Interrogation of cancer gene dependencies reveals paralog interactions of autosome and sex chromosome-encoded genes

Genetic networks are characterized by extensive buffering. During tumor evolution, disruption of functional redundancies can create de novo vulnerabilities that are specific to cancer cells. Here, we systematically search for cancer-relevant paralog interactions using CRISPR screens and publicly available loss-of-function datasets. Our analysis reveals >2,000 candidate dependencies, several of which we validate experimentally, including CSTF2-CSTF2T, DNAJC15-DNAJC19, FAM50A-FAM50B, and RPP25-RPP25L. We provide evidence that RPP25L can physically and functionally compensate for the absence of RPP25 as a member of the RNase P/MRP complexes in tRNA processing. Our analysis also reveals unexpected redundancies between sex chromosome genes. We show that chrX- and chrY-encoded paralogs, such as ZFX-ZFY, DDX3X-DDX3Y, and EIF1AX-EIF1AY, are functionally linked. Tumor cell lines from male patients with loss of chromosome Y become dependent on the chrX-encoded gene. We propose targeting of chrX-encoded paralogs as a general therapeutic strategy for human tumors that have lost the Y chromosome.