SUMMARY Cyclin D-dependent kinases (CDK4 and CDK6) are positive regulators of cell cycle entry, and they are overactive in the majority of human cancers. However, it is currently not completely understood by which cellular mechanisms CDK4/6 promote tumorigenesis, largely due to the limited number of identified substrates. Here we performed a systematic screen for substrates of cyclin D1-CDK4 and cyclin D3-CDK6. We identified the Forkhead Box M1 (FOXM1) transcription factor as a common critical phosphorylation target. CDK4/6 stabilize and activate FOXM1, thereby maintain expression of G1/S phase genes, suppress the levels of reactive oxygen species (ROS), and protect cancer cells from senescence. Melanoma cells, unlike melanocytes, are highly reliant on CDK4/6-mediated senescence suppression, which makes them particularly susceptible to CDK4/6 inhibition.

The two fields of machine learning and graphical causality arose and are developed separately. However, there is, now, cross-pollination and increasing interest in both fields to benefit from the advances of the other. In this article, we review fundamental concepts of causal inference and relate them to crucial open problems of machine learning, including transfer and generalization, thereby assaying how causality can contribute to modern machine learning research. This also applies in the opposite direction: we note that most work in causality starts from the premise that the causal variables are given. A central problem for AI and causality is, thus, causal representation learning, that is, the discovery of highlevel causal variables from low-level observations. Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities.

Cyclin C was cloned as a growth-promoting G1 cyclin, and was also shown to regulate gene transcription. Here we report that in vivo cyclin C acts as a haploinsufficient tumor suppressor, by controlling Notch1 oncogene levels. Cyclin C activates an “orphan” CDK19 kinase, as well as CDK8 and CDK3. These cyclin C-CDK complexes phosphorylate Notch1 intracellular domain (ICN1) and promote ICN1 degradation. Genetic ablation of cyclin C blocks ICN1 phosphorylation in vivo, thereby elevating ICN1 levels in cyclin C-knockout mice. Cyclin C ablation or heterozygosity collaborate with other oncogenic lesions and accelerate development of T-cell-acute lymphoblastic leukemia (T-ALL). Furthermore, the cyclin C gene is heterozygously deleted in a significant fraction of human T-ALL, and these tumors express reduced cyclin C levels. We also describe point mutations in human T-ALL that render cyclin C-CDK unable to phosphorylate ICN1. Hence, tumor cells may develop different strategies to evade cyclin C inhibitory function.