated genes MMP1 and MMP3, but not CHI3L1, in primary cultured fibroblasts (figure 4C). TNF induces the PFD-associated signature in primary fibroblasts. (A) qRT-PCR for LTB in peripheral T cells unstimulated or stimulated with TNF for 48 hours. Paired t-test, parametric, **p≤0.01, n=4. (B) Flow cytometry plots of peripheral T cells stimulated with anti-CD3 or anti-CD3+TNF for 48 hours and stained with a-LTα and LTβR-Fc. Double-positive cells correspond to LTα1β2+ T cells. Plots are from a representative replicate, and percentages of LTα1β2+ T cells comparing the two conditions are represented on the right. Paired t-test, parametric, ns=p≥0.05, n=4. (C) qRT-PCR of primary intestinal fibroblasts stimulated with TNF (20 ng/mL) for 24 hours. Data is shown as a Log2FC comparing TNF-stimulated versus unstimulated fibroblasts. One sample t-test, parametric, ****p<0.0001, n=8. (D) Venn diagram of the DEGs in the primary fibroblasts stimulated with LTα1β2 (orange), TNF (green) or TGFβ (purple). (E) Dot plot with the top 10 genes induced in primary fibroblasts stimulated with LTα1β2, TNF and TGFβ. Dot size indicates adjusted p value in the logarithm base 10, while colour scale indicates the magnitude of Log2FC. (F) Pathway analysis from the RNA sequencing data of the primary fibroblasts stimulated with LTα1β2, TNF and TGFβ. Colour scale indicates adjusted p value obtained from the enrichment test for biological pathway analysis by ‘ToppFun’ in a logarithm base 10, scaled for each treatment. Data is represented as a Z-score. (G) Heatmap representing the qRT-PCR data of primary intestinal fibroblasts stimulated with LTα1β2 (100 ng/mL) and/or TNF (0.5 ng/mL) for 24 hours. Data is shown as a Log2FC comparing stimulated versus unstimulated fibroblasts. LTα1β2 and TNF conditions: one sample t-test, parametric, *p<0.05, **p<0.01, ***p<0.001. TNF versus LTα1β2+TNF condition: paired t-test, #p<0.05, ##p<0.01, n=5. CD, Crohn’s disease; DEGs, differentially expressed genes; Log2FC, logarithm of the fold change in base 2; PFD, perianal fistulising disease; qRT-PCR, quantitative reverse transcription polymerase chain reaction; TGFβ, transforming growth factor beta; TNF, tumour necrosis factor. To better characterise the response of intestinal fibroblasts to TNF and LTα1β2, we sequenced RNA from fibroblasts stimulated with each cytokine. Likewise, TGFβ-stimulated fibroblasts were also sequenced for signature comparison. A total of 1276 genes were significantly regulated in response to TNF (822 up and 454 down), while 316 genes responded to TGFβ (205 up and 111 down) and 41 genes were significantly upregulated by LTα1β2 (online supplemental table 4). Among the LTα1β2-induced genes, 40 genes (>97%) overlapped with the TNF-induced signature on fibroblasts (p=6.14e−55, Fisher’s exact test), and no overlap was found between LTα1β2 and TGFβ-induced signatures (figure 4D). In contrast, a partial but significant overlap between TNF-induced and TGFβ-induced genes was observed (p=8.57e−27, Fisher’s exact test). Genes regulated by both TNF and LTα1β2 included NFκB subunits (NFKB2, RELB), genes related to interferon signalling (OASL) and immune responses (IL34, CXCL10, VCAM1) (figure 4E and online supplemental table 4). Conversely, sharing signatures between TNF and TGFβ were involved in ECM organisation and MAPK cascade pathways (figure 4F). Hence, while all three cytokines are fibroblast modulators, they each imprint a unique response, with TNF showing the most overlap with both LTα1β2 and TGFβ. While our in vitro data shows that LTα1β2 and TNF could each independently induce the fibroblast signature associated with PFD, most patients in the CD+PFDun group (5/7) were under anti-TNF treatment at the time of inclusion (online supplemental table 1), and none had rectal involvement (previous or current luminal disease). This suggests that the PFD signature in this group may be driven predominantly by TL1A/LTα1β2. Nonetheless, a role for TNF in promoting fistula formation cannot be ruled out in the context of intestinal inflammation. In fact, TNF, even at low concentrations (0.5 ng/mL), synergises with LTα1β2 to induce common target genes (ie, IL6, CXCL2), including genes from the PFD signature (ie, MMP1 and MMP3) (figure 4G). Altogether, our results suggest that the TNF and the TL1A/LTα1β2 axis exert independent and synergistic effects in promoting fibroblast activation in PFD. PFD is associated with an inflammatory-like signature in the rectal epithelium The current hypothesis postulates that transitional cells lead to the formation of fistula tracks, as a result of epithelial cells undergoing EMT.5 Hence, we analysed the activity of the EMT pathway in the epithelial cells of our scRNA-seq data but found no evidence of activated EMT signature when comparing CDun and CD+PFDun patients (online supplemental figure 6A). However, our analysis revealed a large number of DEGs between these patients, particularly within the undifferentiated transit amplifying (TA) cells and the top-most differentiated AQP8+ enterocytes (online supplemental figure 1D, online supplemental table 3). Pathway analysis showed the upregulation of inflammation-related processes, such as the defence response to bacterium or biotic stimulus, the response to cytokines such as type II, or the production of TNF superfamily cytokines in AQP8+ enterocytes from CD+PFDun samples (figure 5A). Genes involved in these processes include COX5B, CYBA and DUOXA2. Indeed, DUOXA2, the top fifth gene overexpressed in AQP8+ enterocytes from CD+PFDun patients, was also significantly upregulated in whole tissue rectal biopsies from those patients compared with the CDun group (online supplemental figure 6B). Similar processes were observed in the TA cells of CD+PFDun patients, and upregulated genes involved in the immune response include the TL1A receptor TNFRSF25, TNFRSF1A and CXCL2. In addition, there was a strong enrichment in processes involving the endoplasmic reticulum and the unfolded protein response (XBP1), suggesting that these cells are undergoing high translational activity (figure 5B). Interestingly, we also observed that the most upregulated (highest FC) gene in CD+PFDun patients was OLFM4 (figure 5B,C). Olfactomedin-4 is a glycoprotein involved in mucosal defence responses specifically produced by undifferentiated epithelial cells in the lower crypt, whose expression dramatically increases in the inflamed epithelium.24 We confirmed the upregulation of Olfactomedin-4 by protein staining (figure 5D) and bulk RNA analysis of rectal biopsies (online supplemental figure 6B) in the absence of previous or ongoing disease activity, suggesting inflammation-independent mechanisms are modulating the epithelial response in CD+PFDun patients. The epithelium of CD+PFDun patients presents activation downstream of IL-22 signalling. (A and B) On the left, volcano plot representing the DEGs in the AQP8+enterocytes (A) or the TA cells (B) between CDun and CD+PFDun patients. A two-sided Wilcoxon rank-sum test was applied. Genes with an adjusted p value < 0.05, and an avgLog2FC ˃ log2 (1.2) or avgLog2FC ˂ −log2 (1.2) were considered upregulated (UPP) or downregulated (DWW), respectively. Genes with a nominal p value ˂ 0.05 are also indicated in light colours if avgLog2FC ˃ log2 (1.2) (UP) or avgLog2FC ˂ −log2 (1.2) (DW). On the right, a bar plot indicating pathways upregulated in the AQP8+ enterocytes (A) or TA (B) of CD+PFDun patients generated with SCPA. Bar colour indicates the logarithm of the fold change, while the bar length indicates the q value score from SCPA, which signifies the size distribution of each biological process. (C) Violin plot of OLFM4 expression in epithelial cells. P value was obtained by DEGs analysis. * = adjusted p value < 0.05 and avgLog2FC ˃ log2 (1.2), # = nominal p value ˂ 0.05 and avgLog2FC ˃ log2 (1.2). (D) OLFM4 immunohistochemistry staining of CDun and CD+PFDun formalin-fixed rectum biopsies. Representative images of tissue sections per each group are shown. Percentage of OLFM4+ cells were scored with the software QuPath. Unpaired t-test, ***p<0.001. Scale bar: 100 µm. (E) qRT-PCR of organoid-derived two-dimensional monolayer (left) and cultured tissue explants (right) after 24 hours cultured with or without TL1A (300 ng/mL). Data are represented as the FC relative to the unstimulated condition. One sample t-test, *p<0.05, ns=p>0.05. (F) Dot plot indicating top five cytokines explained by CytoSig for undifferentiated and absorptive cells. Dot size indicates adjusted p value in logarithm base 10, while colour scale indicates the difference in Z-score between CD+PFDun and CDun. (G) Violin plot of IL22 expression in the IL22+ expressing cell types (ILC3, Th17, CD4 effector memory and CD4 effector TNF). P value was obtained by DEG analysis considering only IL22 expressing cells. (H) Heatmap indicating the difference score of IL-22 pathway activity by AddModuleScore of CD+PFDun relative to CDun patients. The Wilcoxon rank-sum test (two-sided) with Holm correction was applied to compare CDun and CD+PFDun patients. Asterisks indicate significant differences (adjusted p value < 0.05). CD, Crohn’s disease; DEGs, differentially expressed genes; FC, fold change; IL, interleukin; PFD, perianal fistulising disease; OLFM4, Olfactomedin 4; TA, transit amplifying; TL1A, tumour necrosis factor-like ligand 1A; TNF, tumour necrosis factor; qRT-PCR, quantitative reverse transcription polymerase chain reaction; SCPA, single-cell pathway analysis. Changes in PFD-associated epithelium are driven by TL1A-dependent IL-22 Increased expression of the receptor for TL1A (TNFRSF25) in TA epithelial cells (figure 5B) suggested a potential effect of TL1A on the epithelium in CD+PFDun patients. To test this, we stimulated with TL1A intestinal two-dimensional (2D) epithelial monolayers or tissue explants from non-inflamed rectal biopsies. While TL1A did not regulate OLFM4 or DUOXA2 expression on epithelial cultures, it significantly induced the expression of both epithelial genes in tissue explants (figure 5E), suggesting that TL1A might drive its effect in the epithelium by acting on non-epithelial cells. To investigate potential upstream regulators of the PFD-associated epithelial signature, we applied CytoSig to the epithelial compartment.25 Several cytokines were significantly predicted to drive the PFD epithelial signature, including IL-12, IL-2 and IL-13 on absorptive cells, and IL-12, IL-2 and IL-22 on undifferentiated cells (figure 5F). Among them, IL-13, IL-2 and IL-22 have previously been described to be induced by TL1A,20 26–28 but only the receptors for IL-13 and IL-22 were expressed on epithelial cells (online supplemental figure 6C). Hence, we looked for differences between groups in the expression of these cytokines in our dataset. While no differences were detected for IL13 (online supplemental figure 6D), we observed that CD+PFDun patients expressed higher levels of IL22 in IL-22-producing cell types (ILC3, Th17, CD4 effector memory and CD4 effector TNF, online supplemental figure 6E) compared with CDun individuals (figure 5G). Mapping the epithelial-specific IL-22 signature from an in-house microarray dataset to our scRNA-seq confirmed that this pathway was more active in most epithelial cell types from CD+PFDun patients, and it was significantly enriched in 10 out of 15 cell types (figure 5H). This was further confirmed in a 2D-monolayer epithelial culture, where IL-22 induced the significant overexpression of DUOX2 and OLFM4, along with other genes upregulated in CD+PFDun patients (online supplemental figure 5F). Moreover, when we measured IL-22 protein production in the supernatants of the TL1A-stimulated tissue explants, we observed a significant correlation between DUOX2 and IL-22 levels, although this did not reach significance for OLFM4 (online supplemental figure 6G). Finally, we confirmed that TL1A, but not TNF, could induce the secretion of IL-22 by anti-CD3-stimulated peripheral T cells (online supplemental figure 6H). Overall, these results suggest that increased TL1A signalling in patients with PFD drives increased levels of IL-22, which in turn can induce an anti-bacterial response on the epithelial compartment of these patients. Discussion PFD in CD remains a significant clinical challenge with poorly understood pathophysiology. Most previous studies have focused on fistula or peri-fistula tissues to identify potential therapeutic interventions aimed at treating already established fistulising disease. In contrast, few have investigated the predisposing molecular mechanisms that drive fistula development. In that sense, several genetic variants have been found that increase the risk of perianal disease in patients with CD,29 supporting the existence of patient-specific predisposing factors. Beyond genetic predisposition, to identify novel mechanisms of fistula formation we examined tissues distant from the fistula tract but within a relevant anatomical context (ie, rectal mucosa) of patients with CD with and without PFD and used an unbiased approach that relied on whole-transcriptome sequencing of all rectal mucosa cells. Using this strategy, we identified changes in cell abundances and in transcriptional signatures across cell types, with the most pronounced alterations observed in lamina propria fibroblasts, T cells and epithelial cells. The fibroblast signature associated with the perianal phenotype was at odds with the classical view that implicated TGFβ in the pathophysiology of PFD,4 5 30 but was in line with recent studies using high throughput analyses of fistula tracts, which did not report any higher activity in TGFβ signalling either.11 31 In contrast, our data supports the role of TNF family members (LTα1β2 and TNF) in modulating the transcriptional states of fibroblasts in the lamina propria of patients with PFD. While the role TNF plays in driving fistula development22 and treating established tracts21 has been well described, no previous data had addressed LTα1β2 in the pathophysiology of PFD. Lymphotoxin β (LTB), which we found to be significantly overexpressed in mucosal lymphocytes from patients with PFD both in our dataset and Kong et al’s dataset, forms a surface-bound heterotrimer with lymphotoxin α (LTα1β2), which exclusively binds to the lymphotoxin β receptor (LTBR) expressed by stromal, epithelial and some myeloid cells.32 Largely known for its role in organising and maintaining the structures of lymphoid tissues, LTα1β2-LTβR signalling is also involved in initiating immune responses, such as those protecting against microbial pathogens in mucosal barriers.33 34 Based on our scRNA-seq data, all mesenchymal cell types express LTBR, being the interaction between LTB-LTBR predicted to be more predominant in CD+PFDun patients. In line with its known function, LTα1β2 induced genes associated with the organisation of the lymphoid follicle and with immune responses (CXCL2, IL6 or VCAM1), significantly overlapping with the TNF-induced gene signature. Notably, LTα1β2 also induced some of the genes overexpressed in the lamina propria S1 fibroblasts of CD+PFDun patients, as well as MMP1, MMP3 and CHI3L1, supporting a role for LTα1β2-LTβR signalling in PFD. Interestingly, a preprint from Constable et al31 reports that while fistula tissue from patients with CD and patients with cryptoglandular disease were transcriptionally similar, LTB and LTA were two of the few genes overexpressed in CD, strengthening our hypothesis. While our data also supports a role for TNF in imprinting the stroma of patients with PFD, TL1A, but not TNF, was identified as the upstream inducer of LTα1β2 expression on activated T cells, thus revealing a novel therapeutic target independent of TNF. Despite TL1A (TNFSF15) not being increased in patients with PFD, the immune response enrichment analysis of CD4+ T cells, as well as the overexpression of both LTB and IL22, strongly reflects higher downstream TL1A activity in CD+PFDun patients. Importantly, the cell interaction analysis also points towards ILC3s as important sources of LTB. Given that ILC3s also express high levels of TNFRSF25 and IL22, a role for ILC3s —on top of CD4+ T cells— in promoting this phenotype cannot be ruled out. TL1A is currently under investigation for its potential therapeutic value in IBD.35 A role for TL1A has been proposed as both an inflammation promoting agent and as a mediator of fibro-stenosis based on preclinical studies.16 The potential involvement of TL1A in complicated phenotypes is supported by genome-wide association studies describing susceptible loci for TNFSF15 in European and Japanese cohorts.36 37 In a Korean cohort, another non-risk single nucleotide polymorphism allele of TNFSF15 (rs4574921) was also associated with a higher cumulative risk for developing PFD.38 However, to our knowledge, no previous study had functionally involved TL1A in the pathophysiology of PFD. Beyond its novel effects on LTα1β2 production by CD4+ T cells, we also confirmed the ability of TL1A to induce IL-22 expression by lymphoid cells.20 26 A role for IL-22 in established PFD has recently been suggested, including a role in tissue remodelling6 31 and EMT.6 While we found no evidence of EMT activation in the rectal epithelium of CD+PFDun patients, immune and anti-bacterial responses, including IL-22 targets, were upregulated in these patients, despite the non-involved and non-inflamed nature of the tissue. The overall epithelial response, imprinted by IL-22, suggests an increase in bacteria recognition, which would be supported by recent data showing an increased detection of serum antibodies against bacterial antigens such as anti-ASCA, anti-OmpC or anti-flagellins in CD+PFD patients.29 39 While our study represents a step towards identifying the molecular mechanisms driving the development of perianal fistulae in CD, it has some limitations. First, the number of subjects per study group is limited and while we have validated some of our findings in Kong et al’s dataset, a larger validation cohort with rectum samples will be required. However, the TL1A-LTα1β2/IL-22 axis was validated in primary intestinal cultures. Most importantly, regardless of the upstream drivers (TL1A/LTα1β2 or TNF), the functional implications of the PFD-associated signature remain unknown. Based on our analysis, we propose a potential metabolic reprogramming of CD+PFDun lamina propria fibroblasts, with a deregulation in cellular respiration and lipid metabolism processes. Data from fibrotic skin fibroblasts shows that a metabolic switch from fatty acid oxidation to glycolysis can promote ECM accumulation.40 Likewise, it has been observed that fibroblasts can also sense mechano-physical alterations in the ECM, as a softer matrix promotes cholesterol and neutral lipid synthesis, whereas stiff ECM sustains glycolysis.41 Based on our data, one could hypothesise that the rectum of CD+PFDun patients might present an altered ECM stiffness, which could be sensed and sustained by lamina propria S1 fibroblasts. This hypothesis is consistent with the upregulation of MMP1 and MMP3 in CD+PFDun patients, involved in matrix degradation. Nonetheless, further studies are required to demonstrate a direct link between fibroblast metabolic changes and fistula development. To conclude, our data provides novel insights into the molecular mechanisms underlying the development of PFD and points to TL1A as a key upstream regulator in this complication. While our results also support a role for TNF in this phenotype, we identified a new signalling axis independent of, and potentially complementary to, TNF. This raises new questions, including how the changes we uncovered lead to the formation of fistulous tracts and, more importantly, whether interfering with TL1A signalling, alone or even in combination with other biologics like anti-TNF, could prevent this process. Supplementary material Acknowledgements We thank the Single Cell Genomics Group at CNAG (Barcelona), the Cytometry Core Facility of the IDIBAPS for their technical help and Joe Moore for editorial assistance. We would also like to thank Dr Xavier Ramnik, Dr Marco Colonna and Dr Khai M Nguyen for assisting us with the metadata information from the Kong et al dataset. Footnotes Funding: This work was supported by the Leona and Harry Helmsley Charitable Trust (Grant #2008-04050) and the PID2021-123918OB-I00 grant, supported by MCIN/AEI/10.13039/501100011033 and “FEDER”. EM-A was funded by the fellowship RH042155 (RTI2018 096946- B-I00) from the Ministerio de Ciencia e Innovación, Gobierno de España. AM, CL and ME received funding from the Department of Health of Catalonia: Programme for the Incorporation of Support Staff into Research Groups (Code: SLT028/23/000194). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: The study was approved by the Ethics Committee of Hospital Clinic Barcelona (HCB/2020/0035 and HCB/2016/0389) and the Hospital Universitari Mutua Terrassa (CI201901). Participants gave informed consent to participate in the study before taking part. Data availability free text: The raw data generated in this work will be publicly available upon acceptance of the manuscript in the Gene Expression Omnibus (GEO). Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Data availability statement Data are available in a public, open access repository. References Associated Data Supplementary Materials Data Availability Statement Data are available in a public, open access repository.