The CD49b (ITGA2) collagen receptor excludes CD8⁺ T cell infiltration in pancreatic ductal adenocarcinoma

Summary

Pancreatic ductal adenocarcinoma (PDAC) features a dense desmoplastic stroma that promotes tumor progression and immune evasion. We examined how the collagen receptor CD49b (ITGA2) shapes anti-tumor immunity and stromal architecture in PDAC. CD49b was highly expressed in tumors compared with normal pancreas and peripheral blood mononuclear cells. CD49b loss impaired PDAC cell adhesion to collagen, delayed tumor growth, and increased CD8⁺ T cell infiltration; depletion of CD8⁺ T cells in vivo restored tumor growth in CD49b knockout tumors. CD49b knockout tumors secreted more CXCL10, a T cell chemoattractant, and showed reduced FAK signaling, while FAK overexpression rescued tumor growth. CD49b loss also altered collagen fiber alignment, generating a microenvironment permissive to immune infiltration. Pharmacological CD49b blockade with vatelizumab disrupted tumor cell adhesion, reduced tumor progression, and enhanced CD8⁺ T cell infiltration. These findings identify CD49b as a regulator of collagen-mediated immune exclusion in PDAC and support its evaluation as a therapeutic target.

Graphical abstract

Highlights

• •ITGA2/CD49b is a dominant collagen receptor expressed by PDAC tumor cells
• •ITGA2/CD49b loss delays PDAC tumor growth and enhances CD8⁺ T cell infiltration
• •ITGA2/CD49b regulates collagen architecture and FAK-mediated tumor progression
• •Vatelizumab blocks CD49b function and promotes anti-tumor immunity

Immunology; Oncology; Tumor microenvironment; Mechanism of action; Cell biology

Introduction

PDAC is a highly aggressive malignancy characterized by an extensive fibrotic and desmoplastic tumor microenvironment (TME).^1,2 This dense, collagen-rich TME plays a pivotal role in shaping an immunosuppressive landscape, often leading to immune exclusion and limited infiltration of cytotoxic T lymphocytes (CTLs) into the tumor parenchyma.^3,4 Although significant progress has been made in cancer immunotherapy over the past decade, most clinical trials targeting PDAC have reported limited efficacy, underscoring an urgent need to better understand the immune dynamics including impact of the desmoplastic TME.^5,6,7,8

A defining feature of the PDAC TME is the abundance of extracellular matrix (ECM) proteins, collectively known as the “matrisome”, comprised of collagen types I and III, fibronectin, and elastin.⁹ The deposition of these ECM components is largely driven by cancer associated fibroblasts (CAFs), which constitute a major cellular component of the PDAC stroma.¹⁰ While therapeutic strategies aimed at targeting the ECM or depleting CAFs have been pursued, these approaches have yielded inconsistent and, in some cases, detrimental outcomes in clinical trials. For example, inhibition of SHH-SMO signaling and depletion of hyaluronic acid were associated with either insufficient efficacy or, paradoxically, shortened patient survival.^{11,12,13,14,15,16} These findings highlight the dual roles of CAFs in both tumor promotion and suppression, emphasizing the complexity of PDAC stroma.^{11,17,18,19,20,21,22,23}

Collagen type I has received significant attention for its role in modulating the TME. Collagen type I has long been perceived as a physical barrier that restricts immune cell infiltration, particularly of CD8⁺ T cells.²⁴ Collagen type I depletion unexpectedly reduce T cell infiltration, indicating that the interplay between the ECM and immune cells is more nuanced.^25,26 This ongoing debate underscores the need for a deeper understanding of the dynamic interactions between collagen type I and cellular compartments in PDAC.

Collagen type I mediates its effects through interactions with cellular receptors, including integrins (α1 [ITGA1, CD49a], α2 [ITGA2, CD49b], α10 [ITGA10], α11 [ITGA11], and β1 [ITGB1, CD29]), discoidin domain receptors (DDR1, DDR2), and leukocyte-associated immunoglobulin-like receptor 1 (LAIR1). Integrins in particular are key mediators of cell-ECM adhesion and are known to bind specifically to collagen type I, triggering signaling cascades that promote tumor cell survival, proliferation, and invasion. Similarly, DDR1 and DDR2 receptors also recognize collagen type I and influence tumor cell behavior.^27,28,29

There are several collagen receptors that can bind to collagen, and critical questions remain unanswered. Which collagen receptors are most highly expressed by PDAC cancer cells? Do these receptors modulate tumor-stroma interactions, immune cell infiltration, and therapeutic resistance? Additionally, how does targeting these receptors affect the overall architecture and immunogenicity of the TME? To address these questions, we explored collagen receptor expression using PDAC mouse models and clinical PDAC samples. Through RNA sequencing (bulk and singlecell) and spatial transcriptomic analyses, we identified CD49b (α2 integrin) as a key collagen receptor predominantly expressed by PDAC tumor cells. Notably, knocking out CD49b correlated with increased CD8⁺ T cell infiltration, suggesting that CD49b plays a functional role in maintaining immune exclusion. Targeting CD49b with vatelizumab suppressed tumor growth in vivo and increased CD8⁺ T cell infiltration. Our findings provide new insights into the role of collagen type I interactions in shaping the immune landscape of PDAC and highlight CD49b as a potential therapeutic target to enhance anti-tumor immunity.

Results

Comprehensive profiling of collagen receptor expression in pancreatic TME

PDAC exhibits a collagen-dense desmoplastic stroma, with COL1A1 and COL3A1 emerging as the dominant collagen subtypes. Transcriptomic data from TCGA revealed that these fibrillar collagens account for nearly 70% of total collagen expression in PDAC, ranking it second in COL1A1 and COL3A1 expression among cancers, following sarcoma (Figures S1A and S1B). Given the prominence of collagen in PDAC, we investigated how cellular compartments interact with collagen through specific receptors. In our analysis of 27 pancreatic cancer cell lines from the cancer cell line encyclopedia (CCLE) database.^30,31 ITGA2, DDR1, and ITGB1 were the only receptors consistently and highly expressed across pancreatic cancer cell lines. In contrast, ITGA1, DDR2, LAIR1, ITGA10, ITGA11, MRC2, GP6, and TMEM119 exhibited negligible expression (Figure 1A). Further analysis revealed that pancreatic cancer cell lines ranked among the top three for ITGA2 mean expression, using CCLE database includes transcriptomic profiles of 1,673 cell lines from various tissues, following ampullary carcinoma and cutaneous squamous cell carcinoma (Figure S2). Phenotypic analysis confirmed prominent expression of CD49b and CD29 in human pancreatic cancer cell lines and mouse KPC-derived lines (Figures 1B, 1C, and S3A), aligned with western blot analysis (Figure 1D). To explore spatial localization and cellular specificity, we analyzed scRNA-seq data from mouse KPC models. ITGA2 and DDR1 were highly expressed in cancer cells (KRT19), while ITGA1 and DDR2 were enriched in CAFs (ACTA2 and COL1A1). LAIR1 was expressed in myeloid cells (ITGAM), and ITGB1 is broadly expressed across different cell types (Figures 1E, 1F, and S3B). Further analysis by volcano plot showed that ITGA2 was associated with cancer cell clusters expressing cytokeratin markers, while ITGA1 was linked to fibroblast markers (Figure 1G).

Figure 1.

The expression of collagen receptors in pancreatic cancer

(A) Log2 expression levels of collagen receptors in human PDAC cell lines based on CCLE RNA-seq data (RMA-normalized).

(B) The percentage of total human PDAC cell lines stained with CD49b, CD29, CD49a, DDR1, and LAIR-1 by FACS.

(C) Representative FACS histogram of KPC cell lines stained with CD49b, CD29, and CD49a.

(D) Representative western blot of CD49b, CD29, and DDR1.

(E–G) Collagen receptors expression across cell clusters in mouse scRNA-seq (GEO: GSE166298).

(F) ITGA2 expression within various cell clusters.

(G) Volcano blot of ITGA2 vs. ITGA1 clusters ((logFC) cutoff ±1).

(H) Heatmap of scRNA-seq showing the expression of collagen receptors among cellular clusters in normal pancreas (GSE159343) vs. KPC mouse model (GSE166298).

(I) C57Bl6J mice were inoculated with KPC cell line (Hy15549) in the right flank. (I) Bar graph of CD49a⁺ CD29⁺ and CD49b⁺ CD29⁺ bar graph in tumor, pancreas, and spleen.

(J) Representative dot plot comparing the CD49b⁺ CD29⁺ on EPCAM⁺ cells in tumor and pancreas.

(K) Representative immunofluorescence (IF) image of tumor stained with CD49b (yellow), Col1 (red), and DAPI (blue). Data are presented as means ± SD of at least three replicates. Statistical significance (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

We next examined whether CD49b expression is restricted to cancer cells or also present in normal pancreas. Comparative scRNA-seq analysis of normal pancreas and KPC mouse models confirmed that CD49b is expressed in cancer cells, with negligible expression in normal pancreatic tissue (Figure 1H). To validate these findings, we examined collagen receptor expression in flank tumors derived from KPC mice (Ptf1-Cre; LSL-KRAS^G12D; Trp53^Lox/+) on a C57BL/6 background, a model recapitulating key PDAC mutations (Figure S4). Flow cytometry of tumor-infiltrating lymphocytes (TILs) revealed strikingly high CD49b⁺ CD29⁺ expression within the EPCAM⁺ tumor cell population, reaching 70%–80%, while expression in other cell populations, including dendritic cells (CD11b⁺ CD11c⁺), macrophages (CD11b⁺ F4/80⁺), CD4⁺ T cells (CD3⁺ CD4⁺), and CD8⁺ T cells (CD3⁺ CD8⁺), was significantly lower (1%–5%) (Figure 1I). CD49a⁺ CD29⁺ expression within the EPCAM⁺ population was minimal (∼10%), and expression of CD49b⁺ CD29⁺ and CD49a⁺ CD29⁺ in normal pancreas and the spleen was similarly low (1%–5%) (Figure 1I). Notably, CD49b⁺ CD29⁺ was highly enriched in EPCAM⁺ tumor cells compared to EPCAM⁺ cells in normal pancreas (Figure 1J). Immunofluorescent staining confirmed robust CD49b expression in the TME, colocalizing with cytokeratin (Figure 1K). These findings underscore the tumor-specific localization and thereby potential functional significance of CD49b in the PDAC microenvironment.

Distinctive expression of CD49b in the naive PDAC patients

We investigated CD49b in human PDAC. scRNA-seq analysis of tumor (n = 16) and normal tissue (n = 3) from untreated patients revealed high CD49b expression in cancer cells (KRT19) with negligible expression in normal tissues (Figures 2A and S5A). Interestingly, scRNA-seq of human peripheral blood mononuclear cells (PBMCs) from PDAC patients and healthy controls revealed that CD49b is not expressed in any blood-derived cellular clusters, regardless of disease status (Figure S5B). These findings were confirmed in an independent scRNA-seq dataset of 18 treatment-naïve patients, confirming the expression of ITGA2 (CD49b) in cancer cell in human PDAC (Figures 2A and S6). Comparison of scRNA of normal mouse pancreas (C57BL/6J), KPC mouse tumors (C57BL/6J), and human PDAC revealed that ITGA2 is not expressed in normal pancreas and is expressed in cancer cells (Figure S7). To validate the scRNA-seq findings, we performed phenotypic analysis using flow cytometry on untreated, resected tumor and adjacent non-tumor specimens. Cell populations were gated into CD45⁺ (immune cells), EPCAM⁺ (epithelial cells), CD45⁻ EPCAM⁻ CD31⁺ (endothelial cells), and CD45⁻ EPCAM⁻ CD31⁻ (CAFs) (Figure S8). Distinct patterns of collagen receptor expression were observed across these populations. Using tSNE flow analysis, we localized collagen receptor expression within single cells, revealing that the EPCAM⁺ CD45⁻ population predominantly expressed CD49b and CD29, while CD45⁺ immune cells primarily expressed LAIR1. DDR1 expression was rare across all populations (Figure 2B). However, we observed elevated expression of CD49b⁺CD29⁺ cells within the EPCAM⁺ population, comprising approximately 70% of tumor samples compared with ∼25% in adjacent non-tumor tissue (Figure 2C). Notably, some CD49b⁺CD29⁺ cells were detected in adjacent normal samples, which differs from our single-cell RNA-seq analysis showing minimal CD49b expression within CK19⁺ epithelial cells. This difference may reflect heterogeneity within adjacent normal tissues.

Figure 2.

Colocalization of CD49b and CK19 in human PDAC Tissue

(A) Collagen receptors expression among various cell clusters from scRNA-seq datasets of treatment naive human PDAC patients (GEO: GSE166298 ((n = 16 patients); with matched adjacent normal pancreas (n = 3 patients) and GSE155698 (n = 18 treatment naive patients).

(B) tSNE FACS clustering shows the collagen receptors expression in resected human PDAC specimens (adjacent-tumor and tumor).

(C) Bar graph comparing the collagen receptors expression among the immune cells (CD45⁺), epithelial cells (EPCAM⁺), endothelial cells (CD45⁻ EPCAM⁻ CD31⁺), and other cells including CAFs (CD45⁻ EPCAM⁻ CD31⁻).

(D) Kaplan-Meier survival analysis (TCGA, n = 177) stratified into two groups based on the median CD49b expression level. Log rank (Gehan-Breslow-Wilcoxon) test was used.

(E) Spatial transcriptomics in PDAC patients (GSE211895) showing clusters of epithelium and stroma, ITAG2 is localized in epithelium clusters.

(F) Immunofluorescence staining of human PDAC tissue CD49b (green) and CK19 (red). Scale bars, 250 μm.

(G) Pearson’s correlation test between CD49b, KRT19 and ACTA2 (TCGA dataset). Data are presented as means ± SD of at least three replicates. Statistical significance (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

In contrast, CD49b⁺ CD29⁺ expression was low in other cell populations. Consistently, higher CD49b transcript levels, among other collagen receptors, were correlated with significantly worse survival outcomes in PDAC patients within the TCGA cohort (150 PDAC patients) (Figures 2D and S9A). Spatial transcriptomic of human PDAC confirmed the location of expression of CD49b is in epithelial clusters rather than the stroma (Figure 2E). Immunofluorescence staining confirmed the localization of CD49b with cytokeratin 19 in treatment-naive PDAC patient samples, indicating its specific expression in tumor-associated epithelial cells (Figures 2F and S9B). Notably, the CD49b data also revealed a positive correlation between CD49b and KRT19 expression. Conversely, ACTA2 expression correlated with ITGA1 and DDR2 (Figures 2G, S9C, and S9D). These findings collectively demonstrate that CD49b is primarily expressed within the cancer cells within PDAC TME.

Knockout of CD49b or CD29 increases T cell infiltration and hinders pancreatic cancer tumor growth

Given the elevated expression of CD49b and CD29 within the PDAC TME, we investigated their functional roles, particularly proliferation and adhesion to extracellular matrix (ECM) components. Our initial experiments focused on the adhesion of PDAC cell lines to collagen type I (COL1), which is expressed predominantly in the PDAC stroma. All PDAC cell lines showed strong adhesion to both COL1 and fibronectin (Figure 3A). Increasing concentrations of COL1 influenced cellular adhesion (Figure S10A). We next assessed the impact of collagen on PDAC cell proliferation. Collagen promoted proliferation regardless of its concentration, indicating a non-linear relationship between collagen abundance and proliferative response (Figure S10B). To further elucidate the regulation of CD49b and CD29 expression, we explored the influence of various cytokines, including interferon-alpha (IFNα), interferon-beta (IFNβ), transforming growth factor β (TGFβ), interleukin-7 (IL-7), and interleukin-2 (IL-2). TGFβ was identified as the most potent inducer of CD49b and CD29 expression, emphasizing its regulatory role in shaping the adhesion dynamics of PDAC cells within the TME (Figure S10C). To delineate the specific roles of CD49b and CD29 in collagen adhesion, we generate KPC cell line with single knockouts for CD49b (KPC^49bKO) and CD29 (KPC^29KO), as well as double knockouts (KPC^49bKO29KO) (Figure 3B). Knockout did not affect cellular proliferation in vitro (Figure 3C). Loss of CD49b and CD29 significantly impaired adhesion to COL1, even at higher collagen concentrations, while adhesion to fibronectin remained unaffected (Figure 3D). Pan02^CD49bKO knockout cells failed to adhere to COL1 while they adhered to fibronectin (Figure S10D). To further examine these findings, we utilized blocking antibodies targeting CD49b. These antibodies effectively inhibited adhesion to COL1 (Figure S10E). Collectively, these results reveal distinct roles for CD49b and CD29 in collagen interactions.

Figure 3.

Knockout of CD49b impair of PDAC adhesion to collagen and delayed tumor growth In vivo

(A) PDAC cell adhesion measured after 30 min on COL1 (50 μg/mL) or fibronectin (50 μg/mL) coated plates.

(B) Representative FACS dot plot showing the (KPC^wt, KPC^49bKO, KPC^29KO, KPC^49b29KOcell lines) stained with CD49b and CD29.

(C) Cell proliferation over 72 h.

(D) Representative images for the adhesion of KPC^wt, KPC^49bKO, KPC^29KO, KPC^{49bKO29 KO} on COL1 (50 μg/mL) or Fibronectin (50 μg/mL) coated plates. The cell adhesion were also measured on different concentrations of COL1 coated plates.

(E) C57BL/6J mice were subcutaneously inoculated with KPC^wt, KPC^49bKO, KPC^29KO, and KPC^49bKO29KO cells (1 × 10ˆ6 cells/100 μL PBS) on the right flanks. Tumor growth was measured three times per week. On day 45 for knockout cell lines and day 12 for KPC^wt (300 mm³ to match tumor sizes), TILs were harvested and analyzed.

(F) Representative dot plot of CD8⁺ and CD4⁺ T cells gated on CD45⁺CD3⁺.

(G) Bar graph showing the percentages of different T cell populations.

(H) Bar graph showing the percentages of myeloid cells. Data are presented as means ± SD of at least three replicates. Statistical significance (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

We compared the tumor growth of each knockout cell line (KPC^49bKO, KPC^29KO, and KPC^49bKO29KO) to the tumor growth of the parental KPC cell line (KPC^wt). KPC^49bKO, KPC^29KO, and KPC^{49b KO29KO} cell lines exhibited significantly slower and delayed tumor growth compared to the wild-type cell line. This impaired tumor growth resulted in a notable improvement in survival rate (Figures 3E and S10F). This same effect was observed in Panc02 mouse PDAC tumors (Figure S10G). These observations suggest that knockout of receptor CD49b and its heterodimer partner CD29 has a notable inhibitory effect on tumor growth in an immunocompetent mouse model, and suggests potential interactions between the receptors and the immune response in the TME.

To study mechanisms through which these receptors influence tumor progression, we conducted an assessment of immune cell infiltration within the TME following the evaluation of tumor growth. A significant increase in infiltration of CD3⁺ T cells into the tumor site was observed in the knockout groups (KPC^49bKO, KPC^29KO, and KPC^49bKO29KO) and included a significant increase in CD8⁺ and CD4⁺ T cells compared to the parental KPC cell line (KPC^wt) (Figures 3F and 3G). Furthermore, CD8⁺ T cells exhibited signs of activation including IFNγ expression as well as CD44⁺ without change in PD1 expression indicating their active participation in the immune response within the TME (Figure 3G). An analysis of myeloid cell populations, including macrophages (CD11b⁺F4/80⁺), M2 macrophages (CD11b⁺CD206⁺), M1 macrophages (CD11b⁺CD206^-), MDSCs (CD11b⁺GR1⁺), dendritic cells (CD11c⁺CD103⁺), and (CD11c⁺XCR1⁺), revealed no statistically significant differences among the groups (Figure 3H). Knockout of receptor CD49b and its heterodimer partner CD29 had a significant impact on the immune landscape within PDAC with recruitment of CD8⁺ T cells and slowed tumor growth. CD49b loss leads to enhanced CD8⁺ T cell infiltration and functional activation, which slows tumor growth and prolongs survival, but does not lead to permanent tumor eradication. The phenotype reflects strong immune-mediated suppression and delayed kinetics, not absolute loss of tumor-forming capacity. These findings suggest a role for CD49b and CD29 in modulating the infiltration of CD8⁺ T cells within the TME.

CD49b excludes CD8⁺ T cell infiltration

Given the ubiquitous expression of CD29, we focused on CD49b, which exhibits notably high and specific expression on PDAC cancer cells. We evaluated the tumor growth of KPC^49bKO and KPC^wt in nude mice, which are deficient in T cells. KPC^49bKO cells formed tumors similar to the parental cell line (KPC^wt) in nude mice, indicating that the absence of CD49b did not hinder tumor development in the absence of an intact immune system (Figure 4A). As a complementary experimental approach we used an immunocompetent tumor model in which we depleted CD8⁺ T cells. We again observed that the KPC^49bKO cells exhibited unimpeded growth like the parental cell line (KPC^wt) (Figure 4B). Together, the results suggest that CD8⁺ T cells play an important role in the slowed tumor growth observed in the absence of CD49b receptor.

Figure 4.

CD8⁺ T cells control tumor growth in CD49b knockout cell line

(A) Tumor growth following the subcutaneous injection of KPC^wt or KPC^49bKO cells (1 × 10⁶ cells) into immunodeficient nude mice (12 mice for each group).

(B) Tumor growth in C57BL/6 mice with prior treatment with anti-IgG or anti-CD8 antibody (n = 5 tumors per group).

(C) Pearson correlation analysis between collagen receptors expression and CD8a⁺ levels in the TCGA PDAC dataset.

(D) Spatial transcriptions spots of KRT18, CD49b and CD8a in PDAC patients (GSE235315).

(E) Donut chart showing the spatial spots that express CD49b⁺ CD8a⁺, CD49b⁺ CD8a⁻, CD49b⁻CD8a⁺, and CD49b⁻ CD8a⁻.

(F) The spots that express either CD49b⁺ CD8a⁻, CD49b⁻CD8a⁺ are correlated and represented as percentage per total spots for each patient.

(G) HE, trichrome and multiplex immunofluorescent staining of naive PDAC specimen stained with aCD49b, aCD8, and COL1. Bar graph represents the quantification of CD8⁺ T cells and CD49b expression levels in both the tumor core and margin regions.

To further investigate the relationship between CD8⁺ T cells and the collagen receptor, we conducted an analysis of TCGA. We observed a significant negative correlation between CD8⁺ T cell and expression levels of CD49b (ITGA2), while a positive correlation was evident with CD49a (ITGA1) and DDR2 (Figure 4C). These findings affirm the potential regulatory role of CD49b in CD8⁺ T cell infiltration. This suggests that CD49b may have a suppressive effect on CD8⁺ T cells infiltration within the TME.

To assess T cell exclusion in human PDAC, we measured the expression levels of CD49b and CD8a from spatial spots using SpatialGE and categorized them into four groups: spots co-expressing ITGA2 and CD8a (ITGA2⁺ CD8a⁺), spots expressing neither (ITGA2^- CD8a⁻), spots expressing ITGA2 but not CD8a (ITGA2⁺ CD8a⁻), and spots expressing CD8a but not ITGA2 (ITGA2^- CD8a⁺). The analysis stratified patients based on expression patterns, revealing that 3 patients had high ITGA2 and low CD8a expression, while 4 patients exhibited high CD8a and low ITGA2 expression, confirming an inverse correlation between ITGA2 and CD8a (Figures 4D–4F and S11). We explored all the correlation between CD8a and collagen receptors, revealed that ITGA2 and DDR1 is inversely correlated with CD8a spot (Figure S11). Moreover, we examined the localization of CD8⁺ T cells and CD49b expression in resected PDAC specimens from untreated patients. Our result confirmed that CD8⁺ T cells were predominantly localized at the margin of CD49b expression (Figure 4G). However, there was a notable decrease in the abundance of CD8⁺ T cells within the tumor core with high expressed CD49b providing additional evidence to support the notion that the CD49b receptor plays a role in limiting T cell accessibility to the tumor core (Figure 4G). These findings collectively point to a potential immunomodulatory role for CD49b receptor in regulating CD8⁺ T cell infiltration and antitumor immune responses.

CD49b knockout suppress the FAK pathway and increase CXCL10

RNA seq was conducted on KPC^wt, KPC^49bKO, KPC^29KO, and KPC^49bKO29KO. Differential gene expression analysis was carried out using DESeq2, comparing gene expression profiles (Figure 5A). For the comparison between KPC^wt versus KPC^49bKO, KPC^29KO, and KPC^49bKO29KO a total of 367 DEGs were identified, including 52 upregulated and 315 downregulated genes (Figure 5B). The global transcriptional change across the groups compared was visualized by a volcano plot (Figure 5B). Significantly differentially expressed genes were clustered by their gene ontology and the enrichment analysis revealed that upregulated DEGs significantly activated transcriptional programs related to cell adhesion and FAK signaling (Figure 5C). When compared to KPC^wt, knockout of CD49b, CD29, or both led to a significant downregulation of focal adhesion and cell adhesion pathways (Figure 5D). These integrins are known to play a crucial role in maintaining adhesion-related transcriptional networks. CD49b and CD29 knockout cells displayed a significant decrease in the levels of FAK and pFAK (Figure 5E). The assessment of the cytokine and chemokine profiles of KPC parental cell line and KPC^49bKO conditioned media (CM) revealed distinctive alterations induced by the absence of CD49b. Notably, the concentration of CXCL10 increased over 1.5-fold compared to KPC^wt cells (Figure 5F). Transwell migration assays of activated mouse splenic T cells demonstrated that the conditioned medium of KP^49bKO cancer cells (as compared with that of KPC^wt cancer cells) elicited more T cell migration, which was effectively abrogated by CXCL10 neutralizing antibody (Figure 5G). Canonically, CXCL10 expression is induced by inflammatory cytokines through coordinated activation of NF-κB and STAT1/IRF signaling pathways. CD49b knockout tumors showed elevated NF-κB signaling, increased phosphorylation of IRF3 (p-IRF3), and upregulation of STAT1 and phosphorylated STAT1 (p-STAT1) (Figure 5H). These results support a role for CD49b in regulating FAK-mediated signaling, thereby connecting integrins (specifically CD49b and CD29 receptors), FAK-mediated signaling, and CD8⁺ T cell infiltration. FAK signaling has been associated with immune evasion mechanisms in cancer cells,³² and the observed decrease in FAK levels may contribute to enhanced immunogenicity, rendering the knockout cell lines more susceptible to immune recognition and response.

Figure 5.

Knockout of CD49b suppress FAK downstream pathway

(A–D) RNA-seq analysis on for KPC^wt, KPC ^49bKO, KPC^29KO, and KPC^49b29KO cancer cells.

(A) The heatmap showed differentially expressed genes.

(B) Volcano plots showed significantly upregulated genes comparing KPC^wt to KPC^49bKO, KPC^29KO, and KPC^49b29KO based on fold change and an adjusted p value (C) Gene ontology enrichment (top 40 terms; Fisher’s exact test, GeneSCF v1.1-p2).).

(D) GSEA highlights focal-adhesion and cell-adhesion pathways. Normalized enrichment score (NES), and p value were shown.

(E) Representative western blot of KPC^wt, KPC^49bKO, KPC^29KO, and KPC^49b29KO cancer cells treated with Col1 for 16 h.

(F) Proteome profiler mouse cytokine and chemokine array in the cell culture supernatants of KPC^wt, KPC^49bKO cells cultured for 16 h. The fold change for each cytokine and chemokine in KPC 49bko cells compared to KPC^wt cells was calculated.

(G) CD8⁺ splenocytes migration assay, CD8⁺ T cells activated with aCD3/CD28 for 24 h before seeded in the upper chamber, conditioned media from KPC^wt, KPC^49bKO, treated with or without anti-CXCL10 antibody, in the lower chamber. Bar graph shows relative CD8⁺ T cell migration and CXCL10 neutralization.

(H) Representative western blot of KPC^wt, KPC^49bKO cancer cells treated with Col1 for 16 h. Data are presented as mean ± SEM. Statistical significance (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

FAK expression restores the tumor growth in the CD49b knockout tumors

FAK and CD49b are both key regulators of cell adhesion, migration, and survival. Our results and published data suggest that they may interact to influence tumor growth and immune evasion mechanisms. We examined whether FAK overexpression in KPC^49bKO could abrogate the tumor inhibition observed with CD49b knockout. Enrichment analysis revealed that FAK adhesion signaling is significantly upregulated in the KPC^49bKoFAK cell line compare to the KPC^49bKO cell line (Figure 6A). Assessment of immunogenic markers, MHCI and PDL1, showed no significant differences, suggesting that FAK does not directly modulate these pathways in our model (Figure S12A). KPC^wt, KPC^49bKO, and KPC^49bKOFAK cells were subcutaneously injection into mice flank (Figure 6B) or orthotopically implanted into the pancreatic tail of syngeneic mice (Figure 6C). FAK overexpression rescued tumor growth in the CD49b knockout context in subcutaneous model. Quantitative analysis of tumor weights of orthotopic model at day 15 revealed a significant reduction in tumor weight in the KPC^49bKO group relative to (Figure 6B). Extended analysis of KPC^49bKO tumors at day 35 further confirmed sustained suppression of tumor growth. These results confirm that FAK plays an essential role in tumor rescue. Additionally, immune cell profiling showed a significant increase in CD8⁺ T cell infiltration in KPC^49bKO. Additionally, there is a significant reduction of CD8⁺ T cell as well as CD8⁺ CD69⁺ in KCP^49bKOFAK compared to KPC^49bKO (Figure 6C). To determine whether CD49b deletion also affected extracellular matrix organization, collagen deposition was markedly reduced in KPC^49bKO tumors and was restored upon FAK re-expression (Figure 6E).

Figure 6.

FAK expression restore the tumor growth in CD49b knockout tumors

(A) GSEA of RNA-seq of KPC^wt, KPC^49bKO, and KPC^49bKOFAK cells showing focal adhesion pathway.

(B) Tumor growth following subcutaneous injection of KPC^wt, KPC^49bKO, and KPC^{49bKO FAK} cell lines (1 × 10⁶ cells) on the right flanks.

(C) Representative images showing tumor size at day 15 following orthotopic implantation of KPC^wt, KPC^49bKO, and KPC^49bKOFAK cells into the pancreatic tail of mice. Tumors were excised at day 15 and weighed, and tumor weights are presented as a bar graph. An additional bar graph shows tumor weights of KPC^49bKO tumors harvested at day 35 post-implantation.

(D) Quantitative analysis of TILs of orthotopic tumors derived from KPC^wt, KPC^49bKO, and KPC^49bKOFAK cells cells.

(E) Representative immunofluorescence staining of orthotopic tumors derived from KPC^wt, KPC^49bKO, and KPC^49bKOFAK cells, showing collagen (green) and CD49b (magenta). Nuclei are counterstained with DAPI (blue). Scale bars represent 150 μm.

(F and G) RNA-seq analysis for KPC^wt, KPC^49bKO, and KPC^49bKOFAK after in vivo tumor growth.

(F) Gene ontology enrichment using Fisher exact test (GeneSCF v1.1-p2). GSEA enrichment of showing ECM receptor interaction ad collagen biosynthesis pathways.

(G) Log2-normalized gene expression of ECM selected genes comparing KPC^wt, KPC^49bKO, and KPC^49bKOFAK after in vivo tumor growth.

(H) Representative images of collagen in tumor sections from KPC^wt, KPC^49bKO, and KPC^49bKOFAK tumor. Collagen was stained with collagen-mimetic peptide (CMP) conjugated with FITC (green), and DAPI (blue). Visualized and quantitatively analyzed of collagen fiber alignment was analyzed with CT-FIRE and presented as histograms. Scale bars represent 150 μm.

(I) Collagen was stained with collagen-mimetic peptide (CMP) conjugated with RFP (Red), CD8a (green), DAPI (blue). Scale bars represent 150 μm. Statistical significance (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

Analysis of the collagen in in vivo enrichment analysis revealed that extracellular matrix and collagen biosynthesis signaling is significantly upregulated in KPC^49bKOFAK compared to KPC^49bKO (Figure 6F). The most notable differences were in ECM related genes which includes collagen genes (Col1a1, col3a1), elastin (Eln), thrombospondin-1 (THBS1), and tenascin-C, which were downregulated in KPC^49bKO and restored by FAK overexpression (Figure 6G). Hence, we investigated the collagen fiber alignment in the CD49b knockout tumors. We observed that in absence of CD49b, collagen fibers are scattered and thinner compared to the thicker collagen in the parental tumor (Figure 6H, S12B, S12C, and S13A). Comprehensive quantification of collagen architectural parameters were evaluated including collagen area fraction, alignment index, fiber diameter, pore diameter, and porosity. The results showed that KPC^49bKO tumors display significantly reduced the alignment index, as long with reduced median fiber diameters and increased pore diameter comparing to KPC^wt while KCP^49bKOFAK tumors restores this effect comparable to KPC^wt. Additionally, we evaluated CD8⁺ T cell localization to assess potential immune modulation effects. FAK overexpression did not produce a significant change in CD8⁺ T cell localization (Figure S13B). Moreover, we stained CD8⁺ T cells, collagen and evaluated the alignment index in both aligned and non-aligned collagen. We demonstrated that non-aligned collagen exhibits higher CD8⁺ T cell infiltration compared to aligned collagen (Figure 6I).This finding confirms that KPC^49bKO tumor which is characterized by non-aligned and thinner collagen fibers, shows increased T cell infiltration. Pharmacologic inhibition of focal adhesion kinase (FAK) using defactinib resulted in effective pFAK in both KPC^wt and PANC-1 cell lines, with robust inhibition observed at 24 h and earlier kinetics in PANC-1 cells (Figure S14A). FAK inhibition leads to an upregulation of CXCL10 expression (Figure S14B). Collectively, our results suggest a potential compensatory role for FAK in the absence of CD49b. Rescue by FAK promotes PDAC tumor growth perhaps by maintaining the structural integrity and organization of collagen that inhibits movement of immune cells.

CD49b loss suppresses a tumor derived paracrine secreted genes that maintains CAF activation, and FAK overexpression restores it

To investigate the molecular basis by which tumor-intrinsic CD49b regulates stromal remodeling, we examined whether CD49b controls the expression of secreted paracrine factors known to sustain CAF activation in pancreatic ductal adenocarcinoma. Marker gene enrichment scoring of bulk RNA-seq data revealed that all four CAF subtype signatures (myCAF, iCAF, apCAF, and vCAF) showed reduced enrichment scores in KPC^CD49bKO tumors relative to KPC^wt, with partial recovery upon FAK overexpression (Figure S15A). Consistent with this, bulk RNA-seq analysis of KPC tumor cells revealed broad, concordant suppression of 40 curated secreted factors spanning five functional categories: TGF-β/activin, Wnt ligands, PDGF/VEGF, ECM/CTGF, and COX-2/MMP (Figure S15B). Effect sizes were large and consistent across categories; among the most suppressed were Vegfc (log₂FC= −4.14, padj <0.001), Pdgfd (log₂FC= −3.71, padj <0.001), Wnt7b (log₂FC= −3.47, padj <0.001), and Tgfb2 (log₂FC= −2.73, padj <0.001). FAK overexpression in CD49bKO tumor cells substantially restored expression of these secreted factors toward wild-type levels, demonstrating that FAK acts downstream of CD49b to maintain this pro-stromal secreted gene (Figure S15B).

To determine whether suppression of tumor-derived ligands was accompanied by a corresponding and directionally coherent loss of CAF activation, we performed ligand-target concordance scoring using 15 manually curated ligand/CAF target gene pairs from the primary PDAC literature, each linking a tumor-secreted factor to a fibroblast gene known to depend on that signal for example, Tgfb2/Acta2, Pdgfd/Pdgfra, and Vegfc/Pdgfrb. In the KPC^CD49bKO vs. KPC^WT comparison, all 15 of 15 ligand/target pairs were concordantly suppressed: every secreted gene that declined in KPC^CD49bKO tumor cells was accompanied by a significant reduction in its cognate CAF target gene (Figure S15C). Representative examples include Tgfb2 (log₂FC= −2.73) paired with Acta2 (log₂FC= −0.44), and Pdgfd (log₂FC= −3.71) paired with Pdgfra (log₂FC= −2.63; padj <0.001 for all ligands). FAK overexpression reversed this suppression in 14 of 15 pairs, with both ligands and their CAF targets recovering toward wild-type levels (Figure S15D). The full bidirectional pattern across both comparisons is visualized in the concordance scatterplot, where KPC^CD49bKO pairs cluster in the third quadrant and KPC^CD49bKOFAK pairs cluster in the first quadrant, with concordance rates of 15/15 and 14/15, respectively (Figure S15E). To validate these transcriptional findings at the protein level, CAFs were co-cultured with KPC^wt, KPC^CD49bKO, or KPC^CD49bKOFAK tumor cells, or exposed to their respective CM. Western blot analysis demonstrated that both Col1a1 and lox protein levels in CAFs were reduced upon co-culture or conditioned-media treatment from KPC^CD49bKO tumor cells, and were partially restored in KPC^CD49bKOFAK (Figure S15F). These findings functionally corroborate the RNA-level concordance data and confirm that CD49b dependent tumor fibroblast communication regulates CAF collagen-remodeling capacity at the protein level. However, we have to acknowledge that. First, this is bulk RNA-seq, and we cannot confirm that the ligand changes are occurring specifically in tumor cells versus other cell types in the tumor. Second, concordance in expression does not prove causation, it is consistent with the ligand driving the target, but the relationship could be indirect or co-regulated by a shared upstream factor.

Targeting CD49b suppresses PDAC tumor growth and enhances T cell infiltration

To evaluate the therapeutic potential of targeting CD49b in PDAC, we examined the effects of vatelizumab on cell adhesion across various PDAC cell lines (Figure 7A). Specifically, we assessed their adhesion to collagen types I and IV. Treatment with vatelizumab significantly suppressed cell adhesion on both collagen types I and IV, indicating that inhibiting CD49b with vatelizumab effectively disrupts PDAC cell adhesion (Figure 7A). To verify the specificity of vatelizumab, we tested its effect on the mouse KPC cell line. Notably, vatelizumab failed to inhibit cell adhesion in KPC cells, whereas mouse-specific anti-CD49b antibodies effectively suppressed adhesion (Figure 7B). We further confirmed the role of human CD49b by overexpressing it in CD49b deficient KPC cells (KPC^49bKOh49b) (Figure 7C). Overexpression of human CD49b in KPC^49bKO cells restored adhesion to collagen I, demonstrating the functional contribution of human CD49b to cell adhesion (Figure 7D). Next, we evaluated the antitumor efficacy of vatelizumab in vivo. Since tumor growth in KPC^49bKO mice remained delayed for over 30 days, we initiated treatment when tumors reached ∼100 mm³. Vatelizumab treatment significantly suppressed tumor growth compared to the untreated group (Figure 7E). This reduction in tumor growth was accompanied by a marked increase in CD8⁺ T cell infiltration (Figure 7F). Lastly, we assessed the effects of anti-mouse CD49b treatment. Consistent with the results for vatelizumab, anti-mouse CD49b treatment significantly suppressed tumor growth (Figure 7G) and increased T cell infiltration compared to untreated controls (Figure 7H). These findings suggest that targeting CD49b can effectively inhibit PDAC cell adhesion or enhance anti-tumor immune responses.

Figure 7.

Targeting CD49b delays the tumor growth in vivo

(A) Representative images of human PDAC cell lines cultured on COLI or COLIV-coated plates, with or without vatelizumab (10 μg/mL), following a 2-h incubation.

(B) Representative images of mouse KPC cells treated with or without vatelizumab (10 μg/mL) compared to anti mouse-CD49b antibody (10 μg/mL) on COL I and COL IV coated substrates, following a 2-h incubation. Scale bars, 600 μm.

(C) Histogram showing human CD49b expression in KPC^wt and KPC^49bKO.

(D) Representative images of mouse KPC cells expressing human CD49b treated with or without vatelizumab (10 μg/mL) on COL I- and COL IV-coated substrates after 2-h incubation. Scale bars, 600 μm.

(E) C57BL/6J mice were subcutaneously inoculated with KPC^49bKOhCD4b cells (1 × 10⁶ cells). Tumors were allowed to grow before mice were randomly assigned to PBS or vatelizumab (200 μg) intratumoral treatment groups. Tumor growth was measured three times per week.

(F) Bar graph shows immune cell infiltration in tumors, (G) C57BL/6J mice were subcutaneously inoculated with KPC^wt cells (1 × 1⁶ cells) on the right flank. Mice were randomly grouped to IgG control or amCD49b (200 μg) intratumoral treatment groups. Tumor growth was measured three times per week.

(H) Bar graph shows immune cell infiltration in tumors. Statistical significance: (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). Data are shown as mean ± SEM.

Discussion

In this study, we demonstrated that CD49b is significantly and selectively expressed in PDAC cancer cells. Prior studies have reported overexpression of CD49b in pancreatic cancer; however, our data highlight the spatial and cellular specificity of CD49b expression by demonstrating that it is predominantly localized within cancer cells of the PDAC TME.^33,34 This contrasts with previous investigations, which evaluated CD49b expression based on total tissue levels. Our observation that CD49b expression was absent in normal pancreatic tissue and PBMCs further defines its specificity to the TME. Prior studies have demonstrated that hypomethylation of ITGA2 correlates with elevated gene expression and are associated with poorer survival outcomes in PDAC.³⁵ Importantly, we also confirmed our findings in human PDAC tumors.

scRNA-seq, spatial transcriptomics, and FACS analysis of human PDAC samples further confirmed the high expression levels of ITGA1 and DDR2 in (CAFs). DDR2 has been reported to be upregulated in pancreatic, colorectal, and breast cancers.^36,37,38 Although DDR1 transcript levels are high and spatially enriched in PDAC cancer cells, protein-level detection via western blotting and histology shows minimal expression, likely due to proteolytic cleavage by membrane-type MMPs.^39,40,41 Our study detected high ITGB1 expression across various cellular compartments in the TME, aligning with findings that all beta integrins (ITGB1-8) are significantly overexpressed in PDAC and normal tissue.⁴²

Knocking out CD49b suppressed adhesion to collagen, even though CD29 is expressed. Similarly, CD29 knockout also abrogated collagen adhesion, despite the presence of intact CD49b. These results indicate that both CD49b and CD29 are indispensable for collagen-binding, as they function together as a heterodimeric receptor and essential for engaging collagen type I, a major component of the PDAC TME.

The knockout of either CD49b or CD29 not only suppressed tumor growth but also led to a significant increase in intratumoral CD8⁺ T cell infiltration. Tumor growth suppression was rescued following the depletion of CD8⁺ T cells. This suggests that disrupting integrin-mediated adhesion to collagen weakens the immunosuppressive TME, thereby enhancing antitumor immune responses. CD49b knockout tumors displayed disrupted collagen fiber alignment, characterized by thinner, scattered collagen fibrils. The functional consequences of this disorganization are unknown, but conceivably this alteration could modulate physical barriers and facilitate immune cell migration and enhance immune accessibility to the tumor. This is supported by the observed increase in CD8⁺ T cell infiltration and improved anti-tumor immune responses. Immunophenotyping revealed an increase in activated CD8⁺ T cells (CD44⁺ IFNγ⁺) without corresponding changes in immune checkpoint marker PD1 expression. The correlation between CD49b expression and the reduced presence of CD8⁺ T cells within tumor cores suggests that CD49b facilitates barriers to T cell infiltration. Spatial transcriptomics analysis further confirmed an inverse correlation between ITGA2 expression and CD8⁺ T cell levels. Additionally, TCGA analysis supported this finding. Similar exclusion of CD8⁺ T cells by DDR1 in breast cancer has been previously reported.⁴³ Our study demonstrates that CD49b plays a comparable role in excluding CD8⁺ T cells in PDAC.

Concurrently, CD49b-deficient cells exhibited reduced activation of FAK signaling, a pathway linked to immune evasion. Decreased FAK phosphorylation in knockout models suggests that CD49b-FAK signaling may regulate immunosuppressive properties. Notably, FAK overexpression in CD49b knockout cells restored tumor growth, suggesting that FAK signaling compensates for the loss of CD49b-mediated effect. These findings resonate with prior work indicated that FAK activation supports immune evasion, suggesting a broader role for FAK-mediated signaling in tumor progression.³² The CM from CD49b knockout cells contained elevated levels of CXCL10, a chemokine associated with T cell recruitment. These findings point to a dual role for CD49b in regulating both physical stromal architecture and chemotactic signals. CXCL10 elevation were reported to enhance CD8⁺ T cell infiltration when adhesion-related pathways are disrupted.^44,45 CXCL10 is canonically induced by inflammatory signaling through NF-κB and STAT1/IRF transcriptional programs, downstream of cytokine and innate immune pathways.⁴⁶ FAK signaling has been confirmed as a regulator of multiple signaling pathways during tumor progression, including chemokine transcription.⁴⁷ In line with these observations, our data show that CD49b knockout increases NF-κB, p-IRF3, STAT1, and p-STAT1 signaling, providing a mechanistic basis for elevated CXCL10 expression. Together, these findings support a model in which CD49b and FAK signaling restrains inflammatory chemokine programs through modulation of NF-κB/STAT/IRF pathways. Our findings align with prior work identifying tumor-cell-intrinsic FAK signaling as a key regulator of the pancreatic tumor microenvironment.⁴⁷ FAK inhibition in the KPC model supported by orthotopic and subcutaneous systems reduced fibrosis and enhanced CD8⁺ T cell infiltration, with CXCL12 mediating fibroblast driven matrix remodeling.⁴⁷ In our study identifies a complementary mechanism in which CD49b-dependent FAK signaling regulates CXCL10 expression in PDAC cells. CD49b knockout reduces FAK activation, increases CXCL10, and promotes a less fibrotic, immune-permissive tumor microenvironment.

Our findings demonstrate that tumor intrinsic CD49b maintains a broad pro-stromal paracrine secretome that sustains CAF activation and collagen-remodeling programs in PDAC. The concordant suppression of 15 ligand–CAF target pairs upon CD49b loss, spanning TGF-β/activin, Wnt, PDGF/VEGF, ECM/CTGF, and COX-2/MMP categories, is consistent with the established roles of these axes in driving CAF heterogeneity and myofibroblastic differentiation in pancreatic cancer.^48,49,50,51 FAK overexpression restored both the tumor secretome and downstream CAF activation markers, positioning FAK as a key effector linking CD49b-mediated adhesion signaling to stromal reprogramming. Reduction in COL1A1 and LOX protein in CAFs exposed to KPC^CD49bKO CM further validates this paracrine axis at the protein level.

Targeting CD49b using vatelizumab (anti-human CD49b) significantly impaired PDAC cell adhesion and suppressed tumor growth. However, previous reports using vatelizumab showed only modest tumor growth reduction in orthotopic xenografts.^52,53 Our results demonstrate significantly stronger antitumor activity, likely due to the presence of T cells in our model. In murine models, anti-mouse CD49b antibodies similarly reduced tumor progression and increased CD8⁺ T cell and dendritic cell infiltration. These findings highlight CD49b as a promising therapeutic target in PDAC, particularly in combination with strategies aimed at disrupting FAK signaling or enhancing T cell recruitment. Future studies should explore the clinical efficacy of CD49b inhibitors and their impact on immune cell dynamics within the TME.

Our comprehensive profiling of collagen receptors reveals their potential role in PDAC tumor progression and immune modulation, and specifically CD49b. CD49b′s influence on collagen architecture, immune evasion, and signaling pathways positions it as a key mediator of tumor-immune interactions. Targeting CD49b offers a novel approach to enhance anti-tumor immunity and disrupt PDAC progression, warranting further clinical investigation.

Limitations of the study

The central objective of the present study is to establish that tumor intrinsic CD49b regulates collagen architecture and immune exclusion, a relationship that is robust and reproducible across our transcriptional, co-culture, and structural analyses. However, several limitations remain. Our bulk RNA-seq data identify a broad CD49b dependent paracrine secreted genes concordantly suppressed with downstream CAF activation markers, but cannot definitively attribute these changes to tumor cells specifically, as single-cell resolution would be required to confirm cellular sources. As a consequence, while transcriptional concordance is consistent with a ligand-driven CAF suppression model, it does not establish formal causation. Unbiased proteomic profiling of CM, followed by ligand-specific perturbation experiments, would be needed to identify the precise soluble mediators responsible. Downstream of CAF suppression, KPC^CD49bKO tumors display disrupted collagen fiber alignment with thinner, scattered fibrils; however, the functional consequences of this matrix disorganization remain to be directly demonstrated, though we postulate that reduced collagen density may lower physical barriers to immune cell infiltration. Finally, how FAK transcriptionally regulates CXCL10 as part of this axis remains incompletely defined, and whether this relationship is direct or mediated through intermediate regulators will require dedicated mechanistic investigation.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Kenneth K. Tanabe (ktanabe@mgh.harvard.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

Bulk RNA-seq data were deposited at NCBI Gene Expression Omnibus (GEO) (GSE331086). Single-cell RNA-seq data analyzed in this study were derived from existing publicly available datasets accessible (for more information on STAR Methods). This paper does not report original code. All bioinformatic analyses were performed using publicly available software packages as described in the STAR Methods section. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We would like to thank Maris A. Handley and her team at the HSCI-CRM Flow Cytometry Core Facility at Massachusetts General Hospital (MGH) for their expert technical support. We also gratefully acknowledge the Center for Comparative Medicine (CCM) Animal Facility at MGH for their excellent care and support in animal husbandry and experimental procedures.

Author contributions

Conceptualization, I.R.E. and K.K.T.; methodology, I.R.E.; investigation, I.R.E.; writing – original draft, I.R.E.; writing—review and editing, I.R.E. and K.K.T.; funding acquisition, K.K.T.; resources, G.M.B., S.C., and T.S.; supervision, K.K.T.

Declaration of interests

The authors declare no conflicts of interest.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
TruStain FcX™ (anti-mouse CD16/32)	Biolegend	BioLegend Cat# 101320, RRID:AB_1574975
FITC anti-mouse CD3	Biolegend	BioLegend Cat# 100204, RRID:AB_312660
PE anti-mouse CD49a	Biolegend	BioLegend Cat# 142604, RRID:AB_10945158
PerCP/Cyanine5.5 anti-mouse CD279 (PD-1)	Biolegend	BioLegend Cat# 135208, RRID:AB_2159184
PE/Cyanine7 anti-mouse CD49b	Biolegend	BioLegend Cat# 103517, RRID:AB_2566103
APC anti-mouse CD4	Biolegend	BioLegend Cat# 100412, RRID:AB_312696
APC/Cyanine7 anti-mouse/rat CD29	Biolegend	BioLegend Cat# 102225, RRID:AB_2128076
Brilliant Violet 510™ anti-mouse CD69	Biolegend	BioLegend Cat# 104531, RRID:AB_2562326
Brilliant Violet 605™ anti-mouse CD45	Biolegend	BioLegend Cat# 103139, RRID:AB_2562341
Alexa Fluor® 700 anti-mouse CD8a	Biolegend	BioLegend Cat# 100730, RRID:AB_493703
Alexa Fluor® 647 anti-mouse/human CD44	Biolegend	BioLegend Cat# 103018, RRID:AB_493680
Brilliant Violet 421™ anti-mouse CD62L	Biolegend	BioLegend Cat# 104435, RRID:AB_2562560
Zombie UV™ Fixable Viability Kit	Biolegend	BioLegend Cat# 423108
PE anti-mouse F4/80	Biolegend	BioLegend Cat# 123109, RRID:AB_893498
PerCP/Cyanine5.5 anti-mouse CD11c	Biolegend	BioLegend Cat# 117328, RRID:AB_2129641
Alexa Fluor® 647 anti-mouse CD206 (MMR)	Biolegend	BioLegend Cat# 141712, RRID:AB_10900420
PE/Cyanine7 anti-mouse Ly-6G/Ly-6C (Gr-1)	Biolegend	BioLegend Cat# 108416, RRID:AB_313380
Alexa Fluor® 700 anti-mouse I-A/I-E	Biolegend	BioLegend Cat# 107622, RRID:AB_493727
APC/Cyanine7 anti-mouse/human CD11b	Biolegend	BioLegend Cat# 101225, RRID:AB_830641
Brilliant Violet 510™ anti-mouse/rat XCR1	Biolegend	BioLegend Cat# 148218, RRID:AB_2565231
Brilliant Violet 421™ anti-mouse CD103	Biolegend	BioLegend Cat# 121421, RRID:AB_10900074
FITC anti-mouse CD45	Biolegend	BioLegend Cat# 103108, RRID:AB_312972
PE anti-mouse CD45	Biolegend	BioLegend Cat# 103106, RRID:AB_312971
PerCP/Cyanine5.5 anti-mouse CD45	Biolegend	BioLegend Cat# 103132, RRID:AB_893340
APC anti-mouse CD45	Biolegend	BioLegend Cat# 103112, RRID:AB_312976
PE/Cyanine7 anti-mouse CD45	Biolegend	BioLegend Cat# 103114, RRID:AB_312978
Alexa Fluor® 647 anti-mouse CD45	Biolegend	BioLegend Cat# 103124, RRID:AB_493533
Alexa Fluor® 700 anti-mouse CD45	Biolegend	BioLegend Cat# 103128, RRID:AB_493715
Brilliant Violet 421™ anti-mouse CD45	Biolegend	BioLegend Cat# 103133, RRID:AB_10899570
Brilliant Violet 510™ anti-mouse CD45	Biolegend	BioLegend Cat# 103133, RRID:AB_10899570
APC/Cyanine7 anti-mouse CD45	Biolegend	BioLegend Cat# 103115, RRID:AB_312981
PerCP/Cyanine5.5 anti-human CD49a Antibody	Biolegend	BioLegend Cat# 328321, RRID:AB_2876635
PE/Cyanine7 anti-human CD49b Antibody	Biolegend	BioLegend Cat# 359313, RRID:AB_2820012
APC/Cyanine7 anti-human CD29 Antibody	Biolegend	BioLegend Cat# 303014, RRID:AB_493580
PE anti-human CD167a (DDR1) Antibody	Biolegend	BioLegend Cat# 334005, RRID:AB_2261656
Alexa Fluor® 647 anti-human CD305 (LAIR1) Antibody	Biolegend	BioLegend Cat# 342802, RRID:AB_2133480
Alexa Fluor® 700 anti-human CD31 Antibody	Biolegend	BioLegend Cat# 303133, RRID:AB_2566326
FITC anti-human CD326 (EpCAM) Antibody	Biolegend	BioLegend Cat# 324204, RRID:AB_756078
Pacific Blue™ anti-human CD45 Antibody	Biolegend	BioLegend Cat# 304029, RRID:AB_2174123
Rabbit phospho-ERK1/2	Cell Signaling	Cat# 4376, RRID:AB_331772
Rabbit ERK	Cell Signaling	Cat# 9102, RRID:AB_330744
Rabbit phospho-AKT	Cell Signaling	Cat# 4060, RRID:AB_2315049
Rabbit AKT	Cell Signaling	Cat# 4691, RRID:AB_915783
Rabbit beta-actin	Cell Signaling	Cat# 4970, RRID:AB_2223172
Rabbit DDR1	Cell Signaling	Cat# 5583, RRID:AB_10694842
Rabbit phospho-FAK	Cell Signaling	Cat# ab39967, RRID:AB_955850
Rabbit FAK	Cell Signaling	Cat# 3285, RRID:AB_2269034
Rabbit alpha2 integrin	Abcam	Cat# ab133557, RRID:AB_2833020
Rabbit beta1 integrin	Santa Cruz	Cat# sc-8978, RRID:AB_2130101
Rabbit CXCL10	Cell Signaling	Cat# 42912T, RRID: AB_2798668
Rabbit NF-κB p65	Cell Signaling	Cat# 8242S, RRID:AB_10859369
Rabbit Stat1	Cell Signaling	Cat# 14994, RRID: AB_2799965
Rabbit pSTAT1	Cell Signaling	Cat# 9167, RRID:AB_561284
Rabbit IRF-3	Cell Signaling	Cat# 4302, RRID: AB_1904036
Rabbit pIRF3	Cell Signaling	Cat# 29047, RRID:AB_2773013
Bacterial and virus strains
psPAX2	Addgene	Plasmid #12260
pMD2.G	Addgene	Plasmid #12259
FAK expression plasmid	Addgene	Plasmid #50515
Biological samples
Tissue sections of human PDAC	MGH	IRB (DF/HCC 02–240)
Chemicals, peptides, and recombinant proteins
Phosphatase inhibitor	Roche	Cat# 04906845001
Protease inhibitor	Roche	Cat# 4693116001
Purified anti-mouse CD49b Antibody	Biolegend	BioLegend Cat# 103501, RRID:AB_313024
Purified anti-mouse/rat CD29 Antibody	Biolegend	BioLegend Cat# 102201, RRID:AB_312878
Vatelizumab	MedChemExpress	HY-P99524
Type I collagen solution from rat tail	Sigma-Aldrich	Cat# C3867
Fibronectin (pure)	Sigma-Aldrich	11051407001
Critical commercial assays
Opal 3-Plex Manual Detection Kit (formerly Opal 4-Color Manual IHC Kit)	Akoya	NEL810001KT
Tumor Dissociation Kit, mouse	Miltenyi Biotec	130-096-730
Tumor Dissociation Kit, human	Miltenyi Biotec	130-095-929
SuperSignal™ West Pico PLUS Chemiluminescent Substrate	Thermoscientific	34580
MitoSpy™ Red CMXRos	Biolegend	424801
Deposited data
Bulk RNA sequencing of KPCwt and KPC49bKO	The current study	GEO: GSE331086
Single-cell RNA-sequencing of mouse PDAC tumors	Chen et al.26	GEO: GSE166298
Single-cell RNA-sequencing of human PDAC tumors	Peng et al., 2019	GSA: CRA001160
Single-nucleus RNA-sequencing of human PDAC tumors	Hwang et al., 2022	GEO: GSE202051
Spatial transcriptomics human PDAC	Kim, Seongryong et al., 2024	GEO: GSE235315
Spatial transcriptomics human PDAC	Lyubetskaya, Anna et al., 2022	GEO: GSE211895
TCGA pancreatic adenocarcinoma cohort survival and gene expression data (GDAC Firehose PAAD)	Broad Institute	http://gdac.broadinstitute.org/runs/stddata__2016_01_28/data/PAAD/20160128/
CCLE Database	Broad Institute	https://depmap.org/portal/download/all/
Experimental models: Cell lines
Hy15549 (KPC1 cell line)	Gift from Dr Nabeel Bardeesy at Harvard Medical School (Boston, MA, USA)	Derived from primary pancreatic tumors from Ptf1-Cre; LSL-KRAS-G12D; p53 Lox/+ mice; C57BL6J mice background
Han4.13 (KPC2 cell line)	Gift from Dr Nabeel Bardeesy at Harvard Medical School (Boston, MA, USA)	Derived from primary pancreatic tumors from Ptf1-Cre; LSL-KRAS-G12D; p53 Lox/+ mice; C57BL6J mice background
AKA4.4	Gift from Dr Nabeel Bardeesy at Harvard Medical School (Boston, MA, USA)	Derived from primary pancreatic tumors from a Ptf1-Cre/LSL-KrasG12D/Trp53Lox/+ mouse; FVB background
Experimental models: Organisms/strains
Wild-type C57BL/6	Jackson Laboratory	RRID: IMSR_JAX:000664)
NU/J nude homozygous Foxn1 mice	Jackson Laboratory	RRID: JAX:002019
Oligonucleotides
The ITGA2 sgRNA CRISPR/Cas9 All-in-One Lentivector set	abm	Cat# 25149114
ITGB1 sgRNA CRISPR/Cas9 All-in-One Lentivector set	abm	Cat# 25165114
A non-targeting sgRNA	abm	Cat# K010
Recombinant DNA
TaqMan probe anti human CXCL10	Thermo scientific	Hs00171042_m1
TaqMan probe antimouse CXCL10	Thermo scientific	Mm00445235_m1
Software and algorithms
Flowjo v10.7.1	Flowjo, L.L.C.	RRID: SCR_008520
Prism v8.0.0	GraphPad Software Inc.	RRID: SCR_002798
Fiji v2.0.0	ImageJ	RRID: SCR_002285
EVOS M7000	Thermo scientific	N/A
cBioportal v2.2.0	MSK Center for Mol Onc	https://www.cbioportal.org/
Trailmaker (Cellenics) scRNA seq analysis website	Harvard Medical School and Parse Biosciences	https://scp.biomage.net/| https://app.trailmaker.parsebiosciences.com/
Spatial transcriptomics analysis	Moffitt cancer center	https://spatialge.moffitt.org/

Experimental model and study participant details

Mice

All animal experiments were approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee (IACUC) and conducted according to institutional ethical guidelines. Female mice aged 8–10 weeks were used in all experiments. Wild-type C57BL/6 (RRID: IMSR_JAX:000664) and NU/J nude homozygous Foxn1 mice (Jackson Laboratory, 002019) were used. Animals were randomized before treatment experiments. For flank tumor experiments, tumor cells (KPC^wt, KPC^49bKO, KPC^29KO, KPC^49bKO29KO cell lines, Panc02^wt, Panc02^49bKO, Panc02^49bCD29KO cell lines) were injected subcutaneously (1 × 10⁶/100 μl PBS) into the flanks of C57BL/6J or NU/J nude mice. The flank tumor experiments were conducted until the tumors reached a humane endpoint, defined as maximum size of 1500 mm³. Tumor growth was regularly monitored using calipers. Tumor size was calculated using the formula L ∗W ²/2 where L represents the length and W represents the width of the tumor. Animals were euthanized when tumors reached the specified size to prevent unnecessary suffering. This approach aligns with ethical considerations to minimize distress and discomfort for the animals involved in the study. For FAK in vivo experiments, KPC^wt, KPC^49bKO, and KPC^49bKOFAK cells were injected subcutaneously, and tumor growth was monitored as described. For anti-mouse CD49b treatment, KPC^wt cells were injected subcutaneously, and mice were randomly assigned to treatment groups when tumors reached approximately 100 mm³. The control group received isotype IgG intratumorally (Ultra-LEAF Purified Armenian Hamster IgG Isotype Ctrl Antibody, Biolegend 400940), while the treatment group received anti-CD49b antibodies (Ultra-LEAF Purified anti-mouse IgG, Biolegend, 103526)) at a dose of 200 μg per tumor, administered intratumorally twice consecutively. For Vatelizumab treatment. Mice were injected subcutaneously with KPC^49bKOh49b cells and monitored until tumors formed (approximately 35 days). Mice were then randomly assigned to treatment groups when tumors reached approximately 100 mm³. The treatment group received intratumoral injections of Vatelizumab (GBR500, MedChemExpress HY-P99524) at 200 μg per tumor, administered three consecutive times. The control group received isotype IgG intratumorally. Human IgG4 (S228P) kappa, Isotype Control (MedChemExpress, HY-P99003). For pre-tumor treatments, mice were randomized to ensure comparable initial tumor sizes and body weights across groups. Tumors were collected for downstream analyses, including immunophenotyping, immunohistochemistry (IHC), and RNA sequencing (RNA-Seq). Survival outcomes were defined by tumor volumes exceeding 1,000 mm³, animal death, or distress. For post-tumor treatments. Body weights and tumor volumes were measured at the indicated time points. All experiments were conducted to minimize distress and discomfort to the animals. Euthanasia was performed in accordance with the approved protocol and institutional animal care and use guidelines to ensure humane treatment and confirm death.

For orthotopic tumor implantation; KPC^wt, KPC^CD49bKO, and KPC^CD49bKOFAK tumor cells were harvested, washed in ice-cold PBS, and viability confirmed at ≥ 90% by trypan blue exclusion. C57BL/6 mice (6–8 weeks) were anesthetized with ketamine/xylazine (100/10 mg/kg, i.p.), and 5 × 10⁴ tumor cells in 50 μL matrigel were injected into the pancreatic tail parenchyma via a 27-gauge needle under sterile surgical conditions. Analgesia was provided with buprenorphine extended-release (Ethiqa XR, 3.25 mg/kg, s.c.). Mice were monitored daily by palpation and body weight measurement and sacrificed at predefined or humane endpoints. Tumors were harvested, weighed, and processed for flow cytometry and formalin-fixed paraffin-embedded histological analysis. For the In vivo depletion of CD8⁺ T cells; C57BL/6 mice (RRID: IMSR_JAX:000664) were subjected to intraperitoneal (i.p.) administration of 200 μg per mouse of either anti-mouse CD8 antibody (clone 2.43, BioxCell, BE0061) or IgG2b isotype control (clone LTF-2, BioxCell, BE0090). This treatment regimen started two days prior to tumor inoculation and continued with twice-weekly administrations throughout the experimental period. Efficacy and specificity of CD8⁺ T cell depletion were assessed through flow cytometry analysis, ensuring the reliable interpretation of experimental outcomes. Because only female mice were used in these experiments, the influence of sex on the results could not be directly assessed.

Human tumor specimens

The study involved (7) treatment-naïve patients diagnosed with PDAC who underwent surgical resection. Tumor tissue specimens were collected and confirmed by a pathologist. Samples were obtained from both the tumor core and adjacent histologically normal pancreatic tissue (referred to as adjacent tumor) by the pathologist. The inclusion criteria required histopathologically confirmed PDAC. Tissue collection was performed following institutional review board (IRB) approval (DF/HCC 02–240), with written informed consent obtained from each patient prior to surgery.

Cell lines

Human PDAC cell lines were used including AsPC-1 (RRID: CVCL_0152), BxPC-3 (RRID: CVCL_XZ31), Capan-2 (RRID: CVCL_0026), CFPAC-1 (RRID: CVCL_1119, MiaPaca-2 (RRID: CVCL_0428), Panc-1 (RRID: CVCL_0480). Human PDAC cell lines were not independently authenticated after receipt. The mouse pancreatic cancer cell line KPC (Hy15549) (C57BL/6 background) was established from a tumor induced in an LSL-Kras^G12D/LSL-Trp53^R172H/Pdx1-Cre mouse. The mouse pancreatic cancer cell line AK4.4 (FVB background) was established from a tumor induced in a Ptf1-Cre/LSL-Kras^G12D/Trp53^Lox/+ mouse and were kindly provided by Nabeel Bardeesy, Massachusetts General Hospital. The authentication of mouse cell lines was performed using the 10000-MOUSE CELL LINE PLUS panel for pathogen screening via (VRL MARYLAND, LLC). Primary human and murine pancreatic stellate cells (PSCs) were obtained from tumor tissue sampled from surgical resection specimens with PDAC or murine tumor. Human and mouse pancreatic stellate cells were isolated and cultured by the outgrowth method as previously described.⁵⁴ The purity of the PSCs was assessed by morphology and demonstration of α-SMA and vimentin expression. Cell lines were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. Cell Cultures were maintained in a humidified incubator at 37°C with 5% CO2. Viability was assessed using trypan blue exclusion. Routine testing for mycoplasma was conducted by MycoAlert Mycoplasma Detection Kits (Lonza, 75870-454).

Method details

Generation of ITGA2 and ITGB1 knockout pancreatic tumor cell lines

To generate CD49b (ITGA2) and CD29 (ITGB1) knockout (KO) pancreatic tumor cell lines, CRISPR/Cas9-mediated gene editing was performed on KPC (Hy15549) and Panc02 (CVCL_D627) cell lines. The ITGA2 sgRNA CRISPR/Cas9 All-in-One Lentivector set (abm, 25149114) and the ITGB1 sgRNA CRISPR/Cas9 All-in-One Lentivector set (abm, 25165114) were used according to the manufacturer’s instructions. The sgRNA sequences used for ITGA2 and ITGB1 targeting were as follows: ITGA2 sgRNA sequences: (Target1: GGTTCGGCATAGCAGTTCTT, Target 2: GTTACTTCTGTTCCA TCGAC, Target 3: GGATCGGGTTAAGTGAAGTT). ITGB1 sgRNA sequences:(Target 1–116: TACAAGCA GGGCCAAATTGT, Target 2–280: AATGTCACCAAT CGCAGCAA, Target 3–500: CGGATTTGATGAATGAAATG). A non-targeting sgRNA (abm, K010) was used as a negative control for both ITGA2 and ITGB1 knockouts. Lentivirus production was performed by co-transfecting HEK293T cells (ATCC) with the respective KO vectors and the packaging plasmids psPAX2 and pMD2.G using Lipofectamine 2000 (Life Technologies, 11668027). After 48 h, the lentivirus-containing supernatant was harvested and used to transduce the target tumor cell lines. Double lentiviral transduction was performed to simultaneously knock out ITGA2 and ITGB1. Single-cell clones were selected using puromycin and expanded. The selected clones were further sorted using flow cytometry to enrich for successful knockouts. The knockout efficiency was confirmed using flow cytometry (FACS staining). All experiments using lentivirus were conducted following ethical guidelines and biosafety protocols. To evaluate downstream signaling, focal adhesion kinase (FAK) was overexpressed in KPC^49bKO cells using the FAK expression plasmid (Plasmid #50515, Addgene), which encodes a GFP-tagged FAK. Transfections were performed using Lipofectamine 2000 (ThermoFisher Scientific/Invitrogen) according to the manufacturer’s protocol. Successfully transfected clones were selected based on GFP expression and sorted using FACs and further validated by immunoblotting to confirm FAK overexpression.

Humanized KPC^49bKOh49b to evaluate vatelizumab efficacy

To assess the efficacy of Vatelizumab, a therapeutic antibody targeting human CD49b (ITGA2), human ITGA2 was overexpressed in KPC^49bKOmouse KPC cells (referred to as KPC^49bKOh49b). Human ITGA2 ORF mammalian expression plasmid C-OFPSpark/RFP tag (HG13024-ACR) was used to transfect KPC^49bKO using Lipofectamine 2000 cells. Successfully transduced clones were selected using hygromycin and expanded. The selected clones were sorted using flow cytometry (FACS) to enrich for populations expressing human CD49b. Expression of human CD49b and knockout of mouse CD49b (mCD49b) were confirmed by FACS co-staining, demonstrating successful overexpression of human ITGA2 and the complete loss of mouse ITGA2 expression. This humanized KPC^49bKOh49b cell line was established to specifically evaluate the binding and functional efficacy of Vatelizumab against human CD49b expressed in the mouse pancreatic tumor cell line.

In vitro CD8⁺ T cells isolation and migration assay

CD8⁺ T cells were isolated from splenocytes of naive C57BL/6 mice using the BioLegend MojoSort Mouse CD8⁺ T cell Isolation Kit (BioLegend, 480053) according to the manufacturer’s instructions. Prior to the assay, KPC^wt and KPC^49bKOcells were starved for 18–24 h in serum-free RPMI-1640 media. The migration assay was performed using the Cell Migration/Chemotaxis Assay Kit (96-well, 8 μm pore size; Abcam, ab235673). CD8⁺T cells (1 × 10⁶) in RPMI-1640 supplemented with 10% fetal 2-mercaptoethanol were added to the upper chamber and mixed with 100 ng of Ultra-LEAF Purified anti-mouse CXCL10 (IP-10) Antibody (BioLegend, 604453) as neutralizing antibody. Migration conditions included KPC^wt tumor-conditioned medium or KPC^49bKO tumor-conditioned medium in the bottom chamber. Cell migration was analyzed by reading fluorescence (Ex/Em = 530/590 nm) in a plate reader after 24 h.

Cancer co-cultured with CAF

For conditioned medium experiments, KPC^wt, KPC^CD49bKO, and KPC^CD49bKOFAK were cultured in serum-free medium for 48 h. Conditioned medium was collected, centrifuged at 300 × g for 5 min to remove cell debris, filtered through a 0.22 μm membrane, and added to mouse CAFs at a 1:1 ratio with complete fibroblast medium for 24–48 h. For direct co-culture experiments, KPC tumor cells and CAFs were co-seeded at a 1:3 tumor-to-fibroblast ratio cultured for 48 h under standard conditions (37°C, 5% CO₂). All experiments were performed in biological triplicate with at least three technical replicates per condition.

Immune phenotyping of the collagen receptors in PDAC by flow cytometry

The study involved treatment-naïve patients diagnosed with PDAC who underwent surgical resection. Tumor tissue specimens were collected and confirmed by a pathologist. Samples were obtained from both the tumor core and adjacent histologically normal pancreatic tissue (referred to as adjacent tumor) by the pathologist. The inclusion criteria required histopathologically confirmed PDAC. Tissue collection was performed following institutional review board (IRB) approval, with written informed consent obtained from each patient prior to surgery. Tissue specimens were processed based on the experimental approach. For histological analyses, tissues were fixed in 10% neutral buffered formalin and embedded in paraffin. For flow cytometry analysis, tissues were freshly digested using enzymatic dissociation protocols to obtain single-cell suspensions for fluorescence-activated cell sorting (FACS). Tumor core and adjacent histologically normal tissue were dissociated into single-cell suspensions using the Miltenyi Human Tumor Dissociation Kit (Miltenyi; 130-095-929). Cells were washed with PBS containing 5% FBS, (0.1%) sodium azid, and incubated with Fc receptor blocking reagent for 10 min at 4°C. Cells were stained with fluorescently conjugated antibodies targeting collagen receptors CD49a (ITGA1), CD49b (ITGA2), CD29 (ITGB1), DDR1, DDR2, and LAIR1. To define cellular subpopulations, markers for CD45 (immune cells), EPCAM (epithelial cells), CD31 (endothelial cells), and CD45⁻ EPCAM- CD31⁻ (stromal cells, including fibroblasts) were included. A viability dye was used to exclude dead cells. Flow cytometry was performed using [BD FACSAria II] with compensation applied for spectral overlap. Data acquisition and analysis were performed using FlowJo v10. Gating strategies included exclusion of debris and doublets, live single-cell selection, and CD45-based separation of immune from stromal and epithelial cells. Collagen receptor expression was quantified in epithelial (EPCAM+), endothelial (CD45⁻ EPCAM- CD31⁺), and stromal (CD45⁻ EPCAM- CD31⁻) compartments. The analysis conduct t-distributed stochastic neighbor embedding (t-SNE) visualization of the flow cytometry data by the FlowJo software.

Tumor-infiltrating lymphocytes (TILs) were isolated from mouse flank tumors using the Tumor Dissociation Kit (Mouse) from Miltenyi Biotec. Tumors were harvested and minced into small pieces before being enzymatically dissociated according to the manufacturer’s protocol. The resulting cell suspension was filtered through a 70 μm cell strainer to obtain a single-cell suspension and resuspended in PBS containing 5% FBS, (0.1%) sodium azid. After dissociation, cells were washed, Fc blocked, stained for flow cytometry analysis of collagen receptors on EPCAM+, CD45⁺, CD11b+, CD11c+, and T cells. All experiments were performed in accordance with ethical guidelines and institutional regulations governing the use of animal samples. For analysis of collagen receptors in human PDAC cell lines, cells were harvested and stained for CD49a, CD49b, CD29, DDR1, and LAIR1. Flow cytometry analysis was performed using FlowJo software. Due to the lack of commercially available mouse-specific FACs antibodies for DDR1 and LAIR1, KPC cell lines were harvested and stained for CD49a, CD49b, and CD29 only.

Histological and immunofluorescence

PDAC tissues and KPC flank tumors were fixed in 10% buffered formalin (Avantik, Cat # GR2030-40), paraffin-embedded, and sectioned at 5 μm. Sections were deparaffinized, rehydrated, and subjected to antigen retrieval in AR6 buffer (pH 6.0, Akoya Biosciences) for 20 min. After blocking with 10% normal goat serum in PBS for 1 h, sections were incubated overnight at 4°C with primary antibodies anti-CD8α (1:25; CST, Cat #98941), anti-CD49b (1:100; Abcam), and anti-CK19. After washing, Opal fluorophore-conjugated secondary antibodies corresponding to Opal 520 (GFP-equivalent), Opal 570 (Cy3-equivalent), and Opal 650 (Cy5-equivalent) were applied. Between each staining cycle, fluorophores were stripped by autoclaving at 95°C in AR6 buffer. Slides were mounted with Fluoromount-G Mounting Medium, with DAPI (Thermofisher; 00-4959-52). Fluorescent images were captured using the EVOS M7000 Imaging System (Thermo Fisher Scientific) at appropriate wavelengths: Opal 520 (494/525 nm), Opal 570 (550/570 nm), and Opal 650 (627/650 nm).

Segmentation of collagen fibers and quantification with CT-FIRE

Tumor tissues were fixed in formalin, embedded in paraffin, and sectioned at 5 μm thickness. For collagen staining, tissue sections underwent antigen retrieval by heating in citrate buffer (pH 6.0) at 95 °C for 20 min. Sections were washed in phosphate-buffered saline (PBS) and incubated incubated overnight at 4 °C with 5-FAM-conjugated Collagen CMP F-CHP (3Helix, FLU60) as per manufacturer instructions. After incubation, slides were rinsed three times in PBS and mounted with Fluoromount-G Mounting Medium, with DAPI (Thermofisher; 00-4959-52). Whole-slide fluorescence imaging was performed using the EVOS M7000 automated microscope. Images were analyzed using CT-FIRE (version 2), an open-source software that segments individual fibers and quantifies descriptors associated with single fibers (length, width, straightness) and the fiber network (alignment and density).⁵⁵ Fiber density was calculated by dividing the number of fibers by the image area, and straightness was defined as the ratio of end-to-end distance to fiber length, with values ranging from 0 (curved) to 1 (straight). Alignment was computed using circular descriptive statistics, generating a value between 0 and 1, where 1 indicates all fibers are oriented in the same direction. CT-FIRE preprocessing steps included background thresholding, noise reduction, and fiber edge enhancement. The software segmented individual fibers and reconstructed their structure while excluding short fibers below the length threshold.

Cell viability

The cell viability assay was conducted using the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) colorimetric method. Cells were seeded into 96-well plates and subjected to experimental treatments, with untreated cells serving as controls. Following the specified treatment period, MTT reagent (Invitrogen; M6494) was added to each well, and the cells were incubated to allow the conversion of MTT into formazan by metabolically active cells. The formazan crystals were then solubilized using dimethyl sulfoxide (DMSO), and the absorbance of the resulting solution was measured spectrophotometrically at 570 nm, with a reference wavelength of around 630 nm. The absorbance values were correlated with the number of viable cells, providing a quantitative assessment of cell viability. The assay was performed in triplicate to ensure reproducibility, and standard laboratory safety protocols were adhered to throughout the experimental procedure.

Western blotting

Cells were harvested and lysed in RIPA buffer (Boston BioProducts) supplemented with phosphatase (Roche, Cat #04906845001) and protease inhibitors (Roche, Cat #4693116001). Equal protein amounts were mixed with 6× reducing SDS Laemmli sample buffer (Boston BioProducts), denatured at 95°C for 20 min, and separated by SDS-PAGE. Proteins were transferred to PVDF membranes and blocked with 5% BSA to prevent nonspecific binding. Membranes were probed overnight at 4°C with primary antibodies, including actin (1:2000), AKT (1:1000), phospho-AKT Ser473 (1:500), ERK1/2 (1:500), phospho-ERK1/2 Thr202/Tyr204 (1:500), FAK (1:500), phospho-FAK Tyr397 (1:500), integrin α1 (1:300), α2 (1:2000), and β1 (1:200). After washing, membranes were incubated with HRP-conjugated secondary antibodies and visualized using enhanced chemiluminescence (ECL). All experiments were performed in triplicate with consistent results.

CAF co-culture and conditioned media experiments

Mouse CAFs were seeded and allowed to adhere overnight, then co-cultured with KPC KPC^wt, KPC^CD49bKO, or KPC^CD49bKOFAK tumor cells in direct contact for 48 h. For conditioned media (CM) experiments, tumor cells were cultured in serum-free media for 24 h; the resulting CM was collected, centrifuged to remove cellular debris, and applied to CAFs for 48 h. Following treatment, CAFs were lysed and protein was extracted for western blot analysis. Protein lysates were resolved by SDS-PAGE, transferred to PVDF membranes, and probed with antibodies against Col1a1, LOX, and β-actin as a loading control. Band intensities were quantified by densitometry. All co-culture and CM experiments were performed in at least three independent biological replicates.

Cytokine array analysis

KPC^wt or KPC^49bKO cells were cultured to 70–80% confluence, washed with PBS, and incubated in serum-free medium for 12 h. Conditioned media were collected, centrifuged at 1,000 × g for 10 min to remove debris, and stored at −80°C. Cytokine levels were measured using a mouse cytokine array kit (R&D Systems, Cat #ARY006) according to the manufacturer’s instructions. Membranes were blocked, incubated with CM overnight at 4°C, washed, and incubated with detection antibodies and streptavidin-conjugated reagents. Chemiluminescent signals were captured using the Odyssey XF Imaging System (LI-COR Biosciences), and spot intensities were analyzed with Image Studio Software, normalized to positive controls.

Collagen coating and cell adhesion

96-well plates were coated with Collagen I, rat tail (Gibco, Cat #A1048301) or Collagen IV (Sigma, Cat #CC076) at 100 μg/mL in 0.02 M acetic acid at 37°C for 1 h, with uncoated wells as controls. After two washes with 0.1% BSA in DMEM (washing buffer), wells were blocked with 0.5% BSA in DMEM (blocking buffer) at 37°C for 45–60 min, washed again, and chilled on ice. Cells were counted, adjusted to 4 × 10⁵ cells/mL, and 100 μL was added per well. Plates were incubated at 37°C for 30 min, shaken at 2,000 rpm for 10–15 s, and washed 2–3 times. Cells were fixed with 4% paraformaldehyde for 10–15 min, washed, and stained with Crystal Violet for 10 min. After rinsing with water and air-drying, 2% SDS was added, and plates were incubated at room temperature for 30 min before measuring absorbance at 550 nm.

In vitro blocking of CD49b and CD29

KPC cells were harvested and incubated with αCD49b (BioLegend, Cat #103526) or αCD29 (BioLegend, Cat #102202) blocking antibodies for 1 h on ice. For collagen coating, Collagen I (Gibco, Cat #A1048301) or Collagen IV (Sigma, Cat #CC076) was diluted to 100 μg/mL in 0.02 M acetic acid and incubated at 37°C for 1 h. Excess solution was removed, and wells were washed with PBS and air-dried. Cells were plated on the coated wells and allowed to adhere for 30 min. Non-adherent cells were removed by washing with PBS, and adherent cells were fixed with 4% paraformaldehyde at room temperature for 10 min. Fixed cells were stained with Crystal Violet (5 mg/mL in 2% ethanol) for 20 min. Excess stain was removed by rinsing with water, and plates were air-dried. The stain was solubilized with 2% SDS for 30 min at room temperature, and absorbance was measured at 550 nm using a microplate reader.

Assessment of vatelizumab-mediated inhibition of cell adhesion

Human PDAC cell lines (BXPC3, hPSC, CAPAN1, CAPAN2, CFPAC1, and ASPC1) as well as KPC mouse cell line were harvested and incubated with Vatelizumab (GBR500, MedChemExpress, Cat #HY-P99524) at a concentration of 10 μg/mL for 2 h on ice. For collagen coating, Collagen I (Gibco, Cat #A1048301) or Collagen IV (Sigma, Cat #CC076) was diluted to 100 μg/mL in 0.02 M acetic acid and incubated at 37°C for 1 h, and excess solution was removed before washing with PBS and air-drying. After incubation with Vatelizumab, cells were seeded onto collagen-coated wells and allowed to adhere for 30 min. Non-adherent cells were removed by washing with PBS, and adherent cells were fixed with 4% paraformaldehyde for 10 min at room temperature. Fixed cells were stained with Crystal Violet (5 mg/mL in 2% ethanol) for 20 min, and excess stain was rinsed off with water. Plates were air-dried, and the stain was solubilized with 2% SDS for 30 min at room temperature. Absorbance was measured at 550 nm using a microplate reader.

Bulk RNA sequencing

Bulk RNA-seq was conducted on in vitro-cultured KPC^wt, KPC^49bKO, KPC^29KO and KPC^{49b 29 KO}, KPC^49bKOFAK) as well as KPC^wt, KPC^49bKO and KPC^49bKOFAK tumors transplanted into C57BL/6 hosts. Sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality using fastp v.0.23.1. UMI-based de-duplication was performed using fastp v.0.23.1 simultaneously. Trimmed and de-duplicated reads were then mapped to the Mus musculus GRCm38 ERCC reference genome available on ENSEMBL using the STAR aligner v.2.5.2b. The STAR aligner is a splice aligner that detects splice junctions and incorporates them to help align the entire read sequences. BAM files were generated as a result of this step. Unique gene hit counts were calculated by using featureCounts from the Subread package v.1.5.2. After extraction of gene hit counts, the gene hit counts table was used for downstream differential expression analysis. Using DESeq2, a comparison of gene expression between the customer-defined groups of samples was performed. The Wald test was used to generate p-values and log2 fold changes. Genes with an adjusted p-value <0.05 and absolute log2 fold change >1 were called as differentially expressed genes for each comparison.

scRNA-seq, TCGA and microarray database analysis

scRNA-seq analysis was performed to investigate collagen receptors expression in PDAC using GEO datasets GSE166298 (KPC mouse models), GSE202051 (13 naive PDAC patients), and GSE155698 (6 adjacent normal vs. tumor),. Data were processed using Trailmaker (Parse Biosciences, 2024), formerly Cellenics, following a 7-step pipeline. Steps 1–5 involved quality control filtering based on mitochondrial content, gene counts, and UMI thresholds. In step 6, batch effects were corrected, and dimensionality reduction was performed. Step 7 involved embedding (UMAP) and clustering to identify cell populations. Localization of collagen receptor expression (e.g., ITGA2, DDR1) across cell types. For the analysis of differential gene expression in microarray datasets, GEO2R, a web-based tool available on the Gene Expression Omnibus (GEO), was employed. The datasets GSE28735 and GSE62452 were selected for this analysis. For gene expression analyses among TCGA PDAC cohort samples, data were derived from the RNA Seq V2 RSEM dataset, obtained through cBioPortal.

Spatial transcriptomics data analysis using spatialGE

We analyzed PDAC Visium 10× datasets (GSE235315 and GSE211895) using SpatialGE to explore the spatially-resolved expression of collagen receptors in PDAC tissues.⁵⁶ Raw 10× Visium data were imported and converted into STlist objects, followed by quality control filtering based on gene counts and unique molecular identifiers (UMIs). Normalization was performed to standardize expression across sections. Spatial gene expression patterns were visualized using STplot and gene_interpolation, while spatial autocorrelation metrics, such as Moran’s I and Geary’s C, were computed with SThet to assess clustering or dispersion. Tissue domains were identified through spatially-informed hierarchical clustering using STclust. We focused on the expression of ITGA2 and confirmed its localization within tumor regions by identifying KRT18-positive spots. Correlation analyses were conducted to examine the relationship between CD49b and the cancer cell marker KRT18, as well as between CD8A, a T cell marker. To evaluate T cell exclusion, we extracted the expression of ITGA2 and CD8a spots using SpatialGE and categorized them into four groups: spots expressing both ITGA2 and CD8a (ITGA2⁺ CD8A⁺), spots expressing neither (ITGA2^- CD8a⁻), spots expressing ITGA2 but not CD8a (ITGA2⁺ CD8a⁻), and spots expressing CD8a but not ITGA2 (ITGA2^- CD8a⁺). To calculate the number of spots expressing each receptor per the total number of spots, you can use the following formula: Percentage of spots expressing a receptor = (Number of spots expressing the receptor/total number of spots)×100. The total number of spots in each category was calculated and visualized using a donut chart using GraphPad Prism.

Bulk RNA-seq CAF subtype enrichment scoring

To estimate the relative transcriptional representation of CAF subtypes from bulk RNA-seq data, a marker-gene enrichment scoring approach was applied, conceptually analogous to MCPcounter (Becht et al., Genome Biology, 2016). Curated marker gene sets were defined for five CAF signatures based on published single-cell RNA-seq studies of PDAC stroma: myCAF (Acta2, Thy1, Pdpn, Col1a1, Col1a2), iCAF (Il6ra, Has2, Has3, Pdgfra, Cxcl12, Lrrc32), apCAF (Fap, Ly6c1, Pdpn, Ptprc), vCAF (Pdgfrb, Pecam1, Notch3, Acta2), and a general CAF signature (Acta2, Fap, Pdpn, Col1a1, Col1a2, S100a4, Vim). For each gene in a given signature, normalized expression counts were Z score transformed across all samples from the three experimental groups (KPCˆWTˆ, KPCˆCD49bKOˆ, and KPCˆCD49bKOFAKˆ). An enrichment score for each CAF subtype per sample was then computed as the mean Z score of all marker genes belonging to that signature. Enrichment scores were visualized as a heatmap with samples grouped by experimental condition.

Curated secretome analysis and visualization

A panel of 40 secreted factors with established roles in CAF activation and stromal remodeling was manually curated from the PDAC literature and organized into five functional categories: TGF-β/Activin (Tgfb1, Tgfb2, Tgfb3, Inhba, Inhbb, Bmp4), Wnt ligands (Wnt5a, Wnt7b, Ctnnb1), PDGF/VEGF (Pdgfa, Pdgfb, Pdgfd, Vegfa, Vegfc), ECM/CTGF (Fn1, Ctgf, Postn, Lox, Loxl2, Cyr61, Sparc, Thbs2), and COX-2/MMP (Ptgs2, Mmp2, Mmp3, Mmp9, Mmp11, Mmp14). For each gene, DESeq2-derived log₂ fold-change values and Benjamini–Hochberg-adjusted p-values were extracted from two pairwise comparisons: KPC^CD49bKO vs. KPC^WT and KPCCD^49bKOFAK vs. KPC^CD49bKO. To enable cross-gene comparison independent of absolute expression magnitude, gene-level z-scores were computed from group mean normalized expression values across the three conditions and displayed as a flanking heatmap. Dot plots of log₂FC values for both comparisons were generated with genes grouped by functional category. Statistical significance: (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

Ligand–target concordance scoring

To assess whether suppression of tumor-derived paracrine signals in CD49b-KO tumor cells was accompanied by directionally coherent changes in cognate CAF target genes, a ligand–target concordance scoring framework was applied. Fifteen ligand → CAF target gene pairs were manually curated from primary PDAC literature (Biffi et al. 2019; Elyada et al. 2019; Öhlund et al. 2017; Laklai et al. 2016; Bhattacharya et al. 2020), with each pair linking a secreted factor expressed by tumor cells to a fibroblast gene whose expression is known to depend on that upstream signal. Curated pairs included, for example: Tgfb2/Acta2 (TGF-β2 drives αSMA expression in myCAFs), Pdgfd/Pdgfra (PDGF-D activates iCAFs via PDGFRα), and Vegfc/Pdgfrb (PDGFRβ defines the vCAF subtype). For each pair, DESeq2 log₂ fold-change values were extracted for both the ligand gene and its CAF target gene in two comparisons: KPC^CD49bKO vs. KPC^WT and KPC^CD49bKOFAK vs. KPC^CD49bKO. A pair was scored as concordant if the ligand and its target changed in the same direction (both negative or both positive) within a given comparison. Concordance rate was reported as the number of concordant pairs out of 15 for each comparison. Results were visualized as paired horizontal bar plots in which full bars represent ligand log₂FC and shaded bars of the same color represent CAF target log₂FC, grouped by functional category. All statistical analyses and visualizations were performed in R (version 4.3) using the ggplot2, ComplexHeatmap, and DESeq2 packages.

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism and R version 4.3, as indicated for specific analyses. Flow cytometry, immunostaining, cell adhesion, viability, cytokine array, and tumor growth quantifications were analyzed using unpaired two-tailed Student’s t tests, one-way ANOVA followed by Tukey’s multiple-comparison test, or Fisher’s exact test, as appropriate for the experimental design and data type. For RNA-seq differential expression analysis, statistical testing was performed using DESeq2 with the Wald test. p values were adjusted for multiple comparisons using the Benjamini-Hochberg method, and genes with adjusted p < 0.05 and absolute log₂ fold change >1 were considered differentially expressed. Unless otherwise stated, data are presented as mean ± SEM. Exact n values and definitions of n are provided in the figure legends. Statistical significance thresholds were defined as: Statistical significance: (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).