ChIP-exo and ChIP-seq of human KRAB-ZNFs transduced in HEK 293T cells

Kruppel-associated box (KRAB)-containing zinc-finger proteins (KZFPs) represent the largest family of transcription factors encoded by higher vertebrates. Together with their heterochromatin-inducing cofactor KAP1, many act as repressors of endogenous retroelements (EREs) during early embryonic genome reprogramming, yet their widespread expression in adult tissues and enrichment at other genetic loci indicate additional roles. By characterizing the protein interactome of 101 of the ~350 human KZFPs, we uncovered that it is highly diversified for a subset of those. Most of placental mammal- or primate-restricted KZFPs, which primarily bind EREs and constitute 90% of human KZFPs, essentially recruit KAP1 and effectors. However, certain KZFPs conserved in placental mammals and more distant clades associated with factors related to functions such as genome architecture or RNA processing. Additionally, members of the most conserved human KZFPs did not appear to interact with KAP1. Nevertheless, KZFPs from coelacanth, our most distant KZFP-encoding relative, bind KAP1 from this species. We thus propose that KZFPs first emerged as ERE-controlling repressors, had their pool continuously renewed by turn-over of their hosts’ ERE loads, thereby yielded largely species-specific sets of transcriptional controllers, yet occasionally produced derivatives that escaped this evolutionary flushing by development of novel functions and exaptation. We transduced HEK 293T cells with doxycycline-inducible plasmids encoding for human KRAB-ZFPs fused with 3 HA tags at the C-terminus. We next performed ChIP-exo and ChIP-seq and prepared a barcoded Illumina library for each sample. For the ChIP-exo, the libraries were pooled and sequenced 12 per lane on an Illumina HiSeq 2500 to a minimum depth of around 15 million 100 bp single-end reads. For the ChIP-seqs, all samples were pooled and sequenced on the same lane on an NextSeq 500 to a minimum depth of around 27 million 85 bp single-end reads. The reads were mapped to the hg19 assembly of the human genome and peaks were identified using MACS. For the ChIP-exos, we used a total input derived from random sampling from all samples in the study, as it was found to perform better than either the standard total input or the exonuclease treated total input. For the ChIP-seqs, we used a total input sample that had a matching sonication profile to the IP sample.