Splice variants of the SWR1-type nucleosome remodeling factor Domino have distinct functions during Drosophila melanogaster oogenesis

The local replacement of nucleosomal histone H2A by variants is an evolutionarily conserved principle that endows chromatin with structural and functional diversity (Bönisch and Hake, 2012; Talbert and Henikoff, 2010). Two H2A variants can be considered as universal since they are found in all eukaryotes: H2A.Z and H2A.X. Currently, H2A.X is best known for the role of its phosphorylated form in DNA damage signaling. H2A.Z most likely has a role in regulating transcription, as it marks the nucleosomes next to promoters. Curiously, Drosophila melanogaster only has a single H2A variant, H2A.V, which combines the functions of H2A.Z and H2A.X as an architectural element downstream of active promoters, in heterochromatin organization and, in its phosphorylated form (γH2A.V), in the DNA damage response (Baldi and Becker, 2013).

H2A variants are incorporated by evolutionarily conserved SWR1-like remodeling enzymes, which use the energy freed by ATP hydrolysis to disrupt canonical nucleosomes. SWR1 enzymes share an N-terminal HSA domain and a spacer region that splits the conserved ATPase domain, enabling unique regulation mechanisms (Morrison and Shen, 2009). Mammals utilize two SWR1-like enzymes: p400 (or Ep400) and SRCAP (Cai et al., 2005; Eissenberg et al., 2005; Jin et al., 2005; Ruhl et al., 2006). By contrast, the D. melanogaster genome only contains a single SWR1-like enzyme gene: domino (dom) (Ruhf et al., 2001). Alternative splicing of the dom transcript produces two major isoforms, DOM-A and DOM-B (Ruhf et al., 2001), which differ in their C-termini. The DOM-A isoform features poly-glutamine (poly-Q) stretches and a SANT domain, whereas the DOM-B C-terminus is largely unstructured (Eissenberg et al., 2005; Ruhf et al., 2001). Early reports found DOM-B to be rather ubiquitously expressed, but DOM-A was detected only in the embryonic nervous system, larval salivary glands and S2 cells (Eissenberg et al., 2005; Messina et al., 2014; Ruhf et al., 2001), suggesting distinct functions for the two isoforms.

SWR1-like remodelers typically reside in multi-subunit complexes (Morrison and Shen, 2009), but little is known about DOM-containing complexes. DOM-A has been purified from S2 cells as part of a 16-subunit assembly that also contains the acetyltransferase TIP60, apparently combining features of the yeast SWR1 remodeling and NuA4 acetyltransferase complexes (Kusch et al., 2004). Likewise, our understanding of the roles of DOM enzymes in H2A.V exchange in vivo is anecdotal. DOM has been linked to H2A.V incorporation at the E2f promotor (Lu et al., 2007). It has also been suggested that H2A.V exchange requires prior acetylation by TIP60 (Kusch et al., 2004, 2014).

Previous genetic analyses characterized dom as required for cell proliferation and viability, homeotic gene regulation and Notch signaling (Eissenberg et al., 2005; Ellis et al., 2015; Gause et al., 2006; Kwon et al., 2013; Lu et al., 2007; Ruhf et al., 2001; Sadasivam and Huang, 2016; Walker et al., 2011), but did not attempt to resolve distinct functions of the two isoforms. dom is essential for fly development (Ruhf et al., 2001) and dom mutants die during pupariation (Braun et al., 1998; Ruhf et al., 2001). Furthermore, oogenesis in adult flies is strongly perturbed, causing sterility (Ruhf et al., 2001).

D. melanogaster oogenesis provides an excellent opportunity to study self-renewal and differentiation of germline and somatic stem cells (GSCs and SSCs, respectively) in the context of egg chamber morphogenesis (Hudson and Cooley, 2014; Ting, 2013; Yan et al., 2014). The formation of eggs starts in the germarium at the anteriormost end of an ovariole. There, two to three GSCs divide asymmetrically to self-renew and shed a daughter cystoblast. This cell initiates four mitotic divisions with incomplete cytokinesis to form an interconnected 16-cell cyst. One particular cell of a cyst is determined to become the oocyte and the remaining 15 cells transform into polyploid nurse cells by endoreplication. In parallel, SSCs produce somatic follicle cells, which encapsulate cysts to form individual egg chambers. Oogenesis further runs through 14 stages of egg chamber development to produce a functional egg.

It is not surprising that nucleosome remodeling factors have been found to be important for oogenesis given the widespread requirement for chromatin plasticity during development (Chioda and Becker, 2010; Ho and Crabtree, 2010; Iovino, 2014). Interestingly, an RNA interference (RNAi) screen identified DOM as important for GSC self-renewal and cystoblast differentiation (Yan et al., 2014). DOM function is also essential for SSC self-renewal (Xi and Xie, 2005). The mechanisms of DOM function in these processes are unclear, but involvement of H2A.V exchange has been suggested. H2A.V signals show a modest decrease in dom mutant GSC clones in D. melanogaster testes (Morillo Prado et al., 2013) and are not detectable in mutant germline clones for MRG15, a DOM-A/TIP60 complex subunit (Joyce et al., 2011). Yet, direct evidence for a role of DOM – and specific roles for each isoform – in H2A.V incorporation during oogenesis is lacking.

We used a cell type-specific RNAi approach to dissect the functions of DOM-A and DOM-B during D. melanogaster oogenesis. Our analysis suggests non-redundant requirements for both DOM isoforms in several cell differentiation programs and for H2A.V exchange.

DOM is ubiquitously expressed in the female germline of D. melanogaster ovarioles (Yan et al., 2014), but other studies suggest expression of only DOM-B in soma and germline cells (Ruhf et al., 2001; Xi and Xie, 2005). Owing to the lack of specific DOM-A and DOM-B antibodies for immunofluorescence microscopy we generated transgenes expressing GFP-3×FLAG-tagged DOM from a recombined fosmid using the flyfosmid recombineering technique (Ejsmont et al., 2009). This way, we obtained fly lines expressing N-terminally tagged DOM (GFP-dom) and C-terminally tagged DOM-A (dom-A-GFP) and DOM-B (dom-B-GFP) from its chromosomal regulatory context (Fig. 1A,B). The latter two constructs express one tagged isoform, while the other is untagged. As a control, we replaced the dom locus with the GFP-3×FLAG cassette (Δdom-GFP) (Fig. 1B).

We complemented lethality and sterility of two dom alleles to assess the functionality of the dom transgenes. Viable homozygous dom¹ or dom⁹ fly lines were obtained by complementation with the dom transgenes (Fig. S1A). By contrast, the Δdom-GFP transgene did not rescue pupal lethality in the dom¹ allele and compromised viability in the dom⁹ allele (Fig. S1A). Furthermore, a characteristic larval phenotype of dom¹ mutants (Braun et al., 1998; Ruhf et al., 2001) was completely rescued by the dom transgenes (Fig. S1B,C). Importantly, egg laying capacity was restored to wild-type levels in homozygous dom⁹ females complemented with the dom transgenes, but not by the Δdom-GFP transgene (Fig. 1C). Our analysis provides the first comprehensive validation of dom allele phenotypes using complementation with dom transgenes.

Western blotting using the FLAG antibody showed expression of dom transgenes in ovaries (Fig. 1D) and larval brains (Fig. S1D) (Ruhf et al., 2001). We confirmed the ubiquitous expression of DOM-B in germline and somatic cells of ovarioles with the GFP antibody (Fig. 1F) (Ruhf et al., 2001; Xi and Xie, 2005). The specificity of antibody was confirmed as wild-type ovarioles showed only background staining (Fig. 1H). Interestingly, we also observed expression of DOM-A in germline and somatic cells of ovarioles (Fig. 1E). Both DOM isoforms were present in GSCs, cystoblasts, SSCs and follicle cells, with enrichment in oocyte nuclei in comparison to nurse cells (Fig. 1E,F). Nuclear localization was not due to the GFP tag since GFP expressed from the Δdom-GFP transgene showed only cytoplasmic signal (Fig. 1G). In summary, DOM-A and DOM-B are expressed in germline and somatic cells during oogenesis.

To dissect the cell type-specific requirements for DOM-A and DOM-B we induced RNAi by expressing small hairpin (sh) RNA directed against dom under the control of the Gal4/UAS system (Ni et al., 2011). We also generated transgenic flies expressing shRNA directed selectively against each individual DOM isoform and compared the resulting phenotypes with corresponding depletions of H2A.V and TIP60. GFP shRNA served as a control for non-specific effects.

The effect of ubiquitous depletion of DOM should resemble the lethal phenotype of the dom¹ allele (Ruhf et al., 2001). Indeed, we observed pupal lethality upon expression of dom-1 shRNA with an actin-Gal4 driver, whereas siblings lacking the driver were not affected (Fig. 2A). Remarkably, individual depletion of DOM-A or DOM-B led to pupal lethality, indicating essential functions for each isoform during fly development (Fig. 2A).

DOM and other TIP60 complex subunits are well known as positive regulators of Notch signaling in wing formation (Eissenberg et al., 2005; Ellis et al., 2015; Gause et al., 2006; Kwon et al., 2013). We also scored typical Notch phenotypes, such as defective wing margins and wing nicks, with high penetrance upon expression of two dom shRNAs from the C96-Gal4 driver (Fig. S2). Interestingly, the two DOM isoforms appear to have distinct functions in wing development since dom-A shRNA led to wing margin phenotypes with high penetrance, whereas dom-B shRNA had only minor effects (Fig. S2).

To document the efficiency and specificity of RNAi depletion we raised monoclonal antibodies against peptides of DOM-A (DOA1) and DOM-B (DOB2) for western blot detection. No DOM signals were detected in larval brain extracts with DOA1 and DOB2 antibodies upon dom-1 shRNA depletion driven by elav-Gal4 (Fig. 2B). Furthermore, signals for multiple DOM-A bands and DOM-B were strongly reduced upon depletion with dom-A and dom-B shRNAs, respectively (Fig. 2B). Multiple DOM-A bands have been reported previously (Kusch et al., 2004; Ruhf et al., 2001) and might reflect additional minor DOM-A isoforms or low-level degradation. As a control, levels of the SWI2/SNF2 ATPase ISWI were unaltered (Fig. 2B).

Despite all efforts, the monoclonal DOM-A and DOM-B antibodies could not be used for immunofluorescence microscopy applications. Alternatively, we combined GFP-tagged dom transgenes with shRNA constructs expressed from the traffic jam-Gal4 driver, which is specifically expressed in somatic follicle cells but not in germline nurse cells (Olivieri et al., 2010). As a control, we detected the dom transgenes in follicle and nurse cells using the GFP antibody (Fig. 2C,I). We did not detect GFP signal in follicle cell nuclei upon dom-1, dom-A and dom-B shRNA depletion in the respective GFP-dom, dom-A-GFP and dom-B-GFP transgenes (Fig. 2D,E,G,I). However, GFP signal was detected in nurse cell nuclei of the same egg chambers (Fig. 2D,E,G). Importantly, DOM-A or DOM-B depletion did not affect the localization and levels of the other isoform (Fig. 2F,H,I and Fig. S3), arguing for an independent localization of the two DOM isoforms. Furthermore, TIP60 depletion did not affect DOM-A or DOM-B localization (Fig. S4). We conclude that shRNAs against dom, dom-A and dom-B specifically and efficiently deplete their corresponding target protein.

Previous analysis could not reveal the specific contributions of DOM-A and DOM-B function to germline development (Ruhf et al., 2001; Yan et al., 2014). We confirmed the importance of DOM for fertility by scoring the egg laying capacity of F1 females from crosses of different dom shRNAs with the germline-specific MTD-Gal4 driver (Fig. 3A) (Yan et al., 2014). We found the egg laying capacity strongly impaired upon separate depletion of DOM isoforms, an effect that was similar if the H2A.V variant was depleted (Fig. 3B,C). Tip60 shRNAs showed moderately decreased egg laying capacity (Fig. 3C).

We stained ovaries of knockdown animals for the germline-specific cytoplasm marker Vasa and the follicle cell-specific adhesion molecule Fasciclin III (Fig. 3D-I). As expected, depletion with dom-1 shRNA led to an agametic phenotype with characteristic small ovaries and defects in cyst differentiation in the germarium (Fig. 3E) (Yan et al., 2014). The effects of depleting DOM-A and H2A.V resembled this phenotype (Fig. 3F,H). By contrast, germline development in the germarium appeared normal upon DOM-B and TIP60 depletion (Fig. 3G,I), but many defective late egg chambers were observed (Fig. 3G,I). We conclude that the two DOM isoforms are both required for proper oogenesis, but that they differ in their requirements: DOM-A is essential for early germline development, whereas DOM-B is crucial for egg chamber development at later stages of oogenesis.

DOM-B was required in germline cells for proper egg chamber development. The abundant defects in the germarium upon DOM-A depletion precluded the assessment of a similar role during later stages of oogenesis. Therefore, we combined the matα4-Gal4 driver, which is specifically expressed in the germline from stage 1 egg chambers onwards (Fig. 4A) (Yan et al., 2014), with different dom, H2A.V and Tip60 shRNAs and stained for Vasa and the cytoplasmic oocyte marker ORB. Notably, matα4-Gal4-directed knockdown did not affect the development of germarium and ovariole, but resulted in defective late egg chambers with short eggs and non-fragmented nurse cells characteristic of the ‘dumpless’ egg phenotype (Fig. 4B-G). Interestingly, individual loss of either DOM isoform or TIP60 led to ‘dumpless’ eggs comparable to DOM or H2A.V depletion (Fig. 4H,I) and accordingly to significantly reduced egg laying (Fig. 4J,K). We conclude that both DOM isoforms are required in germline cells outside of the germarium for egg chamber formation.

DOM function is not only required in the germline but is also essential for somatic follicle cell development (Xi and Xie, 2005). To address whether both DOM isoforms are also required for follicle cell function, we used the traffic jam-Gal4 driver (Fig. 5G) in combination with different shRNAs. We scored severe packaging phenotypes upon DOM, DOM-A, DOM-B, TIP60 or H2A.V depletion, ranging from two cysts in a single egg chamber to complete ovariole fusions (Fig. 5A-F). Furthermore, many germaria showed additional stalk-like structures without germline cysts (Fig. 5H-K), indicating defects in the coordination of follicle cell proliferation with cyst differentiation. Remarkably, we observed egg chambers with abnormal cyst numbers ranging from one to more than 16 cells (Fig. 5A-F), suggesting a soma-dependent proliferation defect in the germline. Defective late egg chambers showed signals for activated Caspase 3 (Fig. 5L-O). As a result, egg laying capacity was drastically impaired (Fig. 5P). We validated packaging phenotypes and sterility with another somatic driver line, c587-Gal4, which is expressed in somatic escort cells and early follicle cells in the germarium (Fig. S5) (Eliazer et al., 2011; Kai and Spradling, 2003). We conclude that DOM-A and DOM-B function is crucial in somatic follicle cells for the proper packaging of egg chambers.

In order to explore the roles of DOM isoforms in H2A.V incorporation into chromatin and in γH2A.V turnover in vivo, we raised a specific polyclonal antibody against an H2A.V peptide. We found H2A.V to be ubiquitously expressed in somatic and germline cells of the germarium with a notable enrichment in GSCs (Fig. 6A,B). Remarkably, H2A.V and γH2A.V signals were only detected in nurse cells up to stage 5 of oogenesis (Fig. 6C), and both signals were absent in later nurse cell nuclei (Fig. 6D). However, H2A.V was present in the surrounding follicle cells at all stages, providing convenient staining controls (Fig. 6A-D). These observations are in general agreement with previous studies (Joyce et al., 2011), except that we did not detect signals for H2A.V in oocytes, possibly owing to differences in H2A.V epitope sequence.

Upon depletion of DOM and DOM-B with MTD-Gal4 we found that H2A.V was lost specifically in germline cells of the germarium (Fig. 6F,H,L,M). H2A.V was essentially absent since the signals were comparable to those of H2A.V shRNA (Fig. 6I,L,M). As a control, levels of H2A.V were unaltered in somatic follicle cells, documenting the germline specificity of the effect (Fig. 6E-M). In remarkable contrast, loss of DOM-A or TIP60 did not affect levels of H2A.V (Fig. 6G,K-M). Likewise, depletion of ISWI ATPase, which causes similar germline phenotypes (Ables and Drummond-Barbosa, 2010; Xi and Xie, 2005; Yan et al., 2014), did not affect H2A.V levels (Fig. 6J,L,M). We conclude that DOM-B, but not DOM-A, is required for global incorporation of H2A.V into germline chromatin of the germarium.

To monitor H2A.V levels in germline cells outside of the germarium we again employed the matα4-Gal4 driver, which is specifically expressed from stage 1 of oogenesis onwards (Fig. 4A and Fig. S6) (Yan et al., 2014). As expected, matα4-Gal4-directed knockdown of DOM, DOM-A and DOM-B did not affect H2A.V levels in the germarium (Figs S7, S8). As a control, H2A.V shRNA knockdown specifically depleted H2A.V and γH2A.V in germline nurse cells of stage 3 egg chambers (Fig. 7A,E,G). Surprisingly, however, H2A.V and γH2A.V were unaltered in germline nurse cells up to stage 4 egg chambers upon DOM, DOM-A, DOM-B or TIP60 knockdown (Fig. 7B-D,F,G and Figs S7, S8). Notably, corresponding depletion of another SWI/SNF remodeler, INO80, with two different Ino80 shRNAs (Bhatia et al., 2010; Yan et al., 2014) did not affect H2A.V localization in ovarioles (Fig. S9). This suggests that global H2A.V incorporation into germline chromatin at stage 3 might be independent of DOM.

We next assessed how depletion of individual DOM isoforms affects global H2A.V and γH2A.V in nurse cells of later egg chamber stages. Remarkably, H2A.V and γH2A.V signals persisted in stage 8 egg chambers upon DOM or DOM-A depletion (Fig. 8B,C,H,I,M and Figs S7, S8). Additionally, a moderate but significant increase of H2A.V as well as γH2A.V foci was detected upon TIP60 depletion (Fig. 8F,L,M). By contrast, H2A.V was removed in the absence of DOM-B as in wild-type ovarioles (Fig. 8A,D,G,J,M and Figs S7, S8). Evidently, DOM-A/TIP60 are specifically involved in removing H2A.V from germline cells by stage 5 of oogenesis, illustrating once more the functional diversification of the two splice variants.

To address whether other cell types also utilize DOM for H2A.V incorporation, we used the somatic driver traffic jam-Gal4 for knockdown in follicle cells. As in germline cells of the germarium, we observed that depletion of DOM and DOM-B led to complete absence of H2A.V signals in follicle cell nuclei (Fig. 9A,B,D,H,I), when compared with the corresponding H2A.V knockdown (Fig. 9E,H,I). As before, dom-A and Tip60 shRNAs did not affect H2A.V levels in follicle cells (Fig. 9C,G,H,I) and neither did depletion of ISWI (Fig. 9F,H,I), which causes similar packaging phenotypes (Börner et al., 2016). Furthermore, canonical histone H2A (Fig. 9J,K) and heterochromatin-associated HP1 protein (Fig. 9L,M) were unaltered upon DOM depletion, indicating a globally intact chromatin structure in the absence of H2A.V. This finding highlights the specific requirement of the DOM-B isoform for H2A.V incorporation not only in the germline but also in dividing somatic follicle cells.

In summary, our systematic analysis showed that the nucleosome remodeling factor isoforms DOM-A and DOM-B have non-redundant functions in germline and soma in the formation of egg chambers. We further demonstrated that the two DOM isoforms have distinct functions in H2A.V exchange during D. melanogaster oogenesis.

In D. melanogaster the properties of the two ancient, ubiquitous histone H2A variants H2A.X and H2A.Z are combined in a single molecule, H2A.V (Baldi and Becker, 2013; Talbert and Henikoff, 2010). Given that H2A.V carries out functions as a DNA damage sensor and architectural element of active promoters (Madigan et al., 2002; Mavrich et al., 2008), as well as having further roles in heterochromatin formation (Chioda et al., 2010; Hanai et al., 2008; Qi et al., 2006), this histone appears loaded with regulatory potential. Accordingly, placement of the variant, either randomly along with canonical H2A during replication or more specifically through nucleosome remodeling factors, becomes a crucial determinant in its function. Mechanistic detail for replacement of H2A-H2B dimers with variants comes from the analysis of the yeast SWR1 complex, which incorporates H2A.Z in a stepwise manner at strategic positions next to promoters (Luk et al., 2010; Mizuguchi et al., 2004; Ranjan et al., 2013).

So far, the published phenotypes associated with dom mutant alleles have not been systematically complemented (Braun et al., 1997; Eissenberg et al., 2005; Ruhf et al., 2001). Our comprehensive complementation analysis shows that dom mutant phenotypes are indeed due to defects in the dom gene. Remarkably, dom lethality and sterility can be partially rescued by complementation with the orthologous human SRCAP gene, providing an impressive example of functional conservation of SWR1-like remodelers (Eissenberg et al., 2005). The contributions of the two splice variants DOM-A and DOM-B had not been assessed. We now demonstrate that both isoforms are essential for development, suggesting non-redundant functions. The DOM-A isoform contains a SANT domain followed by several poly-Q stretches, which are widely found in transcriptional regulators, where they may modulate protein interactions (Gemayel et al., 2015). By contrast, SANT domains are thought to function as histone tail interaction modules that couple binding to enzyme catalysis (Boyer et al., 2004). Therefore, the SANT domain in DOM-A could mediate specificity towards H2A.V eviction depending on particular functional contexts. These features are also present in the C-terminus of p400 (EP400), the second human SWR1 ortholog, but are absent in either DOM-B or SRCAP (Eissenberg et al., 2005). Remarkably, p400 interacts directly with TIP60 (Jha et al., 2013) and the SANT domain of p400 inhibits TIP60 catalytic activity (Park et al., 2010), providing an interesting lead for further investigation of DOM isoforms and TIP60 interactions.

We speculate that distinct functions of p400 and SRCAP in humans might be accommodated to some extent by the two DOM isoforms in flies. Accordingly, it will be interesting to explore whether the two isoforms reside in distinct complexes. Previous affinity purification of a TIP60-containing complex using a tagged pontin subunit apparently only identified DOM-A, but not DOM-B (Kusch et al., 2004). Following up on the initial observation of early defects in GSCs and cyst differentiation upon loss of DOM (Yan et al., 2014), we now find that this phenotype is exclusively caused by loss of DOM-A. Interestingly, studies with human embryonic stem cells show that p400/TIP60 (KAT5) integrates pluripotency signals to regulate gene expression (Chen et al., 2013; Fazzio et al., 2008), suggesting similar roles for DOM-A in GSCs. This is in contrast to requirements for both isoforms for germline development outside of the germarium, highlighting a developmental specialization of the two DOM remodelers.

DOM is also involved in the differentiation and function of SSCs in the germarium (Xi and Xie, 2005). Our data now document non-redundant requirements of both DOM isoforms in somatic cells for proper coordination of follicle cell proliferation with cyst differentiation. Failure to adjust these two processes leads to 16-cell cyst packaging defects that manifest as compound egg chambers. These rare phenotypes had previously only been described upon perturbation of some signaling pathways, such as Notch, or Polycomb regulation (Jackson and Blochlinger, 1997; Narbonne et al., 2004).

Because the phenotypes of DOM depletions resemble those of H2A.V depletion we favor the idea that many of the cell-specification defects are due to compromised H2A.V incorporation, depriving key promoters of the H2A.Z-related architectural function. Alternatively, scaffolding activities might partially explain some roles of chromatin remodelers, as suggested for SRCAP (Bowman et al., 2011). So far, our knowledge of the mechanisms of H2A.V incorporation has been anecdotal (Joyce et al., 2011; Kusch et al., 2014; Lu et al., 2007). Our comprehensive analysis revealed a specific involvement of DOM-B for the incorporation of H2A.V into chromatin at the global level. The N-termini of SWR1 and DOM-B harbor the HSA and ATPase spacer domains (Morrison and Shen, 2009), with interaction surfaces for further complex subunits (Billon and Côté, 2012; Gerhold and Gasser, 2014), and an additional H2A.Z-H2B dimer binding site (Wu et al., 2005, 2009). Given the requirement for both isoforms for cell specification during oogenesis, we speculate that DOM-B might serve to incorporate bulk H2A.V into chromatin similar to SWR1, whereas DOM-A would be more involved in the regulatory refinement of location.

Although the failures in cell specification and egg morphogenesis are likely to be explained by loss of the H2A.Z-related features of H2A.V, ablation of DOM might also compromise the DNA damage response, which involves phosphorylation of H2A.V (γH2A.V). Conceivably, the role of γH2A.V as a DNA damage sensor might be best fulfilled by a broad distribution of H2A.V throughout the chromatin (Baldi and Becker, 2013). Such an untargeted incorporation may be achieved by stochastic, chaperone-mediated incorporation during replication (Li et al., 2012) or by an untargeted activity of DOM-B. We observed DOM-independent incorporation in endoreplicating polyploid nurse cells of stage 3 egg chambers, where global H2A.V and γH2A.V signals did not depend on DOM. Immunofluorescence microscopy may lack the sensitivity to detect DOM-dependent incorporation of H2A.V at some specific sites. Nevertheless, DOM-independent incorporation of H2A.V might serve to cope with many naturally occurring DNA double-strand breaks during the massive endoreplication of nurse cells.

There is some evidence that nucleosome remodelers not only incorporate H2A variants but can also remove them. In yeast, the genome-wide distribution of H2A.Z appears to be established by the antagonistic functions of the SWR1 and Ino80 remodeling complexes, where Ino80 replaces stray H2A.Z-H2B with canonical H2A-H2B dimers (Papamichos-Chronakis et al., 2011). A recent study identified the vertebrate-specific histone chaperone ANP32E as part of a TIP60/p400 complex that facilitates the eviction of H2A.Z-H2B dimers from chromatin (Obri et al., 2014). Remarkably, in D. melanogaster a TIP60/DOM-A complex is involved in a similar reaction. The TIP60/DOM-A complex acetylates γH2A.V at lysine 5 to facilitate exchange of γH2A.V by unmodified H2A.V during the DNA damage response (Kusch et al., 2004). Furthermore, it has been speculated that H2A.V and γH2A.V could be actively removed from nurse cells, since corresponding signals are absent from stage 5 onwards (Jang et al., 2003; Joyce et al., 2011). We now demonstrate that depletion of DOM-A and TIP60 leads to the persistence of H2A.V and γH2A.V in nurse cells of late egg chambers, clearly documenting the ability of the remodeler to remove bulk H2A.V and variants modified during DNA damage induction.

Our findings highlight the specific requirements of DOM splice variants for the incorporation and removal of H2A.V during D. melanogaster oogenesis. It remains an interesting and challenging question how DOM-A and DOM-B complexes are targeted genome-wide and function in vivo to establish specific H2A.V patterns in different cell types during development.

The genomic region of dom was modified by recombineering in Escherichia coli using pRedFLP4 recombination technology (Ejsmont et al., 2009). Details are provided in the supplementary Materials and Methods and in Tables S1 and S2. All fosmid, UAS-shRNA and Gal4 driver lines are listed in the supplementary Materials and Methods. For ovary analysis, UAS-shRNA males were crossed with Gal4 driver virgins at 29°C and 5- to 7-day-old F1 females were used, unless stated otherwise.

Males and female virgins of dom¹/bcg; dom fosmid/+ or of dom⁹/bcg; dom fosmid/+ genotype were crossed. F1 offspring were analyzed for bcg phenotype and rescue of pupal lethality was determined as percentage of observed to expected frequency of homozygous dom¹ or dom⁹ offspring. Mean values in percentage with s.d. from three biological replicates were calculated.

Necrotic lymph glands were scored in third instar larvae of homozygous w¹¹¹⁸, dom¹, dom¹;GFP-dom, dom¹;dom-A-GFP and dom¹;dom-B-GFP. Mean values in percentage with s.d. from three biological replicates were calculated.

Homozygous female virgins for w¹¹¹⁸, dom⁹, dom⁹;GFP-dom, dom⁹;dom-A-GFP, dom⁹;dom-B-GFP and dom⁹;Δdom-GFP were collected for 2-3 days at 25°C. Females of each genotype were mated with w¹¹¹⁸ males in vials with yeast paste for 2 days. Females were placed in individual vials for 3 days without males and the egg laying capacity determined as the number of eggs laid per female per day. The data were averaged and mean values with s.d. from three biological replicates calculated.

UAS-shRNA males for GFP, dom-1, dom-A and dom-B were crossed with actin-Gal4/CyO driver virgins at 25°C. Pupae were transferred to new vials and numbered. A 1:1 ratio of CyO and actin-Gal4 F1 siblings was expected. CyO phenotype was counted and pupal hatching was determined as the observed to expected frequency of CyO and actin-Gal4 F1 siblings. Mean values in percentage with s.d. from three biological replicates were calculated.

Homozygous UAS-shRNA males for GFP, dom-1, dom-2, dom-A and dom-B were crossed with MTD-, matα4-, c587- or traffic jam-Gal4 driver virgins at 29°C. F1 female virgins of each genotype (2-3 days old) were mated with w¹¹¹⁸ males in vials with yeast paste for 2 days. Females were then placed in individual vials without males and the egg laying capacity determined as described above.

UAS-shRNA males for GFP, dom-1, dom-2, dom-A and dom-B were crossed with C96-Gal4 driver virgins at 29°C. Wing phenotypes were scored in F1 offspring and relative mean values in percentages with s.d. from three biological replicates were calculated.

DOA1, DOB2 and H2A.V antibodies were made for this study as described in the supplementary Materials and Methods.

Immunofluorescence microscopy was performed using standard procedures and a Leica TCS SP5 II confocal microscope. Images were processed using ImageJ (NIH) and Adobe Photoshop. For details, including the antibodies used, see the supplementary Materials and Methods.

Twelve pairs of ovaries or 20 brains of third instar larvae were dissected, homogenized in Laemmli buffer and incubated at 95°C for 5 min. Western blot was performed using standard procedures; antibodies are listed in the supplementary Materials and Methods.

Two-tailed Student's t-test was used. P-values are given in the figure legends.

Some stocks were obtained from the Bloomington Drosophila Stock Center [NIH P40OD018537]. We thank the TRiP at Harvard Medical School [NIH/NIGMS R01-GM084947] for transgenic RNAi fly lines. The monoclonal antibodies ORB 4H8 and 6H4, 7G10 anti-Fasciclin III, Spectrin 3A9 (323 or M10-2), HP1 C1A9 and UNC93-5.2.1 were obtained from the Developmental Studies Hybridoma Bank. We thank A. Eberharter, J.-R. Huynh, J. Jiang, E. Kremmer, J. Müller, H. Saumweber, A. Spradling and P. Tomancak for providing reagents; and S. Baldi, D. Jain, N. Steffen and S. Vengadasalam for stimulating discussions.