Cutting The Trunk Is Not Enough!

If cancer is the tree trunk, then metastasis was postulated to be its branches, thereby the idea of treating metastasis by cutting the trunk of the tree, i.e. successful surgery, was born (1). However, even after that, the tumors would return in different places, sometimes in months and sometimes in years. Medical oncologists had no clue as to the why and when.





Size Does Not Matter !

Also, it was thought that only when a tumor was significantly large did it become metastatic, which was due to the overwhelming size and additional genetic changes. However, recent data have shown that even when the tumor is very small in size (< 103 cells), some cells would disseminate into the blood and there were not much genetic differences between these primary and secondary tumor cells (2,3). Given that PET scan could detect only around 109 cells and liquid biopsy sensitivity was no less than 105 cells, for most patients, the inevitable have had happened even before the tumor was diagnosed. This early dissemination also explains why all targeted therapeutics around metastasis, starting from matrix metallo-proteases to integrin inhibitors had failed in the clinic. It was too late!!

“To-EMT or Not To EMT”

The biggest challenge, or controversy, that have marred the metastasis field was “to EMT or not to EMT” (4). Many scientists believed that epithelial to mesenchymal transition (EMT) was key to metastasis, whereas others pointed out that mesenchymal cells were not enough to induce metastasis. Recently, some pioneering work has suggested that EM plasticity is the key and not only the ability to change to mesenchymal cells was important, but the ability to change back to epithelial form was equally important (5).

Key unanswered questions

The question is how much of epithelial to mesenchymal transition is required? How much of the reverse mesenchymal to epithelial form is necessary and sufficient? What properties of metastasis are complimented by the epithelial forms and what would require the mesenchymal forms? How much of the complex metastasis biology is cancer agnostic? Is there any pattern across different carcinomas? Can we help the medical oncologist in understanding which of the patients have a higher probability of metastasis in the future, based on the functional properties of their primary tumors? Can we exploit the unique properties of mesenchymal cells for robust anti-metastasis drug discovery?

Our Approach

Mestastop has created three unique and novel platform that defines the complex biology of cancer metastasis.

The first platform, METAssayTM  completely dissects metastasis biology into multiple in vitro phenotypic assays, each mimicking a particular facet of metastasis, capable of differentiating between a growing and a moving tumor cells belonging to the same tumor. The platform is available for multiple solid tumors.

METVivoTM a highly efficient metastasis animal model platform that speeds up the generation of an animal model for in-vivo testing. Mestastop has engineered cancer cells to make them more metastatic, and as a result, liver metastasis for colorectal cancer orthotopic models takes just six weeks. The take rate for mesenteric lymph node metastasis is 100%, and that for liver metastasis is 90%, with rare metastasis events like abdominal wall and bone also observed.

METSCANTM is a unique platform that integrates all the data generated from the METAssayTM platform and normalizes it with actual patient sample data, thereby distributing weightages to the steps in the metastasis cascade that are more relevant for metastasis success vis- a-vis patient perspective. All experimental data is input into METSCANTM, after which machine learning algorithms are applied to identify the most weighted steps. Once these weightages are in place, patient samples are analyzed to understand their respective position against the baseline of the moving and growing cell phenotypes, thereby attributing a summative score for each patient and determining their probability of metastasis. Currently, supervised learning using support vector machine algorithms has given an accuracy of 92.3% in the prediction. The first trial with a small cohort of patients is currently ongoing, which did not show any false negatives (as per their node status), and three predictions have already matched. There are a few false positives, which can actually be due to the higher sensitivity of the platform than the traditional node status guidelines.

Publications (Click the links for the abstract and the poster)