Safeguarding Drosophila female germ cell identity depends on an H3K9me3 mini domain guided by a ZAD zinc finger protein

H3K9me3-based gene silencing is a conserved strategy for securing cell fate, but the mechanisms controlling lineage-specific installation of this epigenetic mark remain unclear. In Drosophila, H3K9 methylation plays an essential role in securing female germ cell fate by silencing lineage inappropriate phf7 transcription. Thus, phf7 regulation in the female germline provides a powerful system to dissect the molecular mechanism underlying H3K9me3 deposition onto protein coding genes. Here we used genetic studies to identify the essential cis-regulatory elements, finding that the sequences required for H3K9me3 deposition are conserved across Drosophila species. Transposable elements are also silenced by an H3K9me3-mediated mechanism. But our finding that phf7 regulation does not require the dedicated piRNA pathway components, piwi, aub, rhino, panx, and nxf2, indicates that the mechanisms of H3K9me3 recruitment are distinct. Lastly, we discovered that an uncharacterized member of the zinc finger associated domain (ZAD) containing C2H2 zinc finger protein family, IDENTITY CRISIS (IDC; CG4936), is necessary for H3K9me3 deposition onto phf7. Loss of idc in germ cells interferes with phf7 transcriptional regulation and H3K9me3 deposition, resulting in ectopic PHF7 protein expression. IDC’s role is likely to be direct, as it localizes to a conserved domain within the phf7 gene. Collectively, our findings support a model in which IDC guides sequence-specific establishment of an H3K9me3 mini domain, thereby preventing accidental female-to-male programming.


Reviewer #1
This is a well-written paper with a series of logical, well executed experiments that support the author's conclusion that the ZAD zinc-finger protein IDC plays a direct role in establishing the H3K9me3 domain that regulates the expression of the phf7 gene in the ovary. The importance of this finding is well described in the manuscript: while a lot is known about how H3K27me3-repressive chromatin domains are established, little is known about how small H3K9me3 domains are established. Making good use of public data, the authors clearly show that the piRNA pathway, important in forming large H3K9me3 domains, is not required for the formation of the small H3K9me3 domain that covers the male-specific phf7 promoter. The authors identified the ZAD zinc-finger gene IDC in a previous RNAi screen as a candidate for a phf7 transcriptional regulator. Here they show that (1) knocking down idc with RNAi in ovaries leads to the production of the male-specific phf7 RNA and production of the phf7 protein in ovaries 2) H3K9me3 levels are reduced over the phf7 gene in idc germline mutant clones and (3) IDC binds to the phf7 gene. This paper is made even more interesting by the potential similarities between the recruitment of the H3K9me3 methyltransferase by the KRAB-zinc finger family in mammals and the ZAD zinc-finger family in flies.
Thank you.
I have a few questions: 1) Regarding Fig. 1A and the discussion of the repetitive element: Does the phf7 gene from the related Drosophila species also have a repetitive element (perhaps unrelated)?
Interestingly, we did not find a repetitive element in the introns of either D. similans or D. yakuba.
2) Regarding Fig. 2A. It's not clear to me exactly what is in the transgene. From the diagram it looks like TSS1 and part of the non-coding region upstream of TSS2 are in the transgene, however, the title of the figure is "Non-coding sequences within the first intron are sufficient for H3K9me3 deposition." This is important because IDC binds to TSS1. If the exons are included then change the title to "Non-coding sequences within the first intron are required for H3K9me3 deposition" Sorry for the confusion. The first exon is included in the transgenes. To clarify, we have modified Fig 2. We have added details about the construction in Fig S2. We have also changed the title of And modified the text to read: (Lines 133-137) To complement these studies, we used a transgenic approach to identify the sequences capable of promoting H3K9me3 when inserted into a heterologous genomic location on the 3 rd chromosome using site specific integration. We first created a transgene, that includes a portion of the phf7 DR mutant allele, extending from the first male specific exon to the beginning of the open reading frame in exon 2 (S2 Fig). 3) Fig. 3A, please put a diagram of the vector used to do this experiment. Where were the fragments cloned in? I can't really visualize it based on the description in the methods. We can't come up with a reference; as far as we know, our work provides the first example… we were just being cautious in saying one of the first examples. We have therefore changed the text to read the first example (line 265).

Reviewer #2
In this manuscript, Shapiro-Kulnane et al. take advantage of the phf7 locus to investigate how H3K9me3 mediated silencing is promoted at protein coding genes. This builds on prior work from the lab that had identified H3K9me3 deposition over a testis-specific transcription start site for phf7 as important for silencing expression of the male-specific protein.
In a recently published screen, the authors identified CG4936 (IDC) as a putative regulator of phf7 expression. Here, they link these two studies by directly showing that IDC binds near the male-specific TSS of phf7 and is essential for H3K9me3. Overall, the manuscript is clear, and the experiments are rigorously performed. However, the impact beyond what has been previously shown by the lab is somewhat limited and additional experiments could strengthen the conclusions and the model. Thank you. We note that our previously published screen simply identified CG4936 as one of the 8 ZAD-ZNF genes with functions in the female germline, not as a putative phf7 regulator. In addition to demonstrating that CG4936, now called idc, is required for H3K9me3-mediated silencing of phf7, we identify required cis-acting sequences, and demonstrate that the piRNA pathway is not involved in silencing. We feel that this work represents a major advance in our understanding of how silencing occurs.
We have modified the text to make it clear that our recently published screen simply identified CG4936 as a gene required for oogenesis, not as a putative regulator of phf7 expression.
(Lines [188][189][190][191] We recently tested the function of 68 of the 93 ZAD-ZNF encoding genes in female germ cells by performing an RNAi screen [55]. This screen identified eight ZAD-ZNF genes required for oogenesis. Here, we focus on CG4936 which we name identity crisis (idc).
Major issues: 1. The major novel finding in this work is the role of IDC in transcriptional repression in the germline, which was already hinted at by the results of the RNAi screen. To further support this conclusion, the authors should validate the specificity of their RNAi construct and/or use an orthogonal approach to confirm the specificity.
We apologize for this omission. We now present confocal images to show that the TRiP short hairpin line we used produces an effective protein knockdown in Figure S4 and modified the body of the text to read:

(Lines 202-203) We demonstrated knockdown efficiency by showing that idc GLKD significantly reduces IDC protein levels in female germ cells, but not in the surrounding somatic cells (S4 Fig).
The idc loss of function is lethal, therefore our analysis of idc function in the adult germline necessitates the use of RNAi reagents. We cannot validate our findings by using a different line because there is only one germline optimized shRNA line in existence. We also note that UP-TORR indicates that there are no predicted off-target effects (see screen shot below for HMC05569). We have added a note to the methods section that reads: (Lines 323-324). Although off-target effects remain a concern for RNAi-induced knockdown studies, neither RNAi line is predicted to have off-target activity (https://www.flyrnai.org/up-torr/UptorrFly.jsp).
2. The expression pattern of IDC as a whole is not clearly explained. Is IDC expressed broadly in all tissues or only in the germline?
We apologize for not providing enough information about the expression pattern. To make clear that it is likely that IDC is ubiquitously expressed, but that we have limited our analysis to its expression pattern in the ovary, we begin this section with the following statement:

(Line 220-221) Published modENCODE RNA-seq data sets indicates that the idc RNA is broadly expressed throughout development [62]. To examine IDC protein expression in the ovary,…
And modified the description to read: (Fig 6A). Notably, the IDC-GFP protein is nuclear. Furthermore, IDC-GFP is tightly associated with the nurse cell polytene chromosomes, consistent with its presumed DNA binding activity (Fig 6B).

(Lines 226-231) We observed IDC-GFP in the somatic cells, the nurse cells and the oocyte
In the germarium, at the anterior end of the ovariole, we observed prominent IDC-GFP staining in the nucleus of only 3 to 6 germ cells (Fig 6A).
And we end with:

(Lines 242-243) Together, these observations indicate that the nuclear IDC protein is broadly expressed, with prominent expression in the GSCs and their immediate daughter cells.
In the Discussion, the authors mention that IDC is also expressed in male germ cells, suggesting that the simple model presented in Figure 7B is misleading.
This comment led us to redraw the model in Fig 7B. The authors have all the tools in hand to test if IDC binds phf7 in the male germline. The authors discuss a potential candidate for mediating the female-specific functions (STWL). It is unclear why they do not directly test this hypothesis.
These are interesting and important questions, but they are beyond the scope of our current work.
3. To better understand the function of IDC in the female-specific silencing, it would be useful to determine IDC binding beyond the phf7 locus both in the male and female germlines. This would provide insights into additional targets and possibly more generally into the role of the ZAD-ZNF gene family.
We agree that a better understanding of IDC function will require a global analysis. However, we feel that this line of investigation is well beyond the scope of this work which is focused on phf7 silencing.
Along these same lines, in Figure 6 IDC-GFP appears to be broadly expressed in the female germline. It would be useful for the authors to test and/or discuss in which cell types IDC is required.
We generated an idc null allele and found that the animals do not survive to adulthood (see materials and methods). As our focus is on how phf7 is silenced in the female germline, we did not look beyond its requirement in the female germline.
Does overexpression of IDC lead to H3K9me3 deposition over phf7?
IDC is broadly expressed, so I am not sure what the impact of overexpression would be. 4. A major conclusion from the manuscript is that the determinants required for H3K9me3 deposition at phf7 are within element A of the 1st intron. However, additional experiments would strengthen this conclusion, which is currently based on the fact that deletion of a set of repeats does not disrupt H3K9me3 while a deletion that also includes the A region does. With the data presented, it remains possible that these two regions are redundant. Deletions of the A region alone should be tested. Likewise, it would be useful to test sufficiency. Does insertion of the A region in a transgene lead to H3K9me3 in the female germline?
These are excellent points. We have modified our conclusions to read: (Lines 141-150). As expected, we found that these sequences included in the DR transgene promotes H3K9me3 accumulation. (Fig 2A and Fig 2B). These data reinforce our conclusion that the tandem repeats (element R) are not required for H3K9me3 accumulation. In contrast, we found that H3K9me3 did not accumulate on a second transgene in which both elements A and R were removed (Fig 2A and Fig 2B). We therefore conclude that the sequences that remain, including the first exon and region B, are not sufficient for H3K9me3 recruitment. These results also establish that element A contains cis-regulatory determinants required for H3K9me3 recruitment. It remains to be determined whether the element A determinants are sufficient for H3K9me3 recruitment or function redundantly with sequences within element R. .

5.
As written, some of the data seems contradictory. The A region in the first intron appears important for H3K9me3 deposition, but IDC appears to bind the exon ( Figure 7A). Do the authors think that there are additional factors that recognize the A region? At the very least, the authors should discuss these conflicting data.
We elaborate on this excellent point in the discussion.
(Lines 267-275) Although our work is consistent with a simple model in which the SETDB1 H3K9me3 methyltransferase is recruited to phf7 by IDC, the mechanism by which IDC guides the methylation machinery to phf7 remains an open question. For example, it remains unclear whether recruitment is direct, as our attempts to co-immunoprecipitate SETDB1 and IDC were unsuccessful. Furthermore, while we establish that IDC is required for H3K9me3 recruitment, our chromatin transgenic reporter assays show that the region to which it binds, the conserved first exon, is not sufficient. This observation, together with our identification of a second conserved cis-regulatory element within the adjoining intron invites speculation that IDC works in conjunction with other sequence-specific recruitment factors.

Minor issues:
In line 125 it says the p value is 0.0003, however in the corresponding figure it says the p value is 0.003 Thank you for catching this error. Fixed.
All of the comparisons made with RT-qPCR and ChIP-qPCR should have statistics. Figure 4 should include scales for the y-axes for the genome browser tracks.

Done
Fixed.