UC San Diego

Metastases arise from a small subset of cells, less than 1%, in the primary tumor. Metastasis is responsible for most cancer-related deaths and plays a significant role in therapy resistance, yet this hallmark of cancer is poorly understood. Metastasis is not determined by a genetic coding pattern among metastatic tumors pan-cancer. Identifying roots of metastatic potential is crucial for long-term research for the field; to this end, we must identify metastatic phenotypes for targeted therapy. Instead of the therapies we have now, metastatic cells can still avoid. First, we employ a generative neural network in combination with supervised machine learning to classify patient-derived melanoma xenografts as “efficient” or “inefficient” metastatic, validate predictions regarding melanoma cell lines with unknown metastatic efficiency in mouse xenografts, and use the network to generate in silico cell images that amplify the critical predictive cell properties. These exaggerated images unveiled metastatic phenotypes of pseudopodial extensions and increased light scattering as hallmark properties of metastatic cells. We found from our mouse experiments that the metastatic melanoma samples with TPMs had increased metastasis in the mice and were predicted to have increased metastatic potential from our deep learned model. Recently, sequencing the non-coding regions of patient metastatic tumors reveals two monoallelic TPMs, C228T and C250T. TPMs are found in the primary tumor but become highly enriched in metastatic tumor sites suggesting a functional role of the TPMs in metastatic potential. To determine the functional role of TPMs in metastasis, we have taken a phenotypic-driven interrogation. We have shown through bioinformatic analysis that C250T TPM has increased metastatic potential and penetrance in mice studies. We observed that TPMs increase the collective cell migration rate and distance with significant spatiotemporal features. We highlighted the increased gene expression heterogeneity due to the TPMs. We have shown differences between the TPMs and WT through machine learning using imaging cytometry focusing on morphological and light scattering features. This shows the trend that C250T has an advantage over C228T and WT, while C228T has an advantage over WT Telomerase in metastatic tumors.