Article Text
PDF
Abstract
Background One reason patients with cancer cannot benefit from immunotherapy is the lack of immune cell infiltration in tumor tissues. Cancer-associated fibroblasts (CAFs) are emerging as central players in immune regulation that shapes tumor microenvironment (TME). Earlier we reported that integrin α5 was enriched in CAFs in colorectal cancer (CRC), however, its role in TME and cancer immunotherapy remains unclear. Here, we aimed to investigate the role for integrin α5 in fibroblasts in modulating antitumor immunity and therapeutic efficacy combined with checkpoint blockade in CRC.
Methods We analyzed the CRC single-cell RNA sequencing (scRNA-seq) database to define the expression of ITGA5 in CRC tumor stroma. Experimentally, we carried out in vivo mouse tumor xenograft models to confirm the targeting efficacy of combined α5β1 inhibition and anti-Programmed death ligand 1 (PD-L1) blockade and in vitro cell-co-culture assay to investigate the role of α5 in fibroblasts in affecting T-cell activity. Clinically, we analyzed the association between α5 expression and infiltrating T cells and evaluated their correlation with patient survival and immunotherapy prognosis in CRC.
Results We revealed that ITGA5 was enriched in FAP-CAFs. Both ITGA5 knockout fibroblasts and therapeutic targeting of α5 improved response to anti-PD-L1 treatment in mouse subcutaneous tumor models. Mechanistically, these treatments led to increased tumor-infiltrating CD8+ T cells. Furthermore, we found that α5 in fibroblasts correlated with extracellular matrix (ECM)-related genes and affected ECM deposition in CRC tumor stroma. Both in vivo analysis and in vitro culture and cell killing experiment showed that ECM proteins and α5 expression in fibroblasts influence T-cell infiltration and activity. Clinically, we confirmed that high α5 expression was associated with fewer CD3+ T and CD8+ T cells, and tissues with low α5 and high CD3+ T levels correlated with better patient survival and immunotherapy response in a CRC cohort with 29 patients.
Conclusions Our study identified a role for integrin α5 in fibroblasts in modulating antitumor immunity by affecting ECM deposition and showed therapeutic efficacy for combined α5β1 inhibition and PD-L1 blockade in CRC.
- Immunotherapy
- Biomarkers, Tumor
- CD8-Positive T-Lymphocytes
- Immunomodulation
- Tumor Microenvironment
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Previous reports on integrin α5 have focused on vascular endothelial cells, but clinical trials targeting integrin α5β1 by inhibiting angiogenesis failed to show significant benefit in solid tumors. We recently reported that integrin α5 was enriched in cancer-associated fibroblasts (CAFs) in colorectal cancer (CRC), however, its role in tumor microenvironment and cancer immunotherapy remain unclear.
WHAT THIS STUDY ADDS
we revealed that ITGA5 was enriched in FAP-CAFs, and we identified a role for integrin α5 in fibroblasts in modulating antitumor immunity and showed therapeutic efficacy for combined α5 targeting and checkpoint blockade.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Our study highlights the potential therapeutic value of targeting α5 in combination with immunotherapy to inhibit CRC, which could inform future patient studies to explore the potential of α5β1 in fibroblasts, lead to the development of more effective treatments and improved patient outcomes.
Introduction
The development of colorectal cancer (CRC) is a complex process involving interactions between cancer cells and the tumor microenvironment (TME). The TME, also known as the tumor stroma, consists of fibroblasts, immune cells, blood and lymphatic vessels, and the extracellular matrix (ECM). It is important in regulating cancer growth and therapeutic response through cytokines, cell–cell interactions, and mechanical sensing.1 2 Accumulating evidence suggests that the features of tumor-infiltrating immune cells are correlated with the development and progression of cancer; patients with high proportions of infiltrating CD8+ and CD4+ T cells have a better prognosis in CRC.3–5 Additionally, over 85% of CRC cases are microsatellite stable tumors with low tumor mutational burden, which are immune desert or immune excluded tumor types with poor immune cells infiltration, where immune checkpoint blockade fails to elicit a response.6–8 Thus, it is essential to identify regulatory pathways that alter immune cell infiltration.
Recent studies have found that interactions between immune cells and mesenchymal cells in the TME are more predictive of tumor immunotherapy prognosis and response.9–12 Cancer-associated fibroblasts (CAFs) are one of the most abundant cell types in the tumor stroma, and their accumulation is often associated with poor prognosis in many tumors.10 13 Moreover, CAFs are capable of secreting various chemokines, cytokines, interferons (IFNs), and Transforming growth factor-β (TGF-β) to regulate immune cell migration.12 Some studies suggest that CAFs may impede antitumor immune responses.10 Furthermore, depleting FAP-expressing fibroblasts enhances T cell-dependent tumor regression in mouse subcutaneous models of lung and pancreatic cancer.14 In pancreatic cancer, FAP-positive CAFs can produce the chemokine CXCL12, and inhibiting CXCL12 promotes T-cell infiltration in tumors sensitive to anti-Programmed cell death protein-1 (PD-1) and anti-Cytotoxic T-lymphocyte antigen 4 (CTLA-4) treatments.15 Therapeutic co-administration of TGF-β-blocking and anti-Programmed death ligand 1 (PD-L1) antibodies reduces TGF-β signaling in stromal cells, facilitates T-cell penetration into the center of tumors, and provokes vigorous antitumor immunity and tumor regression.16 The use of TGF-β inhibitors is currently being evaluated in multiple clinical trials.17 Moreover, CAFs can regulate the ECM by secreting large amounts of collagen and fibronectin to promote extratumor matrix solidification and inhibit the infiltration of immune cells and penetration of antitumor drugs.18 The ECM in the tumor stroma can influence antitumor immunity by regulating the localization and migration of T cells.19–21 Understanding the mechanisms underlying the generation of this fibrous network has the potential to contribute to the development of new therapies.
Integrin subunit α5 (ITGA5) combines with the β1 integrin subunit (ITGB1) to form the α5β1 complex and acts as a receptor for fibronectin.22 Previous reports on integrin α5 have focused on cancer cells and vascular endothelial cells, which are known to participate in tumor-promoting processes such as angiogenesis, proliferation, and metastasis.22 Based on understanding its role in pro-angiogenesis, integrin α5β1 has been chosen as a target in clinical trials. However, phase I and II trials using the small peptide antagonist of integrin α5β1 ATN-161 (Ac-PHSCN-NH2) or α5β1 chimeric monoclonal antibody volociximab have not shown significant benefits in solid tumors.23 Recent studies found that ITGA5 was a prognostic biomarker associated with immune infiltration in gastrointestinal tumors.24 Reduced expression of ITGA5 in pancreatic ductal carcinoma can downregulate the differentiation of pancreatic stellate cells into CAFs, which are the main source of CAFs.25 Our previous study also showed that integrin α5 expression was significantly elevated in intestinal cancer tissues,26 and it was found that integrin α5 expression was enriched in CAFs, knockout of α5 in fibroblasts inhibited fibroblast-induced cancer cell migration and invasion by downregulating fibronectin assembly.26 However, the role of α5 expression in CAFs in TME and CRC immunotherapy remains unclear.
Given the improved understanding of the TME and immune system, this study aimed to investigate the role for integrin α5 in fibroblasts in modulating antitumor immunity and the impact of targeting α5 in CAFs on CRC inhibition by synergizing with immunotherapy.
Materials and methods
In vivo tumor studies
For subcutaneous colon cancer tumors with fibroblasts, MC38 cells (5×105) with or without mouse fibroblasts (vector control or α5 KO, 5×105 cells separately) were injected subcutaneously into the flanks of C57B6/L mice aged 6–8 weeks (one injection/mouse). Mouse fibroblasts alone (5×105) were used as control. Tumor volumes were measured and calculated two times per week using the following modified ellipsoid formula: 0.5×(length×width2). When the tumor volume reached ~100 mm3 (about 8 days after inoculation), 140 µg of anti-mouse PD-L1 antibody (BP0101, Bio X Cell) or isotype control antibody, was intraperitoneally administered every 4 days until the end of the experiment. Tumors >2000 mm3 were considered progressed and animals were removed from the study. After 23 days, the mice were sacrificed and the tumor samples were collected, and used for further blind analysis.
For subcutaneous colon cancer tumors, age-matched and sex-matched 6–8 weeks old BALB/c mice were subcutaneously inoculated in the right unilateral flank with 5×105 CT26.WT tumor cells. The mice were randomized into different groups. When the tumor volume reached ~100 mm3 (about 8 days after inoculation), 120 µg of anti-mouse PD-L1 antibody (BP0101, Bio X Cell) or 100 µg of anti-mouse α5 antibody (103910, BioLegend) (in the other mouse model, 100 mg/kg integrin α5β1 inhibitor ATN-161 (S8454, Selleck)27 was used) or the combined antibody, was intraperitoneally administered four times every 4 days until the end of the experiment.
All mouse experiments were performed in strict accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and approved by the Experimental Animal Ethical Committee of Shanghai Tenth People’s Hospital (SHDSYY-2022–4156). The study used the ARRIVEreporting guidelines.28
Statistics
R, SPSS V.22.0 and GraphPad Prism V.8 were used for statistical analysis. Single-cell analyses were performed mainly with Seurat. All plots were produced with R software and refined with Adobe Illustrator. Each experiment was repeated independently a minimum of three times in the same conditions. Acquired data were presented as mean values, and error bars represent the SEM. Student’s t-test was used for comparison between two groups, and a one-way analysis of variance (ANOVA) was used to compare multiple groups with Bonferroni’s multiple comparison correction. The relationships between tumor volume curves were compared using two-way ANOVA. The Kaplan-Meier method was performed to estimate overall survival (OS) and progression-free survival (PFS) with a log-rank test. Patients’ characteristics were compared between two groups using χ2 or Fisher’s exact test wherever appropriate. A Cox proportional hazards model was used for the univariate and multivariate analyses to estimate HRs and 95% CIs. For all analyses, two-tailed p values below 0.05 were considered significant: *p<0.05, **p<0.01, ***p<0.001.
Data availability
All data supporting this paper are present within the paper and the online supplemental material. The original data sets are also available from the corresponding author on request.
Supplemental material
Results
ITGA5 was enriched in FAP-CAFs in CRC tumor stroma
We previously found that ITGA5 was mainly expressed in CAFs in human CRC samples. To further confirm its expression in tumor stroma, we analyzed Single-cell RNA sequencing (scRNA-seq) database on human CRC tissue samples (n=27; nine samples from the tumors (core and border regions for each tumor) and adjacent non-malignant tissue from the same resection specimen) from the Gene Expression Omnibus (GEO) data set GSE144735. We found that ITGA5 was mainly enriched in fibroblasts and endothelial cells (online supplemental figure S1A, B). To further investigate the role of ITGA5 in fibroblasts, we identified nine subtypes of fibroblasts, including two crypt fibroblasts (WNT2B, RSPO3), two villus fibroblasts (WNT5B), two myofibroblasts (ACTA2, TAGLN) and three CAFs (FAP) (online supplemental figure S1C). ITGA5 mainly enriched in CAFs, which was characterized by the expression of FAP, and higher expression level of FN1 and COL3A1 (figure 1A and online supplemental figure S1D). And CAFs expanded specifically in tumor border and tumor core, but not in normal tissues (figure 1B).
Targeting integrin α5 in fibroblasts potentiates colorectal cancer response to PD-L1 blockade. (A) Violin plots showing ITGA5 expression levels in CAF subclusters of fibroblasts in colorectal cancer scRNA-seq database. (B) Proportions of different fibroblast populations in normal tissues, tumor border and tumor core. (C–E) α5 deletion in fibroblasts enhances responsiveness to anti-PD-L1 treatment. Mouse MC38 cells were co-transplanted with mouse fibroblasts (vector control or α5 KO) subcutaneously into C57BL/6 mice on day 0, anti-PD-L1 antibody or isotype IgG were administrated four times every 4 days starting from day 8, followed by examining tumor growth. (C) Average tumor growth curves (n=5 mice per group; *p<0.05, two-way ANOVA with Bonferroni’s post-tests). The x axis represents days after implantation. (D) Photographs of dissected tumor samples, and (E) tumor weight on day 23 after implantation. n=5 mice per group. Error bars, mean±SEM; *p<0.05, **p<0.01, ***p<0.001; ns, not significant; one-way ANOVA. (F, G, H) Tumor regression following therapeutic anti-α5 and anti-PD-L1 treatment in CT26.WT tumors. CT26.WT cells were injected subcutaneously, and tumor volume was monitored two times a week. Mice were grouped when the tumor volume reached approximately ~100 mm3, anti-PD-L1 and anti-α5 were administered intraperitoneally two times a week from day 8 to day 21. (F) Tumor growth curves (n=7 per each group). (G) Average tumor growth curves (n=7 mice each group; mean±SEM; ns, not significant; **p<0.01; two-way ANOVA with Bonferroni’s post-tests). (H) Analysis using SynergyFinder V.3.0 showed synergistic combinations of anti-PD-L1 and anti-α5 on day 21. ANOVA, analysis of variance; CAFs, cancer-associated fibroblasts; CR, complete response; PD-L1, programmed death ligand 1; scRNA-seq, single-cell RNA sequencing.
Survival analysis of a set of 1273 CRC samples from GEO database (GSE14333, GSE17536, GSE31595, GSE33113, GSE38832, GSE39084, GSE39582) showed that high ITGA5 level was associated with decreased patient survival, accordingly we found that high gene ratio of ITGA5/Col3a1 or high gene ratio of ITGA5/PECAM1 were associated with decreased patient survival (online supplemental figure S1E–G), suggesting a differential expression of ITGA5 in CRC with good and poor outcome.
Additionally, we analyzed scRNA-seq database on human inflammatory bowel diseases tissue samples (n=18; including healthy controls (HC, n=6), ulcerative colitis (n=6) and Crohn’s disease (CD, n=6)) from the GEO data set GSE214695. We found that ITGA5 was enriched in endothelial cells of HC and CD samples, and mainly in fibroblasts of CD samples (online supplemental figure S1H–J).
Targeting integrin α5 in fibroblasts potentiates colorectal cancer response to PD-L1 blockade
Since ITGA5 is mainly expressed in CAFs and considering CAFs plays a central role in regulating tumor growth and antitumor immunity, we hypothesized that ITGA5 in fibroblasts may regulate antitumor immunity. To test the hypothesis, we knocked out ITGA5 in immortalized mouse colon fibroblasts (mFib) and obtained two fibroblast cell lines: α5 depleted cells (mFib-α5KO) and vector control cells (mFib-Vector). ITGA5 depletion showed significantly downregulated expression of ECM genes in messenger RNA (mRNA) level and reduced expression of fibronectin and FAP in protein level (online supplemental figure S2A,B), while maintained comparable expression levels of fibroblast markers (PDGFRβ and α-SMA) (online supplemental figure S2C).
Then mFib-Vector or mFib-α5KO were co-transplanted with mouse colon cancer cells-MC38 into subcutaneous of C57BL/6 mice. The results showed that mFib-α5KO cells failed to promote tumor growth compared with the mFib-Vector cells (online supplemental figure S2D–F). Thus, α5 in fibroblasts is necessary to promote CRC growth in mouse tumors.
Previous studies have shown that FAP+ myofibroblast depletion or LRRC15+ CAF ablation can improve anti-PD-L1 responsiveness.15 29 To determine whether α5 KO in fibroblasts can disrupt immunosuppressive TME and render CRC tumors shrink by immunotherapy, MC38 cells were co-injected with mFib-Vector or mFib α5KO. Then the mice were administered with anti-PD-L1 antibody or corresponding isotype IgG intraperitoneally, and tumor growth was evaluated (figure 1C–E). Responsiveness to anti-PD-L1 treatment was significantly potentiated in MC38 /α5 KO subcutaneous tumors, as indicated by a substantial reduction in tumor burden (figure 1C). Both MC38/Vector and MC38 /α5 KO injected mice treated with anti-PD-L1 showed significantly lower tumor weights than mice treated with IgG separately, and MC38 /α5 KO subcutaneous tumors treated with anti-PD-L1 showed further reduction compared with MC38/Vector tumors, reflecting the tumor volume kinetics observed during treatment (figure 1C–E). This result suggests that inhibiting integrin α5 in CAFs potentiate the antitumor efficacy of anti-PD-L1. We further confirmed this by using ATN161, a small peptide antagonist of integrin α5β1 and a blocking antibody-anti-α5 (clone 5H10-27). We evaluated the antitumor activity of ATN161 or anti-α5 antibody combined with anti-PD-L1 antibody in CT26 tumor‐bearing BALB/c mice, as CT26 tumor mice are less sensitive to anti-PD-L1 blockade compared with MC38. The results showed that the therapeutic blockade of PD-L1 or α5 alone had little effect; however, mice treated with reagents against both PD-L1 and α5 exhibited a significant reduction in tumor burden (online supplemental figure S2G–K and figure 1F–H). Our result suggests that directly targeting α5 in vivo could achieve inhibition on the TME and tumor growth in CRC by synergizing with immunotherapy.
Targeting integrin α5 in fibroblasts combine with anti-PD-L1 treatment enriched CD8+ T cells in subcutaneous tumors
To investigate the mechanism underlying the synergistic effect of inhibiting ITGA5 and anti-PD-L1 treatment, we used flow cytometry to analyze infiltrating T cells in tumor samples (online supplemental figure S3A). In experiments where MC38 cells were co-injected with either mFib-Vector or mFib-α5KO cells, we observed that anti-PD-L1 treatment significantly increased the frequency of CD3+ T cells (figure 2A) and CD8+ effector T cells (CD8+ IFN-γ+) (figure 2B). When integrin α5 was depleted, there appeared to be a trend towards increased numbers of tumor-infiltrating T cells and CD8+ effector T cells. When α5 depletion was combined with anti-PD-L1 treatment, the number of CD8+ effector T cells was further increased compared with either treatment alone (figure 2A,B). Additionally, the frequency of CD4+ T regulatory (Treg, CD4+Foxp3+) cells were significantly lower (figure 2C), resulting in a higher ratio of effector CD8+ T cells to Treg cells (figure 2D) in MC38 /α5 KO subcutaneous tumors after anti-PD-L1 treatment compared with that in the control group of MC38/Vector tumors. Quantitative histopathology demonstrated that T-cell distribution was significantly changed following α5 KO in fibroblasts and after anti-PD-L1 treatment, with the mean distance from the tumor border increasing (figure 2E,G). We also observed lower total collagen levels in tumors with α5 KO fibroblasts and those treated with anti-PD-L1 (figure 2F,H).
Targeting integrin α5 in fibroblasts combine with anti-PD-L1 treatment enriched CD8+ T cells in subcutaneous tumors. Flow cytometry quantification of (A) the frequency of CD3+, (B) CD8+IFN-γ+ and (C) CD4+Foxp3+of total CD45+ cells, and (D) ratio of effector CD8/Treg in tumor tissues. (E, F) Representative IHC staining of CD3 and Masson’s trichrome staining of collagen in tumor periphery (E) and tumor center (F). Scale bar: 100 µm. (G) Quantification of tumor-infiltrating CD3+ T lymphocytes localization by IHC. (H) Quantification of per cent collagen area per field of tumor tissues by Masson’s trichrome staining. n=5 mice per group. Error bars, mean±SEM; *p<0.05, **p<0.01, ***p<0.001; ns, not significant; one-way ANOVA. (I–K) Flow cytometry quantification of the frequency of (I) CD8+IFN-γ+, (J) CD4+Foxp3+ and ratio of (K) effector CD8/Treg in tumor tissues following therapeutic anti-α5 and anti-PD-L1 treatment in CT26.WT tumors. n=7 mice each group. Error bars, mean±SEM; **p<0.01, **p<0.001; ns, not significant; one-way ANOVA. ANOVA, analysis of variance; IFN, interferon; IHC, immunohistochemistry; PD-L1, programmed death ligand 1;Treg, regulatory T cell.
We observed comparable alterations in CT26.WT tumors treated with therapeutic α5 antagonist (anti-α5 antibody or ATN161) in combination with anti-PD-L1. In vivo mouse experiments consistently showed a significantly increased frequency of tumor-infiltrating CD8+ effector T cells (CD8+ IFN-γ+) (figure 2I) or CD8+ T cells (online supplemental figure S3I) and a decreased frequency of Treg cells (figure 2J and online supplemental figure S3J), resulting in a significantly increased effector CD8/Treg (figure 2K) or CD8/Treg ratio (online supplemental figure S3K) of tumor-infiltrating lymphocytes in the combination group of mouse tumors. Targeting α5 alone or in combination with anti-PD-L1 had no effect on the number of CD4+ T cells, macrophages (CD11b+ F4/80+), or dendritic cells (CD11b+ CD11c+) in the tumor (online supplemental figure S3B–D) and none of the immune cells in the spleen (online supplemental figure S3E–H). The above results showed that α5 inhibition potentiated the ability of anti-PD-L1 to enhance antitumor immunity, resulting in optimal T-cell infiltration and tumor regression.
Taken together, these results suggest that α5 in fibroblasts plays a role in regulating the infiltration and activity of T cells in tumors, which help explain the observed synergistic effect between inhibiting integrin α5 and anti-PD-L1 treatment.
α5 in CAFs affects the levels of ECM proteins deposition in CRC stroma
To further investigate the underlying mechanism that affecting T-cell infiltrating in the subcutaneous tumors, we performed pathway and functional enrichment analysis with the upregulated genes in ITGA5+CAFs using the CRC scRNA-seq database, and found that many of the identified biological processes were involved in ECM organization and collagenous ECM generation (figure 3A,B), suggesting that ITGA5+CAFs formation was associated with ECM deposition. To validate the transcriptomic findings, we performed immunohistochemistry (IHC) staining to further assessed the expression levels of integrin α5, CAFs markers (FAP, fibronectin, PDGFRβ, α-SMA, and β1), and the ECM protein-collagen in tissue microarray of selected 29 eligible samples diagnosed with colorectal adenocarcinoma (figure 3C and online supplemental figure S4A). We found that the expression levels of α5 were significantly and positively correlated with FAP, fibronectin, and collagen; and the two ECM proteins fibronectin and collagen were also positively correlated (figure 3D). No correlation was observed between α5 expression and PDGFRβ, α-SMA, and β1 (online supplemental figure S4B). Pre-existing normal stromal fibroblasts could potentially convert into CAFs in response to TGF-β, specifically during the course of tumor progression.30 31 We then used TGF-β to induce activation of human colonic fibroblast cells-CCD-18Co to examine whether α5 expression affects fibroblast activation markers. The presence of gradually elevated concentrations of TGF-β dramatically enhanced the protein levels of α5, fibronectin and FAP, and the depletion of α5 substantially reduced the expression of fibronectin and FAP (online supplemental figure S4C). These data suggest that α5 expressing fibroblasts were associated with ECM proteins and may affect FAP-high-expressing CAFs in CRC tissues.
α5 in fibroblasts affects the levels of ECM proteins deposition in CRC stroma. (A) Top biological processes of GO analysis with the upregulated gene clustered in ITGA5+CAFs compared with that of non-CAFs in CRC scRNA-seq database. (B) Top significant ECM genes in ITGA5+CAFs population. (C) Representative IHC staining of α5 and CAFs markers in colorectal adenocarcinoma tissues. Scale bar: 200 µm. (D) Pearson’s correlation between the expressions of α5 and CAFs or ECM markers in 29 colorectal adenocarcinoma tissues evaluated by IHC. (E) Heatmap showing association of mRNA expression in 517 colorectal adenocarcinomas extracted from TCGA database between ITGA5 and 18 ECM genes that are statistically significant (p<0.05) by Pearson’s correlation. (F) Differential mRNA expression of ECM genes in CCD-18Co cells after ITGA5 depletion evaluated by qPCR in triplicate. Each gene expression was relative to vector control (set at 1.0), statistically significant (p<0.05) by Student’s t-test. (G) Representative Masson’s trichrome staining of collagen in tumor tissues approaching tumor border following therapeutic anti-α5 and anti-PD-L1 treatment in CT26.WT tumors. Scale bar: 200 µm. (H) Quantification of per cent collagen area per field of tumor tissues using Masson’s trichrome staining. n=7 mice per group. Error bars, mean±SEM; *p<0.05, **p<0.01; ***p<0.001; ns, not significant; one-way analysis of variance. CAFs, cancer-associated fibroblasts; CRC, colorectal cancer; ECM, extracellular matrix; GO analysis, gene ontology analysis; IHC, immunohistochemistry; mRNA, messenger RNA; PD-L1, programmed death ligand 1; scRNA-seq, single-cell RNA sequencing. TCGA,
We further investigated the association of integrin α5 with ECM by analyzing the correlation between ITGA5 and ECM gene expression in the CRC RNA-sequencing database extracted from the Cancer Genome Atlas (TCGA) cohort. The heatmap showed that ITGA5 was significantly correlated with the 18 ECM-related genes (figure 3E and online supplemental figure S4D) and the mRNA levels of the 13 related ECM genes were significantly downregulated after ITGA5 depletion in CCD-18Co fibroblasts after TGF-β treatment (figure 3F). ONCOMINE database analysis showed that these ECM-related genes were significantly upregulated in various human cancers, including CRC (online supplemental figure S4E). Subsequently, we performed Masson’s trichrome analysis of subcutaneous tumor tissues in immunocompromised (nu/nu) mice post-implantation with colorectal adenocarcinoma cells (SW480), alone or with TGF-β pretreated CCD-18Co (with or without α5 expression). Our findings showed that wild-type fibroblasts significantly upregulated collagen deposition, whereas ITGA5 depletion in fibroblasts did not (online supplemental figure S4F,G). Immunofluorescence staining also revealed that the level of fibronectin showed a similar trend as collagen fibers (online supplemental figure S4H,I). Moreover, to determine whether α5-targeting is relevant to ECM deposition in tumor tissues following therapeutic anti-α5 and anti-PD-L1 treatment in CT26.WT tumors, we evaluated collagen density in tumor tissues and observed a thick layer of collagen in the tumor approaching the tumor border in the IgG and saline group, which was significantly decreased in the Combo Group (figure 3G,H). These results suggest that α5 expression in fibroblasts can significantly alter ECM protein loading in tissues, and α5 may affect tumor growth by influencing ECM protein deposition.
ECM density affects the tissue localization and activity of T cells
The migration of T cells and access to tumor antigens are of utmost importance for the induction of protective antitumor immunity. Recent studies have highlighted the impact of the matrix architecture in solid tumors on immune cells, especially T cells, in which the motile behavior of T cells and their capacity to infiltrate tumor cell regions are controlled by the density and orientation of ECM fibers.32 To investigate whether the expression of α5 in fibroblasts and ECM proteins is relevant to the immunosuppressive TME, we further assessed the expression levels of integrin α5, CD3, fibronectin and collagen in CRC tissue microarray (figure 4A). The α5 expression level was significantly associated with fibronectin and collagen deposition (figure 4A). Additionally, high deposition of fibronectin and collagen correlated with fewer infiltrated CD3+ T cells (figure 4B,C). Further, we performed immunofluorescence staining to examine the spatial organization of α5, ECM deposition, and CD3+ T cells in human CRC tumor stroma. The α5 expression level (green, stained for α5) was observed to be correlated with the density of ECM fibers composed of fibronectin or collagen (red, stained for fibronectin or collagen), whereas T cells (white, stained for CD3) were preferentially distributed within the areas with sparse fibronectin and collagen fibers (figure 4D,E). Moreover, in some areas, T cells were in contact with α5 expression cells and fibronectin fibers (figure 4F). This implies that α5 and ECM proteins may affect T-cell migration and activity.
ECM deposition affects the tissue localization and activity of T cells. (A) Representative of IHC and Masson’s trichrome staining of α5, fibronectin, CD3 and collagen in human colorectal adenocarcinoma tissues. Scale bar: 50 µm. (B) Pearson’s correlation between the level of fibronectin and CD3+ T-cell count in 29 colorectal adenocarcinoma tissues evaluated by IHC. (C) Pearson’s correlation between the level of collagen and CD3+ T-cell count in 29 colorectal adenocarcinoma tissues. Quantification of collagen deposition using Masson’s trichrome staining, quantification of CD3 positive cells number per field by IHC. Data were analyzed using log2 values. (D, E, F) Distribution of T cells in relation to α5 expression and ECM fibers in human colorectal adenocarcinoma tissues. Representative immunofluorescence staining of human colorectal adenocarcinoma sections with fibronectin (D) or collagen (E) (red), CD3 (T cells, white) and α5 (cancer-associated fibroblasts, green), with white dashed lines showing more aggregation of T cells in areas with sparse fibronectin/collagen fibrils, with white arrows showing T cells were directly in contact with α5 expression cells and fibronectin fibers (F). Scale bar: 100 µm. (G) Schematic model of the collagen matrix culture system. (H–L) Flow cytometry quantification of the percentage of CD8+ PD-1+ Tim-3+ (H) CD4+ PD-1+ Tim-3+ (I) CD8+ CD137+ (J) CD4+ CD137+ (K) CD8+IFN-γ+ (L) of CD3+ T cells harvested from 2D culture or collagen matrix. Error bars, mean±SEM; *p<0.05, **p<0.01, ***p<0.001; ns, not significant; one-way analysis of variance. 2D, two-dimensional; DAPI, 4′,6-diamidino-2-phenylindole; ECM, extracellular matrix; FACS,flow cytometry analysis; IFN, interferon; IHC, immunohistochemistry; PD-1, programmed cell death protein 1; Tim-3, T-cell immunoglobulin and mucin domain 3.
Some studies have shown that collagen concentrations affect T-cell activity.33 34 In our study, CD3+ T cells were isolated from healthy donors and embedded in a collagen matrix of high concentration (2.4 mg/mL, high density) or low concentration (0.6 mg/mL, low density) or inoculated on conventional tissue cell culture dishes (two-dimensional (2D) culture) in the presence of CD3/CD28 stimulus (figure 4G). After 4 days of culture, high viability of over 85% was observed in 2D culture and different collagen densities (results not shown). A significant reduction in T-cell proliferation was observed when cells were cultured in high-concentration collagen compared with low-concentration collagen or 2D culture (online supplemental figure S5A). To investigate whether T-cell activity is affected by collagen, we examined the expression levels of a set of T-cell exhaustion markers (PD-1+ T-cell immunoglobulin and mucin domain 3 (Tim-3) +), activity markers (CD137 and CD69), and Treg markers (CD4+CD25+) (online supplemental figure S5B). The results showed that differential culture conditions did not change the ratio of CD4+ to CD8+ T cells (online supplemental figure S5C). CD3+ T cells cultured in a high-concentration collagen matrix showed a significantly increased proportion of CD8+ PD-1+ TIM-3+, CD4+ PD-1+ TIM-3+ T cells (figure 4H,I), and CD4+ CD25+ T cells (online supplemental figure S5D) compared with those cultured under 2D conditions or low-concentration collagen. In contrast, CD3+ T cells cultured in a high-concentration collagen matrix maintained low expression of CD8+CD137+, CD4+CD137+ (figure 4J,K), and CD4+CD69+, CD8+CD69+ (online supplemental figure S5E,F) and a significantly reduced proportion of effector T cells (CD8+IFN-r+) (figure 4L). Heatmap analysis showed that ITGA5/ECM genes and Treg marker genes (FOXP3, TGFB1, and IL-10) were significantly correlated (online supplemental figure S5G). These results suggest that high concentrations of collagen decreased T-cell activity and induced T-cell exhaustion.
Collagen has been shown to cause T-cell failure through the SHP-1 signaling pathway by binding to the T-cell surface receptor LAIR1(33). Our study showed that LAIR1 gene expression was significantly upregulated in T cells under collagen matrix culture conditions (online supplemental figure S5H). Moreover, heatmap analysis revealed a significant correlation between ITGA5/ECM genes and T-cell failure marker genes (HAVCR2, TIGIT, LAG3, PDCD1, and BTLA), as well as receptor LAIR1 in TCGA CRC samples (online supplemental figure S5I). These results suggest that ECM collagen might cause pathway alterations by binding to the T-cell surface receptor LAIR1 protein, thus leading to T-cell failure.
α5 knockout impairs the tumor-promoting effects of fibroblasts and affects T-cell killing function
Tumor-infiltrating lymphocytes play a crucial role in tumor regression by carrying out their functional activities. However, studies have shown that the tumor environment can compromise antitumor immunity. To investigate the function of α5 in fibroblasts on antitumor immunity, we first constructed three-dimensional organotypic co-culture systems of colorectal tumor organoids with TGF-β pretreated CCD-18Co fibroblasts, and found that co-culturing with wild-type CCD-18Co (vector control) resulted in larger organoids (figure 5A,B) but did not affect their number (data not shown). However, integrin α5 knockout in CCD-18Co (α5 KO) diminished the effect of fibroblasts on organoid size (figure 5A,B). To investigate the impact of fibroblasts on T cell-killing ability, we used this system to generate organoids with or without fibroblasts, which were then co-cultured for 72 hours in the presence of T cells. The killing assay showed that the presence of wild-type fibroblasts (vector control) significantly increased the percentage of viable tumor cells from approximately 40%–55% compared with the organoid group formed by tumor cells alone (figure 5C,D), indicating that fibroblasts may effectively prevent the killing effect of T cells on tumor cells. In contrast, after integrin α5 depletion (α5 KO), CCD-18Co did not significantly increase the percentage of viable tumor cells (figure 5C,D). These results suggest that the presence of fibroblasts promotes organoid growth and suppresses T-cell killing capability, which is diminished by α5 depletion.
α5 KO impairs the protumor growth effect of fibroblasts in organoids and affects the killing function of T cells. (A) Representative image of human organoid and co-cultures in the presence of fibroblasts CCD-18Co cells at day 14. Scale bars: top row, 200 µm; bottom row: 100 µm. (B) Diameters of human organoids cultured with or without fibroblasts (n=5; 10 organoids for each well randomly were measured). (C, D) Quantification of the tumor organoid killing assay. (C) Flow cytometry showing tumor cell activity after 72 hours incubation with CD3+ T-cell populations, in the presence or absence of fibroblast CCD-18Co. Live tumor cells were assessed as the percentage of cells defined as Organoids cells-CellTracker Red+; Zombie-NIR-A−. Gating strategy is shown. (D) Flow cytometry quantification of the percentage of live tumor cells. Organoids cells, single tumor cells digested from organoids; (E) schematic model of fibroblasts and CD3+ T cells co-culture system and human cancer cell killing assay. (F–G) Flow cytometry quantification of the frequency of CD4+CD137+ (F) and CD8+CD137+(G) of CD3+ T cells after 4 days culture alone or with fibroblasts-CCD-18Co. (H) Flow cytometry quantification of the percentage of live cancer cells. (I–K) Flow cytometry quantification of the frequency of CD8+IFN-γ+ (I) CD8+ PD-1+ Tim-3+ (J) CD4+ PD-1+ Tim-3+ (K) of CD3+ T cells harvested from co-cultures with HTC116. Vector ctrl, CCD-18Co cells transfected with plasmid vector control; α5 KO, CCD-18Co cells with α5 knockout. Error bars, mean±SEM; *p<0.05, **p<0.01, ***p<0.001; ns, not significant; one-way analysis of variance.FACS, flow cytometry analysis; IFN, interferon; PD-1, programmed cell death protein 1; Tim-3, T-cell immunoglobulin and mucin domain 3.
Further, we co-cultured CCD-18Co (vector control or α5 KO) with T cells for 4 days and tested T-cell activation markers to investigate the effect of α5 expression in fibroblasts on T-cell activity (figure 5E). The results showed that the presence of fibroblasts could promote the expression of T-cell activation markers compared with T cells alone. In addition, α5 depletion cells caused a higher increase in the expression of CD4+ CD137+ and CD8+ CD137+ compared with vector control cells (figure 5F and G). Subsequently, we harvested the co-cultured T cells or T cells alone and added them to wells with human colon cancer cells-HCT116 in the presence of CD3/CD28 T-Cell Activator (figure 5E). After 2 days, the presence of T cells killed almost all cancer cells, leaving only approximately 4% viable cells compared with HCT116 cells alone, with almost 90% cell viability (figure 5H). The killing ability of T cells from vector control CCD-18Co co-culture wells was significantly reduced, leaving approximately 15% viable cells, and the killing of T cells from α5 KO co-culture wells was like that of T cells cultured alone (figure 5H). Additionally, we assessed IFN-γ protein expression in CD8+ T cells. Compared with CD8+ T cells alone, IFN-γ expression was significantly increased in the presence of HCT116 cancer cells. Nonetheless, the function of T cells from CCD-18Co co-culture wells was reduced, and with a higher expression in T cells from the α5 KO co-culture wells compared with that from vector control co-culture wells (figure 5I). On the contrary, the proportion of CD8+ PD-1+ TIM-3+ and CD4+ PD-1+ TIM-3+ T cells from CCD-18Co co-culture wells were increased, with a lower proportion in T cells form α5 KO co-culture wells compared with that from vector control co-culture wells (figure 5J and K). These results demonstrated that α5+ fibroblasts play a role in suppressing intratumoral CD8+ T-cell function and showed that α5+ fibroblasts can directly limit CD8+ T-cell effector potential.
Additionally, we isolated primary CAFs separately from human patients with CRC, and immunofluorescence staining showed similar expression levels of fibroblast markers and integrin α5 compared with CCD-18Co (online supplemental figure S6A). RNA-sequencing data showed that the primary isolated CAFs maintained high expression of CAFs canonical marker gene-FAP and TWIST1 as defined in online supplemental figure S1C. CCD-18Co cells maintained high expression of crypt fibroblasts marker gene- WNT2B and RSPO3, while had high expression of ACTA2, TAGLN and FAP after TGF-β treatment (online supplemental figure S6B). Following α5 knockdown in CAFs and their co-cultured with CD3+ T cells, more T cells were in direct contact with CAFs with α5 knockdown (CAFs-siRNA) compared with the wild-type control (CAFs-ctr) (online supplemental figure S6C,D). Then we examined the expression levels of a set of T-cell activity markers (CD69 and CD137) and exhaustion markers (PD-1+TIM-3+). The results were consistent with those of CCD-18Co cells, with only CAF-siRNA cells caused a significant increase in the expression of CD4+CD69+, CD8+CD69+, CD4+CD137+, and CD8+ CD137+ compared with T cells alone (online supplemental figure S6E–H). CD3+ T cells co-cultured with control CAFs had a significantly increased proportion of CD4+ PD-1+ TIM-3+ and CD8+ PD-1+ TIM-3+ T cells compared with those cultured alone, whereas CD3+ T cells co-cultured with CAFs-siRNA did not show a significant increase (online supplemental figure S6I,J). These results further support that the expression of α5 in CAFs could affect T-cell activity.
α5 expression level is significantly associated with T-cell infiltration in CRC and correlates with immunotherapy prognosis
To further investigate the clinical relationship between α5 expression level and T-cell infiltration among CRC stroma, we selected 237 eligible Chinese patients diagnosed with colorectal adenocarcinoma and assessed the level of integrin α5 expression and the number of T cells (CD3, CD8, Foxp3) in tumor tissues. The clinicopathological parameters of this independent group were summarized based on the staining status of α5, CD3, CD8, Foxp3, and CD45 in online supplemental table S1. The α5 expression level showed a significant inverse correlation with CD3+ T-cell numbers and CD8+ T-cell numbers (figure 6A), and a positive correlation with Foxp3+ T-cell numbers (figure 6A). No correlation was observed between the α5 expression level and CD45+ total immune cell number (figure 6A).
α5 expression is significantly correlated with T-cell infiltration in colorectal cancer (CRC) and associated with immunotherapy prognosis. (A) Pearson’s correlation of α5 expression with immune infiltration level in CRC evaluated by IHC. Data were analyzed using log2 values. (B, C) Overall survival analysis of patients with CRC (n=237). (D) Tumor progression-free survival analysis of patients with CRC after PD-1 blockade therapy (n=29). Survive curves were analyzed using the Kaplan-Meier method and compared among groups using the log-rank test. *p<0.05, **p<0.01. ns, not significant. (E) Rate of response to anti-PD-1 treatment in α5 low/CD3 high group and the others in the CRC cohort (n=26). Patients without signs of progression within 4 months after PD-1 blockade treatment were defined as responders, while patients who progressed within 4 months after PD-1 blockade treatment were defined as non-responders. α5 low/ CD3 high group (5 responders and 1 non-responders), the others (7 responders and 13 non-responders). Data were analyzed by a two-sided Fisher’s exact test. FITC, fluorescein; FACS, flow cytometry analysis; IFN, interferon; PD-1, programmed cell death protein 1.
In our previous study, we showed that α5 expression in the tumor stroma serves as a potential prognostic marker in colorectal adenocarcinoma,26 which was further confirmed (online supplemental figure S7A). Additionally, we analyzed the association between T-cell counts and patients’ OS. A statistically significant difference was observed in the OS between the two groups of patients divided by the CD8+ T-cell count. Patients with high CD8+ T-cell infiltration had significantly longer OS than those with low CD8+ T-cell infiltration (figure 6B). There was no significant difference in OS between the groups divided by CD3+ T cell or Foxp3+ T-cell counts (figure 6B). We subsequently divided these patients into four groups: α5 low/CD3 low, α5 high/CD3 low, α5 low/CD3 high, and α5 high/CD3 high. Patients in the α5 low/CD3 high group had significantly longer OS than others (figure 6C). Accordingly, we also observed that the α5 low/CD8 high group had significantly longer OS than those in the α5 high/CD8 low group, and the α5 low/Foxp3 low group had significantly longer OS than patients in the α5 high/Foxp3 low group (figure 6C). Additionally, multivariate Cox regression analysis demonstrated that tumor stage and immune status (α5 low/CD3 high vs others) were independent prognostic factors in this cohort (online supplemental table S2). Thus, tumor stromal α5 expression was clinically correlated with T-cell infiltration, and α5 expression in combination with CD3+ T-cell count was predictive of patient survival. We also analyzed gene expression data from TCGA database with 517 colorectal adenocarcinoma cases that had complete follow-up information, the result further confirmed that α5 low/CD3 high group had significantly longer OS than those in the α5 high/CD3 low group, and the α5 low/CD8 high group had significantly longer OS than patients in the α5 high/CD8 low group (online supplemental figure S7B).
Subsequently, we analyzed additional clinical samples from 29 patients with colorectal adenocarcinoma who received PD-1/PD-L1 blockade therapy to evaluate whether the expression of α5 in the tumor stroma could affect the response to immunotherapy. We stained α5 and CD3 in tumor tissues before PD-1/PD-L1 blockade treatment. The cases were divided into two groups based on the expression status of α5 (high or low) or CD3 count (high or low) (online supplemental table S3), and no significant difference in PFS was observed between the two groups treated with PD-1 blockade (figure 1D). We then divided these patients into four groups: α5 low/CD3 low, α5 high/CD3 low, α5 low/CD3 high, and α5 high/CD3 high. Patients in the α5 low/CD3 high group treated with PD-1 blockade had significantly longer PFS than those in the α5 low/CD3 low group (figure 6D). In addition, the responder rate was higher in the α5 low/CD3 high group (83.3%) than in the other groups (35%) (figure 6E), suggesting that α5 expression in the tumor stroma combined with CD3+ T-cell infiltration in CRC tissues may be associated with the therapeutic effect of PD-1/PD-L1 blockade.
Discussion
Stromal CAFs play a critical role in promoting ECM remodeling and immunosuppressive TME in CRC. This study showed that α5 was enriched in CAFs and correlated with ECM deposition, tumor infiltration of T cells, and patient survival. Targeting α5 eliminated the ability of CAFs to promote tumor growth and altered the deposition of ECM in the TME, thus affecting T-cell infiltration and activity in mouse subcutaneous tumors. Our study revealed a potential method to change the tumor stroma architecture and disrupt the physical barrier that impedes T-cell infiltration within tumor tissues by targeting α5 together with immunotherapy (online supplemental figure S8).
In this study, we initially analyzed ITGA5 expression in tumor stroma using CRC scRNA-seq database and found that ITGA5 was mainly enriched in endothelial cells and FAP-CAFs (online supplemental figure S1A–D and figure 1A,B). CAFs are the predominant component in the TME, and previous studies have shown the heterogeneity and potential plasticity of CAFs originating from endothelial cells by endothelial-mesenchymal transition,35 36 thus our study mainly focused on the role of α5 in fibroblasts. Our previous study also underlines the essential role of α5 in fibroblasts in promoting CRC development,26 which was further confirmed by α5 deletion in fibroblasts in the in vitro and in vivo experiments in this study. However, further research is needed to define the role of α5 in various CAFs and its relationship with FAP.
In mice subcutaneous tumor models, we observed that α5 depletion in fibroblasts blocked the tumor-promoting role of wild-type fibroblasts when co-transplanted with MC38 cancer cells in vivo and improved the response to immunotherapy (online supplemental figure S2D–F and figure 1C–E). We also found that the combined administration of the α5β1 antagonist (ATN161 and anti-α5 antibody) and anti-PD-L1 antibody significantly blocked tumor growth compared with either treatment alone (online supplemental figure S2G–K and figure 1F–H). The result was likely due to α5 inhibition in fibroblasts favoring CD8+ T-cell infiltration in tumors (figure 2 and online supplemental figure S3I–K). These data suggest the potential for targeting α5+ CAFs in CRC immunotherapy, highlighting the role of CAFs in impeding antitumor immune responses, as previously reported.10 15 16 29
Subsequently, Gene Ontology (GO) analysis with the upregulated gene cluster in ITGA5+CAFs showed biological processes that involved in ECM organization and collagenous ECM generation (figure 3A,B). We provided direct evidence that α5 expression level was positively correlated with FAP and ECM proteins (collagen and fibronectin) in CRC tissues (figure 3C,D and online supplemental figure S4C,S2B), and α5 deletion in fibroblasts correlated significantly with reduced ECM deposition (figure 3E–H and online supplemental figure S4F–I and S2A). Previous studies have shown that excessive ECM deposition and remodeling characterize the tumor stroma, which transmits signals affecting T-cell activity and creates environmental obstacles to T-cell traffic.20 21 32 34 Consistent with this, our study showed that ECM deposition affected T-cell location and activity, as CD3+ T cells were preferentially distributed within the areas with sparse fibronectin and collagen fibers in human CRC tissues and T cells cultured in high concentrations of collagen showed reduced cell proliferation and activity (figure 4). Moreover, we demonstrated that α5 deletion impaired the tumor-promoting effect of fibroblasts and affected T-cell killing function in both the CRC organoids-fibroblast co-culture system and fibroblasts T-cell preculture system (figure 5), emphasizing the essential role of α5 in regulating the activity of fibroblasts and T cells.
Clinically we showed that high expression of α5 correlated with fewer CD3+ T or CD8+ T cells in CRC tissues (figure 6A), which was consistent with previous studies that reported a significant association between ITGA5 expression and immune infiltration levels of various immune cells in gastrointestinal tumors and gliomas.24 37 We also showed that tissues with low α5 and high CD3+ T were associated with better patient survival and immunotherapy response in CRC (online supplemental figure S7 and figure 6C,E), which supports the study that showed therapeutic potential for combining ITGA5 targeting with immunotherapy37 and administration of integrin-targeted immunotherapy together with an anti-PD-1 antibody further improves responses.38 Moreover, previous studies focusing on α5β1 have based on its role in angiogenesis,22 23 our study showed a novel role of α5 in CAFs in modulating immunosuppressive TME during CRC development, and therapeutic targeting of α5 might have a direct effect in decreasing CAFs and antiangiogenic activity, which might potentiate anti-PD-L1 responsiveness during CRC immunotherapy. Additionally, this study demonstrated that tissues with low α5 and low Foxp3 levels showed slightly better patient survival than other groups, which is consistent with the finding that Foxp3 Tregs coexisting with CAFs are correlated with a poor outcome in lung adenocarcinoma.39
In summary, our study highlighted the role of α5 in fibroblasts in modulating the tumor immune microenvironment by regulating ECM protein deposition and influencing T-cell activity. This suggests a potentially novel therapeutic approach that targets α5 in combination with immunotherapy to inhibit CRC, paving the way for larger-scale patient studies to explore the potential of α5β1 as a prognostic biomarker in solid tumors by multiple promising therapeutic combinations.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Ethics approval
The study was approved by the Ethical Committee of Shanghai Tenth People’s Hospital (SHSY-IEC-4.1/19–180/02). Participants gave informed consent to participate in the study before taking part.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
LL, YG and DH contributed equally.
Author Contributions LL, YG and HQ designed the research study. LL, YG and DH performed majority of the experiments and data analysis. ML, HL, LZ and WW helped with experiments. SW and JW assisted with in vivo experiments and part of the research design. DH and DB helped to download and analyze the scRNA-seq and TCGA data. JZ, HL and JW contributed to specimen preparation. YZ and FS assisted with confocal microscopy and helped with interpreting results. The manuscript was drafted by LL and edited by YG, DY, DH, HQ and QW. QW is responsible for the overall content as the guarantor. All authors approved the final version of the manuscript.
Funding This work was supported by the Fundamental Research Funds for the Central Universities (22120220644), the National Natural Science Foundation of China (81700478, 82273366 and 82072634), Science and Technology Commission of Shanghai (20ZR1442800) and Clinical research plan of SHDC (No. SHDC2020CR2069B and No. SHDC2020CR5006–002).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.