TSPAN6 is a suppressor of Ras-driven cancer

Oncogenic mutations in the small GTPase RAS contribute to ~30% of human cancers. In a Drosophila genetic screen, we identified novel and evolutionary conserved cancer genes that affect Ras-driven tumorigenesis and metastasis in Drosophila including confirmation of the tetraspanin Tsp29Fb. However, it was not known whether the mammalian Tsp29Fb orthologue, TSPAN6, has any role in RAS-driven human epithelial tumors. Here we show that TSPAN6 suppressed tumor growth and metastatic dissemination of human RAS activating mutant pancreatic cancer xenografts. Whole-body knockout as well as tumor cell autonomous inactivation using floxed alleles of Tspan6 in mice enhanced KrasG12D-driven lung tumor initiation and malignant progression. Mechanistically, TSPAN6 binds to the EGFR and blocks EGFR-induced RAS activation. Moreover, we show that inactivation of TSPAN6 induces an epithelial-to-mesenchymal transition and inhibits cell migration in vitro and in vivo. Finally, low TSPAN6 expression correlates with poor prognosis of patients with lung and pancreatic cancers with mesenchymal morphology. Our results uncover TSPAN6 as a novel tumor suppressor receptor that controls epithelial cell identify and restrains RAS-driven epithelial cancer.

replicated, with biological and technical replicates, as detailed in the figure legend, with similar results.
After splitting, cells were counted and 50,000 cells/ well were transferred into a 96-well plate.
After adhering, 0.1ml of 0.1 mCi/mL 3 H-thymidine was added into each well. Cells were incubated with 3 H-thymidine for 8 hours and uptake determined using a scintillation counter. reactions were performed with 20 pmol of gene-specific primers and iQ SYBR Green Supermix (Bio-Rad) using PCR conditions and primers as described previously 9 . Threshold cycle numbers (Ct) were determined with C1000 Thermal Cycler (Bio-Rad) and the relative quantities of mRNA per sample were calculated using the ΔΔCt method with GAPDH as the calibrator gene. The relative levels of mRNA were determined by setting the mRNA expression level of the first sample as previously described 9 . These experiments were independently replicated, with biological and technical replicates, as detailed in the figure legend, with similar results.
Generation of TSPAN6 over-expressing human pancreatic cancer cells.

Co-immunoprecipitation Analysis
For co-immunoprecipitation with the anti-EGFR antibody, PANC1 and MIA-PACA2 cell lines were cultured in tissue culture flasks until they reach 80% confluency. Subsequently, each 1 x These experiments were conducted once.

TSPAN6 correlation to EMT signature in NSCLC.
Scoring of patient tumor samples based on an EMT and epithelial signature was performed based on NSCLC patient cohort analysis performed by Tan and colleagues 10 . TSPAN6 expression between two patient groups of mesenchymal-like and epithelial-like samples was analysed. EMT scores closer to +1 are considered mesenchymal-like tumors, EMT scores closer to -1 are considered epithelial tumors. In the GSE10245 cohort 5 , 35 NSCLC samples were scored as epithelial and 23 as mesenchymal on the hgu133plus2 platform using two probe-sets (209108-at and 209109-s-at).

TSPAN6 correlation to EMT signature in pancreatic cancer.
The Collison dataset was processed from normalized gene expression data of the PDA UCSF tumors defined in GSE17891 with clinical survival and censoring information for 27 patients 2 .
Similar analysis was conducted on the GSE11838 cohort 1 . Patients with status "alive" were assigned as censored with days to last follow up. We performed a log-rank test using the survival R package 11 . The minimal p-value was defined based on a log-rank test to split the patient cohorts into a TSPAN6 low and TSPAN6 high expression group.

Statistics.
Statistical analysis using gene expression data was performed with R version 2. 5