Machine Learning Identifies Stemness Features Associated with Oncogenic Dedifferentiation

Cancer progression involves the gradual loss of a differentiated phenotype and acquisition of progenitor and stem-cell-like features. Here, we provide novel stemness indices for assessing the degree of oncogenic dedifferentiation. We used an innovative one-class logistic regression (OCLR) machine-learning algorithm to extract transcriptomic and epigenetic feature sets derived from non-transformed pluripotent stem cells and their differentiated progeny. Using OCLR, we were able to identify previously undiscovered biological mechanisms associated with the dedifferentiated oncogenic state. Analyses of the tumor microenvironment revealed unanticipated correlation of cancer stemness with immune checkpoint expression and infiltrating immune cells. We found that the dedifferentiated oncogenic phenotype was generally most prominent in metastatic tumors. Application of our stemness indices to single-cell data revealed patterns of intra-tumor molecular heterogeneity. Finally, the indices allowed for the identification of novel targets and possible targeted therapies aimed at tumor differentiation.

Intestinal stem cells differentiation and its application in cancer therapy

Maxim Norkin

Colorectal cancer (CRC) is one of the deadliest cancers worldwide. Despite extensive research in the field, new therapies targeting CRC are in urgent need. CRC is a highly heterogeneous disease with high levels of intra- and inter-patient heterogeneity. Cancer stem cells are the driving force of tumor development and metastasis. One of the approaches to treat CRC is differentiation therapy thereby induction of cancer stem cell differentiation may lead to tumor regression. Intestinal normal and cancer organoids represent an ideal model for studying the differentiation process ex vivo. Gene expression profiling of drug-treated organoids can be used for quantification of the differentiation process with high precision. RNA-seq based drug-screening platforms in organoids are not yet established. Here we developed a high-throughput high-content targeted RNA-seq-based platform (TORNADO-seq) for monitoring gene expression of large gene signatures in intestinal organoids which allowed us to quantify and evaluate complex cellular phenotypes and cellular alterations upon drug treatment. We applied TORNADO-seq in drug screening in wt and cancer intestinal organoids. We found many differentiation-inducing drugs in wt organoids, including some drugs specifically enriching for certain cell phenotypes, e.g. enteroendocrine cells. Further, we applied identified differentiation-inducing drugs to cancer organoids (APClof;KRASG12D;TP53lof) and discovered that the majority of those differentiation-inducing drugs also targeted cancer organoids. Differentiation induction seems to be a common mode of action of cancer-targeting drugs. Based on gene expression profiles of treated cancer organoids, we were able to propose and confirm the mechanism of action of several drugs indicating that cancer cells may to be sensitive to alterations in the cholesterol biosynthesis pathway. In general, utilization of TORNADO-seq allows predicting connections between cellular phenotypes and signaling pathways in organoids which can be used for hypothesis generation in a variety of biological systems. The second part of the thesis is devoted to the investigation of mechanisms underlying the accumulation of genetic alterations in CRC tumors. It is not known whether cancer driver genes affect the rate of mutation acquisition. Here we performed exome sequencing of single cell-derived organoids obtained from tumors with either APClof, APClof;KRASG12D or APClof;KRASG12D;TP53lof mutation phenotypes that reflect early stages of CRC development. We discovered that presence of KRAS and P53 mutations didn't affect the number of single nucleotide substitutions (SNSs) in tumors, while the latter was increased in APC mutated tumors compared to wt cells and was highly correlated with the tumor age. P53 inactivation was highly correlated with the presence of large-scale (>10 Mb) copy number alterations (CNAs).

EPFL2021

The niche under siege: novel targets for metastasis therapy

Albert Santamaria Martinez

Metastasis is an inefficient process and most cancer cells fail to colonize secondary sites. There are several possible reasons for this. First, the nature of the infiltrating cells is important as a small population of cancer stem cells has been shown to have exclusive metastasis-initiating potential. Secondly, supportive niches are required to promote the outgrowth of disseminated tumour cells. S uch niches are either produced prior to the arrival of cancer cells in the target organ or are induced ad hoc upon cell infiltration. Components of the extracellular matrix (ECM) have been found to play a role in establishing these niches. This has highlighted the importance of the ECM for metastatic progression, and suggests that such components may provide alternative targets for treatment of metastastic disease. © 2012 The Association for the Publication of the Journal of Internal Medicine.

Wiley-Blackwell2013

Intestinal stem cells differentiation and its application in cancer therapy

Maxim Norkin

EPFL2021