Search

Simple Summary Different types of cells with tumor-initiating cell (TIC) activity contribute to colorectal cancer (CRC) progression and resistance to anti-cancer treatment. In this study, we aimed to understand whether different cell types exist within a patient-derived tumor culture, distinguishable by different patterns of their gene expression. By mRNA sequencing of patient-derived CRC cultures at the single-cell level, we defined expression programs that closely resemble differentiated cell populations of the normal intestine. Here, cell type-associated subpopulations showed differences in functional properties such as cell growth and energy metabolism. Subsequent functional analyses in vitro and in vivo demonstrated that metabolic states are linked to TIC activity in primary CRC cultures. We also show that TIC activity is dependent on oxidative phosphorylation, which may therefore represent a target for novel therapies. Abstract Intra-tumor heterogeneity of tumor-initiating cell (TIC) activity drives colorectal cancer (CRC) progression and therapy resistance. Here, we used single-cell RNA-sequencing of patient-derived CRC models to decipher distinct cell subpopulations based on their transcriptional profiles. Cell type-specific expression modules of stem-like, transit amplifying-like, and differentiated CRC cells resemble differentiation states of normal intestinal epithelial cells. Strikingly, identified subpopulations differ in proliferative activity and metabolic state. In summary, we here show at single-cell resolution that transcriptional heterogeneity identifies functional states during TIC differentiation. Furthermore, identified expression signatures are linked to patient prognosis. Targeting transcriptional states associated to cancer cell differentiation might unravel novel vulnerabilities in human CRC.

Summary Novel treatment options for metastatic colorectal cancer (CRC) are urgently needed to improve patient outcome. Here, we screen a library of non-characterized small molecules against a heterogeneous collection of patient-derived CRC spheroids. By prioritizing compounds with inhibitory activity in a subset of—but not all—spheroid cultures, NCT02 is identified as a candidate with minimal risk of non-specific toxicity. Mechanistically, we show that NCT02 acts as molecular glue that induces ubiquitination of cyclin K (CCNK) and proteasomal degradation of CCNK and its complex partner CDK12. Knockout of CCNK or CDK12 decreases proliferation of CRC cells in vitro and tumor growth in vivo. Interestingly, sensitivity to pharmacological CCNK/CDK12 degradation is associated with TP53 deficiency and consensus molecular subtype 4 in vitro and in patient-derived xenografts. We thus demonstrate the efficacy of targeted CCNK/CDK12 degradation for a CRC subset, highlighting the potential of drug-induced proteolysis for difficult-to-treat types of cancer.

Tumor progression in colorectal cancer (CRC) is driven by a subpopulation of cells with tumor-initiating cell (TIC) activity. In the past, bulk experiments provided strong evidence for functional heterogeneity of the CRC TIC compartment with distinct types of TICs driving CRC progression. However, in these retrospective experiments direct assignment of functional states to individual functionally relevant cell types in CRC has not been possible. We here asked whether functional states of individual cells can be assigned to specific CRC cell subpopulations. Therefore, 4,663 single-cell profiles were generated by single-cell RNA-sequencing (scRNA-seq) of patient-derived CRC spheroid cultures (n=12). Non-negative matrix factorization identified 13 gene expression modules linked to cell states or cell types that resemble differentiation states of normal intestinal epithelial cells. To evaluate their presence in CRC patient tumors, the identified transcriptional programs were applied to publicly available expression data (TCGA cohort; n=459). Clustering of samples based on the contribution of individual expression signatures revealed six patient clusters with significantly different progression-free survival (p=0.043) and a trend towards different overall survival (p=0.12), indicating a relevance for patient outcome. Comparing cell state meta-signature expression in cell type-associated subpopulations revealed differences in functional states, i.e. proliferative and metabolic activity. Interestingly, stem-like cells showed significantly increased oxidative phosphorylation (OXPHOS) and decreased glycolysis scores, suggesting a link between metabolic states and cell differentiation. scRNA-seq of patient-derived organoid and xenograft models as well as primary tumors revealed similar trends. To assess whether metabolic states are linked to TIC activity, CRC spheroid cells were sorted according to mitochondrial membrane potential (MMP), and sphere forming cell (SFC) as well as TIC frequency were assessed by limiting dilutions. Strikingly, MMPhigh cells exhibited a pronounced increase of SFC frequencies in vitro compared to MMPlow cells in 80% of patients (n=5). In addition, TIC frequencies were strongly increased in the MMPhigh compared to the MMPlow subpopulation in vivo (n=2; 1/46,535 vs. 1/211,305 and 1/249 vs. 1/2,089), demonstrating that TIC activity relies on OXPHOS. In line with this, pretreatment with the OXPHOS inhibitor CCCP decreased SFC frequencies (n=3), indicating targetable metabolic vulnerability of CRC spheroid cells. In summary, we here show that transcriptional heterogeneity identifies functional states during CRC TIC differentiation. Targeting context-dependent regulation of tumor cell differentiation might unravel novel vulnerabilities for therapeutic intervention in human CRC. Citation Format: Martina K. Zowada, Stephan M. Tirier, Sebastian M. Dieter, Teresa Krieger, Ava Oberlack, Robert L. Chua, Mario Huerta, Foo Wei Ten, Karin Laaber, Jeongbin Park, Katharina Jechow, Torsten Müller, Mathias Kalxdorf, Mark Kriegsmann, Katharina Kriegsmann, Friederike Herbst, Jeroen Krijgsveld, Martin Schneider, Roland Eils, Hanno Glimm, Christian Conrad, Claudia R. Ball. Transcriptional heterogeneity identifies functional states of tumor-initiating cell differentiation in human colorectal cancer [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 1491.

Background Glucocorticoids are commonly used as adjuvant treatment for side-effects and have anti-proliferative activity in several tumors but, on the other hand, their proliferative effect has been reported in several studies, some of them involving the spread of cancer. We shall attempt to reconcile these incongruities from the genomic and tissue-physiology perspectives with our findings. Methods An accurate phenotype analysis of microarray data can help to solve multiple paradoxes derived from tumor-progression models. We have developed a new strategy to facilitate the study of interdependences among the phenotypes defined by the sample clusters obtained by common clustering methods (HC, SOTA, SOM, PAM). These interdependences are obtained by the detection of non-linear expression-relationships where each fluctuation in the relationship implies a phenotype change and each relationship typology implies a specific phenotype interdependence. As a result, multiple phenotypic changes are identified together with the genes involved in the phenotype transitions. In this way, we study the phenotypic changes from microarray data that describe common phenotypes in cancer from different tissues, and we cross our results with biomedical databases to relate the glucocorticoid activity to the phenotypic changes. Results 11,244 significant non-linear expression relationships, classified into 11 different typologies, have been detected from the data matrix analyzed. From them, 415 non-linear expression relationships were related to glucocorticoid activity. Studying them, we have found the possible reason for opposite effects of some stressor agents like dexamethasone on tumor progression and it has been confirmed by literature. This hidden reason has resulted in being linked with the type of tumor progression of the tissues. In the first type of tumor progression found, new cells can be stressed during proliferation and stressor agents increase tumor proliferation. In the second type, cell stress and tumor proliferation are antagonists so, therefore, stressor agents stop tumor proliferation in order to stress the cells. The non-linear expression relationships among DUSP6, FERMT2, FKBP5, EGFR, NEDD4L and CITED2 genes are used to synthesize these findings.