|Home | About | Journals | Submit | Contact Us | Français|
Specific chromatin characteristics, especially the modification status of the core histone proteins, are associated with active and inactive genes. There is growing evidence that genes that respond to environmental or developmental signals may possess distinct chromatin marks. Using a T cell model and both genome-wide and gene-focused approaches, we examined the chromatin characteristics of genes that respond to T cell activation.
To facilitate comparison of genes with similar basal expression levels, we used expression-profiling data to bin genes according to their basal expression levels. We found that inducible genes in the lower basal expression bins, especially rapidly induced primary response genes, were more likely than their non-responsive counterparts to display the histone modifications of active genes, have RNA polymerase II (Pol II) at their promoters and show evidence of ongoing basal elongation. There was little or no evidence for the presence of active chromatin marks in the absence of promoter Pol II on these inducible genes. In addition, we identified a subgroup of genes with active promoter chromatin marks and promoter Pol II but no evidence of elongation. Following T cell activation, we find little evidence for a major shift in the active chromatin signature around inducible gene promoters but many genes recruit more Pol II and show increased evidence of elongation.
These results suggest that the majority of inducible genes are primed for activation by having an active chromatin signature and promoter Pol II with or without ongoing elongation.
The timed and coordinated regulation of gene expression is important at every developmental stage of a multicellular organism as well as in the response of the organism to environmental changes. One of the central regulators of eukaryotic gene transcription is the organization of the genome into chromatin. Histone proteins are key components of chromatin, forming the basic nucleosome packaging structure. Over the past decade, the post-translational modification of histone proteins has been shown to have a complex role in controlling gene expression (reviewed in [1,2]). In general, actively transcribed genes are associated with lysine acetylation on histones H3 and H4 and with methylation of histone H3 on lysine 4 (H3K4me). On the other hand, methylation of lysine 9 (H3K9me) or lysine 27 (H3K27me) on H3 is associated with repression. Many protein complexes responsible for adding or removing these modifications have been isolated and shown to play important roles in controlling gene expression (reviewed in ).
In terms of chromatin packaging, these histone modifications are considered to be important in inter-nucleosome interactions and higher order chromatin packaging . In relation to gene transcription, they can form important binding surfaces on nucleosomes for chromatin binding proteins that play key roles in gene transcription (reviewed in ). These observations have led to the idea of a 'histone code' that marks chromatin domains in the eukaryotic nucleus and either plays a role in controlling gene transcription or is a result of the transcriptional activity of that locus.
Although the 'histone code' that marks active and inactive genes has now been characterized in some detail, there is less information in regard to the chromatin status of inducible genes prior to activation. Of particular interest in this regard are recent genome-wide studies of histone marks in mouse pluripotent embryonic stem cells that have defined a class of developmentally regulated genes as 'bivalent' - genes marked with both active (histone H3 lysine 4 trimethyl (H3K4me3)) and repressive (histone H3 lysine 27 trimethyl (H3K27me3)) histone modifications [4-6] Furthermore, many of these bivalent genes are found to have RNA polymerase II (Pol II) located at their promoters in what is proposed to be a poised state . The existence of a bivalent state has also been shown on some genes in other types of stem cells and in more differentiated cells, implying that this chromatin state may be involved in controlling genes that respond to developmental or environmental signals in all cell types [8-11]. Sequential chromatin immunoprecipitation (ChIP) has been used in a couple of cases to clearly show the bivalent nature of specific genes [5,8]. Following differentiation, it has been shown that these genes often resolve into a monovalent state for expression or repression [5,9,10]. Whether genes that respond rapidly to cellular activation signals also display bivalent chromatin marks remains to be examined.
It has long been known that certain inducible genes, such as the heat shock genes [12-14] and some oncogenes [15,16], have Pol II paused or stalled close to the start of gene transcription and that an increased elongation rate plays a role in their response to signaling. Not only inducible genes but many other genes also show evidence of pausing even with detectable transcription, implying that this constitutes a common mechanism to control the transcription rate . More recently, genome-wide studies in mouse and human embryonic stem cells and differentiated human cells have identified large numbers of genes where Pol II is located at the promoter in the absence of ongoing transcription and these genes are often referred to as poised [5,17,18]. In yeast, Pol II was constitutively bound to hundreds of promoter regions that are activated immediately following exit from stationary phase . Recent genome-wide studies in Drosophila have also defined groups of genes with promoter-enriched Pol II, a feature that is postulated to facilitate rapid induction of transcription of these genes [19-21]. These studies have led to the definition of three classes of genes based on Pol II location [17,22]. Genes in the first class lack Pol II and are considered as inactive. The second class includes active genes where Pol II can be detected at both the promoter and in the body of the gene, but it should be noted that, in general, the level of Pol II in the body of the gene is lower than at the promoter or the 3' end. The third class consists of those genes where Pol II is detected at the promoter but not in the body of the gene and are considered potentially active. Genes in this third class are generally referred to as poised genes and are enriched for developmental control genes and genes that respond to developmental or environmental signals [20,21]. Recent evidence in Drosophila suggests that genes with promoter-proximal enrichment of Pol II can span a wide range of expression levels, supporting the idea that promoter proximal pausing is a common mechanism used to control transcription rate [20,23]. These data in turn suggest that the regulation of elongation may play an important role in the response of genes to environmental signals.
The mature cells of the immune system represent an exquisitely poised system for rapid response to pathogens and thus can be used to investigate the chromatin characteristics of genes that respond rapidly to extracellular signals. Recent genome-wide studies in human T cells have extensively characterized a large number of histone modifications using ChIP combined with massively parallel sequencing (ChIP-Seq) and identified modification patterns associated with enhancers, promoters, other genomic control regions as well as conserved domains [24-28]. These studies have also defined histone modification patterns associated with active and inactive genes, but the patterns associated with inducible genes were not examined in any detail [24-28]. Earlier studies have shown that many new regions of acetylation appear in response to T cell activation, suggesting that inducible genes may change their chromatin signature in response to activation [26,29].
Using three approaches - ChIP combined with microarray technology (ChIP-on-chip), mining of ChIP-Seq data and ChIP with quantitative PCR (ChIP-qPCR) - for individual genes, we sought to define the chromatin signature of inducible genes in T cells. To facilitate comparison of genes with similar basal expression levels, genes were binned according to their basal expression levels determined from expression profiling studies. Our results show that inducible genes in the lower basal expression bins, especially rapidly induced primary response genes, were more likely to display the chromatin characteristics of active genes than their non-responsive counterparts.
To ask whether T cell inducible genes have a defined chromatin signature, genome-wide approaches were used to both identify inducible genes and to examine the chromatin characteristics of these genes. First, expression profiling was performed on non-stimulated or phorbol 12-myristate 13-acetate and ionomycin (P/I)-treated (4 h) EL-4 T cells with or without cycloheximide (CHX) treatment, and inducible genes were identified (false discovery rate (FDR) <0.1) and grouped into primary (539 genes; those genes whose expression was not inhibited by CHX and thus do not need new protein synthesis for expression) and secondary (1,238 genes; those genes whose expression was inhibited by CHX and thus require new protein synthesis for expression) gene groups dependent on their response to CHX treatment. Both of the gene groups displayed a wide spread of basal mRNA expression levels but, on average, the primary and secondary groups displayed higher basal expression levels compared with the unchanged group or all genes (Additional data file 1a), implying that many inducible genes are already producing detectable transcripts. Therefore, to ensure comparison of genes with similar basal expression levels, the primary, secondary and unchanged groups were binned according to their basal mRNA expression levels (Table (Table1).1). The numbers of primary response genes in the lower expression bins (Log2 3 to 4 and 4 to 5) were small and thus could not be treated in a sound statistical manner (Additional data file 7; noted as NA or not applicable).
ChIP-on-chip experiments on unstimulated EL-4 cells were performed using H3K9ac and H3 antibodies and Affymetrix mouse promoter arrays (1.0R) and the data were analyzed using the model-based analysis of tiling array (MAT) algorithm . The promoter region of a gene was defined as -1.2 kb to +0.6 kb from the transcriptional start site (TSS) and the highest score of any overlapping H3K9ac or H3 region detected by MAT was used as the score for that gene. As expected from previous studies showing an association between gene expression and H3K9ac [28,31], all gene groups showed an increase in the median H3K9ac MAT region score as their basal mRNA expression levels increased (Figure (Figure1a)1a) but control immunoprecipitations did not show this pattern (Additional data file 1b). In general, both the primary and secondary gene groups displayed significantly higher median levels of H3K9ac compared to the unchanged gene group (Figure (Figure1a)1a) with the statistical significance of the differences decreasing with increasing basal expression (Additional data file 7; compare log2 5 to 6 with log2 9 to 10 for primary or secondary versus unchanged). In addition, the primary response genes were significantly more acetylated than the secondary response genes in some but not all basal expression bins (Figure (Figure1a;1a; Additional data file 7). Because the underlying histone density can vary across the genome, especially at promoter regions, the H3K9 acetylation values were also calculated relative to the total histone H3 scores with very similar results (Figure (Figure1b;1b; Additional data file 7).
Within each binned gene group there was a considerable spread of acetylation values, so we next asked if the percentage of genes above a specific acetylation score threshold was higher for the inducible gene groups. If a MAT score of 35.2 (FDR of 0.05) was set as a threshold and genes above this score designated as acetylated, then a significantly greater percentage of primary and secondary response genes were acetylated compared with the unchanged genes in the log2 5 to 6, 6 to 7 and 7 to 8 expression bins (Figure (Figure1c;1c; Additional data file 7). These data suggest that inducible genes with lower basal expression have relatively high levels of acetylation in the basal state compared with non-responsive genes and may be primed for activation.
We verified these results using ChIP-qPCR for a number of genes from the basal expression log2 5 to 6 bin (Figure (Figure1d).1d). The PCR data agreed with the predictions from the array studies, with the primary genes having the highest ratio of H3K9ac:H3, followed by the secondary response genes and then the unchanged genes (Figure (Figure1d;1d; Additional data file 2). We also selected a group of previously well characterized inducible genes and examined the H3K9ac status of their promoters in non-stimulated cells. The induction levels, response to CHX and basal expression levels for this gene group are shown in Additional data file 3. Four (Fos, Nfkbia, Tnfaip3 and Tnfsf9) out of the five primary response genes displayed relatively high levels of acetylation whereas those of the secondary response group were generally lower (Figure (Figure1e;1e; Additional data file 2). Several control genes, the active Gapdh (log2 13.9) and the inactive Rho (log2 4.4), Snail, Slc22a13 and Col11a1 displayed the expected pattern for active and repressed genes, respectively (Figure (Figure1e;1e; Additional data file 2).
We next mined a genome-wide ChIP-Seq data set from human primary CD4+ lymphocytes  to find the number of H3K9ac tags that overlapped with the promoter regions (-1 kb to +1 kb of the annotated TSSs) of the human orthologs of the mouse genes. The basal expression level bins were adjusted using expression profiling data available for human CD4+ lymphocytes  from the same investigators (Table (Table2).2). The stimulation used in the aforementioned paper was longer than the 4 h stimulation used in this study, so we used a data set ([GEO:GSE3720] ) from human γδT lymphocytes stimulated for 4 h with P/I to establish if the inducible genes in EL-4 T cells were also induced in human primary lymphocytes. For the primary and secondary response genes with basal expression less than log2 6, 52% and 39% of the genes, respectively, were induced compared to 25% for the unchanged group (P < 0.002). The profile of H3K9ac was very similar to that derived from the mouse ChIP-on-chip studies, with significantly higher median levels of acetylation for the primary response genes compared with the unchanged genes in the log2 3 to 4, 4 to 5 and 5 to 6 expression bins (Figure (Figure1f;1f; Additional data file 7). Secondary response genes also showed some evidence of increased acetylation compared with unchanged genes in the lower basal expression bins and in some bins there were significant differences between primary and secondary genes (Figure (Figure1f;1f; Additional data file 7). The human data set contains information about a number of other acetylation marks and we found that the majority of the acetylation marks showed a similar pattern to H3K9ac, with H2AK9ac, H2BK20ac, H3K36ac and H4K16ac showing the most significant differences between the inducible and unchanged gene groups (Additional data file 4a-d). Once again, the primary response gene groups generally showed a stronger trend than the secondary response gene groups (Additional data file 4a-d).
Thus, all three approaches show that inducible genes, especially primary response genes with lower basal expression, are more likely than their non-responsive counterparts to have a histone acetylation profile that resembles active genes.
Previous studies have shown that promoters without CpG islands are less likely to have acetylated histones than those with CpG islands . We therefore divided the gene groups into those with and without CpG islands (Figure (Figure2a)2a) and asked if the presence of a CpG island correlated with the H3K9ac pattern. As expected, for the genes with CpG islands, acetylation levels were generally higher and a higher percentage of the genes were acetylated than those without CpG islands across all of the gene groups (Figure 2b-e). However, in both CpG and non-CpG island promoter groups, the inducible gene groups had significantly higher median acetylation scores than the unchanged genes in the lower basal expression bins (Figure 2b, c; Additional data file 7) and a significantly higher percentage of them was acetylated (Figure 2d, e; Additional data file 7). These data show that while GC content influences the level of acetylation across the entire gene set, the difference between inducible and non-inducible genes was not directly related to GC content.
Active genes have been shown to display a high level of H3K4 trimethylation (H3K4me3) whereas inactive genes have low levels of H4K4me3 but high levels of H3K27me3 [25,33]. In genome-wide studies in embryonic stem cells, genes with CpG islands that are destined to be activated later in development display both active (H3K4me3) and inactive (H3K27me3) histone marks and have been described as 'bivalent' [5,6]. Therefore, we next examined the patterns of the permissive H3K4me3 and the repressive H3K27me3 marks from the ChIP-Seq data set in human CD4+ T cells. As expected, these two methylation marks showed a reciprocal pattern across the range of expression bins, with H3K4me3 being strongest for the highest expression bins and H3K27me3 strongest for the lowest expression bins (Figure 3a, b). The permissive H3K4me3 mark was significantly higher for both the primary and secondary gene groups compared to the unchanged group in the log2 3 to 4, 4 to 5 and 5 to 6 basal expression bins (Figure (Figure3a;3a; Additional data file 7) while H3K27me3 displayed a reciprocal pattern across these expression bins with the exception of the secondary response genes in the lowest expression bin (Figure (Figure3b;3b; Additional data file 7). While the H3K4me1 and me2 patterns were very similar to the H3K4me3 pattern and the H3K27me2 and me3 patterns resembled each other and were reciprocal to the H3K4 marks, the H3K27me1 mark displayed a pattern very similar to the H3K4 marks (Additional data file 5a-d). There was no significant difference between primary and secondary gene groups for these methylation marks.
We next verified the genome-wide findings by determining the status of these two chromatin marks on our selected gene groups. All of the primary response genes, in either the gene group selected from genome-wide data or the well known primary response gene group, displayed a high level of H3K4me3 and a very low level of H3K27me3, except for Egr2 (Figure 3c, d). The secondary response genes had more variable levels of both marks, but in general the trend was towards lower H3K4me3 and higher H3K27me3 levels (Figure 3c, d). The constitutively active or repressed genes displayed the expected patterns except for Col11a1, where neither mark was detected. A small number of genes, notably Egr2 and Il2, displayed both active and repressive methylation marks and could be classified as potentially bivalent (Figure (Figure3d3d).
We used clustering of all of the methylation marks and the genes in the Log2 3 to 6 basal expression bins to ask whether primary response genes in the lower basal expression bins may be enriched for genes with a 'bivalent' mark (Figure (Figure3e).3e). Only a small subset of the primary response genes were identified as potentially bivalent (Figure (Figure3e,3e, cluster 3), with the majority displaying an active profile (Figure (Figure3e,3e, cluster 2). The genes with potentially bivalent marks did not appear to be enriched for a specific expression bin. Cluster 1 displayed an inactive profile with enrichment for H3K27me3 and me2 marks (Figure (Figure3e).3e). In addition, it can be seen that H3K27me1 did not cluster with the H3K27me3 and me2 marks but was more tightly linked with the H3K4me marks in cluster 3 (Figure (Figure3e).3e). Cluster 4 showed an interesting profile with enrichment for H3K27me1 but lower H3K4 marks.
These data suggest that inducible genes are likely to be marked by active methylation marks in resting cells but that a small number may be in a bivalent state. The implications for expression response for these different gene groups are not yet clear.
Since we have shown that inducible genes with low basal mRNA expression often have an active chromatin signature, we next asked if these genes also had Pol II located at their promoters in non-stimulated cells. Using the human T cell ChIP-Seq data, we found that the median Pol II level was significantly higher at the promoters (-0.25 kb to +0.25 kb) of the inducible gene groups compared with the unchanged group (Figure (Figure4a;4a; Additional data file 7). This was true for the primary response genes across the majority of expression bins but for the secondary response genes in the log2 3 to 4, 4 to 5 and 5 to 6 bins. If promoters with the same or greater number of Pol II tags than the median level of Pol II for unchanged genes in the log2 6 to 7 basal expression bin are plotted, then a similar pattern is seen for the percentage of promoters that reach this threshold (Figure (Figure4b;4b; Additional data file 7). Significantly more of the primary response genes have Pol II at their promoters compared to the secondary genes in some but not all of the basal expression bins (Figure 4a, b; Additional data file 7).
We performed clustering analysis to ask if the genes with the active acetylation and methylation marks were also the genes that had Pol II at their promoters. The ChIP-Seq data from human T cells were used and the primary response genes in the lower basal expression bins (log2 3 to 6; Table Table2)2) were clustered. The chromatin marks used were H3K4me3, H3K9ac, H4K16ac, H3K36ac, H2BK20ac and H2AK9ac as active marks and H3K27me3 as a repressive mark. The largest cluster of these primary response genes was marked by active chromatin (Figure (Figure4c,4c, cluster 2); moreover, all of the genes in this cluster with an active chromatin signature also showed evidence of Pol II at their promoters. Cluster 3 contained genes that were potentially bivalent and these genes displayed lower and more variable levels of Pol II (Figure (Figure4c).4c). As expected, the inactive gene cluster did not display promoter Pol II (Figure (Figure4c,4c, cluster 1). Most importantly, there was little or no evidence for genes with Pol II but without an active or at least bivalent chromatin signature (Figure (Figure4c4c).
We showed above that our selected primary response gene set, with the exception of Egr2, had relatively high levels of active chromatin marks (H3K9ac and H3K4me3) compared to the secondary response group. We therefore asked whether the primary response genes had higher levels of Pol II in the basal state compared with the secondary response genes. Figure Figure4d4d shows that Pol II levels were higher on those primary response genes with an active chromatin signature (Tnfaip3, Nfkbia, Fos and Tnfsf9) and lower on Egr2 (which did not have an active chromatin signature) and also on the secondary response genes. These data support the findings from the human ChIP-Seq data clustering and again link the presence of promoter Pol II with active promoter chromatin.
Thus, we have shown that inducible genes, especially primary response genes, are more likely to have Pol II at their promoter regions and the presence of promoter Pol II is strongly associated with the presence of active chromatin marks.
There has been considerable interest in the nature of Pol II at gene promoters that respond to developmental or environmental signals [20-22,34]. We therefore asked whether the enrichment of Pol II at inducible gene promoters was associated with the enrichment of an elongation signature. H3K36me3 is a mark of elongation and can be used as an indicator of active gene transcription . Hence, we examined the H3K36me3 elongation mark using the human T cell ChIP-Seq data set, and tag counting at 6 to 8 kb downstream of the TSS. While there was a general trend towards higher levels of H3K36me3 in the inducible genes compared with non-responsive genes, this was only statistically significant for the log2 4 to 5, 5 to 6 and 8 to 9 basal expression bins (Figure (Figure5a;5a; Additional data file 7), implying that these genes are more likely to be undergoing elongation. If genes are considered to be H3K36me3 positive if they have the same number of or more tags compared with the average tag count for unchanged genes in the log2 6 to 7 basal expression bin, a similar pattern is seen, although the difference is only significant for primary response genes in the log2 3 to 4, 5 to 6 and 8 to 9 bins and the log2 4 to 5 and 5 to 6 bins for the secondary response genes (Figure (Figure5b;5b; Additional data file 7). The original analysis of the ChIP-Seq data by Wang et al.  showed that in addition to H3K36me3, high levels of H2BK5me1 and H4K20me1 occur in the coding regions of highly expressed genes. Both these marks showed a similar pattern to H3K36me3 in the coding regions of the inducible gene groups (Additional data file 6a, b).
Clustering analysis was used to ask whether these primary response genes with Pol II enrichment in the log2 3 to 6 expression bins could be divided into those with and without evidence of basal elongation. We clustered the three elongation marks described above with the Pol II signal from the human ChIP-Seq data set and found that many genes with promoter Pol II showed evidence of elongation (Figure (Figure5c,5c, clusters 1 and 3). Cluster 3 was enriched for genes in the log2 5 to 6 basal expression bin, which are thus more likely to be producing RNA transcripts in the basal state. A smaller number of genes appeared to have promoter Pol II with little or no evidence of elongation (Figure (Figure5c,5c, cluster 4). It should be noted that 50% of the genes in cluster 4 were from the lowest expression bin (log2 3 to 4) with only two genes from the log2 5 to 6 basal expression bin. Most of these genes (11 of 16) also have active promoter chromatin marks and thus most likely represent a group of poised genes with promoter enriched Pol II, active promoter chromatin but no evidence of elongation or transcript accumulation.
We examined the H3K36me3 levels on six genes in EL-4 T cells, three selected from cluster 3 with clearly detectable levels of this mark and three from cluster 4 with very low levels of this mark. These genes are all inducible in the EL-4 cells (data not shown). We have found that because the level of H3K36me3 varies from one part of the genome to another (data not shown and compare Rho with Gapdh) it is important to compare the level of this mark within the transcribed region and the promoter region of any one gene to gauge the level of enrichment within the gene. The three genes from cluster 3, Gadd45g, Nfkbie and Zswim4, all had higher levels of H3K36me3 in their transcribed regions compared with their promoter regions (Figure (Figure5d).5d). The three genes from cluster 4, Adamts6, Usp54 and Hspa41, however, did not show a significant enrichment of H3K36me3 in their transcribed regions compared with their promoter regions and are similar to the inactive Rho pattern, implying a lack of basal elongation (Figures (Figures5d5d and and6d).6d). The selected primary gene set also displayed an enrichment of H3K36me3 in their transcribed regions with Egr2, the gene with the least promoter Pol II (Figure (Figure4d),4d), also having the lowest H3K36me3 enrichment (Figure (Figure6d).6d). Despite evidence of ongoing elongation as measured by the presence of H3K36me3 in their transcribed regions, these genes display low but variable levels of expression (Additional data file 3), suggesting further post-transcriptional control for at least some primary response genes.
Taken together, these data imply that primary response genes are more likely to have an elongation signature compared with their non-responsive counterparts with comparable basal expression. In addition, we identified a group of primary response genes with active promoter chromatin and promoter Pol II but no or a low number of elongation marks.
We reasoned that if many of the inducible genes, especially the primary response genes, were already in an active chromatin configuration and had Pol II available at their promoters, there may be little or no change in the level of active chromatin marks or Pol II following stimulation. We first examined changes in H3K9ac genome-wide by performing ChIP-on-chip experiments with H3K9ac and H3 antibodies in EL-4 cells stimulated for 0.5 or 4 h with P/I. Acetylation changes were assessed across a +1.2 to -0.6 kb region and genes designated as acetylated if there was a MAT score in this region of >35.2 (FDR <0.1). We found that in the lower basal expression bins (log2 4 to 6), while 25%, 12% and 3% of the primary, secondary and unchanged gene groups, respectively, were acetylated in the unstimulated cells, only 3% (two genes) of the promoters of primary response genes became transiently acetylated at 0.5 h following activation and 5% (14 genes) of the secondary response genes were newly acetylated at 4 h compared with 0.2% (12 and 13 genes at 0.5 h and 4 h, respectively) of the unchanged genes (data not shown). These data imply that the vast majority of inducible genes do not show an increase in H3K9ac in response to activation and may already be sufficiently acetylated for downstream events to occur.
We next examined H3K9ac levels in the selected group of inducible genes following stimulation of EL-4 T cells with P/I. Only Egr2, which had low H3K9ac levels in unstimulated cells, showed consistent evidence of an increase in H3K9ac following activation (Figure (Figure6a).6a). These results are in agreement with our previous studies showing that there was no increase in H3K9ac or other acetylation marks at the promoter regions of the Il2 and GM-CSF genes . These results also agree with the genome-wide ChIP-on-chip studies described above where only a small percentage of genes showed an increase in acetylation. In addition, we did not detect any significant increases in the active methylation mark H3K4me3 (data not shown) as mentioned in the work by Roh et al. .
These data imply that increases in acetylation or active methylation marks are not an essential component of gene activation and that some genes may already be in a sufficiently active chromatin state to allow transcription in response to appropriate signals.
We next determined the changes in Pol II both at the promoter and in the transcribed regions of the selected gene set. For all of the primary response genes, despite detectable levels of Pol II at the majority of the promoters in non-stimulated cells, there was an immediate increase in Pol II at the promoter and an accompanying but smaller increase in the transcribed regions (Figure (Figure6b).6b). In the secondary response group, only three genes, Irf4, Gitr and Il2, showed detectable increases in Pol II and these increases appeared later, in keeping with the delayed expression of these genes (Figure (Figure6b;6b; Additional data file 3). The inability to detect Pol II on some secondary response genes may relate to the affinity of the antibody coupled with the degree of induction.
Additionally, we examined Pol II recruitment to two genes from the cluster analysis in Figure Figure5c5c (Gadd45g from cluster 3 with evidence of basal elongation and Adamts6 from cluster 4 with evidence for a poised polymerase) and an additional constitutively active control gene, Rpl3, in an independent set of quantitative PCR experiments. Both Gadd45g and Adamts6, despite their different basal states, showed evidence of Pol II recruitment at the promoter and an accompanying but lesser increase in the transcribed region while Pol II levels on Rpl3, like those on Gapdh, were not changed (Figure (Figure6c).6c). Tnfaip3, Fos and Tnfsf9 behaved in a similar manner in the two independent experimental sets (Figure 6b, c).
For most genes, with the exception of Il2 and Tnfsf9, the level of Pol II in the transcribed regions of the genes was significantly lower than at the promoter (Figure (Figure6b),6b), making it difficult to assess whether promoter recruitment of Pol II leads to an increase in elongation. We therefore assessed the level of H3K36me3 at the promoter and the transcribed regions of the genes. In non-stimulated cells, the level of H3K36me3 was higher in the transcribed regions than at the promoter for all genes examined, implying, in agreement with the genome-wide results above, that there may be some basal transcription occurring from these genes before activation (Figure (Figure6d).6d). Following activation, all of the genes that had an increased recruitment of Pol II to their promoters also underwent an increase in H3K36me3 levels in the transcribed region of the genes (Figure (Figure6d).6d). In addition, several other secondary response genes also showed increased H3K36me3 following activation (Figure (Figure6d),6d), implying that either a promoter region other than that examined was being used or that Pol II antibody was not as sensitive as the H3K36me3 antibody.
These data suggest that while inducible genes, in particular primary response genes, may be in an active chromatin configuration and have varying levels of Pol II at their promoters in non-stimulated cells, they still recruit more Pol II and increase their rate of elongation, as measured by increased H3K36me3, contributing to the observed increases in mRNA levels.
We have used several approaches to show that, in non-stimulated T cells, inducible genes with lower basal mRNA expression are more likely than their non-responsive counterparts to have an active chromatin signature. This active signature is particularly pronounced for primary response genes. In addition, the presence of this active chromatin signature on a gene is strongly associated with the co-location of Pol II at the gene promoter. A subgroup of genes with an active chromatin signature and promoter Pol II appear to be in a poised state.
Many studies to date, either on individual genes [29,36-38] or genome-wide [17,25,26,33,39-41], have clearly shown that chromatin around active genes is highly acetylated as well as trimethylated at H3K4. Recent ChIP-Seq studies in human CD4+ T cells have identified a chromatin signature for active genes, including acetylation at a number of lysine residues, and mono-, di-, and trimethylation on H3K4 [24,28]. We show here that a similar chromatin signature marks not only active genes but also inducible genes, especially primary response genes, and the enrichment of these marks distinguishes them from their non-responsive counterparts with similar basal expression levels, especially in the lower basal expression bins. Very recently, a similar active chromatin signature has been found on many primary response genes in macrophages . Hargreaves at al.  also showed that their primary response gene group displayed active chromatin marks in several cell types, including embryonic stem cells, and suggested that primary response genes that respond to a variety of signals in many cell types may be tagged in this manner at a very early stage of development. We have also found that our primary response gene set is tagged with active chromatin in macrophages and B cells (data not shown), supporting the data of Hargreaves et al , whereas the secondary response genes that showed cell-specific expression were more variable (data not shown).
Clustering analysis clearly demonstrated that Pol II was associated with the same sets of genes that had an active chromatin signature. We found little or no evidence for the presence of an active chromatin signature in the absence of Pol II or vice versa. The small numbers of genes with such a signature are likely to have alternative TSSs. It is generally accepted that histone modifying complexes are recruited to specific genomic regions through their interaction with transcription factors that recognize their cognate DNA binding sites (reviewed in ). The components of the Pol II initiation complex, likewise, are recruited by directly or indirectly interacting with transcription factors bound to promoter regions [43-45]. Our results support a model of co-recruitment of histone modifying complexes and the Pol II complex since they appear to be collocated on the majority of promoters. Indeed, many transcription factors have been shown to interact with both components of the Pol II initiation complex and with histone modifying complexes (reviewed in ) and so co-recruitment by the same complexes is one possible model. A recent study has shown a possible role for the transcription factor Sp1 in the maintenance of an active chromatin state and the promoter enrichment of Pol II in primary response genes in macrophages . This characteristic was associated with primary response genes with GC-rich promoter regions , and others have also recently shown that GC richness may be an important feature of establishing the permissive chromatin structure of inducible genes . However, we found here that although GC-rich promoters clearly have higher levels of active chromatin marks and promoter Pol II than their non-GC-rich counterparts, the difference between the primary response genes and the non-responsive genes in a given expression bin was evident for both genes with GC-rich promoters and those with non-GC-rich promoters.
Only a small subset of the primary response genes was potentially 'bivalent', having both H3K4me3 and H3K27me3 on the same gene. These genes generally had promoter Pol II but, in general, the Pol II levels were lower and more variable than those of the genes with only active marks. The potentially bivalent genes did not display a clear pattern of active elongation marks and were not associated with either the poised or actively elongating gene sets (data not shown). Perhaps for most primary response inducible genes the permissive chromatin state is established more by the presence of other active histone modifications like H3K9ac, as shown in a study on CD8+ T cells , than by bivalency because of the need to respond rapidly to extracellular signals. The enrichment of active chromatin marks and promoter Pol II and the finding by others of nucleosome depletion  on inducible genes in non-stimulated cells suggests that the promoters of these genes resemble those of active genes regardless of the level of mature mRNA in the basal state.
Several recent genome-wide studies in yeast and Drosophila have shown that Pol II is located at the promoters of large numbers of genes that respond to environmental or developmental signals [17,18,20,21,24,25,28,39,48]. These studies, and others on single genes [49,50], have led to the identification of a group of genes that are described as poised and potentially active [14,22,34,51]. Our data clearly show, in agreement with these studies, that inducible genes, especially primary response genes, are more likely to have Pol II at their promoters than their non-responsive counterparts with comparable expression levels. Even within the lowest basal expression bin (log2 3 to 4) in the human T cell data set, >65% of the genes had evidence of promoter Pol II.
The primary response genes from the log2 3 to 6 basal expression bins with promoter Pol II fell into two distinct categories; those with and without evidence of elongation. The clusters with evidence of elongation are clearly enriched for genes from the higher expression bins (64% of the log2 4 to 5 genes and 77% of the log2 5 to 6 genes fall into cluster 3 in Figure Figure5c).5c). This suggests that many primary response genes are already producing mRNA and induction leads to an increase from that basal level. Many of the independently assessed genes, such as Tnfaip3, Nfkbia, Fos and Tnfsf9, also display evidence of ongoing elongation in the basal state. These genes all display low but variable levels of mRNA, implying that there may be control of mRNA accumulation or stability post-translationally, as has previously been shown for genes such as Fos . While the gene cluster with the strongest elongation signature (cluster 3 in Figure Figure5c)5c) is clearly enriched for genes that are more highly expressed in the basal state (log2 5 to 6), there are also genes in this cluster with lower expression levels (for example, Taf13, Spry4), in which post-translational events may also play a role.
Only a relatively small cluster of genes (cluster 4 in Figure Figure5c)5c) displayed promoter Pol II but had little or no evidence of elongation, and these genes were generally represented in the lowest basal expression bin. Most of these genes (11 of 16) also had an active histone acetylation and methylation signature. It is likely that such genes possess a strong signal for Pol II pausing, but further experiments would be required to determine the nature of that pausing signal. Some factors involved in the pausing of Pol II are the Negative elongation factor (NELF) and also DSIF (5,6-dichloro-1-beta-D-ribofuranosylbenzimidazole (DRB)-sensitivity-inducing factor), both of which have been shown to have repressive effects on Pol II elongation ; these factors have previously been shown to operate in T cells .
All of the primary response genes that we assessed, whether showing evidence of basal elongation or not, recruited more Pol II and displayed evidence of increased elongation following gene activation. These results imply that while the genes may have some level of Pol II and H3K36me3 in the non-stimulated cells, activation leads to an increase in these activities, most likely brought about by the induction of inducible transcription factors that more efficiently recruit and activate Pol II. Many inducible genes in T cells are controlled by inducible transcription factors such as AP-1 and NF-κB, all of which have been shown to play a role in Pol II recruitment and or elongation [42,54]. We found little evidence, however, for an increase in H3K9ac following activation and postulate that many inducible genes already display an active chromatin configuration that does not require further alteration for increased gene activity. Hargreaves et al.  have recently proposed a model whereby constitutive transcription factors such as Sp1 recruit certain histone acetylases, such as p300, to primary response genes to maintain an active chromatin acetylation signature. Inducible transcription factors then recruit different acetylases that modify a different set of lysines on the histone proteins to provide a platform to generate an even more active gene. It would be of interest to examine these latter histone modifications in T cells.
The results presented here show that inducible genes, especially primary response genes, are in a more active chromatin state than their non-responsive counterparts for a given basal expression level. Recent evidence suggests that the permissive state of primary response genes may be present throughout development to allow rapid expression of these genes in many cell types . It will be important to determine the molecular mechanisms that initiate and maintain this permissive state.
All reagents were from Sigma-Aldrich (St Louis, MO, USA) unless otherwise stated. EL-4 T cells were cultured in RPMI 1640 medium with 10 mM HEPES, 10% fetal calf serum (CSL, Parkville, Victoria, Australia), 120 μg/ml penicillin, and 16 μg/ml gentamycin. Cells were pretreated with 10 μg/ml CHX for 30 minutes, and then stimulated with 10 ng/ml phorbol myristate acetate (Boehringer Mannheim, Mannheim, Germany) and 1 μM ionomycin (I; Sigma-Aldrich).
Total RNA was isolated using TRI-reagent, reverse transcribed as previously described and quantitative PCR (qPCR) amplification was performed with SYBR Green as previously described . Amplifications were performed in 384-well optical reaction plates (Applied Biosystems, Foster City, CA, USA) with a 7900 HT Fast Real-Time PCR System (at the ACRF Biomolecular Resource Facility, JCSMR, ANU) using SDS 2.2.2 software to analyze raw data. For mRNA expression, relative mRNA levels were calculated by normalizing Ct values to ubiquitin-conjugating enzyme E2D 2 (UBC) Ct values. For ChIP analysis, relative values were calculated by normalizing immunoprecipitated DNA to total input genomic DNA with subtraction of the no-antibody background. Primers used for ChIP assay detection were designed against the promoter region and 2 kb downstream of the TSS.
ChIP assays were performed as previously described with some modifications . Briefly, cells were harvested and crosslinked with formaldehyde. Cells were lysed and then sonicated using the Bioruptor (Diagenode, Liege, Belgium) to give fragments between 200 bp and 1,000 bp in length. Samples were pre-cleared with protein-A agarose/salmon sperm DNA beads (Upstate, Lake Placid, New York, USA) then immunoprecipitated with 2.5 μg anti-histone-H3 (Abcam, Cambridge, UK), 4 μg anti-acetyl-H3K9 (Upstate), 2 μg anti-trimethyl-H3K4 (Abcam), 8 μg anti-trimethyl-H3K27 (Upstate), 4 μg anti-trimethyl-H3K36 (Abcam), and 6 μg of anti-RNA polymerase II CTD repeat (Abcam, ab817).
Three biological replicates for each treatment were carried out for the expression profiling experiments. Affymetrix Mouse Gene 1.0ST arrays were used as per the manufacturer's instructions. Quantile normalization and robust multichip average (RMA) background correction adjusting for probe sequence (Partek Software, St. Louis, Missouri, USA) was used to generate gene expression levels from the Mouse Gene 1.0ST arrays and an ANOVA test was used to identify genes induced or not induced ('unchanged') with P/I stimulation. Primary response genes were identified as genes with significantly (P < 0.016 equivalent to a FDR <0.1) higher expression in P/I-stimulated cells than in unstimulated cells and these genes also had to have higher or equal expression in CHX-treated P/I-stimulated cells than in just P/I-stimulated cells; that is, genes whose expression increased with stimulation and the increase was not inhibited by CHX. Secondary response genes were identified as genes with significantly (P < 0.016) higher expression in stimulated cells (than unstimulated), whose expression was also significantly (P < 0.024, FDR <0.1) decreased in the CHX-treated P/I-stimulated cells compared to the P/I stimulated cells; that is, genes where expression increased with stimulation and the increase was inhibited by CHX. Genes with P-values > 0.1 for all factors were classified as unchanged genes. Only genes whose promoter was represented on the DNA tiling array were used for further analysis. Groups were then subdivided further depending on their average basal expression level. Raw and normalized data have been deposited in the NCBI Gene Expression Omnibus (GEO) under accession number [GEO:GSE13278].
ChIP DNA was amplified with the Whole Genome Amplification kit from Sigma as per the manufacturer's instructions but with incorporation of dUTP. DNA was fragmented and hybridized to Affymetrix Mouse Promoter 1.0R arrays as per the standard Affymetrix protocol. Three biological replicates were used for data for unstimulated cells and for the effects of stimulation on H3K9ac levels two replicates each for 0-h, 0.5-h and 4-h stimulation were used.
All analysis was performed with NCBI build 36 of the mouse genome. The MAT algorithm  was used to find regions of H3K9ac with a bandwidth of 250 bp and a maximum gap of 150 bp. Either the matching total input DNA or H3 values were used as control samples. A MAT (-10 log p value) region cutoff score of >35.2 corresponds to a FDR <0.05 when three replicates were used (for analysis in unstimulated cells) and a FDR <0.1 when two replicates were used (for analysis of changes with stimulation). The same parameters were used to detect regions of antibody non-specific binding using the no-antibody sample arrays. Raw and normalized data have been deposited in GEO under accession number [GEO:GSE13277].
The TSSs for the genes on the 1.0ST array were obtained from the Affymetrix annotation file for build 36. Duplicate genes were not removed as often they represented pseudo-genes on different chromosomes. There is minimum gene duplication on the 1.0ST array and preliminary analysis indicated that removal of replicates does not affect results. R was used to detect if the regions of H3K9ac overlapped with the promoters of the genes in the expression groups. Where more than one region overlapped with a gene promoter, the region with the highest score was used. The promoter was considered to be from -1.2 kb (upstream) to +0.6 kb (downstream) of the TSS. Either Wilcoxon rank or Fisher exact tests were used to determine statistical significance (using R).
Genome-wide ChIP-Seq data were obtained from Wang et al.  and the matching expression data were downloaded from GEO [GEO:GSE10437]. The raw expression data were normalized using RMA normalization (with the Partek software) as described above as opposed to the original GCOS/MAS5 normalization used in the original paper. The gene co-ordinates for the human orthologs of the mouse genes were obtained using BIOMART. R was used to count the number of sequencing tags overlapping the regions of interest. For clustering, the logs of the sequence counts for the primary, secondary and unchanged genes with basal expression values Log2 3 to 6 were median centered and normalized using the Cluster software (Stanford). Heatmaps were generated by uncentered correlation, complete linkage clustering and viewed with TreeView (Stanford).
ChIP: chromatin immunoprecipitation; ChIP-on-chip: ChIP combined with microarray technology; ChIP-qPCR: ChIP with quantitative PCR; ChIP-Seq: ChIP with massively parallel sequencing; CHX: cycloheximide; FDR: false discovery rate; GEO: Gene Expression Omnibus; MAT: model-based analysis of tiling array; P/I: phorbol 12-myristate 13-acetate and ionomycin; Pol II: RNA polymerase II; qPCR: quantitative PCR; RMA: robust multichip average; TSS: transcriptional start site.
PL carried out the experimental studies and drafted the manuscript. KH carried out the data mining and analysis and drafted the manuscript. KB and CX are responsible for the preliminary studies that contributed to the conception of the study. LM prepared ChIP samples for ChIP-on-chip studies. PK carried out the microarray experiments. MS conceived the study and participated in its design and coordination and drafted the manuscript. All authors read and approved the final manuscript.
The following additional data are available with the online version of this paper: a figure of the density plot of basal expression values for different gene groups and the MAT score from no antibody ChIP-on-chip arrays (Additional data file 1); a figure of ChIP with H3 and H3K9ac in unstimulated EL-4 cells (Additional data file 2); a figure of mRNA expression levels of genes in the gene-focused studies (Additional data file 3); a figure of data mined from human CD4+ ChIP-seq experiments for H2AK9ac, H2BK20ac, H3K36ac and H4K16ac (Additional data file 4); a figure of data mined from human CD4+ ChIP-seq experiments for H3K4me2, H3K4me1, H3K27me2 and H3K27me1 (Additional data file 5); a figure of data mined from human CD4+ ChIP-seq experiments for H4K20me1 and H2BK5me1 (Additional data file 6); a table of the P-values for the epigenomic marks in the different basal expression groups (Additional data file 7).
Density plot of basal expression values for different gene groups and the MAT score from no antibody ChIP-on-chip arrays.
ChIP with H3 and H3K9ac in unstimulated EL-4 cells.
mRNA expression levels of genes in the gene-focused studies.
Data mined from human CD4+ ChIP-seq experiments for H2AK9ac, H2BK20ac, H3K36ac and H4K16ac.
Data mined from human CD4+ ChIP-seq experiments for H3K4me2, H3K4me1, H3K27me2 and H3K27me1.
Data mined from human CD4+ ChIP-seq experiments for H4K20me1 and H2BK5me1.
P-values for the epigenomic marks in the different basal expression groups.
We wish to thank Drs Adele Holloway, David Tremethick and Stephen Smale for useful discussions, and Dr Rohan Williams and Hugh French for assistance with R programming.