ABSTRACT: Mammalian genomes are replicated in a precise, cell-type-specific order during the S phase, a process that correlates with local transcriptional activity, chromatin modifications, and chromatin architecture. However, the causal relationships between these features and DNA replication timing (RT), particularly during cell fate transitions, remain largely unknown. Here, we employed machine learning to quantify chromatin features — including epigenetic marks, histone variants, and chromatin architectural factors — that best predict local RT under steady-state conditions and RT changes during early embryonic stem (ES) cell differentiation. We found that approximately one-third of the genome exhibits RT changes during differentiation. Chromatin features collectively predicted steady-state RT and RT changes with high accuracy. Notably, histone H3 lysine 4 monomethylation (H3K4me1), catalyzed by KMT2C/D, emerged as a top predictor. Genetic deletion of Kmt2c/d (but not Kmt2c alone) or their enzymatic activities erased genome-wide RT dynamics during cell differentiation. Sites that typically gain H3K4me1 in a KMT2C/D-dependent manner during differentiation failed to transition towards earlier RT, often without affecting transcriptional activation. Further analysis at KMT2C/D binding sites revealed a local requirement for KMT2C/D in promoting DNA replication origin firing. Our findings identify KMT2C/D-dependent H3K4me1 as a functional regulator of RT and origin firing, highlighting a causal relationship between the epigenome and DNA replication that is largely independent of transcription. These insights should be relevant to various diseases associated with KMT2C/D mutations. Mammalian genomes are replicated in a precise, cell-type-specific order during the S phase, a process that correlates with local transcriptional activity, chromatin modifications, and chromatin architecture. However, the causal relationships between these features and DNA replication timing (RT), particularly during cell fate transitions, remain largely unknown. Here, we employed machine learning to quantify chromatin features — including epigenetic marks, histone variants, and chromatin architectural factors — that best predict local RT under steady-state conditions and RT changes during early embryonic stem (ES) cell differentiation. We found that approximately one-third of the genome exhibits RT changes during differentiation. Chromatin features collectively predicted steady-state RT and RT changes with high accuracy. Notably, histone H3 lysine 4 monomethylation (H3K4me1), catalyzed by KMT2C/D, emerged as a top predictor. Genetic deletion of Kmt2c/d (but not Kmt2c alone) or their enzymatic activities erased genome-wide RT dynamics during cell differentiation. Sites that typically gain H3K4me1 in a KMT2C/D-dependent manner during differentiation failed to transition towards earlier RT, often without affecting transcriptional activation. Further analysis at KMT2C/D binding sites revealed a local requirement for KMT2C/D in promoting DNA replication origin firing. Our findings identify KMT2C/D-dependent H3K4me1 as a functional regulator of RT and origin firing, highlighting a causal relationship between the epigenome and DNA replication that is largely independent of transcription. These insights should be relevant to various diseases associated with KMT2C/D mutations.
