ZNF146/OZF and ZNF507 target LINE-1 sequences

Abstract

Repetitive sequences including transposable elements and transposon-derived fragments account for nearly half of the human genome. While transposition-competent transposable elements must be repressed to maintain genomic stability, mutated and fragmented transposable elements comprising the bulk of repetitive sequences can also contribute to regulation of host gene expression and broader genome organization. Here, we analyzed published ChIP-seq data sets to identify proteins broadly enriched on transposable elements in the human genome. We show 2 of the proteins identified, C2H2 zinc finger-containing proteins ZNF146 (also known as OZF) and ZNF507, are targeted to distinct sites within LINE-1 ORF2 at thousands of locations in the genome. ZNF146 binding sites are found at old and young LINE-1 elements. In contrast, ZNF507 preferentially binds at young LINE-1 sequences correlated to sequence changes in LINE-1 elements at ZNF507's binding site. To gain further insight into ZNF146 and ZNF507 function, we disrupt their expression in HEK293 cells using CRISPR/Cas9 and perform RNA sequencing, finding modest gene expression changes in cells where ZNF507 has been disrupted. We further identify a physical interaction between ZNF507 and PRMT5, suggesting ZNF507 may target arginine methylation activity to LINE-1 sequences.

Keywords: L1; LINE-1; OZF; PRMT5; ZNF146; ZNF507; transposable element.

Fig. 1.

Identification of TFs that bind transposable elements. a) Schematic of ENCODE TF ChIP-seq peak intersection with Repeatmasker annotated TEs. b) Heatmap of TF ChIP-seq enrichment on abundant TE classes calculated as the log₂ ratio of observed (obs) peak/TE intersections relative to simulated shuffled peaks (expected, exp). c) Density of enrichment across all ChIP-seq experiment peak and TE class comparisons. d) Bar chart of the number of TF ChIP-seq experiments demonstrating 2-fold enrichment binned by TE class. e) Bar chart of peak enrichment on TE classes for GFP-ZNF146 ChIP-seq in HEK293 cells and GFP-ZNF507 in K562 cells. f) Schematic representation of ZNF146 and ZNF507 proteins and the distribution of their C2H2 zinc finger domains.

Fig. 2.

ZNF146 and ZNF507 target LINE-1 sequences. a) Pie chart representing the percentage of peaks in the indicated ChIP-seq experiments overlapping annotated LINE-1 elements. b) Venn diagram of peak interval intersections between the indicated ChIP-seq experiments. c) Meta-analysis of ChIP-seq enrichment at full-length disrupted (n = 13,092) or intact (ORF2 or ORF1 and ORF2) LINE-1 elements (n = 245) normalized to 5 kb. Below is the 100mer mappability score for the same regions (lower scores indicate lower mappability) d, e) Coverage of L1-mapping reads directly mapped to LINE-1 (human specific, PA1) consensus sequence. Below is a schematic of full-length LINE-1. f, g) Enriched motifs (see Materials and Methods) and corresponding LINE-1 (human specific, PA1) consensus sequence determined from ZNF146 and ZNF507 ChIP-seq experiments, respectively. h, i) Heatmap of ChIP-seq enrichment centered around occurrences of identified motifs (FIMO, P < 5e-6) found in annotated LINE-1 elements in their genomic context. On the right is the theoretical mappability of 100 bp reads in the same regions.

Fig. 3.

ZNF507 does not bind at old L1PA elements. a) Heatmap of ChIP-seq enrichment centered around occurrences of identified motifs (FIMO, P < 5e-6) ± 1 Kb after intersection and binning by LINE-1 subfamily. Below the ChIP-seq heatmaps (blue) is the theoretical mappability of 100 bp reads for the same regions (grayscale). b) Heatmaps as above but for LINE-1 subfamilies binned by their ORF2-derived classification. Relationship between the ORF2 and 3’UTR classifications is given in parentheses. c) Multiple sequence alignment of L1 subfamily consensus sequences. Positions are derived from the L1PA1 consensus sequence. Position of the ZNF507 ChIP seq binding motif identified by MEME is indicated above the alignment.

Fig. 4.

Conservation of ZNF146 and ZNF507 zinc finger domains. a) Alignment of protein sequences for ZNF146 and ZNF507 zinc finder domains (indicated by number above alignment) in candidate mammals. Abbreviated common names are used (R. macaque, Rhesus macaque; O. baboon, Olive baboon; Sq. monkey, Squirrel monkey; P. tarsier, Philippine tarsier; G. galago, Garnett’s galago). Black boxes indicate residue differences from human proteins. Gray boxes indicate nonidentical but similar residues. Red arrows indicate residues conserved in monkeys and apes, but not frequently observed in tarsiers, lower primates, or other mammals. b) Approximate timeline of primate evolution relative to the age of L1PA subfamilies (millions of years ago, MYA), adapted from Khan et al. and Konkel et al. Qualitative assessment of ZNF146/ZNF507 binding to L1 subfamilies (inferred from ChIP-seq) is displayed as a gradient below from nonbinding (white) to binding (dark).

Fig. 5.

ZNF146 and ZNF507 are not required to silence LINE-1 elements in HEK293 cells. a) Schematic of CRISPR/cas9 disruption approach. HEK293 cells were transiently transfected with spCas9-sgRNA-puro plasmids with either sgRNAs targeting ZNF146 or ZNF507 respectively or an empty vector (EV) control. Cells were briefly selected with puromycin before outgrowth. Fourteen days post-transfection cells were harvested for western blot analysis (middle panels) to assess disruption of protein expression. RNA was isolated for RNA-sequencing analysis. Disruption and RNA-sequencing were performed in duplicate. b) RNA-sequencing coverage was calculated for reads mapping to ZNF146 (left) or ZNF507 cDNA (right) after filtering for reads marked with or without indels during alignment. Data is presented as the ratio of coverage calculated for reads with indels (determined by CIGAR tags) over total reads in each RNA-sequencing replicate. Triangles in the schematic above represent sgRNA targets. c) Comparison of RNA-sequencing depth in fragments per million (fpm) mapping to individual full-length LINE-1 elements. Outside of the scatter plot are histograms showing most values are at or near zero. d) Comparison of RNA-sequencing depth in fragments per million (fpm) mapping to 1 Kb regions downstream of intact (ORF1/ORF2 or ORF2 only) LINE-1 elements. Outside of the scatter plot are histograms indicating most values are at or near zero. Schematics created with BioRender.com.

Fig. 6.

Disruption of ZNF507 leads to modest changes in gene expression. a) Pie chart of ChIP-seq peak intersection with genomic features. Promoters and enhancers are PHAMTOM-annotated. Genic features are RefSeq-annotated. b, c) Comparison of RNA-sequencing depth in fragments per kilobase per million mapped reads (FPKM) mapping to protein-coding transcripts (top) or lncRNAs (bottom). ZNF146 (left) or ZNF507 (right) disrupted cells are compared to EV control. Dotted gray lines mark 1.5-fold change. Transcripts which were differentially expressed (P < 0.05) are colored blue.

Fig. 7.

Identification of an interaction between ZNF507 and PMRT5. a) Schematic of approach to identify ZNF507-interacting proteins. HEK293 cells were transiently transfected with an 3xFLAG tagged ZNF507 expression plasmid or an empty vector control. After 48 h proteins were extracted and subjected to immunoprecipitation using anti-FLAG beads. Proteins were eluted by competition with 3xFLAG peptide before limited short gel electrophoresis and analysis by mass spectrometry (LC-MS/MS). b) Immunofluorescence analysis of HEK293 cells expressing 3xFLAG-ZNF507. All interphase cells observed expressing 3xFLAG-ZNF507 had nuclear localization. No appreciable signal was observed for cells transfected with empty vector 3xFLAG control plasmid. c) Model for LINE-1 binding by ZNF146 and ZNF507 and potential functions. ZNF146 and ZNF507 bind at thousands of sites in the genome as a result of LINE-1 transposition. Binding by either protein may then recruit various activities to these regions, such as the potential for ZNF507 to recruit protein arginine methylation activity through interaction with PRMT5. These activities may then contribute to context or cell-type dependent regulation of LINE-1 elements or nearby genes. Schematics created with BioRender.com.