Structural Evolution of Gene Promoters Driven by Primate-Specific KRAB Zinc Finger Proteins

Abstract

Krüppel-associated box (KRAB) zinc finger proteins (KZNFs) recognize and repress transposable elements (TEs); TEs are DNA elements that are capable of replicating themselves throughout our genomes with potentially harmful consequences. However, genes from this family of transcription factors have a much wider potential for genomic regulation. KZNFs have become integrated into gene-regulatory networks through the control of TEs that function as enhancers and gene promoters; some KZNFs also bind directly to gene promoters, suggesting an additional, more direct layer of KZNF co-option into gene-regulatory networks. Binding site analysis of ZNF519, ZNF441, and ZNF468 suggests the structural evolution of KZNFs to recognize TEs can result in coincidental binding to gene promoters independent of TE sequences. We show a higher rate of sequence turnover in gene promoter KZNF binding sites than neighboring regions, implying a selective pressure is being applied by the binding of a KZNF. Through CRISPR/Cas9 mediated genetic deletion of ZNF519, ZNF441, and ZNF468, we provide further evidence for genome-wide co-option of the KZNF-mediated gene-regulatory functions; KZNF knockout leads to changes in expression of KZNF-bound genes in neuronal lineages. Finally, we show that the opposite can be established upon KZNF overexpression, further strengthening the support for the role of KZNFs as bona-fide gene regulators. With no eminent role for ZNF519 in controlling its TE target, our study may provide a snapshot into the early stages of the completed co-option of a KZNF, showing the lasting, multilayered impact that retrovirus invasions and host response mechanisms can have upon the evolution of our genomes.

Keywords: Genomics; KRAB zinc finger proteins; gene regulation; primate evolution.

Fig. 1.

Co-evolution of three KZNFs with TEs is paralleled by co-option for direct gene-regulatory properties. A, The top ten promoter binding KZNFs by number of binding sites present in promoter regions (promoter defined as 5,000 bp upstream and 1,000 bp downstream of transcription start site). Adapted from Farmiloe et al. (2020). B, DESeq2 basemean expression values of KZNFs in human and rhesus embryonic stem cells (ESC) and weeks 1–5 wk1-wk5) of cortical organoid development from differential analysis of RNA-seq data. Data points show the average of two replicates for each species and time point. C, ChIP-Seq density plots showing ZNF and KAP1 binding at TEs and gene promoters (ZNF519—MER52; ZNF441—AluY, AluYa5; ZNF468—MER11A). D, The emergence and expansion of the TE classes associated with ZNF441, ZNF468, and ZNF519. Colored sections of the bars show approximately when the TEs emerged and density of color shows peak transposition activity E, evolutionary tree showing approximate time of emergence of TE classes and KZNFs.

Fig. 2.

Transposable element motifs recognized by KZNFs are also seen in TSS binding sites. A, C, E, HOMER de novo motif discovery in DNA sequences ±50 bp from the KZNF summits in TEs and gene promoters. (A) The ZNF519 summits in MER52 elements and promoter regions. (C) The ZNF441 summits in Alu elements and promoters with and without an Alu element. (E) The ZNF468 summits in MER11A elements and promoter regions. B, D, F, KZNF summits ±7 bp lifted over to the UCSC repeat browser, shown at their recognized repeat family's consensus sequence show the presence of the KZNF-bound motif recognized in promoters for (B) ZNF519, (D) ZNF441, and (F) ZNF468.

Fig. 3.

ZNF-bound promoters with hominid-specific insertions or deletions around the summit. Showing ZNF summits and adjacent bases containing indels specific to the hominid line after alignment of the human sequence with that of rhesus, green monkey, and marmoset. Composite values of all promoters are shown in the profile plots below each heatmap (A) ZNF519, (B) ZNF441, (C) ZNF468 (gray = no change, red = insertion in human, blue = deletion in human, black = substitution in human).

Fig. 4.

Collective conservation scores at KZNF summits in bound promoters. (A) Normalized PhyloP conservation scores averaged across all KZNF summits in gene promoter regions ±100 bp and control regions 500 bp downstream of summits. blue = lower, orange = higher normalized PhyloP score than the average across summits for each KZNF. (B–D) Averaged, unnormalized PhyloP scores for ZNF519, ZNF441, and ZNF468 around the ChIP-seq summit and around control regions.

Fig. 5.

Knock out of KZNFs results in changes inbound gene expression. A, C, E, Overview of ZNF519, ZNF441 and ZNF468 loci, gRNAs used for CRISPR-Cas9 KO and RNA of WT and KO hESCs/Day 35 cortical organoids, scaling based on the number of mapped reads. B, D, F, Boxplot showing comparison of log₂ fold change of expressed (baseMean >10), high-confident KZNF-bound genes (gray) compared to unbound genes (white) after (B) ZNF519 KO in 5-week old cortical organoids (bound n = 2,311, unbound n = 10,598), (D) ZNF441 KO in 5-week old cortical organoids (bound n = 850, unbound n = 11,170) (F) ZNF468 KO in 5-day old cortical organoids (bound n = 397, unbound n = 11,345). **** = P < 0.0001, ** = P < 0.01, * = P < 0.05, Wilcoxon rank sum test with continuity correction. Red dashed lines were calculated independently and show the 95% CI of a 10,000 times bootstrapped median of a set of unbound genes with the same sample size as the target genes. Individual data points are not shown.

Fig. 6.

In-depth analysis of ZNF519 KO and overexpression experiments shows reciprocal changes in the regulation of bound genes. (A) Boxplot showing comparison of log₂ fold change of expressed (baseMean >10) high-confident KZNF-bound genes (gray) compared to unbound genes (white) in ZNF519 KO cortical organoids of 2 weeks old (day 14, bound n = 2,293, unbound n = 10,926), (B) ZNF519 KO hESCs (bound n = 2,282, unbound n = 9,937). **** = P < 0.0001, Wilcoxon rank sum test with continuity correction. Red dashed lines were calculated independently and show the 95% CI of 10,000 times bootstrapped median of a set of unbound genes with the same sample size as the target genes. Individual data points are not shown. Mind difference in y-axis. (C) Schematic showing ZNF519 overexpression experiment set up in HEK293 cells. (D) RNA-seq at ZNF519 locus confirms overexpression in HEK293 cells. Mean of three replicates shown, scaled on number of mapped reads (excluding ZNF519 locus). (E) Boxplot showing a comparison of log₂ fold change of expressed (baseMean >10) high-confident ZNF519-bound genes (gray, n = 2,268) compared to unbound genes after overexpression of ZNF519 (white, n = 8,222) after ZNF519 overexpression, **** = P < 0.0001.

Fig. 7.

Promoter escape model of evolution at promoter KZNF binding sites. Promoter escape model of evolution at promoter KZNF binding sites: 1) Binding motif present in TE and recognized by KZNF is also present in the gene promoter region. 2) KZNF also recognizes promoter motifs and binds there, affecting gene expression. 3) The effect of KZNF binding on gene expression exerts a selective pressure at the locus, either leading to a loss of binding site or strengthened binding by the KZNF.