Global Analysis of the Relationship between JIL-1 Kinase and Transcription

Summary

The ubiquitous tandem kinase JIL-1 is essential for Drosophila development. Its role in defining decondensed domains of larval polytene chromosomes is well established, but its involvement in transcription regulation has remained controversial. For a first comprehensive molecular characterisation of JIL-1, we generated a high-resolution, chromosome-wide interaction profile of the kinase in Drosophila cells and determined its role in transcription. JIL-1 binds active genes along their entire length. The presence of the kinase is not proportional to average transcription levels or polymerase density. Comparison of JIL-1 association with elongating RNA polymerase and a variety of histone modifications suggests two distinct targeting principles. A basal level of JIL-1 binding can be defined that correlates best with the methylation of histone H3 at lysine 36, a mark that is placed co-transcriptionally. The additional acetylation of H4K16 defines a second state characterised by approximately twofold elevated JIL-1 levels, which is particularly prominent on the dosage-compensated male X chromosome. Phosphorylation of the histone H3 N-terminus by JIL-1 in vitro is compatible with other tail modifications. In vivo, phosphorylation of H3 at serine 10, together with acetylation at lysine 14, creates a composite histone mark that is enriched at JIL-1 binding regions. Its depletion by RNA interference leads to a modest, but significant, decrease of transcription from the male X chromosome. Collectively, the results suggest that JIL-1 participates in a complex histone modification network that characterises active, decondensed chromatin. We hypothesise that one specific role of JIL-1 may be to reinforce, rather than to establish, the status of active chromatin through the phosphorylation of histone H3 at serine 10.

Introduction

JIL-1 is a ubiquitously expressed, nuclear tandem kinase of Drosophila melanogaster that associates with chromatin at all stages of development. JIL-1 phosphorylates histone H3 in chromatin and the kinase is thought to be responsible for the majority of phosphorylation of histone H3 at serine 10 (H3S10ph) in interphase [1]. At the low level of resolution afforded by staining larval polytene chromosomes JIL-1 is seen to associate with active chromatin such as the decondensed interbands of chromosomes [2]. Ectopic recruitment of JIL-1 to a series of Lac operator repeats by fusion to a lacI DNA binding domain leads to local decondensation of polytene chromatin. This reorganisation of chromatin depends on JIL-1's kinase activity [3]. The essential function of JIL-1 can be explained, at least in part, by the involvement of the kinase in chromosome organisation [4], [5], where it plays a role in maintaining the balance between euchromatin and heterochromatin [6]–[8]. This is supported by the fact that Su(var)3-1 alleles of the JIL-1 gene are strong suppressors of position effect variegation, which provides a sensitive assay for heterochromatin propagation [9]. In the absence of JIL-1 the H3K9me2 mark and HP1 are redistributed genome-wide and tend to be enriched on the X chromosome in both males and females [6]. In addition, the lethal loss-of-function mutation of JIL-1 is rescued by reduced levels of the H3K9 methyltransferase Su(var)3-9 but not by reduced levels of the major HP1 isoform HP1a [10] suggesting that the spreading of H3K9me2 might be the cause of the lethality of the JIL-1 mutants. However, reducing the dose of Su(var)3-7, another essential component of heterochromatin, also rescues the lethality of the JIL-1 mutant. In this case the redistribution of H3K9me2 was still observed [11]. Su(var)3-7 may thus be an essential effector in the pathway of heterochromatin spreading, which is counteracted by JIL-1.

In mammalian cells the H3S10ph mark has been implicated in transcriptional activation, notably in immediate early response to mitogen, stress or steroid hormones, which trigger the transient phosphorylation of H3S10 and H3S28 by the kinases Msk1/2 or PIM-1 (for review see [12]. Whereas some of the inducing effects involve the dissociation of euchromatic HP1 isoforms [13], [14], the H3S10ph mark also serves as docking site for 14-3-3 proteins, which in turn recruit additional activating enzymes [13], [15], [16]. In D. melanogaster, the role of JIL-1 and the H3S10ph mark in transcription regulation is less clear. Corces and colleagues reported that upon heat shock gene activation, H3S10 phosphorylation by JIL-1 leads to recruitment of 14-3-3 and of the elongator protein 3 (Elp3) [17] that in turn would trigger the transition of an initiated RNA polymerase II (Pol II) into the elongation mode [18]. Johansen and co-workers by contrast argued that JIL-1 and JIL-1-dependent Ser10 phosphorylation did not have a direct function in heat shock gene activation [19].

We are interested in the potential involvement of JIL-1 in the process of X chromosome dosage compensation in D. melanogaster, which is hypothesized in light of a number of intriguing links between JIL-1 and the compensated male X chromosome. The process of dosage compensation serves to increase the transcription of genes on the single male X chromosome approximately twofold to match the combined transcription of the two alleles on the female X chromosomes [20], [21]. JIL-1 is enriched on the male X chromosome and this enrichment depends on a functional DCC [2], [22]. Severe reduction of JIL-1 levels leads to a global change in chromatin structure that is best appreciated on polytene chromosomes [1]. The euchromatic interbands appear much reduced in the absence of JIL-1 and replaced by atypical chromatin, in agreement with the above-mentioned role of JIL-1 in restricting the spread of heterochromatin. Interestingly, the dosage-compensated X chromosome shows an enhanced sensitivity to reduced JIL-1 levels, and appears diffuse and lacking polytene banding pattern [1], [4]. The Dosage Compensation Complex (DCC, also known as Male-Specific-Lethal [MSL] complex) associates along the transcribed regions of target genes and may therefore boost transcription elongation [23], [24]. The activation of genes on the male X chromosome involves the DCC subunit MOF on the X chromosome, an acetyltransferase, which selectively acetylates histone H4 at K16 (H4K16ac) [25], [26]. H4K16ac opposes the folding of the nucleosomal fibre in vitro and should render chromatin more accessible [27]. Accordingly, chromatin from the male X chromosome is released more efficiently from nuclei after shearing [28]. However, the precise twofold enhancement of transcription cannot be explained by H4K16ac alone [26].

We now mapped for the first time at high resolution the chromosomal distribution of JIL-1 kinase in SL2 cells using chromatin immunoprecipitation (ChIP) and hybridisation of associated DNA to oligonucleotide tiling microarrays. JIL-1 binding was compared to the distribution of elongating RNA polymerase II and related to steady-state mRNA levels. We also compared them to profiles of a number of histone modifications, and of DCC subunits. The vast majority of active genes is marked by a basal level of JIL-1 on gene bodies, independently of the actual transcription rates. The presence of H4K16ac on X-linked genes correlates with roughly twofold elevated levels of JIL-1. JIL-1 thus marks transcribed genes, but it also senses dosage-compensated chromatin. The phosphorylation of H3S10 is enriched on active genes in the same manner as JIL-1 itself. Perhaps surprisingly, the depletion of JIL-1 by RNA interference (RNAi) has only mild effects of transcription, which suggests that JIL-1 is not absolutely required for transcription in general or for dosage compensation in particular. In the light of the previous genetic analysis our data suggest that JIL contributes to a composite histone mark that may serve to reinforce the active state of chromatin by preventing the inappropriate association of silencing factors.

Results

JIL-1 binds most transcribed genes and is enriched on dosage-compensated genes

We raised two polyclonal antisera (R69 and R70) against a fragment (amino acids 79–571) of JIL-1 and established their specificity. In agreement with previous results [22], we confirmed that JIL-1 is a nuclear protein that associates with active, decondensed chromatin represented by interbands on larval polytenized chromosomes (Figure S1). We also visualized the clear enrichment of JIL-1 on the dosage-compensated, hyperactive male X chromosome. This enrichment on the X chromosomal territory can also be seen in nuclei of the Drosophila SL2 cells, which have a male genotype. By contrast, nuclei of female KC cells display a homogenous staining (Figure S1).

The enrichment of JIL-1 on the male X chromosome and the conflicting reports about the role of JIL-1 in the control of early transcription elongation of heat shock genes [17]–[19] prompted us to generate a high-resolution in vivo chromosomal interaction profile of the kinase and to compare it to elongating RNA polymerase II (ePol). We combined chromatin immunoprecipitation (ChIP) with probing oligonucleotide tiling arrays as before [24], [26], [29], using the JIL-1 antibodies and a well-established monoclonal antibody specific for ePol (H5). This antibody recognises phosphorylated serine 2 in the heptad repeat of the C-terminal domain (CTD) of the largest RNA polymerase II subunit. Data from four biological replicates starting with independent chromatin preparations were combined for each profile. The two JIL-1 antibodies gave very coherent results (Figure S2). We also generated a profile for histone H4 acetylated at K16 (H4K16ac). Figure 1 displays the profiles of JIL-1, ePol and H4K16ac along representative 250 kb portions of the chromosomes X and 3R (Figure 1A, 1B, respectively). The profiles resemble the distribution of histone H3 methylated at lysine 36 (H3K36me3), a modification that is placed co-transcriptionally by the CTD-associated methylase Set2 and therefore serves as a hallmark of transcribed chromatin [30], [31].

Qualitatively, the distribution of JIL-1 and ePol were similar on the X chromosome and on autosomes with a preference for coding sequences. In general, binding of JIL-1 to introns is lower than binding of ePol (Figure 1C). As expected, ePol is mainly found at active genes as assessed by Affymetrix expression profiling (Figure 1D) and there is a good correlation between the density of ePol along genes and the corresponding steady-state mRNA level (Figure S3). The distribution of JIL-1 correlates well with several subunits of the DCC on the X chromosome: at the probe level, the best correlations are observed with the targeting subunit MSL1 [24] and with the spreading subunit MOF, which is the acetyltransferase responsible for H4K16 acetylation (Figure S4). On the X chromosome 1413 genes are detectably transcribed according to ePol occupancy. Among those, the vast majority (93%) have both JIL-1 and MSL1 bound, a minority of 55 genes (4%) show no detectable JIL-1, 86 genes (6%) have no MSL1 bound, and 46 genes (3%) have neither JIL-1 nor MSL1 bound. Remarkably, the densities of ePol on active genes are comparable on autosomes and on the X chromosome (Figure 1D), although the latter ones benefit from dosage compensation (see discussion). By contrast, the mean density of JIL-1 on active genes is 0.92 for autosomal and 1.71 for X-linked genes, thus the calculated enrichment of JIL-1 on X chromosomal genes relative to autosomal ones is about 1.86 (Figure 1E). These results quantitatively reproduces the picture obtained from polytene chromosome staining (Figure S1), where a twofold enrichment had been estimated [22].

The enrichment of JIL-1 at various loci was confirmed by quantitative PCR (qPCR) on the non-amplified immunoprecipitate. Figure 2A shows as example the X chromosomal gene pole hole (phl) where JIL-1 accumulates towards the 3′ end of the gene, together with H4K16ac and H3K36me3. We also confirmed the male-specific enrichment of JIL-1 on several X-linked genes by qPCR of immunoprecipitates of chromatin prepared from sorted male and female flies (Figure 2B). The X-linked genes phl, CG14804, sta, Pgd and RpII215 show elevated levels of JIL-1 in males as compared to females. As an example for autosomal genes, we chose the Cortactin gene, which is similarly bound by JIL-1 in males and females. The H3K36me3 levels at the same genes are rather similar in both sexes, whereas H4K16ac is clearly found there only in males.

We conclude that JIL-1 binds active genes in both sexes and exhibits a twofold enrichment on X chromosomal genes in males.

H3 phosphorylation by JIL-1 contributes to a composite phospho-acetyl mark associated with active chromatin

In order to characterise the potential of JIL-1 to phosphorylate histones we carried out in vitro kinase assays using baculovirus-expressed FLAG-tagged JIL-1. Recombinant histone H3 was phosphorylated, but not histones H2A, H2B, H4 (Figure S5A). Histone H1 prepared from Drosophila embryos, which can be phosphorylated to some extent at serine 10 [32] was not a substrate for JIL-1 in vitro (not shown). We did not detect phosphorylation of H3 in nucleosomes even by increasing the salt concentration in the assay (Figure S5A). A JIL-1 derivative, in which the catalytic domain was impaired by mutating an aspartate to alanine, had no H3 kinase activity (JIL-1^D392A; [33]). We used H3-derived peptides to characterise the substrate specificity further and found that JIL-1 was unable to phosphorylate S28 in the context of the H3_21-34 peptide but efficiently phosphorylated the peptide H3_1–21. The phosphorylation occurred at serine 10 (S10) since it was abolished by incorporation of phospho-serine during peptide synthesis. At saturating amounts of peptide, neighbouring modifications of the H3 tail do not impair S10 phosphorylation (Figure S5).

H3S10 phosphorylation, particularly in the context of K14 acetylation (H3S10phK14ac) had previously been shown to be enriched on the male X chromosome [1]. By ChIP the enrichment of the straight H3S10ph epitope is difficult to document (a 1.09-fold enrichment in the data of Zhang and Oliver [28]), presumably due to the considerable amount of H3S10ph contributed by aurora kinase in mitotic cells (for review, see [34], [35]. Consistent with this, JIL-1 depletion by RNAi in SL2 cells did not lead to reproducible decrease of global H3S10ph levels (Figure S6). In order to estimate the contribution of mitotic H3S10ph in asynchronously growing SL2 cells, we arrested cells in G1/S by combined aphidicholine and hydroxyurea treatment and found the level of H3S10ph 8-fold reduced as compared to asynchronously growing cells. In the presence of the potent aurora kinase inhibitor ZM447439 [36] the global H3S10ph signal was 20 times reduced (Figure S6C). Since the proportion of mitotic SL2 cells in a culture has been estimated to be 4% [37], we conclude that the level of interphase H3S10ph is two orders of magnitude lower than that of mitotic H3S10ph. This may be an overestimate since an H3S10ph profile from ZM44743-treated cells showed very little signal over background (Figure S6E).

The detection of interphase H3S10ph in salivary glands has been shown to greatly depend on the antibody itself as well as on the procedure [19]. In particular, some antibodies may be occluded by neighbouring modifications in the H3 tail and not recognize S10ph as part of a composite epitope [38]. The two antisera used in this study, anti-H3S10ph [39] and anti-H3S10phK14ac [40] were therefore tested for their ability to recognise the S10ph mark in the presence of neighbouring modifications on the H3 tail using a custom-made peptide microarray (Figure S7). Interaction of the anti-S10ph serum, but not of the anti-S10phK14ac serum was adversely affected by the K14 acetylation. However, the latter also detected S10ph in the absence of K14 acetylation. This may explain why upon JIL-1 depletion by RNAi in SL2 cells, no decrease in H3S10phK14ac was detected by Western blotting (Figure S6D). Because the H3S10phK14ac epitope had been found enriched on the X [1] and the anti-H3S10phK14ac had been successfully used in ChIP experiments [40] we generated a global ChIP profile with the latter antibody (Figure 3A). A view of this profile along 140 kb of the X chromosome showed that H3S10phK14ac coincides with the presence of JIL-1. In addition, the H3S10phK14ac signal was enhanced on X chromosomal probes (Figure 3B) and preferentially mapped to the CDS of genes. Average binding profiles revealed that an enrichment of H3S10phK14ac throughout the transcription units can be observed in particular on X chromosomal JIL-1 target genes (Figure 3C). The global depletion of H3S10phK14ac could not be assessed by Western blot (Figure S6D), given the cross-reactivity of the antibody with (mitotic) H3S10ph but, the enrichment at the 3′ end of tested target genes is reduced upon JIL-1 RNAi (Figure 3D). Taken together, the data suggest that JIL-1 contributes to generating the composite S10phK14ac modification on histone H3 at active genes.

JIL-1 increases the transcription efficiency of genes on the male X chromosome

The H3S10ph and H3S10phK14ac marks were correlated with active transcription in various model systems. In order to assess the contribution of JIL-1 to transcriptional output we reduced the levels of the kinase by RNAi in SL2 cells. Efficient depletion (up to 95%) was reproducibly achieved while ePol remained unaffected (Figure 4). This observation is in support of the results from the Johansen laboratory on salivary glands of the JIL-1 null mutant [19]. Removal of JIL-1 did not visibly affect the association of the DCC with the X chromosomal territory, which in Figure 4C is visualised by the diagnostic H4K16ac mark. Although flies with decreasing JIL-1 levels show increasing lethality [1] the proliferation of SL2 and KC cells appears only slightly affected (proliferation was monitored for up to 19 days in RNAi conditions, not shown).

Comparison of steady-state mRNA levels in JIL-1-depleted versus control cells were assessed by Affymetrix profiling. Two sets of experiments each consisting of three biological replicates were conducted with different sets of dsRNA. The statistical analysis of effects incorporates the results of all experiments. Among several thousand active genes bound by JIL-1 in SL2 cells, we found that after JIL-1 depletion the expression level was significantly reduced for 276 genes and increased for 25 genes. Globally, transcription of X-linked genes was significantly reduced as compared to the autosomal genes (p-value 1.2e-06) suggesting that the enrichment of JIL-1 on X-linked genes is functionally relevant for dosage compensation (Figure 4D). Interestingly, the few active genes on the fourth chromosome also appear to be particularly affected. We observed that total amount (Figure 4B) and distribution of ePol (not shown) were not changed after RNAi against JIL-1 and suggest that in SL2 cells, JIL-1 is probably not essential for the release of RNA polymerase II in productive elongation. Alternatively, the few percent of JIL-1 remaining after RNAi may be sufficient to fulfil this function. It is also noteworthy that, contrary to many proteins having a function in transcriptional activation, those genes whose transcription is sensitive to JIL-1 depletion tend to have lower JIL-1 levels to start with (Figure 4E).

JIL-1 association to genes does not follow ePol

To compare the distribution of ePol and JIL-1 along genes, we scaled all genes to equal length and aligned them from transcriptional start sites (TS) to transcriptional termination sites (TT). The genes were furthermore divided into six equally sized groups based on their average ePol enrichment and average distribution profiles were computed for each group (Figure 5A upper panels). The distribution of ePol along the transcribed regions is similar for autosomal and X-linked genes with a modest enrichment towards the 3′ end. Monitoring the binding of JIL-1 along the gene bodies as a function of ePol density, we found that while JIL-1 clearly associated with active genes, its binding was not proportional to ePol levels (Figure 5A centre panels). Rather, three out of the four groups with significant association of ePol showed the same average binding of JIL-1, except that there was roughly twice as much JIL-1 on X-chromosomal genes compared to autosomal genes, as observed before. This property was also illustrated when the density per gene was plotted against the corresponding steady-state transcript levels. Whereas the ePol density and transcript levels correlated well, JIL-1 binding was characterized by two clouds representing autosomal and X chromosomal genes (Figure S3B). These two representations show that the density of JIL-1 on active genes is not proportional to their transcription level. This was also observed for H3K36me3 and MSL1 distributions using the same representations (Figure S3C, S3D; Figure S8B, S8C) whereas the density of H3K4me2 and ePol were roughly proportional (Figure S8D). The discordance between ePol and JIL-1 distributions along gene bodies was even more pronounced when monitored as a function of gene length (Figure 5B). On autosomes, ePol was found enriched towards the 3′ end of short genes, but for genes longer than 1.8 kb the maximal density of ePol gradually shifted towards their 5′ ends as a function of length. This is also illustrated by visual inspection of very long genes both on the X and on autosomes (Figure S9). It is conceivable that the risk of premature transcription termination increases with gene length and thus leads to relative depletion of ePol at 3′ ends. Interestingly, this trend was less pronounced on the X chromosome, in keeping with the attractive hypothesis that dosage compensation facilitates transcriptional elongation through chromatin. By contrast, the relative distribution of JIL-1 along genes of various lengths is comparable on autosomes and the X chromosome, except for the twofold enrichment on the latter (Figure 5B).

JIL-1 recruitment upon gene activation

A recent suggestion that JIL-1-dependent H3S10 phosphorylation was crucial to regulate the release of RNA polymerase into productive elongation in Drosophila in particular for the transcription of heat shock genes has remained controversial [17]–[19]. In our hands, heat shock treatment for 2, 5 and 10 min in SL2 cells triggered a clear enrichment of ePol at the Hsp70 locus as monitored by ChIP without strikingly affecting the distribution of JIL-1. In particular, we found that there was already some ePol and some JIL-1 at that locus in control cells as evidenced by the comparison with the active X-linked locus sta (Figure S10). Western blotting showed that the overall level of ePol drops upon heat shock, whereas the levels of JIL-1 varied only slightly. Because heat shock had been shown to trigger a massive loss of nucleosomes over a large domain [41], and because the genome-wide localization of JIL-1 appeared to correlate best with histone modifications (see below), we speculate that the lack of recruitment might be a consequence of nucleosome depletion.

We next made use of an inducible reporter system in flies ([26], [42]; Figure 6A), which harbours an insertion of a reporter gene cassette consisting of the lacZ gene downstream of five binding sites for the yeast transcription factor Gal4 (‘upstream activating sequences’; UAS_Gal). The reporter was activated by crossing the reporter fly line with a line expressing either Gal4 alone or MOF fused to the Gal4 DNA binding domain (Gal4_DBD). The basal level of β-galactosidase expression in the absence of activation was very low and could be induced about 1000-fold in male and female flies by expression of Gal4 under the control of a tubulin promoter. We monitored the levels of JIL-1, H3K36me3 and H4K16ac at the 5′ and 3′ ends of the lacZ gene by ChIP. As reference for the normalisation of the ChIP signals from different chromatin preparations we used the intergenic locus close to ptr (ptr_IG). In accordance with the results presented in Figure 2C, the X-chromosomal gene phl showed robust enrichment of JIL-1 in female flies and approximately twofold elevated levels in males. Very little recruitment of JIL-1 was observed at the reporter gene upon activation via Gal4 (Figure 6B). However, we did also not detect robust levels of H3K36me3 as compared to the endogenous loci. Since the tubulin-Gal4 driver leads to strong activation, which can be visualized by clear puffing at the level of polytene chromosomes, it is possible that these histone marks have been depleted along with nucleosomes due to vigorous transcription, a situation akin to the heat-induced transcription mentioned earlier. In order to trigger a more modest induction of the reporter gene we expressed the acetyltransferase MOF fused to the Gal4_DBD (Gal4-MOF) at a low level from its endogenous promoter [26]. We limited this analysis to females in order to avoid the complications associated with all other members of the DCC. We had shown earlier that recruitment of Gal4-MOF to the reporter locus in females leads to H4K16ac and to a substantial transcriptional activation [26]. In the current experiment the activation was determined to be about 20-fold in the β-galactosidase assay. This activation was accompanied by an enrichment of JIL-1 at the 3′ end of the lacZ reporter (13-fold over the intergenic locus ptr_IG), which was about twofold higher than the enrichment at the endogenous X-linked locus phl (Figure 6C). In this case of a mild activation, we found an enrichment of H3K36me3 at the 3′ of the LacZ reporter, which was comparable to the endogenous loci phl and Cortactin. Upon induction with MOF, a 7-fold enrichment of H4K16ac is detected at the 5′ end of the LacZ reporter, which decreased towards the 3′ end. The Cortactin gene, localised about 2-3 kb downstream of the UAS_GAL as mapped by PCR-rescue of the transgene, also showed an increased level of H4K16ac upon Gal4-MOF targeting. Interestingly, the presence of H4K16ac in addition to H3K36me3 on the reporter gene LacZ correlated with an enrichment of JIL-1 comparable to that of X-linked genes in males. On the neighbouring Cortactin gene the enrichment was also increased. We found that the enrichment of both H4K16ac and JIL-1 at the Cortactin gene only increased the expression of the gene by a factor 1.2 in females. In accordance with previous data [26], in males H4K16ac spreads out more broadly from the recruitment sites due to the presence of the DCC and the relative expression of the Cortactin gene was increased by a factor 1.6 (Figure S11). Taken together the recruitment experiments suggest that the association of JIL-1 with active genes rather correlates with the presence of H3K36me3 and H4K16ac than with transcription as such. This conclusion can be substantiated by visual inspection of the genome-wide ChIP-chip profiles from SL2 cells, where numerous instances can be found where JIL-1 binding closely parallels the presence of H3K36me3, whereas the presence of high levels of ePol alone and sometimes also significant enrichment of H4K16ac alone do not coincide with JIL-1 binding (see Figure S9 for a selection of views).

JIL-1 recruitment and dosage compensation

In order to further explore the relationship between JIL-1, transcription, H3K36me3 and H4K16ac, we looked in more detail at the relative twofold enrichment of JIL-1 on the male X chromosome. Johansen and colleagues [22] had documented earlier that the enrichment of JIL-1 on the male X chromosome depends on the DCC. Their key experiment was the ectopic expression of MSL2 in female larvae, which leads to inappropriate assembly of the DCC. The association of the DCC with the female X chromosome led to an enrichment of JIL-1, just like in males [22]. In agreement with these data, in similar experiments we found that if high levels of MSL2 are expressed from two NOPU alleles [43], [44] the X chromosome is decorated with DCC and JIL-1 is enriched concomitantly at sites of DCC binding. However, when limiting levels of MSL2 are expressed from the SXB1-2 allele [43], [44] the DCC only associates with about 50 high affinity sites (HAS) on polytene chromosomes. In this case, the enrichment of JIL-1 can only be seen at sites of DCC binding (Figure 7A). This enrichment requires the assembly of a complete DCC, including MOF, as it is not observed in the absence of MLE or MSL3 ([22] and data not shown).

The close relationship between DCC association and JIL-1 was also apparent at the high resolution achieved by ChIP-chip mapping. We had earlier identified 131 HAS that are preferentially bound by limiting amounts of DCC [29]. Plotting the binding of MSL1 to genes as a function of their distance to the closest HAS shows that, in general, the closer a gene resides to a HAS, the more MSL1 and H4K16ac can be found (Figure 7B). The same trend can be observed for JIL-1 as well as for the linked H3S10phK14ac mark: their densities are decreasing with increasing distance to a HAS. This is, however, not true for H3K36me3 or ePol association (Figure 7B).

In summary, diverse experiments establish two hallmarks of active chromatin that correlate with the presence of JIL-1. First, there is an excellent correlation of JIL-1 binding to H3K36me3-marked chromatin. On top of this basal level of binding, the presence of the DCC leads to an approximately twofold enrichment of JIL-1. Indeed, a global comparison of ChIP-chip probe signals (Figure 8) reveals that the correlations of JIL-1 signals with the ones of H3K36me3 (R = 0.758) or H4K16ac (R = 0.76) are very good. Remarkably, the combination of both marks (sum of the intensities of the H3K36me3 and H4K16ac signals) correlated even better with JIL-1 signals (R = 0.899) and displayed an almost linear relationship between signal intensities. This is true if all probes are considered, but also if autosomal or X chromosomal probes are plotted separately (Figure 8). This correlation can be recapitulated to some extent when the distribution of those marks is considered along genes (Figure S12). In contrast, correlation of JIL-1 and ePol (R = 0.69) is less pronounced (Figure 8). This result is consistent with the earlier observation that ePol and JIL-1 are both marks for active genes but that their relative distributions along genes differ (Figure 5).

We conclude that two distinct targeting principles that are represented by H3K36me3 and H4K16ac marks determine the level of JIL-1 recruitment to active genes.

Discussion

Targeting of JIL-1

The genome-wide mapping of JIL-1 and ePol in male SL2 cells together with the reporter assay in flies suggest that JIL-1 is part of a network of factors that collectively define the state of active chromatin. Taken together with previous genetic analyses, our data lead us to consider to a role for JIL-1 not in the establishment of active chromatin, but in the reinforcement of the active state, independently of the extend of transcription.

Although JIL-1 binds the bodies of most active gene and is recruited to a reporter gene upon activation, the degree of binding does not correlate well with ePol occupancy. In this respect JIL-1 belongs to a class of several other active chromatin components whose presence in chromatin does not scale with ePol, such as H3K36me3, the DCC subunit MSL1, or H4K16ac mark. By contrast, H3K4me2 levels are proportional to ePol levels. Conceivably, JIL-1 marks active chromatin independent of whether it is currently transcribed or not. A basal level of JIL-1 binding correlates strongly with H3K36me3, an elongation marker that is placed co-transcriptionally by ePol-associated dSet2 [30]. The turnover time of this modification is not known and hence it is possible that H3K36me3 remains on chromatin between pulses of transcription.

Most recently, high-resolution mapping of 56 chromatin components using Dam-ID revealed five types of chromatin in Drosophila KC cells [45]. A colour code was used to illustrate those types. The two types of active chromatin, ‘red’ and ‘yellow’, differ in that active genes in red chromatin tend to be depleted in H3K36me3 and MRG15, which are abundant constituents of yellow chromatin. Monitoring the relative distributions of ePol and JIL-1 in those types of chromatin we found that whereas ePol is equally distributed in red and yellow chromatin, JIL-1 is clearly mostly found in yellow chromatin (Figure S13). This correlation again highlights the relationship between JIL-1 and H3K36me3. The precise link between H3K36me3 and JIL-1 recruitment remains to be explored (see below).

A second targeting principle is evident from the fact that JIL-1 levels are approximately twofold increased on the male X chromosome, where due to the action of the DCC H4K16ac levels are high. The situation is complex since the recruitment of the DCC to transcribed genes on the X chromosome is promoted by the potential interaction of the DCC subunit MSL3 with H3K36me3 [30]. Because JIL-1 binding is insensitive to different transcription rates on autosomes, it is not plausible that a presumed twofold increase in transcription on the X chromosome is directly responsible for this elevated association. Likewise, we do not think that direct interactions between JIL-1 and the DCC can explain the targeting. Although some interactions have been observed in vitro between JIL-1, MSL1 and MSL3 [22], a quantitative co-purification of endogenous JIL-1 with the DCC was never documented. Indeed, activation of a reporter gene in female flies through recruitment of Gal4-MOF is sufficient to recruit JIL-1 to male levels on the reporter gene and on the adjacent Cortactin gene. Altogether, we favour the hypothesis that the enrichment of JIL-1 on X-linked genes is a consequence of a modulating feature of dosage-compensated chromatin. The genome-wide distribution of JIL-1 indeed correlates best with the combination of the two chromatin modifications, H3K36me3 and H4K16ac, on the X-chromosome suggesting that JIL-1 recruitment occurs downstream of those two histone marks. Interestingly, the correlation still holds for autosomal probes, although autosomal H4K16ac is at least in part placed at the 5′ end of genes by alternative MOF complexes [26], [46], [47]. JIL-1 recruitment has also been linked to the H4K12ac mark, which is placed by the GCN5-containing ATAC complex [48]. However, Workman and colleagues recently found that ATAC also contains an acetyltransferase with specificity for H4K16 (atac2), which may contribute to JIL-1 recruitment [49].

In the cases of very strong transcription activation by high levels of Gal4 activator, we did not see an accumulation of JIL-1 to endogenous levels. Conversely, JIL-1 is also not enriched at developmental puffs on polytene chromosomes where robust transcription takes place according to the strong ePol staining (Figure S1). The massive heat-induced activation leads to extensive chromatin decondensation (puffing) on polytene chromosomes, which is probably accompanied by nucleosome loss [41].

In summary, we favour the idea that the recruitment of JIL-1 relies on several, partly redundant features of active chromatin. JIL-1 harbours an H3-tail binding domain in its C-terminus and another determinant of chromatin targeting in its N-terminal domain [3], but it does not contain any domain known to directly bind modified histone tails. Its recruitment to active chromatin is, therefore, most probably indirect. We know of no histone modification or any other chromatin-associated feature to be enriched on the male X chromosome except for JIL-1 and the JIL-1-dependent phospho-acetyl mark. An appealing model for JIL-1 targeting is that its general recruitment mode (correlating with H3K36me3) would be the same on all chromosomes. The additional presence of H4K16ac might enhance the access of JIL-1 to the first targeting principle, since it has been shown that H4K16ac prevents the folding of the nucleosomal fibre into more compact structures [27].

Effects of H3S10 phosphorylation by JIL-1

It has been observed that histone H3S10ph can enhance acetylation of histone H3K14 [50], [51] and inhibit methylation of histone H3K9 [52]. The biochemical analysis of JIL-1 kinase showed that a number of other modifications of the H3 N-terminus are compatible with the phosphorylation by JIL-1. Therefore, JIL-1 phosphorylation is compatible with the prior existence of other modifications, consistent with a downstream function of JIL-1.

Our data suggest that in Drosophila the function of H3S10ph may at least in part differ from that in other organisms, where it is implicated in the fast and transient induction of promoters in response to various inducers [13], [16], [53], [54]. In Drosophila, JIL-1 is not particularly enriched at promoters, but associates with genes along their entire length. This distribution is more compatible with a role in fine-tuning transcriptional elongation. The effects of JIL-1 may – at least in part – be mediated by 14-3-3 proteins, which are able to recognize the interphase H3S10ph mark [17], [54].

The depletion of JIL-1 has a clear but small effect on overall transcription. The fact that JIL-1 binds gene bodies rather that promoters and the the binding is independent of actual transcription rates leads us to speculate that the kinase may influence transcription efficiency only very indirectly by reinforcing the active chromatin state – once established by co-activators – by protecting it from neighbouring repressive chromatin. This could become important when moderately active genes are not constantly transcribed. Highly transcribed genes experience nucleosome depletion and therefore do not provide anchoring sites for silencing factors. Moderately transcribed genes, however, may experience periods without active elongation [55]-[57]. These ‘gaps’ bear the risk of placement of silencing marks, most notable H3K9me2. In this context JIL-1 may safeguard active genes (marked by H3K36me3 as a sign of recent transcriptional activity). This function may be achieved by phosphorylation of H3S10, which prevents the recognition of HP1 by occlusion of H3K9me2 [58], [59] but possibly also by phosphorylation of other substrates, like Su(var)3,9 [33]. In addition JIL-1 might have a scaffolding function for other factors, like lamins [60]. The observation that genes sensitive to the depletion of JIL-1 tend to have less JIL-1 bound both on the X chromosome and on autosomes also support a model where JIL-1 is associated to active chromatin domains to prevent the spreading of more repressive structures. We speculate that the role of JIL-1 in maintaining the balance between heterochromatin and euchromatin is not only true at the hetero/euchromatin boundary, but also within euchromatin at the level of active chromatin domains.

Materials and Methods

JIL-1 antibodies and kinase assay

Polyclonal antisera were raised in two rabbits against a fragment of JIL-1 corresponding to amino acids 79–571, which was expressed in fusion with an N-terminal MBP-tag in E. coli (Eurogentec). The resulting sera R69 and R70 were either used crude or after affinity purification on a column with covalently coupled antigen according to a standard protocol. For Western blot, the sera were diluted 1/2000 and affinity-purified antibodies were used at 0.5 µg/ml. Recombinant Flag-tagged JIL-1 and kinase assay conditions were as previously described [33].

Reporter lines and β-galactosidase activity measurements

The transgenic line used for the reporter assay in flies was originally published as U/l5 [42]. The P-element insertion site was mapped by PCR-rescue to the 3′end of the Cortactin gene (Figure 6A). To activate the reporter by the Gal4 activator homozygote flies of the reporter line were crossed with males of the activator line, which carry a transgene driving the expression of the Gal4 activator under the control of the tubulin promoter (Bloomington Stock Center #5138). Gal4 is expressed from the cellular blastoderm onwards. The F1 flies were hand-sorted according to their genotype and sex in order to obtain 190–250 mg of flies for chromatin preparation. β-galactosidase activity in extracts prepared from 2–10 flies were measured as previously described [26]. The flies carrying the balancer were used as negative controls for the flies carrying the Gal4 activator. To activate the reporter by Gal4-MOF recruitment, we used a stable line homozygote for the reporter construct and prepared chromatin for ChIP from sorted 200–500 female flies [26].

Chromatin immunoprecipitation

Chromatin from SL2 cells was prepared after crosslinking in 1% formaldehyde as previously described [29]. To preserve the phospho-epitopes and the acetylation status of the chromatin, we used 5 mM NaF, 20 mM β-glycerophosphate, 0.1 mM NaVanadate, 10 mM NaButyrate and a mixture of protease inhibitors (1mM PMSF, 1 µg/ml of each aprotinin, pepstatin and leupeptin) or the complete protease inhibitors (Roche, Cat No. 04693132011). In the particular case of the chromatin used for mapping H3S10phK14ac, 1 µM microcystin was also added. The chromatin from sorted male and/or female flies was done according to following references [26], [61]. Chromatin shearing to 300–1000 bp fragments was assessed on a Bioanalyzer (Agilent) and DNA concentration was measured using NanoDrop (Thermo Scientific). 7.5 to 10 µg of DNA per chromatin was used for ChIP. For JIL-1 ChIPs we used either crude sera R69 and R70 (5 µl/IP) or the affinity-purified antibodies (5 µg/IP). A LiCl wash was done for ChIP-chip analysis. For ePol ChIPs, 5 µl of the anti-CTD S2ph monoclonal antibody Cone H5 (ab24758, Abcam) was first coupled to a 50/50 mix of protein A- and protein G-agarose using a bridging antibody (5 µg of Goat anti-mouse IgM, Jackson ImmunoResearch) and the LiCl wash was omitted. The following volumina of antibodies were used for ChIP: H3S10phK14ac: 3 µl; H4K16ac: 5 µl of antibody #39167 (Active Motif); H3K36me3: 5 µl of ab9050 (Abcam). MSL2 antibodies were used as control as previously published [62].

Quantitative PCR

Quantitative PCR (qPCR) was carried out in an ABI PRISM 7000 Sequence detection system in a 25 µl reaction with Power SYBR Green Master Mix (Applied Biosystems) or in LightCycler (Roche) in 10 µl reactions with Fast SYBR Green Master mix (Applied Biosystems). Input DNA were diluted to 10 ng/µl and ChIPs were diluted 1/30, 5 µl and 2 µl template were used for ABI PRISM 7000 and LightCycler, respectively. The results were comparable on the two machines. Primer sequences are provided in Table S1.

ChIP-chip and data analysis

ChIP-chip data analysis was essentially performed as in [29]. Briefly, input and IP DNA were amplified using the WGA kit (Sigma) according to the online protocol (http://www.epigenome-noe.net/WWW/researchtools/protocol.php?protid=30). Labeling and hybridization to NimbleGen dual-color arrays was carried out at ImaGenes (Berlin, Germany). The layout of the array (approx. 1 probe/100 bases, isothermal probe design) was customized [24]. Data analysis was performed using R/Bioconductor (www.bioconductor.org). Experimental replication was as follows: JIL-1: 4 biological replicates; H3S10phK14ac: 2 biological replicates including dye swap; elongating RNA Polymerase II: 3 biological replicates including one dye swap; H4K16ac: 2 biological replicates including dye swap. Raw signals of corresponding biological samples were log2 transformed and quantile-normalized. Enrichment statistics (IP versus input signals) were computed using the ‘sam’ algorithm within Bioconductor [63]. Fdr values of the sam statistic were determined using ‘locfdr’ [64] region summarization was performed using tileHMM [65] with the following parameters: fragment size of 700, maximal gap of 400. Genes were considered ‘bound’ if covered by at least 4 probes with a posterior probability above 0.5. All data correspond to Drosophila genome version dm3 and annotation version gadfly 5.2.2.

RNAi in SL2 cells and purification of total RNA

SL2 cells were cultured according to standard methods. RNA interference (RNAi) was carried out as published [66] except that 2×10⁶ cells were treated with 10 µg dsRNA and splitted 1/3 after 3 days of treatment. After 7 days cells were collected and counted prior to total RNA purification. For the 2 sets of expression analysis we used different sets of dsRNA for RNAi. In experiment 1, the 5′ coding sequence of JIL-1 was targeted by a corresponding dsRNA (T7_JIL-15′_for: TTAATACGACTCACTATAGGGAGATGAGTCGCTTG, T7_JIL-15′_rev: TTAATACGACTCACTATAGGGAGACATCCTCCTCCA), and the controls involved treating cells with dsRNA targeting GST sequences (T7_GST_for: TTAATACGACTCACTATAGGGAGAATGTCCCCTATACTAGGTTA, T7_GST_rev: TTAATACGACTCACTATAGGGAGAACGCATCCAGGCATTG). In the second set JIL-1 was knock- down by targeting its 3′ coding sequences (T7_JIL-13′_for: TTAATACGACTCACTATAGGGAGACAGCAGCGTCG, T7_JIL-13′_rev:TTAATACGACTCACTATAGGGAGATTGGAACTGAT) and the control dsRNAs represented GFP sequences (T7_GFP_for: TTAATACGACTCACTATAGGGTGCTCAGGTAGTGGTTGTCG, T7_GFP_rev:TTAATACGACTCACTATAGGGCCTGAAGTTCATCTGCACCA). Off-target effects were minimized using http://dscheck.rnai.jp/. Efficiency of the depletion was assessed on whole cell extracts and quantified with the Odyssey imaging system (LI-COR, Biosciences) using anti-tubulin (T9026, clone DM1A, Sigma) or anti-lamin (Clone T40, [67]) as reference. Total RNA was purified from 20-80×10⁶ cells using either the RNA purification kit (RNeasy mini kit, Qiagen) or the lysis reagent (Qiazol, Qiagen) according to manufacturers instructions. The quality of the RNA was determined on the Bioanalyzer (Agilent).

Expression analysis after JIL-1 RNAi

Total RNA was processed for hybridization at ImaGenes (Berlin, Germany). Five biological replicates for each control and JIL-1 RNAi were analysed. Microarray data were processed using R/Bioconductor (www.bioconductor.org). Unless indicated otherwise we used standard parameters in all function calls. Expression values were calculated using ‘gcrma’. Probe sets were kept for differential expression analysis if there were more ‘present’ calls (calculated using ‘mas5calls’) in one of the two treatment groups than non-‘present’ calls and if their expression level variance was higher than 0 across all arrays. “One gene to many probe set“ relationships were resolved by retaining only the probe set with the highest variance across all arrays. Differential expression statistics were obtained using a linear model (library 'limma) including a batch effect coefficient as two sets of arrays were processed at different times. A significant response was defined if the local false discovery (‘locfdr’ package) rate calculated on the moderated t statistic was smaller than 0.2.

Microscopy

Polytene chromosome staining was according to [44]. The affinity-purified antibodies R70 or R69 were used diluted 1/150 and the H5 monoclonal antibody at 1/50. Staining of SL2 cells was according to a standard protocol. In brief, after fixation with 3.7% (w/v) formaldehyde in PBS buffer, the cells were permeabilised with 0.3% triton X-100 (v/v) in PBS in the presence of 1% formaldehyde. The affinity-purified antibodies of R70 or R69 were diluted at 0.4 µg/ml, the anti-H4K16ac antibody #39167 (ACTIVE MOTIF) was diluted 1/1000 for counterstaining of the X-chromosome territory, and the anti-lamin clone T40 [67] was diluted at 1/100 to delimitate the nuclear territory. Epifluorescence microscopy was done on the Axiovert 200 inverted microscope (Zeiss) and confocal images were taken using LSM510 Meta system (Zeiss).

Accession numbers

Microarray data has been deposited at GEO (accession GSE22620). The Nimblegen tiling data have been deposited at GEO (accession GSE22618).

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7

Attachment 8

Attachment 9

Attachment 10

Attachment 11

Attachment 12

Attachment 13

Attachment 14