Machine Learning Links T-cell Function and Spatial Localization to Neoadjuvant Immunotherapy and Clinical Outcome in Pancreatic Cancer.

Tumor molecular data sets are becoming increasingly complex, making it nearly impossible for humans alone to effectively analyze them. Here, we demonstrate the power of using machine learning (ML) to analyze a single-cell, spatial, and highly multiplexed proteomic data set from human pancreatic cancer and reveal underlying biological mechanisms that may contribute to clinical outcomes. We designed a multiplex immunohistochemistry antibody panel to compare T-cell functionality and spatial localization in resected tumors from treatment-naïve patients with localized pancreatic ductal adenocarcinoma (PDAC) with resected tumors from a second cohort of patients treated with neoadjuvant agonistic CD40 (anti-CD40) monoclonal antibody therapy. In total, nearly 2.5 million cells from 306 tissue regions collected from 29 patients across both cohorts were assayed, and over 1,000 tumor microenvironment (TME) features were quantified. We then trained ML models to accurately predict anti-CD40 treatment status and disease-free survival (DFS) following anti-CD40 therapy based on TME features. Through downstream interpretation of the ML models' predictions, we found anti-CD40 therapy reduced canonical aspects of T-cell exhaustion within the TME, as compared with treatment-naïve TMEs. Using automated clustering approaches, we found improved DFS following anti-CD40 therapy correlated with an increased presence of CD44+CD4+ Th1 cells located specifically within cellular neighborhoods characterized by increased T-cell proliferation, antigen experience, and cytotoxicity in immune aggregates. Overall, our results demonstrate the utility of ML in molecular cancer immunology applications, highlight the impact of anti-CD40 therapy on T cells within the TME, and identify potential candidate biomarkers of DFS for anti-CD40-treated patients with PDAC.

Niche-DE: niche-differential gene expression analysis in spatial transcriptomics data identifies context-dependent cell-cell interactions.

Existing methods for analysis of spatial transcriptomic data focus on delineating the global gene expression variations of cell types across the tissue, rather than local gene expression changes driven by cell-cell interactions. We propose a new statistical procedure called niche-differential expression (niche-DE) analysis that identifies cell-type-specific niche-associated genes, which are differentially expressed within a specific cell type in the context of specific spatial niches. We further develop niche-LR, a method to reveal ligand-receptor signaling mechanisms that underlie niche-differential gene expression patterns. Niche-DE and niche-LR are applicable to low-resolution spot-based spatial transcriptomics data and data that is single-cell or subcellular in resolution.

CT109-SN-38, a Novel Antibody-drug Conjugate with Dual Specificity for CEACAM5 and 6, Elicits Potent Killing of Pancreatic Cancer Cells.

BACKGROUND: CEACAM5 and CEACAM6 are glycosylphosphatidylinositol (GPI)- linked members of the carcinoembryonic antigen-related cell adhesion molecule (CEACAM) family, which are frequently upregulated in epithelial cancers where they contribute to invasion, metastasis, immune evasion, and resistance to anoikis. CT109 is a novel antibody with dual specificity to both CEACAM5 and 6. OBJECTIVE: In this study, we aimed to perform the preclinical characterization of CT109 and antibody- drug conjugate (ADCs) derivatives of CT109, focusing on CT109-SN-38. METHODS: CT109's cognate epitope was characterized by scanning mutagenesis. CT109 specificity and internalization kinetics were assessed by immunoblot and flow cytometry, respectively. Cognate antigen expression prevalence in colorectal cancer and normal tissue arrays was determined by immunohistochemistry. CT109 conjugations were generated by the reaction of reduced CT109 cysteines with maleimide-functionalized payload linkers. In vitro cytotoxic activity of CT109 ADCs was characterized on antigen-positive and negative pancreatic ductal adenocarcinoma cell (PDAC) lines using a luminometric viability assay. In vivo efficacy of CT109-SN-38 was assessed on a PDAC tumor xenograft model at 10 and 25 mg/kg concentrations. RESULTS: CT109 showed to bind a glycoepitope centered on N309. CT109 is internalized in the CEACAM5+/CEACAM6+ double-positive PDAC line, BxPC-3, with a t1/2 of 2.3 hours. CT109 ADCs elicit a dose and antigen-dependent cytotoxic effect, with CT109-SN-38 exhibiting an IC50 value of 21 nM in BxPC-3 cells. In a BxPC-3 tumor xenograft model, CT109-SN-38 reduced tumor growth and induced regression in 3/10 mice at 25 mg/kg concentration. CONCLUSIONS: These data suggest that further preclinical and clinical development of CT109-SN-38 is warranted.

Inferring super-resolution tissue architecture by integrating spatial transcriptomics with histology.

Spatial transcriptomics (ST) has demonstrated enormous potential for generating intricate molecular maps of cells within tissues. Here we present iStar, a method based on hierarchical image feature extraction that integrates ST data and high-resolution histology images to predict spatial gene expression with super-resolution. Our method enhances gene expression resolution to near-single-cell levels in ST and enables gene expression prediction in tissue sections where only histology images are available.

SMAD4 loss predicts worse overall and distant metastasis-free survival in patients with resected pancreatic adenocarcinoma.

BACKGROUND: In select patients, pancreatic adenocarcinoma remains a local disease, yet there are no validated biomarkers to predict this behavior and who may benefit from aggressive local treatments. This study sought to determine if SMAD4 (mothers against decapentaplegic homolog 4) messenger RNA-sequencing (RNA-seq) expression is a robust method for predicting overall survival (OS) and distant metastasis-free survival (DMFS) in patients with resected pancreatic adenocarcinoma. METHODS: Utilizing The Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC), 322 patients with resected stage I-III pancreatic adenocarcinoma were identified. In TCGA, multivariable proportional hazards models were used to determine the association of SMAD4 genomic aberrations and RNA-seq expression with OS and DMFS. In the ICGC, analysis sought to confirm the predictive performance of RNA-seq via multivariable models and receiver operator characteristic curves. RESULTS: In TCGA, the presence of SMAD4 genomic aberrations was associated with worse OS (hazard ratio [HR], 1.55; 95% CI, 1.00-2.40; p = .048) but not DMFS (HR, 1.33; 95% CI, .87-2.03; p = .19). Low SMAD4 RNA-seq expression was associated with worse OS (HR, 1.83; 95% CI, 1.17-2.86; p = .008) and DMFS (HR, 1.70; 95% CI, 1.14-2.54; p = .009). In the ICGC, increased SMAD4 RNA-seq expression correlated with improved OS (area under the curve [AUC], .92; 95% CI, .86-.94) and DMFS (AUC, .84; 95% CI, .82-.87). CONCLUSIONS: In patients with resected pancreatic adenocarcinoma, SMAD4 genomic aberrations are associated with worse OS but do not predict for DMFS. Increased SMAD4 RNA-seq expression is associated with improved OS and DMFS in patients with resected pancreatic adenocarcinoma. This reproducible finding suggests SMAD4 RNA-seq expression may be a useful marker to predict metastatic spread.

Virtual biopsies: Proof of concept for a novel quantitative approach to liver biopsy adequacy and pathology education.

OBJECTIVES: To quantitatively measure liver biopsy adequacy requirements and the effect of a teaching intervention that uses a virtual biopsy platform. METHODS: A library of virtual liver biopsies was created using digital whole-slide, trichrome-stained tissue sections from liver resection material and QuPath image analysis software. Blinded participants staged fibrosis on the virtual biopsies before and after a teaching intervention. RESULTS: This platform both modeled adequacy requirements for cirrhosis diagnosis on biopsy material and measured the effect of a teaching intervention on participant performance. Using this platform, diagnostic accuracy for cirrhosis could be modeled according to the function y = λ(1 ‒ e‒x/γ). The platform demonstrated that the relationship between biopsy size and diagnostic accuracy was statistically significant and that biopsies smaller than 6 mm long and 0.8 mm wide were insufficient to diagnosis cirrhosis. The platform also measured improvement in fibrosis staging accuracy among participants following a teaching intervention. CONCLUSIONS: These results provide proof of concept for a virtual biopsy method by which outstanding questions in anatomic pathology can be addressed quantitatively using open source software. Future work is needed to validate these findings in clinical practice.

RALDH1 Inhibition Shows Immunotherapeutic Efficacy in Hepatocellular Carcinoma.

Globally, hepatocellular carcinoma (HCC) is one of the most commonly diagnosed cancers and a leading cause of cancer-related death. We previously identified an immune evasion pathway whereby tumor cells produce retinoic acid (RA) to promote differentiation of intratumoral monocytes into protumor macrophages. Retinaldehyde dehydrogenase 1 (RALDH1), RALDH2, and RALDH3 are the three isozymes that catalyze RA biosynthesis. In this study, we have identified RALDH1 as the key driver of RA production in HCC and demonstrated the efficacy of RALDH1-selective inhibitors (Raldh1-INH) in suppressing RA production by HCC cells. Raldh1-INH restrained tumor growth in multiple mouse models of HCC by reducing the number and tumor-supporting functions of intratumoral macrophages as well as increasing T-cell infiltration and activation within tumors. Raldh1-INH also displayed favorable pharmacokinetic, pharmacodynamic, and toxicity profiles in mice thereby establishing them as promising new drug candidates for HCC immunotherapy.

A missense variant in human perilipin 2 (PLIN2 Ser251Pro) reduces hepatic steatosis in mice.

Background & Aims: Non-alcoholic fatty liver disease (NAFLD) is characterised by the accumulation of lipid droplets (LDs) within hepatocytes. Perilipin 2 (PLIN2) is the most abundant protein in hepatic LDs and its expression correlates with intracellular lipid accumulation. A recently discovered PLIN2 coding variant, Ser251Pro (rs35568725), was found to promote the accumulation of small LDs in embryonic kidney cells. In this study, we investigate the role of PLIN2-Ser251Pro (PLIN2-Pro251) on hepatic LD metabolism in vivo and research the metabolic phenotypes associated with this variant in humans. Methods: For our animal model, we used Plin2 knockout mice in which we expressed either human PLIN2-Pro251 (Pro251 mice) or wild-type human PLIN2-Ser251 (Ser251 mice) in a hepatocyte-specific manner. We fed both cohorts a lipogenic high-fat, high-cholesterol, high-fructose diet for 12 weeks. Results: Pro251 mice were associated with reduced liver triglycerides (TGs) and had lower mRNA expression of fatty acid synthase and diacylglycerol O-acyltransferase-2 compared with Ser251 mice. Moreover, Pro251 mice had a reduction of polyunsaturated fatty acids-TGs and reduced expression of epoxygenase genes. For our human study, we analysed the Penn Medicine BioBank, the Million Veteran Program, and UK Biobank. Across these databases, the minor allele frequency of PLIN2-Pro251 was approximately 5%. There was no association with the clinical diagnosis of NAFLD, however, there was a trend toward reduced liver fat in PLIN2-Pro251 carriers by MRI-spectroscopy in UK Biobank subjects. Conclusions: In mice lacking endogenous Plin2, expression of human PLIN2-Pro251 attenuated high-fat, high-fructose, high-cholesterol, diet-induced hepatic steatosis compared with human wild-type PLIN2-Ser251. Moreover, Pro251 mice had lower polyunsaturated fatty acids-TGs and epoxygenase genes expression, suggesting less liver oxidative stress. In humans, PLIN2-Pro251 is not associated with NAFLD. Impact and Implications: Lipid droplet accumulation in hepatocytes is the distinctive characteristic of non-alcoholic fatty liver disease. Perilipin 2 (PLIN2) is the most abundant protein in hepatic lipid droplets; however, little is known on the role of a specific polymorphism PLIN2-Pro251 on hepatic lipid droplet metabolism. PLIN2-Pro251 attenuates liver triglycerides accumulation after a high-fat-high-glucose-diet. PLIN2-Pro251 may be a novel lipid droplet protein target for the treatment of liver steatosis.

Machine learning links T cell function and spatial localization to neoadjuvant immunotherapy and clinical outcome in pancreatic cancer.

Tumor molecular datasets are becoming increasingly complex, making it nearly impossible for humans alone to effectively analyze them. Here, we demonstrate the power of using machine learning to analyze a single-cell, spatial, and highly multiplexed proteomic dataset from human pancreatic cancer and reveal underlying biological mechanisms that may contribute to clinical outcome. A novel multiplex immunohistochemistry antibody panel was used to audit T cell functionality and spatial localization in resected tumors from treatment-naive patients with localized pancreatic ductal adenocarcinoma (PDAC) compared to a second cohort of patients treated with neoadjuvant agonistic CD40 (αCD40) monoclonal antibody therapy. In total, nearly 2.5 million cells from 306 tissue regions collected from 29 patients across both treatment cohorts were assayed, and more than 1,000 tumor microenvironment (TME) features were quantified. We then trained machine learning models to accurately predict αCD40 treatment status and disease-free survival (DFS) following αCD40 therapy based upon TME features. Through downstream interpretation of the machine learning models' predictions, we found αCD40 therapy to reduce canonical aspects of T cell exhaustion within the TME, as compared to treatment-naive TMEs. Using automated clustering approaches, we found improved DFS following αCD40 therapy to correlate with the increased presence of CD44+ CD4+ Th1 cells located specifically within cellular spatial neighborhoods characterized by increased T cell proliferation, antigen-experience, and cytotoxicity in immune aggregates. Overall, our results demonstrate the utility of machine learning in molecular cancer immunology applications, highlight the impact of αCD40 therapy on T cells within the TME, and identify potential candidate biomarkers of DFS for αCD40-treated patients with PDAC.

Mapping and modeling human colorectal carcinoma interactions with the tumor microenvironment.

The initiation and progression of cancer are intricately linked to the tumor microenvironment (TME). Understanding the function of specific cancer-TME interactions poses a major challenge due in part to the complexity of the in vivo microenvironment. Here we predict cancer-TME interactions from single cell transcriptomic maps of both human colorectal cancers (CRCs) and mouse CRC models, ask how these interactions are altered in human tumor organoid (tumoroid) cultures, and functionally recapitulate human myeloid-carcinoma interactions in vitro. Tumoroid cultures suppress gene expression programs involved in inflammation and immune cell migration, providing a reductive platform for re-establishing carcinoma-immune cell interactions in vitro. Introduction of human monocyte-derived macrophages into tumoroid cultures instructs macrophages to acquire immunosuppressive and pro-tumorigenic gene expression programs similar to those observed in vivo. This includes hallmark induction of SPP1, encoding Osteopontin, an extracellular CD44 ligand with established oncogenic effects. Taken together, these findings offer a framework for understanding CRC-TME interactions and provide a reductionist tool for modeling specific aspects of these interactions.

When You Come to a Fork in the Road, Take It: Wnt Signaling Activates Multiple Pathways through the APC/Axin/GSK-3 Complex.

The Wnt signaling pathway is a highly conserved regulator of metazoan development and stem cell maintenance. Activation of Wnt signaling is an early step in diverse malignancies. Work over the past four decades has defined a "canonical" Wnt pathway that is initiated by Wnt proteins, secreted glycoproteins that bind to a surface receptor complex and activate intracellular signal transduction by inhibiting a catalytic complex composed of the classical tumor suppressor Adenomatous Polyposis Coli (APC), Axin, and Glycogen Synthase Kinase-3 (GSK-3). The best characterized effector of this complex is ß-catenin, which is stabilized by inhibition of GSK-3, allowing ß-catenin entrance to the nucleus and activation of Wnt target gene transcription, leading to multiple cancers when inappropriately activated. However, canonical Wnt signaling through the APC/Axin/GSK-3 complex impinges on other effectors, independently of ß-catenin, including the mechanistic Target of Rapamycin (mTOR), regulators of protein stability, mitotic spindle orientation, and Hippo signaling. This review focuses on these alternative effectors of the canonical Wnt pathway and how they may contribute to cancers.

A fetal wound healing program after intrauterine bile duct injury may contribute to biliary atresia.

BACKGROUND & AIMS: Biliary atresia (BA) is an obstructive cholangiopathy that initially affects the extrahepatic bile ducts (EHBDs) of neonates. The etiology is uncertain, but evidence points to a prenatal cause. Fetal tissues have increased levels of hyaluronic acid (HA), which plays an integral role in fetal wound healing. The objective of this study was to determine whether a program of fetal wound healing is part of the response to fetal EHBD injury. METHODS: Mouse, rat, sheep, and human EHBD samples were studied at different developmental time points. Models included a fetal sheep model of prenatal hypoxia, human BA EHBD remnants and liver samples taken at the time of the Kasai procedure, EHBDs isolated from neonatal rats and mice, and spheroids and other models generated from primary neonatal mouse cholangiocytes. RESULTS: A wide layer of high molecular weight HA encircling the lumen was characteristic of the normal perinatal but not adult EHBD. This layer, which was surrounded by collagen, expanded in injured ducts in parallel with extensive peribiliary gland hyperplasia, increased mucus production and elevated serum bilirubin levels. BA EHBD remnants similarly showed increased HA centered around ductular structures compared with age-appropriate controls. High molecular weight HA typical of the fetal/neonatal ducts caused increased cholangiocyte spheroid growth, whereas low molecular weight HA induced abnormal epithelial morphology; low molecular weight HA caused matrix swelling in a bile duct-on-a-chip device. CONCLUSION: The fetal/neonatal EHBD, including in human EHBD remnants from Kasai surgeries, demonstrated an injury response with prolonged high levels of HA typical of fetal wound healing. The expanded peri-luminal HA layer may swell and lead to elevated bilirubin levels and obstruction of the EHBD. IMPACT AND IMPLICATIONS: Biliary atresia is a pediatric cholangiopathy associated with high morbidity and mortality rates; although multiple etiologies have been proposed, the fetal response to bile duct damage is largely unknown. This study explores the fetal pathogenesis after extrahepatic bile duct damage, thereby opening a completely new avenue to study therapeutic targets in the context of biliary atresia.

Glisson's capsule matrix structure and function is altered in patients with cirrhosis irrespective of aetiology.

Background & Aims: Glisson's capsule is the interstitial connective tissue that surrounds the liver. As part of its normal physiology, it withstands significant daily changes in liver size. The pathophysiology of the capsule in disease is not well understood. The aim of this study was to characterise the changes in capsule matrix, cellular composition, and mechanical properties that occur in liver disease and to determine whether these correlate with disease severity or aetiology. Methods: Samples from ten control patients, and six with steatosis, seven with moderate fibrosis, and 37 with cirrhosis were collected from autopsies, intraoperative biopsies, and liver explants. Matrix proteins and cell markers were assessed by staining and second harmonic generation imaging. Mechanical tensile testing was performed on a test frame. Results: Capsule thickness was significantly increased in cirrhotic samples compared with normal controls irrespective of disease aetiology (70.12 ± 14.16 µm and 231.58 ± 21.82 µm, respectively), whereas steatosis and moderate fibrosis had no effect on thickness (90.91 ± 11.40 µm). Changes in cirrhosis included an increase in cell number (fibroblasts, vascular cells, infiltrating immune cells, and biliary epithelial cells). Key matrix components (collagens 1 and 3, hyaluronan, versican, and elastin) were all deposited in the lower capsule, although only the relative amounts per area of hyaluronan and versican were increased. Organisational features, including crimping and alignment of collagen fibres, were also altered in cirrhosis. Unexpectedly, capsules from cirrhotic livers had decreased resistance to loading compared with controls. Conclusions: The liver capsule, similar to the parenchyma, is an active site of disease, demonstrating changes in matrix and cell composition as well as mechanical properties. Impact and implications: We assessed the changes in composition and response to stretching of the liver outer sheath, the capsule, in human liver disease. We found an increase in key structural components and numbers of cells as well as a change in matrix organisation of the capsule during the later stages of disease. This allows the diseased capsule to stretch more under any given force, suggesting that it is less stiff than healthy tissue.

Nonsteroidal anti-inflammatory drugs sensitize epithelial cells to Clostridioides difficile toxin-mediated mitochondrial damage.

Clostridioides difficile damages the colonic mucosa through the action of two potent exotoxins. Factors shaping C. difficile pathogenesis are incompletely understood but are likely due to the ecological factors in the gastrointestinal ecosystem, mucosal immune responses, and environmental factors. Little is known about the role of pharmaceutical drugs during C. difficile infection (CDI), but recent studies have demonstrated that nonsteroidal anti-inflammatory drugs (NSAIDs) worsen CDI. The mechanism underlying this phenomenon remains unclear. Here, we show that NSAIDs exacerbate CDI by disrupting colonic epithelial cells (CECs) and sensitizing cells to C. difficile toxin-mediated damage independent of their canonical role of inhibiting cyclooxygenase (COX) enzymes. Notably, we find that NSAIDs and C. difficile toxins target the mitochondria of CECs and enhance C. difficile toxin-mediated damage. Our results demonstrate that NSAIDs exacerbate CDI by synergizing with C. difficile toxins to damage host cell mitochondria. Together, this work highlights a role for NSAIDs in exacerbating microbial infection in the colon.

Concomitant western diet and chronic-binge alcohol dysregulate hepatic metabolism.

BACKGROUND AND AIMS: There is significant overlap between non-alcoholic fatty liver disease (NAFLD) and alcohol-associated liver disease (ALD) with regards to risk factors and disease progression. However, the mechanism by which fatty liver disease arises from concomitant obesity and overconsumption of alcohol (syndrome of metabolic and alcohol-associated fatty liver disease; SMAFLD), is not fully understood. METHODS: Male C57BL6/J mice were fed chow diet (Chow) or high-fructose, high-fat, high-cholesterol diet (FFC) for 4 weeks, then administered either saline or ethanol (EtOH, 5% in drinking water) for another 12 weeks. The EtOH treatment also consisted of a weekly 2.5 g EtOH/kg body weight gavage. Markers for lipid regulation, oxidative stress, inflammation, and fibrosis were measured by RT-qPCR, RNA-seq, Western blot, and metabolomics. RESULTS: Combined FFC-EtOH induced more body weight gain, glucose intolerance, steatosis, and hepatomegaly compared to Chow, EtOH, or FFC. Glucose intolerance by FFC-EtOH was associated with decreased hepatic protein kinase B (AKT) protein expression and increased gluconeogenic gene expression. FFC-EtOH increased hepatic triglyceride and ceramide levels, plasma leptin levels, hepatic Perilipin 2 protein expression, and decreased lipolytic gene expression. FFC and FFC-EtOH also increased AMP-activated protein kinase (AMPK) activation. Finally, FFC-EtOH enriched the hepatic transcriptome for genes involved in immune response and lipid metabolism. CONCLUSIONS: In our model of early SMAFLD, we observed that the combination of an obesogenic diet and alcohol caused more weight gain, promoted glucose intolerance, and contributed to steatosis by dysregulating leptin/AMPK signaling. Our model demonstrates that the combination of an obesogenic diet with a chronic-binge pattern alcohol intake is worse than either insult alone.

The enteric nervous system relays psychological stress to intestinal inflammation.

Mental health profoundly impacts inflammatory responses in the body. This is particularly apparent in inflammatory bowel disease (IBD), in which psychological stress is associated with exacerbated disease flares. Here, we discover a critical role for the enteric nervous system (ENS) in mediating the aggravating effect of chronic stress on intestinal inflammation. We find that chronically elevated levels of glucocorticoids drive the generation of an inflammatory subset of enteric glia that promotes monocyte- and TNF-mediated inflammation via CSF1. Additionally, glucocorticoids cause transcriptional immaturity in enteric neurons, acetylcholine deficiency, and dysmotility via TGF-ß2. We verify the connection between the psychological state, intestinal inflammation, and dysmotility in three cohorts of IBD patients. Together, these findings offer a mechanistic explanation for the impact of the brain on peripheral inflammation, define the ENS as a relay between psychological stress and gut inflammation, and suggest that stress management could serve as a valuable component of IBD care.

Myofibroblastoma in the Liver: A Case Report and Review of Literature.

Myofibroblastoma is a rare benign mesenchymal tumor first described in the breast. It is also known as mammary-type myofibroblastoma outside of the breast, more frequently located along the embryonic milk line. Exceptionally, myofibroblastoma can occur at visceral locations. We present a case of myofibroblastoma detected incidentally in the liver. A well-circumscribed mass, grossly measuring 6.2 cm in the liver parenchyma, was found on imaging studies. Histologically, the lesion is characterized by benign spindle cells in a hyalinized collagenous stroma, with positive staining for SMA and ER, focal positivity for CD34, negative for desmin, and loss of RB1. This rare tumor at such an unusual location makes it diagnostically challenging, especially on core biopsy of the lesion. To our knowledge, this is the second case of myofibroblastoma in the liver reported in the English literature and the first such case with a detailed pathology description.

Stiff matrix induces exosome secretion to promote tumour growth.

Tissue fibrosis and extracellular matrix (ECM) stiffening promote tumour progression. The mechanisms by which ECM regulates its contacting cells have been extensively studied. However, how stiffness influences intercellular communications in the microenvironment for tumour progression remains unknown. Here we report that stiff ECM stimulates the release of exosomes from cancer cells. We delineate a molecular pathway that links stiff ECM to activation of Akt, which in turn promotes GTP loading to Rab8 that drives exosome secretion. We further show that exosomes generated from cells grown on stiff ECM effectively promote tumour growth. Proteomic analysis revealed that the Notch signalling pathway is activated in cells treated with exosomes derived from tumour cells grown on stiff ECM, consistent with our gene expression analysis of liver tissues from patients. Our study reveals a molecular mechanism that regulates exosome secretion and provides insight into how mechanical properties of the ECM control the tumour microenvironment for tumour growth.