ChIP-Seq analysis of DHT-treated LNCaP cells shows that about 20% of all AR-occupied regions, including the TMPRSS2
, and KLK3
enhancers, are flanked by AcH3 nucleosomes. Histone modifications surrounding transcription factor binding sites (18
) and local nucleosome positioning and displacement patterns provide information about the functional states of chromatin (13
). In prostate cancer cells, functionally significant chromatin signatures converge at AR-occupied regions, since it is known that the AR drives the disease (6
). We aligned data obtained from a ChIP-Seq analysis of AR occupancy and histone AcH3 modification in LNCaP cells treated for 4 h with the natural AR ligand DHT. A depicts 4,357 highly significant LNCaP cell AR-occupied regions (gray) and the fraction of AR-occupied regions overlapping a significant AcH3 peak at various distances (blue). About 20% of the AR-occupied regions were associated with AcH3; remarkably, a 200- to 400-bp area surrounding the apex of the AR peak displayed a very low number of AcH3 reads, possibly defining an NDR. In contrast, the flanking regions had the highest number of AcH3 reads and showed two well-defined peaks corresponding to two well-positioned acetylated nucleosomes. This pattern is highly reminiscent of that observed in other regulatory regions, such as promoters (36
) and CTCF binding sites (11
), and the fact that it is present in a large number of AR-occupied regions suggests that it may play a role in the expression of target genes controlled by an important group of enhancers.
We selected to study in detail the enhancers of three well-established AR target genes, namely, TMPRSS2
, and KLK3
. The TMPRSS2
enhancer region is located approximately 13.5 kb upstream of the transcriptional start site and contains a noncanonical AR binding site. ChIP-Seq of AcH3 and AR occupancy in DHT-treated LNCaP cells showed two clear AcH3 peaks, flanking the AR binding site (arrow) at the TMPRSS2
enhancer (B). Other androgen-occupied regions at this locus were not studied further. The KLK3
enhancer is located between 5.8 and 3.7 kb upstream of the transcriptional start site, and the KLK2
enhancer spans the region situated between 4.4 and 3.8 kb from the transcriptional start site (27
). As previously reported (3
), and similar to the TMPRSS2
enhancer, two clear AcH3 peaks were found to flank the AR binding site at the KLK2
enhancers (B). These results suggest that, at the enhancers of androgen-responsive genes, AR is bound to an NDR which is flanked by two well-positioned nucleosomes containing AcH3. From these results, however, it is unclear whether the AcH3-free region may contain native unmodified histones or whether the proposed NDR precedes or follows AR binding.
AR recruitment leads to enrichment of AcH3 and H3K4me2 at the TMPRSS2, KLK2, and KLK3 enhancers.
To validate our ChIP-Seq observations, we analyzed gene expression, histone acetylation and methylation, and AR occupancy at the chosen three AR enhancer loci. As expected, 4 h of treatment with DHT significantly increased the TMPRSS2, KLK2, and KLK3 mRNA levels in LNCaP cells compared to those in hormone-deprived cells (C, left panel). Further increases were observed after 16 h of treatment for all three genes (C, right panel).
Site-specific ChIP-qPCR analyses of AR occupancy, AcH3, H3K4me2 modifications, and native histone H3 were conducted at the TMPRSS2, KLK2, and KLK3 enhancers. Five primer sets () were used to map the AR-occupied regions and AcH3 peaks at the enhancers. A displays a detailed image of the ChIP-Seq AcH3 (black ticks) and AR (green ticks) signals in LNCaP cells treated for 4 h with DHT and the relative positions of the five ChIP primer sets (P1 to P5) for each locus. ChIP analysis of AR occupancy (B) in DHT-treated LNCaP cells (green) shows a pattern similar to that observed in our ChIP-Seq data for all three enhancer regions. Significantly higher (P < 0.05) AR occupancy (green bars) was detected in the regions mapped by primer set P3 for at all three loci, than in the regions mapped by primer set P1 (located at the edge of enhancers). In contrast, negligible to no AR occupancy was observed in hormone-deprived cells (, red bars), indicating that AR binds to these loci only in the presence of DHT, a phenotype well known for androgen-dependent LNCaP cells. In agreement with the ChIP-Seq findings, AcH3 increased at the enhancers after DHT treatment (green bars) in a bimodal fashion, with two acetylation peaks in the regions mapped by primer sets P1, P4, and P5 for TMPRSS2 and by primer sets P1, P2, and P5 for KLK2 and KLK3 (). AcH3 enrichment was significantly higher (P < 0.05) in the region mapped by primer set P1 (located at the AR binding site) than in the region mapped by primer set P3 for all three enhancers. Also in agreement with the ChIP-Seq analysis, we found that the area corresponding to AR binding contained lower levels of AcH3. Hormone-deprived cells displayed very low acetylation across the five primer sets (red bars). Similarly, H3K4me2 was enriched in a bimodal pattern only in DHT-treated samples. H3K4me2 enrichment was significantly higher (P < 0.05) in the region mapped by primer set P1 than in the region mapped by primer set P3 for the TMPRSS2 and KLK2 enhancers. Finally, H3 was detected at higher levels in hormone-deprived samples at all three enhancers than in DHT-treated ones. The KLK2 enhancer showed a clear bimodal pattern of distribution of H3 in the androgen-deprived samples. Significant differences were found between primer set P1 and primer set P3, suggesting the presence of an NDR in the hormone-deprived sample. Such a pattern was not readily apparent for the TMPRSS2 and KLK3 enhancers, although significant differences in H3 enrichment were found between primer set P1 and primer set P3 at the KLK3 enhancer. When corrected for H3 levels (B, bottom panels), AcH3 levels were the same across all regions in the absence of DHT for all the enhancers; upon hormone treatment, AcH3 enrichment was observed in the regions flanking the AR binding site. In contrast, the levels of H3K4me2 corrected for H3 levels in the absence or presence of hormone at all enhancers were different than those shown when data were presented as percentages of the input, indicating that the observed depletion of H3K4me2 (when presented as a percentage of the input) simply reflected the presence of histones at the sites. The dramatic increase in AcH3 and H3K4me2 at all sites after DHT treatment more clearly demonstrated that the modifications occurred as a consequence of DHT treatment.
A percentage of TMPRSS2, KLK2, and KLK3 enhancer modules displays NDRs in the absence of ligand.
The results from the KLK2
enhancers were intriguing, particularly since it has recently been proposed that nucleosomes occupy AR enhancer regions in the absence of DHT (14
). A possible explanation for the differences we observed between enhancers in terms of nucleosome positioning as determined by H3 ChIP analysis is the limited resolution level of the assay. To better resolve the pattern of nucleosome positioning and displacement in response to hormone treatment, we performed a high-resolution, single-molecule analysis named NOMe-Seq. This method is based on DNA accessibility to the 40,000-Da M.CviPI molecule, which methylates cytosines at GpC dinucleotides, and has been successfully used to establish nucleosome positions at promoters regions (23
). The advantage of NOMe-Seq analysis over traditional assays is that it allows the investigation of nucleosome position and endogenous methylation in the same molecule in CpG-poor regions (23
) such as enhancers and without the biases previously reported for MNase I (2
). In addition, NOMe-Seq provides a digital readout of nucleosome positioning, thereby allowing direct comparison between treatments. Because NOMe-Seq relies on GC density, loss of resolution may be observed in extremely GC-poor regions. However, the three AR enhancers analyzed in this study have reasonably good GC densities, as do the majority of the gene-rich regions of the genome (28
), and therefore optimal resolution for the NOMe-Seq assay was expected. We designed two overlapping sets of primers () lacking both GpC and CpG sites to eliminate amplification biases, which covered the full enhancer and additional neighboring regions of TMPRSS2
, and KLK3
(A). Primer location was based both on the coordinates of previously characterized enhancers (32
) and on the locations of AR-enriched peaks as determined in our genome-wide ChIP-Seq data (A). The results represent two independent experiments per time point and show a clear M.CviPI-accessible region (teal) in the hormone-deprived sample for all three enhancers (A). However, the degree of accessibility to M.CviPI differed among the regions, with the KLK2
enhancer being the most accessible (41% of enhancer modules being nucleosome depleted), the TMPRSS2
enhancer moderately accessible (26%), and the KLK3
enhancer the least accessible (14%) (A and B). The observed accessible regions were large enough to accommodate at least one nucleosome, showing that even in the absence of androgens, some enhancer modules have NDRs. After 4 h of treatment with DHT (A and B), a significant increase in the number of accessible modules was observed at all three enhancers. Interestingly, the size (footprint) of the NDRs was not appreciably altered. The increase in numbers relative to the control sample was also differential and mirrored the pattern observed for the hormone-deprived samples as follows: TMPRSS2
, 34%; KLK2
, 27.2%; KLK3
, 17.5% (B). In addition, M.CviPI accessibility in both hormone-deprived and DHT-treated samples was observed after 16 h in culture for all three enhancers (C).
Fig. 3. Nucleosome positioning in TMPRSS2, KLK2, and KLK3 enhancers as determined by NOMe-Seq analysis. (A) Single-molecule analysis of nucleosome occupancy for TMPRSS2, KLK2, and KLK3 enhancers. Maps show GpC site density and the location of the enhancer (green (more ...)
To confirm nucleosome depletion at the AR binding site in the absence of hormone, we carried out MNase digestion followed by qPCR analysis of the TMPRSS2
enhancers in LNCaP cells as previously reported (12
). We selected three primer sets (P2, P3, and P5) from the five used for our ChIP analyses (), which covered the nucleosome-depleted (P3) enhancer region and two nucleosome-occupied (P2, P5) enhancer regions, as previously established by NOMe-Seq (). We first determined that the amount of MNase enzyme suitable to enrich for mononucleosomes was 5 IU (A). MNase digestion of LNCaP cells treated with DHT or ethanol for 4 h (after exposure to CSS for 3 days) clearly show a decrease in amplified product in the ethanol-treated samples at both enhancer regions (B). However, MNase digestion was less sensitive than NOMe-Seq in detecting changes in nucleosome depletion as a result of DHT treatment. The P3 primer set used in this analysis was located within the NDR detected by NOMe-Seq, which, according to its size, can accommodate only one nucleosome. Therefore, the decrease in amplified product is likely to reflect true nucleosome depletion, in agreement with the NOMe-Seq results.
In the absence of ligand, a percentage of TMPRSS2 and KLK2 enhancer modules always shows NDRs and this percentage increases after short-term treatment with DHT.
To evaluate the kinetics of nucleosome positioning at AR enhancers, we performed NOMe-Seq analysis after short-term exposure to DHT (). A small increase in TMPRSS2 and KLK2 expression was observed as early as 0.5 h after DHT treatment (A). Results from the NOMe-Seq analysis of TMPRSS2 and KLK2 enhancers (B to E) clearly show 39 to 50% accessibility to M.CviPI (teal) in the ethanol-treated control samples, independently of the time of exposure to ethanol (0.5 h or 2 h). Treatment with DHT induced a significant increase in the number of accessible enhancer modules as early as 0.5 h posttreatment without affecting the size (footprint) of the NDRs, as was observed at later time points. The maximum increase in the number of accessible enhancer modules at the TMPRSS2 enhancer (61%) was observed 0.5 h after exposure to DHT (B and C). At the KLK2 enhancer, this value increased to17.5% after 0.5 h of DHT exposure (B and C) and 21.7% after 2 h of exposure to DHT (D and E), which was similar to that observed after 4 h of treatment. These results suggest that nucleosome depletion in response to DHT treatment occurs shortly after DHT stimulation and coincides with AR occupancy.
Fig. 5. NOMe-Seq analysis of nucleosome occupancy for TMPRSS2 and KLK2 enhancers after short-term DHT treatment. All cultures were exposed to CSS for 3 days prior to the start of the experiments. (A) qPCR analysis of TMPRSS2 and KLK2 mRNA levels after 0.5 h, (more ...) NDRs are present after long-term androgen withdrawal.
The NDR detected in the hormone-deprived sample could be caused by a transient short memory of the AR presence due to recent exposure (3 days) of LNCaP cells to androgens. To evaluate this possibility, we cultured cells in CSS for 7 days (i.e., longer time of hormone deprivation) prior to the 4-h treatment with DHT (). Our results indicate that even after prolonged hormone deprivation, the NDR is maintained at some of the TMPRSS2 enhancer modules in the hormone-deprived sample (A and B), making it unlikely that the NDR is caused by a transient memory of the prior AR presence. In addition, prolonged hormone deprivation did not affect the response to DHT (A and B) and an 18% increase in accessibility relative to that in the hormone-deprived sample was observed.
Fig. 6. Nucleosome positioning in the TMPRSS2 enhancer as determined by NOMe-Seq analysis after extended androgen deprivation. LNCaP cells were cultured in CSS for 7 days and then treated with ethanol (control) or DHT (10 nM) for 4 h. (A) Single-molecule analysis (more ...) Nucleosome positioning analyses of LAPC4 and IMR90 cells.
To determine whether the patterns of nucleosome positioning were conserved among androgen-sensitive cell lines, we performed a NOMe-Seq analysis with LAPC4 cells, which express the wild-type AR (). Cells were treated with DHT or ethanol for 4 h after exposure to CSS for 2 days. The pattern of nucleosome occupancy for the KLK2
enhancer (A and B) was comparable to that of LNCaP cells (A). Importantly, 75% of KLK2
enhancer modules had NDRs in the ethanol-treated control samples. In contrast, the TMPRSS2
enhancer was occupied by nucleosomes (pink bars) in the ethanol-treated sample (A and B), suggesting that accessibility of the enhancers in the absence of ligand may vary among androgen-responsive cell lines. DHT treatment increased the number of enhancer modules displaying NDRs without affecting the footprint size, as was also shown for LNCaP cells. The increase in enhancer accessibility in response to DHT was accompanied by increases in gene expression (C). The same enhancer regions in the human fibroblast cell line IMR90, which expresses neither TMPRSS2
(Oncomine; Compendia Bioscience, Ann Arbor, MI), showed complete nucleosome occupancy (D). On the other hand, the GRP78
promoter, which is expressed in these cells (Oncomine; Compendia Bioscience, Ann Arbor, MI), showed a pattern of M.CviPI accessibility typical of active promoters (D), as was previously reported by our group (12
). The three enhancer regions were CpG poor and, as expected, showed very low endogenous methylation levels within the enhancer in both LNCaP and LAPC4 cells (data not shown).
Fig. 7. NOMe-Seq analysis of nucleosome occupancy for TMPRSS2 and KLK2 enhancers in LAPC4 and IMR90 cells. Enhancer maps show GpC site density and the locations of the enhancers (green bar) in the regions analyzed. (A) LAPC4 cells were maintained in CSS for 2 (more ...) The TMPRSS2, KLK2, and KLK3 enhancers show GATA-2 enrichment in the absence of ligand.
The AR, acting in concert with other (pioneer) transcription factors such as GATA-2, OCT-1, and FOXA1, mediates the expression of androgen-responsive genes. In particular, OCT1 and GATA-2 have been shown to regulate TMPRSS2
), and GATA-2 has also been shown to regulate KLK2
). FOXA1 and GATA-2 act as pioneer factors in the recruitment of AR, although FOXA1 appears to be recruited to a restricted number of sites (43
). In yeast, the transcription factor RSC3 was shown to be required for the maintenance of an NDR in the proximal promoter region of a number of genes that contain the RSC3 binding motif (1
). However, a similar role for transcription factors at promoters or in other regulatory regions has not been described in mammals.
Since pioneer factors are present at AR enhancers in the absence of DHT, we hypothesized that their binding affects the turnover rates of nucleosomes in these regions toward the nucleosome-depleted state, as previously suggested (29
). To test this hypothesis, we used ChIP-qPCR to verify GATA-2 binding to the TMPRSS2
, and KLK3
enhancers by ChIP. The results show that GATA-2 is enriched in all three enhancers irrespective of the presence of DHT (A), which is consistent with the presence of GATA response elements at all three loci (blue rectangles, A). Furthermore, we found that GATA-2 was particularly enriched at the NDRs. For instance, comparing the results of the GATA-2 ChIP analysis and nucleosome positioning at the KLK2
enhancer makes it clear that the regions amplified by P3 and P4 show both the highest signal for GATA-2 binding (A, middle panel) and an NDR (B).
Fig. 8. GATA-2 occupancy at the TMPRSS2, KLK2, and KLK3 enhancers and nucleosome positioning in LNCaP cells after GATA-2 knockdown. (A) GATA-2 occupancy was detected at all three enhancers in both ethanol-treated control cells (red solid bars) and DHT-treated (more ...) GATA-2 is important for NDR maintenance at the TPMRSS2 enhancer in the absence of hormone.
Next, using siRNA to knock down the levels of endogenous GATA-2 in LnCaP cells, we investigated the role of this pioneering factor in maintaining the NDRs at AR enhancers in the absence of hormone. A 70% decrease in GATA-2 mRNA levels (, top panel) and a further decrease in protein levels (B, bottom panel) were observed in LnCaP cells after GATA-2 knockdown. Analysis of nucleosome positioning by NOMe-Seq revealed that knockdown of GATA-2 significantly (P < 0.05) decreased the accessibility of the TMPRSS2 enhancer to M.CviPI by 44% but did not alter the accessibility of the KLK2 and KLK3 enhancers (C and D). However, after GATA-2 knockdown, a new accessible area was observed downstream of the KLK3 enhancer. This area is smaller than a nucleosome and may reflect transcription factor loss (either GATA-2 or others). The results suggest a “pioneering” role for GATA-2 at some, but not all, AR enhancers and provide support for the hypothesis that pioneer factors help maintain enhancer accessibility for subsequent transcription factor binding and action.