Abstract
Peritoneal metastasis (PM) is a debilitating consequence of multiple cancers. As cancer cells lose tonic signaling related to attachment dependence, critical morphologic shifts result in alteration of the transcriptome. Identifying key genes associated with this transformation may lead to targeted therapies for this devastating complication. TC71, CHLA9, PANC1, YOU and HEYA8 cell lines were grown as tumor spheroids in polyHEMA coated plates. Temporal profiling of transcriptomic alterations over 72 hrs was used to develop a comprehensive PM model. We identified transcriptomic outliers using Gaussian mixtures model clustering to identify drivers of spheroid formation. Outliers were validated in The Cancer Genome Atlas (TCGA) and an ovarian tissue microarray (TMA) and by modulation in ovarian cancer models in vitro and in peritoneal xenograft models. Outlier analysis of PM genes identified the gene TXNIP and the TORC signaling as central to PM. Ovarian cancer spheroids isolated from patient ascites had significantly higher TXNIP than their attached counterparts (p = 0.047). TXNIP levels predicted progression-free (log-rank p = 0.026) survival in stage 1/2 ovarian cancer and overall survival (log rank p = 0.047) in stage 3/4 ovarian cancer. In vitro, TXNIP silencing was associated with increased mTOR signaling and enhanced spheroid development which could be overcome by TAK228, a TORC1/2 inhibitor. Similarly, in vivo peritoneal xenograft models of carcinomatosis were prevented by TAK228. PM is driven by TXNIP-associated TORC1/2 signaling. This work provides the first evidence that TORC1/2 inhibition may prevent PM.