HOG, SLS, and PJF conceived and designed the experiments. HOG, SLS, SI, and JLR performed the experiments. HOG, SLS, SI, KB, and PJF analyzed the data. JLR, HYC, and RG contributed reagents/materials/analysis tools. HOG and PJF wrote the paper.
NimbleGen Systems microarrays were used in this study. One of the authors, RG, is an employee of this company. Also, PJF and RG are co-PIs on an NIH grant that partially funded this work.
We performed a genome-scale chromatin immunoprecipitation (ChIP)-chip comparison of two modifications (trimethylation of lysine 9 [H3me3K9] and trimethylation of lysine 27 [H3me3K27]) of histone H3 in Ntera2 testicular carcinoma cells and in three different anatomical sources of primary human fibroblasts. We found that in each of the cell types the two modifications were differentially enriched at the promoters of the two largest classes of transcription factors. Specifically, zinc finger (ZNF) genes were bound by H3me3K9 and homeobox genes were bound by H3me3K27. We have previously shown that the Polycomb repressive complex 2 is responsible for mediating trimethylation of lysine 27 of histone H3 in human cancer cells. In contrast, there is little overlap between H3me3K9 targets and components of the Polycomb repressive complex 2, suggesting that a different histone methyltransferase is responsible for the H3me3K9 modification. Previous studies have shown that SETDB1 can trimethylate H3 on lysine 9, using in vitro or artificial tethering assays. SETDB1 is thought to be recruited to chromatin by complexes containing the KAP1 corepressor. To determine if a KAP1-containing complex mediates trimethylation of the identified H3me3K9 targets, we performed ChIP-chip assays and identified KAP1 target genes using human 5-kb promoter arrays. We found that a large number of genes of ZNF transcription factors were bound by both KAP1 and H3me3K9 in normal and cancer cells. To expand our studies of KAP1, we next performed a complete genomic analysis of KAP1 binding using a 38-array tiling set, identifying ~7,000 KAP1 binding sites. The identified KAP1 targets were highly enriched for C2H2 ZNFs, especially those containing Krüppel-associated box (KRAB) domains. Interestingly, although most KAP1 binding sites were within core promoter regions, the binding sites near ZNF genes were greatly enriched within transcribed regions of the target genes. Because KAP1 is recruited to the DNA via interaction with KRAB-ZNF proteins, we suggest that expression of KRAB-ZNF genes may be controlled via an auto-regulatory mechanism involving KAP1.
Methylation of lysines 9 or 27 of histone H3 (H3me3K9 or H3me3K27, respectively) has been associated with silenced chromatin. However, a comprehensive comparison of the regions of the genome bound by these two types of modified histone H3 has not been performed. Therefore, we compared the binding patterns of H3me3K9 and H3me3K27 at ~26,000 human promoters in four different cell populations. Our studies indicated that the two marks segregate differentially with the two most common types of transcriptional regulators; H3me3K27 is highly enriched at homeobox genes and H3me3K9 is highly enriched at zinc-finger genes (ZNFs). We showed that many of the promoters bound by H3me3K9 are also bound by the corepressor KAP1. A genome-wide screen for KAP1 target genes revealed a difference in the location of KAP1 binding sites in the ZNF genes versus other targets. In general, KAP1 binding sites were localized to core promoter regions. However, KAP1 binding sites associated with ZNF genes are near the 3′ end of the coding region. Our results suggest that the KRAB-ZNF family members participate in an autoregulatory loop involving binding of the KAP1 protein to the 3′ end of the ZNF target genes, resulting in trimethylation of H3K9 and transcriptional repression.
Certain modifications of the core histones have been associated with either active or inactive gene expression. For example, acetylation of histone H3 on lysines 9 and 14 is associated with regions of the chromatin that are undergoing transcription in that particular cell type [
The technique of chromatin immunoprecipitation (ChIP) has been used to demonstrate the presence of H3me3K9 or H3me3K27 at specific human loci [
We have now compared the binding patterns of H3me3K9 and H3me3K27 at ~26,000 human promoters in four different cell populations, identifying thousands of promoters bound by each type of modified histone. Our studies indicate that the two marks segregate differentially with the two most common types of transcriptional regulators. We have also shown that many of the promoters bound by H3me3K9 are also bound by the corepressor KAP1 (also known as TIF1B or TRIM28). Finally, we present a genome-wide screen and characterization of KAP1 target genes in Ntera2 cells. Our results suggest that Krüppel-associated box (KRAB)-zinc finger (ZNF) transcription factors participate in an autoregulatory loop involving the KAP1 protein and trimethylation of histone H3 on lysine 9.
We used ChIP-chip assays with high-density oligonucleotide arrays to analyze the binding patterns of H3me3K9 and H3me3K27 through 5 kb of 26,000 human promoters (see
(A) The Maxfour values from the 5-kb promoter array set (see
(B) Functional categories for the top 2,000 genes bound by either H3me3K9 or H3me3K27 were determined using the program DAVID in Ntera2 cells (black bars), foot fibroblasts (white bars), lung fibroblasts (hatched bars), and foreskin fibroblasts (gray bars). The category “transcription” was significantly enriched for both H3me3K9 and H3me3K27 (see
H3me3K9 and H3me3K27 Predominantly Target Transcription Factors in Four Different Cell Types
Binding patterns of H3me3K9 and H3me3K27 (from 5-kb promoter array ChIP-chip data) in Ntera2 cells are compared at a ZNF gene cluster (1.6 Mb of Chromosome 19, top panel) and at the HoxC cluster (0.15 Mb of Chromosome 12, bottom panel).
We have previously reported that components of the Polycomb repressive complex 2 colocalize with H3me3K27 in F9 cells and mouse embryonic stem cells [
We began by choosing two different antibodies to KAP1, one rabbit polyclonal and one mouse monoclonal antibody. We reasoned that if the same sites were identified using two different antibodies, then we would have confidence that the KAP1 ChIP experiments were identifying true binding sites. Because we did not know if KAP1 prefers to bind to promoter regions or to regions distant from core promoters, we first applied the amplicons made from the KAP1 ChIP samples to ENCODE arrays (which represent approximately 1% of the human genome, including ~400 genes; see
(A) ChIP-chip analysis in Ntera2 cells using two antibodies (rabbit polyclonal, r-KAP1 and mouse monoclonal, m-KAP1) raised against different KAP1 epitopes. Two KAP1 binding sites (one in Enm005, top panel and one in Enr132, bottom panel) were identified using the Tamalpais peak-calling program [
(B) PCR analysis of the two identified KAP1 binding sites in ENCODE regions ENm005 and ENr132 was performed using a third biological replicate of amplicons using the rabbit KAP1 antibody. The enrichment of KAP1 and H3me3K9 is shown in comparison to total chromatin DNA. IgG amplicons were analyzed as a negative control.
With confidence that our KAP1 ChIP-chip assays could identify true KAP1 binding sites, we next performed duplicate ChIP experiments using two independent sets of Ntera2 cells and antibodies to KAP1, SUZ12, H3me3K9, and H3me3K27, prepared amplicons, and performed ChIP-chip experiments using a two-array set of ~26,000 human promoters. As described above, target promoters were identified with the Maxfour peak-calling program. A list of high confidence target genes was generated by selecting the top ranked 2,000 promoters that were bound in the two independent experiments for each antibody. We then determined how many of the SUZ12 or KAP1 targets were also co-occupied by H3me3K9 and H3me3K27. This led to a conservative, but high confidence set of co-occupied targets, since each target had to be identified in four out of four arrays. In support of the hypothesis that the modification of H3me3K9 must be accomplished by a complex other than Polycomb repressive complex 2, we saw no significant overlap between SUZ12 targets and H3me3K9 targets (
Chart showing the percentage of (A) SUZ12 target promoters and (B) KAP1 target promoters that are also bound by H3me3K9 or H3me3K27. For each factor, the top 2,000 target promoters (5-kb array set) from two biological replicate ChIP-chip assays were used. This results in high confidence sets of genes co-occupied by H3me3K9 or H3me3K27 since the target genes had to be present in all four arrays analyzed.
We next compared the top 2,000 KAP1 targets to the top 2,000 H3me3K9 or top 2,000 H3me3K27 targets. We found that a fourth of KAP1 targets also carried the histone modification H3me3K9 but only 11% of KAP1 targets were bound by H3me3K27 (
The significant overlap between KAP1 and H3me3K9 suggests that KAP1 may indeed be functioning as a corepressor in a complex that mediates methylation of lysine 9 of H3. To test this hypothesis, we examined the expression level of different classes of KAP1 and H3me3K9 target genes. To do so, we used NimbleGen expression arrays to analyze RNA levels of the KAP1 target genes in Ntera2 cells, using an average of data obtained from expression arrays probed with RNA isolated from two different cultures of Ntera2 cells. For comparison, we have also analyzed the expression levels of the top 20% of all RNAs on the NimbleGen array. Of the top 2,000 KAP1 and top 2,000 H3me3K9 target genes, 1,952 and 1,842, respectively, were represented on the NimbleGen expression array. As can been seen in
Box plot showing 25th, 50th, and 75th quartile expression levels for the most highly expressed RNAs (the top 20% of all RNAs on the array) and for different target gene categories: all KAP1 target genes, all H3me3K9 target genes, genes co-occupied by KAP1 and H3me3K9, ZNF genes co-occupied by KAP1 and H3me3K9, and KRAB-ZNF genes co-occupied by KAP1 and H3me3K9. Whiskers show the 2.5th and 97.5th percentiles. For these analyses, average RNA expression values from two independent experiments were used.
Previous studies of human transcription factors have indicated that although some transcription complexes are often bound near core promoter regions [
KAP1 Targets Listed by Chromosome in Ntera2 Cells
Inspection of the chromosomal location of the KAP1 binding sites indicated that, in general, the larger chromosomes contained more KAP1 targets than did the smaller chromosomes. However, there were several cases in which large clusters of KAP1 targets resulted in a higher-than-expected number of targets on a particular chromosome. For example, Chromosomes 7 and 19 were highly enriched for KAP1 targets using the whole-genome arrays (
(A). The KAP1 binding patterns along Chromosome 19 (and a 200-kb subregion) containing clusters of ZNF genes are shown. Chromosome positions (Mb) are indicated on the x-axis. The black bar indicates transcribed region of ZNF433 and grey bars represent other transcribed regions (ZNF-like transcripts).
(B) PCR analysis of six KAP1 binding sites was performed using a separate biological replicate of KAP1 amplicons. Four KAP1 binding sites are located within the transcribed region of ZNF genes on Chromosome 19 (ZNF554, ZNF426, ZNF333, and ZNF433) and two KAP1 binding sites are located >100kb from TSS of ZNF genes on Chromosome 7 (ZNF479 and ZNF679). An intergenic region of Chromosome 8 was used as a control for absence of KAP1 binding. The enrichment of KAP1 and H3me3K9 is shown in comparison to total chromatin DNA; IgG amplicons were analyzed as a negative control.
Functional categories for KAP1 target genes were determined using the program DAVID after elimination of genes that had no annotation and genes encoding hypothetical proteins. The top ten categories based on their
Having identified a large set of KAP1 targets in a completely unbiased manner, we could now determine the preferred binding location for KAP1. The distance of KAP1 binding sites to the nearest transcription start site was determined using knownGenes from the University of California Santa Cruz genome browser (
(A) Distribution of the distance between KAP1 binding sites and closest transcription start sites from University of California Santa Cruz KnownGenes (HG17) is shown. Distances are calculated from the center of the KAP1 binding site to the transcription start site and binned in 5-kb intervals between 50 kb upstream and 50 kb downstream of TSS. KAP1 binding sites within each interval are given as a percentage for ZNF target genes (grey) and for KAP1 target genes excluding ZNF genes (black).
(B) Analysis of 277 KAP1 binding sites located within the transcribed regions of ZNF genes. Position of KAP1 binding is plotted relative to the length of the ZNF target gene (5′ end at 0%; 3′ end at 100%).
We have shown that the genes for the two largest classes of site-specific DNA binding transcription factors are bound by distinct histone modifications; trimethylation of histone H3 on lysine 27 is highly enriched at genes encoding homeobox transcription factors (the second largest family of human transcription factors) and trimethyation of histone H3 on lysine 9 is highly enriched at genes encoding ZNF transcription factors (the largest family of human transcription factors). The Krüppel-type ZNF domain is the most common DNA binding domain in the human genome [
Unexpectedly, we have shown that the major targets of KAP1-mediated repression are ZNF genes themselves. For example, in Ntera2 cells KAP1 binds to 60% (212 of 355) of all KRAB-ZNF genes in the genome. This suggests that KAP1 represses transcription of many KRAB-ZNF genes due to its recruitment to their transcribed regions by interaction with the few KRAB-ZNF proteins that are expressed in a cell (
Shown is a model illustrating autoregulation of the family of KRAB-ZNFs by KRAB-ZNF-mediated recruitment of KAP1 to the coding regions of other KRAB-ZNF genes. ZNF426, ZNF333, and ZNF533 are KAP1 targets and are not expressed in Ntera2 cells.
As indicated above, we have shown that the most highly enriched class of target genes for KAP1 is ZNF genes, many of which encode KRAB-ZNF proteins. Interestingly, we find a difference in the location of KAP1 binding sites in the ZNF genes versus other targets. In accordance with our identification of a large set of KAP1 targets using promoter arrays, we find that thousands of the KAP1 targets identified on the whole-genome tiling arrays show localization at the promoter region (defined as 5 kb upstream or downstream of the start site). In contrast, the KAP1 binding sites associated with ZNF genes are predominantly localized within transcribed gene regions, near the 3′ end of the gene. This suggests that the recruitment of KAP1 to the ZNF genes and/or the function of KAP1 in regulating expression of ZNF genes might be different than for other KAP1 targets. The other ~200 DNA binding transcription factors that are KAP1 targets but are not ZNFs show a promoter-localized KAP1 binding pattern (unpublished data). Thus, the 3′ end-localized KAP1 binding pattern is unique to ZNF transcription factors.
A recent study using DamID, rather than ChIP, found that 37% of CBX1 (HP1-BETA) targets and 48% of the SUV39H1 targets correspond to KRAB-ZNF genes on Chromosome 19 [
We have proposed that the general repression of KRAB-ZNF genes is accomplished by recruitment of the KAP1 corepressor to their transcribed regions via interaction with one or more of the KRAB-ZNF proteins that is highly expressed in that particular cell type. It is likely that either simultaneously with or subsequent to KAP1 recruitment a histone methyltransferase such as SUV39H1 or SETDB1 associates with KAP1, resulting in trimethylation of histone H3 at lysine 9. In support of this hypothesis, we have preliminary evidence that reduction of the levels of KAP1 in human 293 cells by stable expression of small interfering RNAs results in a reduction of the levels of H3me3K9 at KAP1 binding sites (
Ntera2 and HEK293 cells were grown in Dulbecco's Modified Eagle Medium supplemented with 10% FBS, 2mM glutamine, and 1% penicillin/streptomycin. The stable KAP1 knockdown cell line K928-cl10 was grown as above, with the addition of 10ug/ml puromycin [
ChIP assays (1 × 107 cells/assay) were performed following the protocol provided at
Amplicons were applied either to ENCODE arrays, 5-kb promoter arrays or to the human genome tiling array set consisting of 38 arrays (see
Sites bound by KAP1 on the ENCODE arrays were identified using the highest stringency level (six consecutive probes above the 98th percentile threshold,
Total RNA was prepared from 5 × 106 Ntera cells using RNAeasy Kit (Qiagen) following the manufacturer's instructions. RNA quality was ensured using the Agilent Systems Bioanalyzer (
(1.2 MB PDF)
(30 KB XLS)
(12 KB XLS)
(22 KB XLS)
(916 KB XLS)
The National Center for Biotechnology Information (NCBI) Entrez (
We thank the members of the Farnham lab for helpful discussion and data analysis and David Schultz for the 293 KAP1 knockdown cells.
chromatin immunoprecipitation
trimethylation of lysine 9 of histone H3
trimethylation of lysine 27 of histone H3
Krüppel-associated box
Su(var), Enhancer of zeste, Trithorax
transcription start site
zinc finger