Abstract
Double strand-breaks (DSBs) of genomic DNA caused by ionizing radiation or mutagenic chemicals are a common source of mutation, recombination, chromosomal aberration, and cell death. Linker histones are DNA packaging proteins with established roles in chromatin compaction, gene transcription, and in homologous recombination (HR)-mediated DNA repair. Using a machine-learning model for functional prioritization of eukaryotic post-translational modifications (PTMs) in combination with genetic and biochemical experiments with the yeast linker histone, Hho1, we discovered that site-specific phosphorylation sites regulate HR and HR-mediated DSB repair. Five total sites were investigated (T10, S65, S141, S173, and S174), ranging from high to low function potential as determined by the model. Of these, we confirmed S173/174 are phosphorylated in yeast by mass spectrometry and found no evidence of phosphorylation at the other sites. Phospho-nullifying mutations at these two sites results in a significant decrease in HR-mediated DSB repair templated either with oligonucleotides or a homologous chromosome, while phospho-mimicing mutations have no effect. S65, corresponding to a mammalian phosphosite that is conserved in yeast, exhibited similar effects. None of the mutations affected base- or nucleotide-excision repair, nor did they disrupt non-homologous end joining or RNA-mediated repair of DSBs when sequence heterology between the break and repair template strands was low. More extensive analysis of the S174 phospho-null mutant revealed that its repression of HR and DSB repair is proportional to the degree of sequence heterology between DSB ends and the HR repair template. Taken together, these data demonstrate the utility of machine learning for the discovery of functional PTM hotspots, reveal linker histone phosphorylation sites necessary for HR and HR-mediated DSB repair, and provide insight into the context-dependent control of DNA integrity by the yeast linker histone Hho1.