Histone locus regulation by the Drosophila dosage compensation adaptor protein CLAMP

CLAMP is recruited to the histone locus via the H3–H4p

While studying CLAMP in the context of male X-chromosome dosage compensation, we noticed distinct CLAMP puncta in the nuclei of early D. melanogaster embryos. Costaining revealed that these CLAMP puncta colocalized with markers of the HLB in embryos (Fig. 1A) and cultured cells (Supplemental Fig. S1A) and on the giant salivary gland polytene chromosomes of third instar larvae (Fig. 1B). A GFP-tagged full-length CLAMP also colocalized with HLB markers on salivary gland polytene chromosomes (Supplemental Fig. S1B). Both CLAMP and GAF recognize GA repeats throughout the genome, often at the same loci (Kasinathan et al. 2014; Kuzu et al. 2016). However, we found that GAF was not present at the HLB (Fig. 1C).

The endogenous Drosophila histone locus on chromosome 2L contains ∼100 tandem copies of a 5-kb gene cluster (Lifton et al. 1978; McKay et al. 2015), each containing a single copy of the five replication-dependent histone genes (Fig. 1D). To determine the exact location of CLAMP binding within the histone locus, we mapped existing Drosophila cell culture CLAMP ChIP-seq (chromatin immunoprecipitation [ChIP] combined with high-throughput sequencing) data from our laboratory (Soruco et al. 2013) and GAF ChIP-seq data (Fuda et al. 2015) to a custom genome containing a single copy of the 5-kb gene cluster (McKay et al. 2015). With this approach, the ChIP-seq signal represents an average binding profile across all ∼100 gene clusters. We found that CLAMP localized precisely to the H3–H4p in both male S2 and female Kc cultured cells (Fig. 1D), which is the same region of the gene cluster that is minimally sufficient for recruitment of HLB components (Salzler et al. 2013). In marked contrast, GAF did not localize to the histone locus (Fig. 1C,D), suggesting that CLAMP provides a unique function at the histone locus, similar to our results on the dosage-compensated X chromosome (Soruco et al. 2013; Kuzu et al. 2016).

A number of factors are present at the HLB constitutively throughout the cell cycle, while others are present only during S phase, when the histone genes are transcriptionally active (Duronio and Marzluff 2017). The scaffolding protein Mxc is present at the HLB throughout the cell cycle but is phosphorylated only when Cyclin E/Cdk2 is active (e.g., during S phase in cultured cells), creating a phosphoepitope recognized by the MPM-2 antibody (White et al. 2011). Therefore, to characterize whether CLAMP localization to the HLB is cell cycle-dependent, we used the MPM-2 antibody to label S-phase HLBs and the Mxc antibody to label all HLBs. Unlike the MPM-2 epitope, CLAMP was present at the HLBs in all cultured cells (Supplemental Fig. S1A), suggesting that CLAMP localizes to the HLB throughout the cell cycle.

A “proto-HLB” composed of FLASH and Mxc forms before the onset of zygotic histone gene transcription (White et al. 2011; Salzler et al. 2013). However, neither FLASH nor Mxc nor any other previously known component of the HLB has been shown to bind DNA and target the HLB to the histone locus. Our observations that CLAMP localizes to the H3–H4p (Fig. 1) and is present at the HLB throughout the cell cycle (Supplemental Fig. S1A) led us to hypothesize that CLAMP may be a factor that is recruited to the histone locus prior to activation of zygotic histone gene expression (i.e., by embryonic nuclear cycle 10) and therefore may be a component of the “proto-HLB.” To test this hypothesis, we performed CLAMP ChIP-seq from pools of 200–400 hand-sorted precisely staged embryos (Blythe and Wieschaus 2015). We identified pools of embryos at each of the nuclear cycles 11–14 using a PCNA-EGFP nuclear reporter. We pooled embryos in cycle 10 and younger to obtain sufficient chromatin from such young embryos for ChIP-seq analysis. We then mapped ChIP-seq reads to the histone gene cluster. CLAMP is present at the embryonic H3–H4p, as observed in cultured cells (Fig. 1D), and was present at all assayed nuclear cycles, including by nuclear cycle 10 (Fig. 1E). In contrast, other HLB components (i.e., Mxc, FLASH, Mute, and U7 snRNP) are not detectable at nuclear foci prior to cycle 10 (White et al. 2007; Terzo et al. 2015). These observations demonstrate that CLAMP is present at the embryonic histone locus prior to zygotic genome activation and suggest that it is recruited to the histone locus before the mature HLB is formed (White et al. 2011; Salzler et al. 2013). Collectively, our observations led us to hypothesize that CLAMP regulates the histone locus during development by recognizing critical cis elements within the H3–H4p.

Conserved GA repeat cis elements in the H3–H4p are required for ectopic HLB formation

We showed previously that a single transgenic copy of the ∼300-bp H3–H4p is necessary and sufficient to recruit HLB components to an ectopic locus. We identified two conserved GA repeat motifs within the H3–H4p as potential CLAMP-binding sites (Fig. 2A,B; Salzler et al. 2013). There is a substantial expansion of one GA repeat in D. melanogaster compared with other drosophilids, including closely related species such as Drosophila simulans (Fig. 2B). Because we reported recently that expanded GA repeats facilitate CLAMP-mediated X-chromosome dosage compensation (Kuzu et al. 2016), we asked whether CLAMP localization to the HLB was specific to D. melanogaster by staining polytene chromosomes from D. simulans (Supplemental Fig. S2A,B) and Drosophila virilis (Fig. 2C,D), which diverged from D. melanogaster >40 million years ago (Russo et al. 1995). The genome of D. simulans contains a single histone locus near the chromocenter, while the genome of D. virilis contains two histone loci in the middles of chromosome arms (Schienman et al. 1998; Berloco et al. 2001). We found that CLAMP is present at the histone locus in both D. simulans and D. virilis. Similar to D. melanogaster (Fig. 1C), GAF did not colocalize with HLB factors in other species. Therefore, CLAMP localization to the histone locus is not specific to D. melanogaster and is not only due to the GA repeat expansion in the D. melanogaster H3–H4p (Fig. 2B).

Our CLAMP ChIP-seq results (Fig. 1D,E) and the sequence conservation of the H3–H4p (Fig. 2B) led us to hypothesize that the GA repeats may function to promote HLB formation. To identify regions in the H3–H4p that are important for HLB factor recruitment, we constructed four transgenes containing deletions in the ∼300-bp H3–H4p. In three of the four constructs, either one or both of the GA repeats are deleted (Supplemental Fig. S3A). We found that HLB factors were efficiently recruited only to the H3–H4p transgene constructs that preserve both GA repeats (Supplemental Fig. S3B–D).

To test directly whether the GA repeats are required for HLB formation, we used mutational analysis of BAC-based transgenes carrying 12 copies of the full 5-kb histone gene cluster (12xHistone^WT). We showed previously that the 12xHistone^WT transgene forms a functional ectopic HLB, recruiting all tested HLB factors in the presence of the endogenous histone locus (Fig. 3A), and rescues the lethality caused by homozygous deletion of the endogenous histone locus (McKay et al. 2015). The 12xHistone^WT transgene also recruits CLAMP (Fig. 3A). We therefore generated a mutant 12x array, 12x2xLacO, in which both GA repeats in the H3–H4p are replaced with LacO sequences (Fig. 3B; promoter sequence in the Supplemental Material). We integrated this transgene into the Drosophila genome at the same site and asked whether it recruits HLB factors in the presence of the endogenous histone locus. The 12x2xLacO transgene fails to recruit CLAMP (Fig. 3B), indicating that the GA repeat sequences are necessary for CLAMP binding to the H3–H4p in the presence of the endogenous histone locus. The 12x2xLacO transgene also failed to recruit Mxc, Lsm10, or Mute to polytene chromosome spreads (Fig. 3B), consistent with our results from the promoter deletion transgene constructs (Supplemental Fig. S3). In contrast to the 12xHistone^WT array (McKay et al. 2015), the 12x2xLacO array in which the GA repeats have been replaced with LacO sequences did not express detectable amounts of Histone2A mRNA (Fig. 3E). Thus, the GA repeats in the H3–H4p are key cis elements that promote HLB formation and histone gene expression.

CLAMP promotes recruitment of HLB components to an ectopic histone locus

We demonstrated previously that CLAMP recognizes GA-rich sequence elements and promotes recruitment of the MSL complex specifically to the male X chromosome (Soruco et al. 2013; Kuzu et al. 2016). We therefore hypothesized that CLAMP functions similarly at the histone locus, recognizing GA repeat cis elements and promoting the recruitment of HLB-specific factors. We performed a tethering experiment using the 12x2xLacO transgenic array described above (Fig. 3B) and a synthetic CLAMP^Q-LacI protein in which the CLAMP zinc finger DNA-binding domain (Larschan et al. 2012) is replaced with LacI (Robinett et al. 1996). The resulting transgene, “CLAMP^Q-LacI,” contains only the N-terminal 287 amino acids of CLAMP, which includes the polyglutamine domain of the CLAMP protein but does not include the DNA-binding domain. We conducted the experiment in this way to avoid the presence of two competing DNA-binding domains on the same protein that could prevent binding to the histone locus (Fig. 3C). We expressed the CLAMP^Q-LacI transgene using a ubiquitin promoter and measured expression of CLAMP^Q-LacI (Supplemental Fig. S4A) by Western blot using protein from third instar larval salivary glands and clamp^q-lacI mRNA levels by quantitative RT–PCR (qRT–PCR) (Supplemental Fig. S4B) using RNA from whole larvae. From this analysis, we determined that the CLAMP^Q-LacI protein is expressed at higher levels than endogenous CLAMP (Supplemental Fig. S4A).

We performed polytene chromosome immunostaining on animals expressing CLAMP^Q-LacI in the presence of the transgenic 12x2xLacO histone array. We found that CLAMP^Q-LacI as well as Mxc, FLASH, and Mute colocalized with the fluorescent in situ hybridization (FISH) signal against the histone gene repeat when paired with the 12x2xLacO histone array (Fig. 3D). In contrast, in the absence of CLAMP^Q-LacI, the 12x2xLacO array does not recruit HLB factors (Fig. 3B). Therefore, CLAMP^Q-LacI binds the 12x2xLacO array and promotes the recruitment of HLB factors. CLAMP^Q-LacI also localized to several additional genomic locations (Fig. 3D), possibly through dimerization with endogenous CLAMP. However, HLB factors were recruited only to the 12x2xLacO locus after CLAMP^Q-LacI expression. Thus, like endogenous CLAMP, CLAMP^Q-LacI is not sufficient to recruit HLB factors to nonhistone loci, suggesting that the H3–H4p may contain other important cis elements and/or recruit additional critical components. Therefore, promoting HLB factor recruitment to the histone locus is a context-specific function of the ubiquitously expressed CLAMP protein, similar to its function in promoting the recruitment of the MSL complex to the dosage-compensated X chromosome.

Although multiple HLB factors were recruited to the 12x2xLacO histone array in the presence of CLAMP^Q-LacI, we did not detect any histone mRNA expressed from the array (Fig. 3E). Thus, HLB factor recruitment can be uncoupled from histone locus transcription, as observed previously in the formation of a “proto-HLB” containing Mxc and FLASH (Salzler et al. 2013). There are several possible explanations for why CLAMP^Q-LacI can recruit HLB factors but not promote transcriptional activation. First, only low levels of HLB factors were recruited to the 12x2xLacO histone array in the presence of CLAMP^Q-LacI (Fig. 3D), compared with the levels recruited to the 12xHistone^WT array in the presence of endogenous wild-type CLAMP (Fig. 3A). Therefore, the stoichiometry of CLAMP^Q-LacI compared with each 12x2xLacO sequence may be insufficient to support transcription. Second, a specific conformation of the endogenous CLAMP protein or the zinc finger domain of CLAMP, which is replaced by LacI in the CLAMP^Q-LacI construct (Fig. 3C), may be required for transcriptional activation of the histone genes but not for recruitment of HLB factors. Alternatively, wild-type CLAMP at the endogenous histone locus may recruit critical limiting factors much more efficiently than ectopically localized CLAMP^Q-LacI. Together, our data indicate that the N-terminal 287 amino acids of CLAMP are sufficient for promoting the recruitment of HLB factors to an ectopic chromosomal location but not for transcriptional activation.

CLAMP regulates histone locus chromatin and histone gene expression

Because CLAMP localizes to the histone locus and binds directly to the essential GA repeat cis elements, we hypothesized that CLAMP regulates histone locus chromatin and histone gene transcription. All five histone transcripts and the clamp transcript are expressed in the oocyte and maternally deposited in 0- to 2-h embryos. Mature HLBs are formed by cycle 11, and the zygotic histone locus is activated by zygotic genome activation (∼2 h after egg lay) (White et al. 2007). To determine how CLAMP functions as the HLB forms and the histone locus becomes transcriptionally activated in the early embryo, we depleted maternally deposited CLAMP protein and mRNA by RNAi using a UAS-driven clamp-specific shRNA (Ni et al. 2008) and a strong GAL4 driver (maternal triple driver [MTD]) that is expressed in the female germline (Staller et al. 2013). As controls, we also analyzed matched MTD controls. We performed qRT–PCR for histone mRNA levels using primers specific to the H3 and H4 transcripts (Bulchand et al. 2010). CLAMP mRNA and protein are nearly completely depleted in 0- to 2-h and 2- to 4-h embryos derived from clamp RNAi mothers (Figs. 4A, 5A). We found that in unfertilized eggs and 0- to 2-h embryos laid by clamp RNAi mothers, both H3 and H4 transcript levels were significantly decreased (Fig. 4A), indicating that CLAMP depletion results in a reduction in the amount of histone mRNA deposited in the egg. Histone transcript levels began to recover in older embryos, likely due to zygotic genome activation of both the embryonic histone locus and the clamp locus. Nearly 100% (99.87%) of clamp RNAi embryos do not hatch, demonstrating that maternally deposited CLAMP is critical for early development.

We measured transcript accumulation from all histone genes by performing mRNA sequencing (mRNA-seq) from the same embryo time points and RNAi conditions that we used in our qRT–PCR assay (Fig. 4A). We found that maternal clamp RNAi resulted in decreased expression of all histone genes compared with MTD controls (Fig. 4B,C; Supplemental Table S1), including H1, H2A, and H2B, which are not adjacent to the site of CLAMP binding within the H3–H4p (Fig. 1D,E). To determine whether this reduction of histone transcripts might be an indirect effect of CLAMP regulating the expression of known HLB components, we compared the levels of mRNAs encoding known HLB components, including FLASH, Mxc, Mute, Lsm10, and Lsm11 (Duronio and Marzluff 2017), from 0- to 2-h and 2- to 4-h clamp RNAi embryos with those from control MTD embryos. The transcript levels of the known HLB factors were not affected by clamp RNAi in the early embryo (Supplemental Table S2). Together, these observations suggest that the effects on histone transcript levels that we observed in the early developing embryo after CLAMP depletion (Fig. 4A–C) are not due to misregulation of genes encoding other HLB components but a specific effect of CLAMP depletion on histone mRNA deposition in the egg.

We next sought to understand how CLAMP might regulate histone gene expression. To determine whether CLAMP promotes chromatin accessibility at the histone locus, we performed clamp RNAi in cultured male S2 cells and measured chromatin accessibility using a previously described micrococcal nuclease (MNase) digestion followed by high-throughput sequencing (MNase-seq) titration approach developed by our group (Mieczkowski et al. 2016). Accessible chromatin is digested even under dilute MNase concentrations, while inaccessible chromatin is protected until it is subjected to higher concentrations. Therefore, sequencing libraries generated from chromatin subjected to a low concentration of MNase are enriched for accessible regions, while libraries from chromatin digested with a higher concentration of enzyme are enriched for more inaccessible regions (Mieczkowski et al. 2016).

Using our chromatin accessibility data, we mapped reads to a single copy of the histone gene array (McKay et al. 2015) and calculated chromatin accessibility as described in Mieczkowski et al. (2016). In control S2 cells subjected to gfp control RNAi, the most accessible regions of the histone gene cluster included promoter and intergenic regions (Fig. 4D). In cells subjected to clamp RNAi, the entire histone gene cluster decreased in accessibility. Overall, our data support the requirement for CLAMP recruitment to the histone locus for both promoting chromatin accessibility across the entire histone locus and activating gene expression of all replication-dependent histone genes (Fig. 4A–D).

CLAMP is specifically retained at the HLB in clamp nulls

The ability of CLAMP to promote HLB factor recruitment to the transgenic array locus (Fig. 3D) and regulate chromatin organization at the histone locus (Fig. 4D) led us to test whether removing CLAMP resulted in loss of endogenous HLB formation in clamp-null larvae and in embryos in which clamp is depleted by RNAi. While almost all embryos laid by mothers with RNAi-reduced clamp levels are inviable, ∼30% of zygotically mutant animals homozygous for the clamp²-null allele survive until the third instar larval stage (Urban et al. 2017). These data indicate that survival of clamp² larvae is likely due to maternal deposition of wild-type clamp mRNA or protein. We reported previously that clamp²-null animals produce undetectable amounts of CLAMP protein by both Western blot of salivary gland tissue and polytene chromosome immunostaining (Urban et al. 2017). However, our previously characterized affinity-purified anti-CLAMP antibody detected many cross-reacting proteins by Western blot (Fig. 5A), and we suspected that it might not be sufficiently sensitive to detect small amounts of CLAMP protein in clamp²-null larvae. Therefore, we developed a new affinity-purified anti-CLAMP antibody (“CLAMP*”)—produced using a different approach (see the Materials and Methods)—that is more specific and does not cross-react with other Drosophila proteins (Fig. 5A).

Using both anti-CLAMP antibodies, we analyzed CLAMP localization and HLB formation on clamp²-null polytene chromosomes from larval salivary glands. In addition, we analyzed early embryos after RNAi depletion of maternally deposited CLAMP protein and mRNA (Figs. 4A, 5A). While these embryos die, it is still possible to analyze CLAMP localization and HLB formation during the early syncytial stages (0–2 h) of embryonic development. Using the CLAMP antibody, we observed loss of CLAMP from all loci on polytene chromosomes from clamp² nulls, as observed previously (Fig. 5B). There were no CLAMP foci in early embryos after CLAMP RNAi, although Mute foci were readily detected (Fig. 5F). However, using the CLAMP* antibody, we detected a modest amount of CLAMP immunostaining, which colocalizes specifically and solely with HLB factors on clamp²-null mutant polytenes (Fig. 5C) and in embryos after clamp RNAi (Fig. 5G). Therefore, the CLAMP* antibody is more sensitive than our previously published CLAMP antibody for immunofluorescence. Multiple HLB factors continue to colocalize on clamp²-null polytene chromosomes, similar to control chromosomes where abundant CLAMP is present (Fig. 5D,E), and in clamp-depleted embryos (Fig. 5H). Because zygotic CLAMP protein is not produced in clamp² homozygous animals (Urban et al. 2017), we conclude that maternally deposited CLAMP from clamp² heterozygous mothers persists at the histone locus throughout development of the larval salivary gland. Furthermore, the strong MTD does not deplete all maternally deposited clamp transcript or protein, and therefore a small amount of CLAMP remains at embryonic HLBs. Some of the ∼200 GA repeat CLAMP-binding sites within the histone gene array are likely capable of recruiting the small amount of maternal CLAMP that remains in clamp²-null third instar larvae and early clamp-depleted embryos.

GAF localizes to the HLB when CLAMP is depleted

Since only trace amounts of maternally deposited CLAMP are present in HLBs in clamp²-null larvae and embryos from clamp RNAi mothers (Fig. 5C,G), we tested whether CLAMP depletion results in novel recruitment of the GAF protein to some H3–H4ps. Consistent with this possibility, we found that GAF colocalized with HLB markers on polytene chromosomes from clamp² homozygous larvae (Fig. 6B) and in syncytial embryos after MTD-driven clamp RNAi (Fig. 6D) even though GAF was absent from the HLBs when CLAMP is present (Figs. 1C, 6A,C). Previous in vitro experiments suggested that GAF is able to bind the GA-rich cis elements in the H3–H4p (Gilmour et al. 1989). However, consistent with our current observations, later in vivo studies suggested that GAF is not present at the histone locus (O'Brien et al. 1995; Bhat et al. 1996). Our results suggest that the 61-kDa CLAMP protein (Urban et al. 2017) is the likely ∼66-kDa protein bound to the H3–H4p in vivo that was identified >25 years ago (Gilmour et al. 1989; Weber and Gilmour 1995).

However, in agreement with previous in vitro observations (Weber and Gilmour 1995), GAF localizes to the histone locus when CLAMP is depleted (Fig. 6B,D), suggesting that it can bind to the H3–H4p when there is not sufficient CLAMP to occupy all GA repeat-binding sites. There are several models that could explain the relationship between CLAMP and GAF at the HLB. For example, residual amounts of CLAMP (Fig. 5) may be sufficient to open the chromatin and initiate HLB formation but not stimulate transcription. In this situation, although GAF can now bind unoccupied GA repeats, it is not able to stimulate histone gene transcription to the level that CLAMP does (Fig. 4A–C). A second possibility is that GAF can open chromatin to promote HLB factor recruitment but cannot stimulate transcription at the histone locus to the same level as CLAMP because it cannot interact with the same cofactors as CLAMP. We discuss possible functional relationships between GAF and CLAMP further in the Discussion.

Regulation of the histone locus

The mature Drosophila HLB is formed by cycle 11 during early embryogenesis, prior to widespread zygotic genome activation (White et al. 2011). Activation of histone genes is critical for organismal development and viability (Saget et al. 1998; Godfrey et al. 2006, 2009; White et al. 2007; Gunesdogan et al. 2010). CLAMP is present at the histone locus when the proto-HLB containing Mxc and FLASH is first detectable (Fig. 1E), prior to zygotic histone gene activation (White et al. 2007). As a DNA-binding transcription factor that directly recognizes the GA repeats that promote HLB formation in the presence of the endogenous locus (Fig. 1D,E), CLAMP represents a potential pioneer factor involved in promoting HLB formation and activation of histone gene transcription. Furthermore, CLAMP facilitates chromatin opening across the entire histone locus repeat (Fig. 4D), suggesting that it could recruit additional factors, such as general transcription factors or cofactors necessary for histone gene regulation.

It is likely that CLAMP affects chromatin accessibility at the histone locus through recruitment of a chromatin-remodeling factor. For example, the similar GAF protein recruits the NURF301 chromatin remodeler (Tsukiyama and Wu 1995). In addition, CLAMP may control histone gene transcription (Fig. 4A–C) through modulating chromatin accessibility, or changes in accessibility may be caused by CLAMP directly regulating histone transcription.

Retargeting the polyglutamine domain of CLAMP to a synthetic histone array transgene rescues recruitment of Mxc (Fig. 3E), the core HLB scaffolding protein (Terzo et al. 2015), and several other HLB components. However, CLAMP^Q-LacI is not sufficient to activate transcription (Fig. 3E). It is possible that transcriptional activation requires additional domains of CLAMP that were removed in our CLAMP^Q-LacI fusion protein, a particular conformation of the wild-type CLAMP protein that is not attainable by CLAMP^Q-LacI, or higher CLAMP occupancy than the LacO/LacI system allows. Overall, we demonstrate that a reduction in wild-type CLAMP levels leads to decreased chromatin accessibility and histone gene expression at the endogenous histone locus. Furthermore, artificially tethering CLAMP^Q-LacI to a synthetic histone gene array recruits HLB-specific factors but does not stimulate histone gene expression.

Interestingly, previous work demonstrated that additional HLB components, including Mute and U7 snRNP, are recruited to the HLB only when transcription is initiated (Salzler et al. 2013) and are present at the mature HLB by nuclear cycle 11, when the histone genes are expressed (White et al. 2007, 2011). However, in the present study, we demonstrate that recruitment of Mute can be uncoupled from histone locus transcription (Fig. 3E), perhaps capturing an intermediate stage in HLB formation.

Conservation of HLB formation

Factors that regulate histone gene expression are well conserved, but current evidence suggests that cis elements that target these factors are less conserved across species. For example, critical components of the HLB that do not directly interact with DNA are conserved across metazoans, including Mxc/NPAT, Mute/YY1-associated protein, FLASH, U7 snRNP, and Coilin (Duronio and Marzluff 2017). In contrast, CLAMP is highly conserved across Diptera (Kuzu et al. 2016), but we did not identify a CLAMP ortholog outside of insects. CLAMP may function together with an early acting Drosophila-specific transcription factor, such as the master zygotic genome activator Zelda, which, although conserved among drosophilids, also has no identified mammalian ortholog (Paris et al. 2013).

The mammalian histone genes are more dispersed and there are no repeated units, making it possible that mammalian cis elements may be within poorly conserved noncoding sequences. Therefore, the plasticity of the components that link highly conserved HLB factors to the histone locus—and the flexibility of their low-complexity domains—may allow for evolutionary variation among critical cis-acting sequences that specify HLB formation. CLAMP binding to GA repeats likely provides this function in Drosophila, while other DNA-binding proteins and cis elements may nucleate the HLB in other species.

A single transcription factor can mediate multiple domains of coordinated gene activation

We originally identified CLAMP as an essential factor (Urban et al. 2017) that is required to nucleate the formation of a domain of coordinated gene activation during male X-chromosome dosage compensation (Larschan et al. 2012; Soruco et al. 2013). GA repeat expansion and elevated GA repeat density on the X chromosome evolved to increase CLAMP occupancy and promote MSL recognition (Kuzu et al. 2016). Here we show that CLAMP is also recruited to the histone locus via GA repeats in the H3–H4p (Figs. 1D,E, 2B), where it promotes the formation of another domain of coordinated gene activation. In contrast to CLAMP, GAF is not required for male X-chromosome dosage compensation (Greenberg et al. 2004) and is dramatically depleted on the X chromosome compared with autosomes (Soruco et al. 2013). GAF is also not present at the histone locus in wild-type situations (Fig. 1C,D). Therefore, we identified the enrichment of CLAMP relative to GAF as a common feature of two different chromatin domains of enhanced chromatin accessibility and active transcription.

Although CLAMP regulates two very different active chromatin domains to which different specific cofactors are recruited, CLAMP is also a transcription factor that binds to thousands of loci throughout the genome. The mechanism by which a single protein fulfills multiple context-specific roles within the genome is not well understood in any system. It is likely that synergy with additional cofactors, such as the MSL complex on the dosage-compensated X chromosome (Soruco et al. 2013) and Mxc at the histone locus, are critical for this process.

In contrast to the histone locus, where HLB-specific factors are conserved and cis sequences are not, the MSL dosage compensation complex components are very poorly conserved even within drosophilids, compared with CLAMP. However, X-enriched GA-rich cis elements are well conserved across insect species (Kuzu et al. 2016). Because CLAMP is less conserved across species compared with HLB-specific factors, it is possible that it provides the plasticity required to couple conserved regulators to rapidly evolving cis elements at the histone locus. In contrast, on the X chromosome, CLAMP functions as a relatively well-conserved factor (compared with the poorly conserved MSL complex) that couples conserved cis elements to diverged cofactors. It is possible that the ancient evolutionary role of CLAMP is as a GA-binding transcription factor, while the context-specific roles of CLAMP at the histone locus and dosage-compensated X chromosome evolved more recently, after GA-rich cis elements became enriched at these locations to promote domains of coordinated gene activation. However, significant further study is required to fully understand how a single protein promotes the formation of multiple active chromatin domains.

The relationship between GA-binding factors at the histone locus

GAF is a ubiquitously expressed transcription factor that binds GA repeats genome-wide and functions as a transcriptional regulator (Horard et al. 2000; Fuda et al. 2015), chromatin modifier (Tsukiyama and Wu 1995), and insulator (Ohtsuki and Levine 1998), depending on the context. Although GAF is excluded from the histone locus in wild-type situations, it localizes to the histone locus when CLAMP is depleted (Fig. 6).

There are several possible models for the relationship between CLAMP and GAF at the HLB: (1) CLAMP and GAF could be partially redundant such that GAF can compensate for some functions normally accomplished by CLAMP, including opening chromatin to minimally recruit HLB factors. However, GAF is not sufficient to fully rescue histone transcription levels when CLAMP is depleted (Fig. 4A–C). (2) CLAMP and GAF could have an antagonistic relationship at the histone locus such that a major function for CLAMP could be to exclude GAF. For example, GAF mislocalization to the histone locus may close chromatin and repress histone gene expression through its well-documented relationship with the repressive Polycomb complex (Horard et al. 2000). (3) Both CLAMP and GAF contain low-complexity glutamine-rich domains (Tariq et al. 2013) that are thought to modulate transcriptional activity (Gemayel et al. 2015) and are also frequently found in a class of proteins thought to undergo biophysical phase transitions to drive NB formation (Zhu and Brangwynne 2015). Therefore, the relative occupancy of CLAMP compared with GAF within a domain may promote or hinder NB formation. (4) GAF binds to unoccupied histone locus GA repeats opportunistically and does not function redundantly or antagonistically with CLAMP at the histone locus. In this model, CLAMP depletion drives the changes that we observed in histone locus regulation (Fig. 4).

Since we were not able to completely remove CLAMP even in the salivary glands of clamp-null mutant larvae, we conclude that CLAMP regulates the histone locus (Fig. 5C,G) but cannot conclude that CLAMP is essential for endogenous HLB formation. In addition, in agreement with previous studies on GAF autoregulation (Bernues et al. 2007), we found that the GAF-encoding mRNA trithorax-like is resistant to depletion by embryonic RNAi using the shRNA system. Therefore, further experimentation is required to determine the functional relationship between CLAMP and GAF at the HLB.

Here, we determined that the same critical cis elements (GA repeats) and trans-acting factor (CLAMP) and the absence of a second factor that binds similar cis elements (GAF) promote the formation of two different domains of coordinated gene activation. Furthermore, both at the histone locus and on the dosage-compensated male X chromosome, CLAMP is required to promote a domain of enhanced chromatin accessibility that extends well beyond its binding sites. Overall, our study provides new insight into the common regulatory mechanisms shared by two different domains of coordinated gene activation.

Drosophila strains

We used the MTD (Bloomington, #31777) and a stock expressing a shRNA against clamp (Bloomington, #57008) made by the Transgenic RNAi Project (TRiP). For the H3–H4p deletion experiments, we inserted promoter sequences into the pMulti-BAC vector containing a single histone repeat unit (McKay et al. 2015) and inserted these transgenes into site 86Fb on chromosome 3 using φC31-mediated integration (Bestgene) (Groth et al. 2004). The full sequences of engineered H3–H4p deletion sequences are in the Supplemental Material. For the LacO array experiments, we synthesized H3–H4ps (Genescript) and used restriction digest cloning to insert the promoter containing LacO sequences in place of the wild-type promoter in the single histone repeat unit. We built an array of 12 histone repeat units in pMulti-BAC for each transgenic promoter and integrated each into site VK33 on chromosome 3 using φC31-mediated integration (Model Systems Injections). The full sequences of engineered H3–H4p with LacO sequences are in the Supplemental Material. We inserted CLAMP^Q-LacI and LacI into the pUbi-GFP (gift from Mark Peifer), in which we swapped LacI for GFP using the LacI-HP1a vector (gift from Lori Wallrath). We amplified the CLAMP polyglutamine domain for Gibson assembly (New England Biolabs) using the primers F (5′-TAGGTCCTGTTCATTGAATGGAAGACCTTACCAAAAAC-3′) and R (5′-GTTACTGGTTTCACCATAGCCACAATTTGCTGAAG-3′). We drove transcription of both CLAMP^Q-LacI and LacI genes using the ubiquitin promoter and integrated these transgenes into site VK20 on chromosome 3 using φC31-mediated integration (Genetivision). To make GFP-CLAMP, we cloned clamp cDNA into a vector containing the ubiquitin promoter (pUbi-GFP; gift from Mark Peifer) and integrated the transgene into site VK33 on chromosome 3.

Promoter alignment

We obtained promoter sequences from D. simulans (DNA Data Bank of Japan [DDBJ] accession no. AB055959) (Tsunemoto and Matsuo 2001), Drosophila erecta (DDBJ accession no. AB073634) (Kakita et al. 2003), Drosophilia pseudoobscura (DDBJ accession no. AB249651) (Nakashima et al. 2016), and D. virilis. We aligned sequences using T-Coffee (Notredame et al. 2000) and formatted the alignment using BoxShade.

FISH and immunofluorescence

We used primary antibodies at the following concentrations: rabbit anti-CLAMP (1:1000; Novus/SDIX) (Larschan et al. 2012), rabbit anti-CLAMP* (1:1000; custom antibody generated by our laboratory through a contract to Abcam; both anti-CLAMP antibodies were raised against the same N-terminal amino acids, CLAMP#22–121), guinea pig anti-Mxc (1:2000) (White et al. 2011), guinea pig anti-Mute (1:5000) (Bulchand et al. 2010), rabbit anti-C terminus FLASH (1:2000) (Yang et al. 2009), rabbit anti-Lsm10 (1:1000), mouse anti-MPM-2 (1:100; Millipore), rabbit anti-GAF (1:1000; gift from Giacomo Cavalli), mouse anti-LacI (1:1000; Millipore), and chicken anti-GFP (1:400; Life Technologies). We used Alexa fluor secondary antibodies (Thermo Fisher Scientific) at a concentration of 1:1000. We detected in situ probes using 15 µg/mL streptavidin-DyLight-488 (Vector Laboratories). To make the FISH probe, we made a PCR product that spanned all five histone genes using a wild-type histone repeat in pMulti-BAC as the template (primers F [AAAGGAGGTTGGTAGGCAGC] and R [ACGCTAGCGCTTTATCTGCA]) (McKay et al. 2015). We made biotinylated FISH probes by nick translation using the purified PCR product: 1 µg of purified PCR product was incubated for 2 h at 15°C in a total of 50 µL containing 1× DNA Pol I buffer (Fisher Optizyme); 0.05 mM each dCTP, dATP, and dGTP; 0.05 mM biotin-11-dUTP (Thermo Scientific); 10 mM 2-mercaptoethanol; 0.004 U of DNaseI (Fisher Optizyme); and 10 U of DNA Pol I (Fisher Optizyme). The reaction was purified on a PCR purification column (Thermo Scientific) and diluted in hybridization buffer (2× SSC, 10% dextran sulfate, 50% formamide, 0.8 mg/mL salmon sperm DNA) to a final volume of 220 µL. We performed FISH according to Grimaud et al. (2005) except that we added hybridization mixture with the probe to the slide before heating. We added a coverslip, sealed it with rubber cement, and heated the slide for 2 min on a 91°C heat block.

We obtained embryos (mixed sex) by mating virgin females, aged 3–4 d, of either genotype (1) homozygous MTD or (2) MTD crossed to Bloomington #57008 for clamp RNAi with w¹¹¹⁸ males. Embryos were fixed in 3.7% formaldehyde in PBS with an equal volume of n-heptane for 20 min, immunostained using the above antibody concentrations, mounted using Prolong Diamond anti-fade reagent (Thermo Fisher), and imaged on a Zeiss laser scanning 510 or 800 confocal microscope equipped with a 63×/1.4 oil immersion plan apochromat objective and Zen software.

We performed polytene chromosome squashes from salivary glands of mixed sex larvae. We passed glands through fix 1 (1.5% formaldehyde, 1% Triton X-100, in 1× PBS) for 1 min, fix 2 (1.5% formaldehyde, 50% glacial acetic acid) for 2 min, and 1:2:3 solution (ratio of lactic acid:water:glacial acetic acid) for 5 min prior to squashing and spreading. Slides were immunostained using the antibody concentrations above and mounted using Prolong Diamond anti-fade reagent (Thermo Fisher), and spreads were imaged on a Zeiss Imager.M1 using a 40×/0.75 plan neofluar objective and AxioVision software.

Western blotting

We conducted Western blotting as in Urban et al. (2017). We collected 2- to 4-h embryos of the relevant genotypes (at least 150 per sample) on grape juice agar plates and washed them briefly with 1× PBS in a cell strainer basket. We dechorionated them for 2 min in 50% bleach and then washed them with several milliliters of PBS before transferring them to lysis buffer (50 mM Tris-HCl at pH 8.0, 150 mM NaCl, 1% SDS, 0.5× protease inhibitor). For salivary glands, we dissected glands from third instar larvae (n = 10 per sample) in cold PBS and froze samples in liquid nitrogen. We extracted total protein from samples by homogenizing the samples in cold lysis buffer using a small pestle. We cleared the samples by centrifuging at 14,000g for 10 min at room temperature. To blot for CLAMP and Actin, we ran 20 µg of total protein on a Novex 4%–12% Tris-glycine precast gradient gel (Life Technologies). We transferred proteins to PVDF membranes using the iBlot transfer system (Thermo Fisher Scientific) and probed the membranes for CLAMP (rabbit anti-CLAMP, 1:1000; rabbit anti-CLAMP*, 1:1000) and Actin (mouse anti-Actin, 1:400,000; Millipore ) using the Western Breeze kit, following the manufacturer's protocol (Thermo Fisher Scientific).

MNase-seq

We maintained S2 cells in standard Schneider's medium (Gibco) with 10% heat-inactivated fetal bovine serum (Thermo Fisher Scientific). We performed RNAi as in Soruco et al. (2013). We performed and analyzed MNase-seq data as in Mieczkowski et al. (2016). We mapped reads to the custom histone locus genome (McKay et al. 2015) using Bowtie aligner with the parameters “-M 5 -k 1 -I 50 -X 500 --solexa-quals --best –chunkmbs 256” (Langmead et al. 2009). We identified genomic positions with abnormally high numbers of mapped reads (Z-score = 7) and discarded tags mapped to such positions. We computed read frequencies in 100-bp nonoverlapping bins and normalized for the library size. We calculated MNase accessibility (MACC) values for each bin by fitting linear regression on the normalized read frequencies computed for each titration point (1.5-, 6.25-, 25-, and 100-U MNase concentrations). We used log scale for the MNase concentrations in the fitting procedure. We applied the GC content correction to obtain the final accessibility scores (MACC values). The chromatin accessibility data are available at NCBI Gene Expression Omnibus (GEO) with series number GSE99894.

Quantitative real-time PCR

We conducted qRT–PCR as described in Urban et al. (2017) using the embryo RNA obtained for mRNA-seq (below) as well as RNA extracted from unfertilized oocytes laid by unmated mothers and collected 0–2 h after egg lay. We used four biological replicates for each genotype and time point. Primers for histone transcripts H3 and H4 and the normalization gene rp49 are listed in Bulchand et al. (2010), while clamp and pka primers can be found in Urban et al. (2017). We normalized histone transcript abundance against rp49 and clamp transcript abundance against pka. We analyzed data using a Student's t-test, comparing transcript abundance between clamp RNAi embryos or oocytes and matched MTD control embryos or oocytes.

Embryo mRNA-seq

We used embryo RNA collected for qRT–PCR (above). As in Wood et al. (2016), we used 100 ng of total RNA as input for the Ovation Universal RNA-seq kit with Drosophila rRNA depletion module (NuGEN). We sequenced libraries on an Illumina HiSeq 2500 in 1 × 50-bp mode. We used at least four individually isolated biological replicates for each time point and RNAi condition. We mapped reads using TopHat version 2.0.13 with default parameters (Trapnell et al. 2009) and counted fragments mapping to histone gene exons (see Supplemental Table S1 for all FPKM [fragments per kilobase per million mapped fragments] values for histone genes). See Supplemental Table S2 for a list of significantly affected genes. The mRNA-seq data are available at NCBI GEO with series number GSE102922.

Staged embryo ChIP-seq

To obtain female embryos, we mated +; SD72/CyO females to 19-3, yw, Rsp[s]-B[s]/Dp(2:y)CB25-4, y+, Rsp[s]B[s]; SPSD/CyO males (both kind gifts from Cynthia Staber) to obtain +/Dp(2:y)CB25-4, y+, Rsp[s]B[s]; SPSD/SD72 males, which we then mated to yw; attP2{ PCNA-EGFP} females (kind gift from Shelby Blythe). We performed 0- to 4-h timed lays and collected and fixed embryos according to Blythe and Wieschaus (2015). We then hand-sorted embryos using a Zeiss Discovery.V8 microscope under GFP excitation using an X-CITE 120Q stereo light source. We pooled 200 (NC 11–14) to 400 (NC < 11) embryos and performed ChIP as in Blythe and Wieschaus (2015) using 3 µL of rabbit anti-CLAMP antibody per sample. We synthesized libraries using the NEBNext ChIP-seq kit (New England Biosystems) and sequenced libraries on an Illumina HiSeq 2500 in 2×100-bp mode. We mapped CLAMP ChIP-seq reads to the custom histone locus genome (McKay et al. 2015), allowing only unique alignments by using Bowtie aligner (Langmead et al. 2009).

Analysis of H2a expression from the ectopic array

We isolated total RNA from larvae (n = 10) of the indicated genotypes by flash freezing samples in liquid nitrogen and homogenizing them with a steel bead using a Retsch MM300 TissueLyser Mixer Mill. We then performed phenol/chloroform (Invitrogen) total RNA extractions. We reverse-transcribed 1 µg of total RNA using the SuperScript II kit (Invitrogen) using random primers according to the manufacturer's instructions. We analyzed expression from the transgenic array Histone2a genes via PCR and restriction digestion as in McKay et al. (2015).