Stonewall and Brickwall: Two Partially Redundant Determinants Required for the Maintenance of Female Germline in Drosophila

Proper specification of germline stem cells (GSCs) in Drosophila ovaries depends on niche derived non-autonomous signaling and cell autonomous components of transcriptional machinery. Stonewall (Stwl), a MADF-BESS family protein, is one of the cell intrinsic transcriptional regulators involved in the establishment and/or maintenance of GSC fate in Drosophila ovaries. Here we report identification and functional characterization of another member of the same protein family, CG3838/ Brickwall (Brwl) with analogous functions. Loss of function alleles of brwl exhibit age dependent progressive degeneration of the developing ovarioles and loss of GSCs. Supporting the conclusion that the structural deterioration of mutant egg chambers is a result of apoptotic cell death, activated caspase levels are considerably elevated in brwl- ovaries. Moreover, as in the case of stwl mutants, on several instances, loss of brwl activity results in fusion of egg chambers and misspecification of the oocyte. Importantly, brwl phenotypes can be partially rescued by germline specific over-expression of stwl arguing for overlapping yet distinct functional capabilities of the two proteins. Taken together with our phylogenetic analysis, these data suggest that brwl and stwl likely share a common MADF-BESS ancestor and they are expressed in overlapping spatiotemporal domains to ensure robust development of the female germline.

modulate transcription of individual genes. Alternatively they can also influence a group of genes simultaneously by regulating chromatin remodeling (Boyer et al. 2004). The Drosophila melanogaster genome consists of fifty-five Myb/SANT like in Adf (MADF) genes (Shukla et al. 2014). In flies, MADF domain containing genes have been implicated in critical biological processes such as hybrid incompatibility (Brideau et al. 2006;Maheshwari et al. 2008), transposition (Casola et al. 2007) and transcriptional regulation Cutler et al. 1998;Timmerman et al. 2013). A subgroup of the MADF family is the MADF-BESS family, which consists of sixteen proteins that contain both, a N-terminal MADF and C-terminal BEAF, Su-Var (3-7), Stonewall (BESS) domain (Bhaskar and Courey 2002). The BESS domain, like MADF, may also be a modified SANT domain (Cutler et al. 1998;Bhaskar and Courey 2002). The MADF domain is a DNA binding domain while the BESS domain is thought to be a proteinprotein interaction domain (Bhaskar and Courey 2002;Maheshwari et al. 2008).
The MADF-BESS family in Drosophila has been proposed as an example of a gene family that has duplicated and expanded (Lander et al. 2001;Shukla et al. 2014) over 40 million years (My) ago and has in the intervening period sub-functionalized (Shukla et al. 2014). We have been exploring the redundancy and sub-functionalization within this family using molecular genetic analysis (Shukla et al. 2014). In an earlier study, we reported that three members of the family (hinge1, hinge2 and hinge3) show overlapping functions during development of the wing-hinge, primarily by negative regulation of Wingless (Wg) expression (Shukla et al. 2014). Moreover, RNAi mediated knockdowns of all sixteen members in different tissues of the body suggested that different subsets of genes function in distinct spatiotemporal domains. For example, only Dlip3, but not any other MADF-BESS gene, is important for eye development  while Coop (Song et al. 2010) is involved in patterning of the wing by regulating Wingless signaling.
Drosophila gene stwl (Clark andMcKearin 1996) (McKearin 1997) encodes for a heterochromatin-associated protein belonging to the MADF-BESS family (Delattre et al. 2004;Delattre et al. 2000). Aside from being the largest (1037 amino acids) member of this family, it also contains a modified SANT domain (Boyer et al. 2004) and has been implicated in regulating chromatin compaction (Yi et al. 2009). Functional characterization of Stwl protein has suggested its possible involvement during position-effect variegation (PEV) similar to Su(var)3-7. Furthermore, it is also thought that Stwl achieves this by affecting methylation of Lys residues in Histone tails (Yi et al. 2009). As would be expected of a pleiotropic chromatin regulator, mutations in stwl display phenotypes in distinct and seemingly unrelated developmental contexts including wing patterning (Shukla et al. 2014;Brun et al. 2006), DNA damage response (Yi et al. 2009) and, female germline development. (Clark and McKearin 1996) (Akiyama 2002).
stwl function seems to manifest at a number of different stages of ovarian development. Immunostaining analysis with anti-Stwl antibodies indicated that Stwl protein is nuclear protein and is specifically detected in the GSCs and their progeny including cystoblasts and cystocytes. Accordingly, Stwl appears to be required in the germline stem cell (GSC) lineage for GSC maintenance as loss of stwl activity leads to corresponding loss of GSCs (Clark and McKearin 1996;Maines et al. 2007;Akiyama 2002). Interestingly requirement of Stwl appears to be specific to GSCs alone as GSCs undergo precocious differentiation in the absence of stwl, however differentiated population i.e., cystocytes, marked by the presence of fusome, are relatively unaffected. Furthermore, oocyte specification is also defective in stwl ovaries resulting in excess number of cells that retain oocyte determinants such as Orb protein (Clark and McKearin 1996). Moreover, a fraction of stwl cysts undergo an extra round of division (Clark and McKearin 1996). Although many of the phenotypic consequences due to loss of stwl function are readily apparent, they are partially penetrant and are also sensitive to external conditions such as temperature. The variability of stwl phenotypes led Clark & McKearin to speculate that other proteins might partially substitute for stwl function. Thus, they in fact, predicted that 'Partial functional redundancy for Stwl protein might account for the variable severity of stwl phenotypes' (Clark and McKearin 1996).
Here we have tested this prediction systematically. The availability of TRiP RNA interference lines (Ni et al. 2009;Ni et al. 2011;Perkins et al. 2015) and germline specific UAS/Gal4 strains (Staller et al. 2013;Rørth 1998) allow for efficient maternal knockdown. Using this strategy, we searched for another member of MADF-BESS family that potentially resembles stwl. Based on female germline specific knockdown and subsequent phenotypic analysis, we have uncovered CG3838, a member of the MADF-BESS family, that has germline-specific functions. Further supporting the claim, transposon insertion mutants spanning the locus are viable but only semi-fertile and fertility of adults deteriorates in an age dependent manner. Moreover, ovaries from mutant females show GSC lineage specific defects. The loss of function phenotypes of CG3838 and stwl thus seem to overlap. Confirming the assertion that the two proteins are partially redundant, Stwl over-expression in the GSC lineage is sufficient to rescue several, but not all, aspects of loss of CG3838 function. Based on our analysis, stwl and CG3838, now christened brickwall (brwl), seem to retain overlapping functions which may reflect the presence of an ancestral gene that existed over 40 My ago.

Drosophila husbandry
All flies were raised at 25°in standard corn meal agar. Crosses were set up at 25°and 29°. The females of the F1 progeny were dissected for the phenotypes in all cases.

Fly Lines
For ovarian somatic and maternal knockdowns, VALIUM10, VALIUM20 and VALIUM22 Transgenic RNAi Project lines procured from Bloomington Drosophila Stock Center (BDSC) were used (Ni et al. 2009) (Perkins et al. 2015) in this study. VALIUM lines used have been listed in Table 1. Here, we have used notations V10/V20/V22 to indicate a VALIUM10/VALIUM20 or VALIUM22 line respectively. For example, Interaction Map), indicates use of C-terminal FLAG-HA tagged DNA clones from the DPiM (Guruharsha et al. 2011) project, procured from Drosophila Genomic Resource Centre (DGRC), Indiana and then injected in the Drosophila Transgenic Facilty (http://www.ccamp.res.in/ Fly-facility) at NCBS Bangalore. FlyORF lines (https://flyorf.ch/) are from the Zurich FlyORFeome project (Bischof et al. 2013). The expression of Brwl in the GSC lineage was monitored by imaging a CG3838:GFP.FTBP fusion available in BDSC (#56150). The fly line was a Bac construct created as part of the modERN project (Flybase FBrf0225001; Personal Communication by A. Victorsen and K. White 2014). Daughters against Dpp (dad)-lacZ expression (Song et al. 2004) was monitored to mark GSCs. The expression of Stwl was monitored using Stwl:GFP protein trap line procured from FlyTrap (Kelso et al. 2004).
Screening MADF-BESS genes with somatic and maternal drivers F1 Ovaries were screened for phenotypes using nos-Gal4 as a maternal driver and c587-Gal4 as a somatic driver. For the screen, 5 F1 females were dissected at 3, 10 and 18 days, ovaries stained with Phalloidin and DAPI and visualized under an epifluorescence microscope. $150 ovarioles were examined by eye for each time point and representative images were stored. V10 lines were tested with the c587-Gal4, V20 lines tested with both somatic (c587-Gal4) and germline (nos-Gal4) drivers and V22 lines were tested only with nos-Gal4.
Transposon insertion Lines brwl KG00824 (BDSC 12901) is P-element insertion line in the 59 of CG3838 and is a part of BDGP Gene Disruption Project collection (Bellen et al. 2004). brwl MI054561 (BDSC 42326) is Minos-element insertion line with the Minos cassette inserted in the 39 of CG3838 and is the part of transposon Minos-mediated integration cassette (MiMic) collection (Venken et al. 2011). Here, these are referred to in short, as brwl KG0 and brwl MI0 .

Phylogenetic Analysis
The coding sequence of MADF and the BESS domains from brwl and stwl genes in 12 Drosophila species was predicted using NCBI Conserved Domain (Marchler-Bauer et al. 2015). The sequences were aligned using MUSCLE (Edgar 2004) implemented in MEGA7.0 (Kumar et al. 2016). Pairwise raw genetic distances were calculated in MEGA7.0 (Kumar et al. 2016). Sequences were partitioned into three codon positions to create a full partition and then the greedy strategy (Lanfear et al. 2012) implemented in IQ-Tree (Nguyen et al. 2015) was used to find the right partitioning scheme based on minimum Bayesian Information Criterion (BIC) (Nei M 2000) (Schwarz 1978). Pairwise codon substitution pattern (v = dN/dS) was determined in DnaSP (Librado and Rozas 2009). Each domain was tested for recombination and appropriate codon substitution model in n Table 1 Summary of a targeted screen setup to evaluate germline and somatic roles for MADF-BESS genes. RNAi lines from the Harvard in-vivo fly RNAi project (TRiP; https://fgr.hms.harvard.edu/fly-in-vivo-rnai) (NI et al. 2009;NI et al. 2011) were used for the experiments. VALIUM10 (V10), VALIUM20 (V20) and VALIUM22 (V22) lines were procured from the Bloomington Drosophila Stock Centre (BDSC) and crossed with nos-Gal4:VP16 (#4937) and c587-Gal4 (#67747). Experiments were done both at 25°C and 29°C. Data shown is for the 25°C cross, which was similar to the results obtained at 29°C. The data for maternal knockdown using both MTD-Gal4 and nos-Gal4;UAS-Dicer2 has been taken from Suppl.
36785 (V22) Phenotype in ovary seen in flies; 7 days post-eclosion 35415 (V22) Strong phenotype in ovary; Seen in female flies immediately after eclosion No Eggs a 'Normal' indicates a normal ovary without any fusion or degeneration related phenotypes in ovarioles from 10 and 18 day old females. We do see minor follicle phenotypes, such as layering, in some of the crosses but these are distinct from the phenotypes seen in brwl mutants. Also seen are significant variations in number of animals emerging from each cross.

RESULTS
A targeted screen to uncover MADF-BESS genes required in the GSC lineage As previously reported (Clark and McKearin 1996;Akiyama 2002), reducing stwl function in the female germline led to ovarian phenotypes. This phenotype is replicated by driving stwl RNAi (V20) in the germline using nos-Gal4 [See Suppl. Figure 1 and (Yan et al. 2014)]. Consistent with the previous reports the phenotypic consequences due to loss of stwl in adult ovaries were severe resulting in very small ovarioles that degenerated within days. Earlier studies have also demonstrated that Stwl is expressed in the GSCs and cysts in the germarium (Clark and McKearin 1996). In order to explore the possible germline specific function(s) of other MADF-BESS domain proteins, we sought to compromise the transcript levels of other MADF-BESS family genes in the female germline using a similar RNAi based approach. By employing the UAS-Gal4 system (Brand and Perrimon 1993;Duffy 2002), we tested MADF-BESS genes for which VALIUM20 (V20) and VALIUM22 (V22) RNAi lines (Hu et al. 2017;Perkins et al. 2015) are available (Table 1). nos-Gal4 was employed as the primary Gal4 driver to knockdown the corresponding MADF-BESS genes specifically in the germline (Staller et al. 2013) by using both V20 and V22 lines. As summarized in Table 1 (column #3) compromising the function of only one other gene (CG3838), christened brickwall (brwl), resulted in ovarian phenotypes ( Figure 1A-B). Interestingly, Yan et al. (2014) had earlier screened available VALIUM lines using both nos-Gal4:VP16;UAS-Dicer2 and Maternal Triple Driver (MTD)-Gal4. Their studies, also summarized in Table 1 (fourth column), indicated that MADF-BESS genes, with the exception of stwl did not affect oogenesis significantly, after maternal knockdown of show phenotypes similar to that of the brwl maternal RNAi knockdowns, with all ovarioles showing degenerating egg chambers by the 18 th day. (L) Quantitative data representing the percentage of ovarioles that show the phenotype on third, tenth (n = 80) and eighteenth (n = 25) day after fly eclosion. Ovaries from Canton-special (CS) animals were used as a control for statistical analyses. Ã , ÃÃ and ÃÃÃ denote P , 0.05, P , 0.01, and P , 0.001 respectively for this and all subsequent figures. N: number of adult females = 3-20 and n: number of ovarioles = 25-80.
transcripts. The study did not uncover CG3838 as a candidate possibly because CG3838/brwl mutant flies are fertile for the first ten days of egg laying, showing age-dependent loss of fertility, as described in the next section.
The ovarian defects induced by the RNAi based knockdown of brwl prompted us to assess the expression of Brwl in the Ovary. Since antibodies against Brwl protein are unavailable, we used a Brwl:GFP fusion (BDSC 56150) line (See Materials and Methods) to monitor the expression of Brwl protein. Brwl:GFP is expressed in the adult ovarioles, where it marks the nuclei of both nurse and follicle cells in developing egg chambers, Stage-1 onwards ( Figure 1C). This is similar to the expression of Stwl:GFP (See Figure 2H, Buszczak et al. 2007). We were unable to detect nuclear expression in the GSCs using the Brwl:GFP line. This was also true for the Stwl:GFP line in our hands and in published literature (Rohrbaugh et al. 2013). The evidence for nuclear GSC staining for Stwl is based on the use of an anti-Stwl antibody (Clark and McKearin 1996;Maines et al. 2007).
Unlike the nos-Gal4 driver, MAT-Gal4 does not express in the GSCs (Perkins et al. 2015;Yan et al. 2014) and drives expression in the cystoblasts, Stage-1 onwards. MAT-Gal4; UAS-Brwl RNAi (V22) females (Suppl. Figure 1E) displayed similar age dependent phenotypes as in the case of nos-Gal4. Taken together, the RNAi based inactivation suggests that brwl activity is likely required in the germline from very early stages of development.
Interestingly, Brwl-GFP is also expressed in the somatic (follicular) nuclei ( Figure 1C, Suppl. Figure 1B). This expression pattern raises the possibility that brwl function is also required in the soma for proper germline development. We tested this possibility by compromising brwl function in the somatic cells using a variety of somatic drivers, namely GR1-Gal4, c587-Gal4, tj-gal4 and hh-Gal4. The VALIUM10 Brwl RNAi line (BDSC 31922) was used for these experiments. Surprisingly however, none of the resulting F1 females showed any significant ovarian phenotypes, even when the experiments were conducted at an elevated temperature (29°). To confirm these observations, we also tested TJ-Gal4 in combination with UAS-Dicer2 to enhance brwl knockdown and could not detect any role for brwl in the soma. Further, we tested all other MADF-BESS genes, using available V10 and V20 lines in combination with c587-Gal4. The lack of phenotypes (Table 1) suggest at best a relatively minor or insignificant role for brwl, as also for other MADF-BESS genes in the somatic cells of the ovary. Thus, while a role for brwl in the soma is possible, our experiments do not support this possibility.
RNAi mediated germline knockdown of CG3838 uncovers age dependent ovarian phenotypes To start characterizing the egg chambers compromised for brwl activity, the ovaries were first stained with DAPI, a DNA dye. Readily visible phenotypes upon brwl RNAi included defects in the ovariole architecture; fused egg chambers and pycnotic nuclei ( Figure 1B). These defects were highly penetrant as substantial number of the ovarioles (50/69; 72%) displayed a combination of these phenotypes. As indicated in the previous section, the phenotypic consequences appeared to be specific as germline knockdown of other MADF-BESS genes did not display such phenotypes (Table 1) while somatic knockdown did not mimic the maternal phenotypes, when tested in flies 3, 10 and 18 days posteclosion (PE).
Subsequent careful characterization revealed that the phenotypic consequences upon loss of brwl in the germline were age dependent. At 25°C, females three days PE showed completely normal ovarian morphology ( Figure 1D). Consistently, these females were fertile, and laid eggs that went through normal development post-fertilization ( Figure 1D). By the 7 th day PE, the abnormalities in the ovarian structure became apparent with fused egg chambers ( Figure 1E). This defect became even more penetrant and severe by the 10 th day PE ( Figure 1F), where multiple instances of a single large ovariole i.e., fused egg chamber with abnormally high number of nurse cells were observed. Moreover, the fused egg chambers showed clear signs of structural abnormalities and the presence of pycnotic nuclei. By the 18 th day PE ( Figure 1G), all the ovarioles were partially or completely degenerated. As many of these phenotypes are reminiscent of stwl loss of function, stwl and brwl appear to be the only two members of the MADF-BESS family with functions in the GSC lineage during female oogenesis.
Transposon insertion alleles of brwl display similar ovarian phenotypes as germline specific knockdown of brwl To extend the RNAi dependent phenotypic analysis further, two independent transposon insertion lines in the brwl locus, referred to as brwl KG0 and brwl MI0 (See Materials and Methods) respectively, were tested for phenotypes during ovarian development. Both brwl KG0 and brwl MI0 females show similar range of ovarian defects when homozygous ( Figure 1I, J). Notably, these defects are also age dependent and at day 10 PE, appear to be equally drastic as compared to those seen with the brwl-RNAi. brwl MI0 is maintained over a balancer and very few homozygotes emerge while brwl KG0 is a homozygous viable allele. Less than 5% of the total ovarioles display ovarian defects in older heterozygous brwl KG0 /+ and brwl MI0 /+ females. By contrast, significant structural deterioration and egg chamber fusions are observed in the ovaries derived from the trans-heterozygous brwl KG0 / brwl MI0 females ( Figure 1K, L). Phenotypic similarities between the germline specific knockdown of brwl function and transposon insertion lines in the brwl locus further supported a germline specific requirement for brwl. We thus sought to analyze this activity further. Both the brwl KG0 /brwl MI0 and brwl MIO /brwl MI0 were employed for subsequent characterization of the ovarian defects.
brwl mutants show a range of ovarian defects The phenotypic consequences due to loss of brwl are highly variable. Thus, to better understand possible function(s) of this novel player, we have undertaken a detailed characterization of the individual phenotypes associated with loss of brwl during different stages of ovarian development. As preliminary data (see above) suggested age dependent progression of the phenotypic severity, we analyzed individual phenotypic traits keeping in mind the age dependence.
Drosophila ovary is made up of 14-16 ovarioles that individually function as an egg production line. The germarium containing GSCs surrounded by the somatic cells is situated at the anterior tip of the ovariole. Typically, the GSCs divide asymmetrically to generate another stem cell and a daughter cell, which undergoes differentiation. Stem cell daughters, or cystoblasts, undergo four rounds of synchronous division to eventually give rise to 16 interconnected cystocytes. While niche derived non-autonomous signaling coordinates GSC specification, cell intrinsic mechanisms contribute to the regulation of both the symmetric as well as asymmetric germline stem cell divisions. The unequal nature of division is reflected in the characteristic morphology of fusomes which are the membranous organelles rich in proteins such as a-Spectrin and, Hu-li tai shao (Hts) (Drosophila homolog of Adducin). Thus, GSCs can be readily identified by the virtue of the presence of a round fusome also referred to as spectrosome, located adjacent to the somatic cap cells. By contrast, the differentiated progeny of GSCs i.e., cystocytes or cystoblasts display diagnostic changes in the shape of the fusome including elongation and bifurcation. Since stwl function is required in the GSCs for their maintenance, we first decided to assess if brwl mutants also show a GSC specific phenotype.
brwl mutants affect cyst architecture and GSC number in the early stages of oogenesis: Preliminary analysis of loss of function mutants in brwl locus as well as germ line specific knock down of brwl function indicated involvement of brwl during several stages of ovarian development. To better understand the function of brwl in the process we decided to take a closer look at the germaria from the brwl mutant females. To this end brwl KG0 /brwl MI0 ovaries were first stained with anti-Vasa antibodies and counterstained with DAPI. Vasa is a maternal effect, posterior group gene and vasa RNA as well as Vasa protein are localized in the pole plasm that is assembled and subsequently anchored at the posterior pole of the oocyte. Vasa protein is enriched in a germline specific manner through the course of development. Accordingly, within early germarium Vasa protein is present in germ cells including GSCs and thus has served as a reliable marker of the germline fate (Lasko and Ashburner 1990). In ten day-old brwl KG0 / brwl MI0 ovaries, Vasa protein appears to be considerably less than the wild type ( Figure 2B). Additionally, the staining pattern also revealed that Vasa positive cysts appear disorganized and are distributed randomly within the germarium ( Figure 2B). By the 18 th day, as the number of cysts degenerate, brwl KG0 /brwl MI0 ovaries show presence of considerably larger than average sized nuclei ( Figure 2C) possibly due to apoptosis (see below). Presence of Vasa around these nuclei also suggests that such cysts might have division defects.
To specifically analyze the influence of Brwl on the GSCs and their progeny, we stained the brwl KG0 /brwl MI0 ovaries with antibodies against a-Spectrin and counted the number of spectrosomes and fusomes ( Figure 2D-E). Typically, wild type germarium consists of 2-3 GSCs, which can be readily identified by the presence of spectrosome. By contrast, in the ovaries derived from 10 days old brwl KG0 /brwl MI0 females, total number of cells with a spectrosome were substantially reduced. At the 18 th day, in several germaria, not even a single cell positive for a-Spectrin could be detected ( Figure 2C, D), indicating a substantial decrease in the total number of GSCs and cystoblasts. A number of fusomes are present in the wild type is between 9-11. Interestingly in the ovaries from the brwl KG0 /brwl MI0 10-day old post-eclosion females, no significant decrease in the number of fusomes was observed. However, this number decreased substantially in the 18 th day old flies (3-5 fusomes) ( Figure 2E). Moreover, unlike the wild type female ovary samples, brwl KG0 / brwl MI0 ovaries also exhibited disorganized cysts that have lost a stereotypical linear arrangement. Instead of canonical 2, 4, 8, and 16 cell cyst organization, they were unevenly distributed in the mutant germarium.
brwl mutants show decrease in the number of GSCs: In wild type females, total number of GSCs decreases with age. A single ovariole generally consists of 2-4 GSCs which are maintained until late in the lifespan and the GSC count decreases by half only by day 63 posteclosion (Casanueva and Ferguson 2004;Pan et al. 2007). brwl MI0 / brwl MI0 ovaries show considerable decrease in the number of spectrosome positive cells ( Figure 2G,H) between day 10 through 18 posteclosion. This observation taken together with disorganized cysts with reduced levels of Vasa protein indicated that brwl activity is likely needed for GSC specification and/or maintenance. To confirm this directly a GSC marker dad (daughters against dpp)-lacZ was used to stain the GSCs. brwl KG0 / brwl MI0 ovaries show a 60% reduction ( Figure  2H) in the dad-lacZ positive cells 10 days after eclosion.
In a wild type germarium, a cyst consists of 16 cells that are interconnected via cytoplasmic bridges also known as 'ring canals'. As the egg chambers derived from brwl KG0 / brwl MI0 adult female ovaries appeared to display variety of structural defects, we decided to characterize the cytoskeletal defects by analyzing ring canals, which are actin rich structures. We stained the brwl mutant ovaries with fluorescently labeled phalloidin ( Figure 2J), which labels actin. brwl KG0 / brwl MI0 ovaries show significant loss of ring canals ($70% reduction as compared to Cantos S), critical for oocyte development from day 10 onwards. Fusome, a branched structure made up of continuous endoplasmic reticulum and cytoskeletal components is responsible in determining the mitotic division plane. Because of this influence, eventually after 4 mitotic cell divisions, only two cells end up with Figure 3 Orb is mislocalized or absent in brwl mutants. (A-C) brwl KG0 / brwl MI0 mutants display defective ovarioles. These defects were classified in two major categories, mis-positioned oocytes and no-oocytes, based on actin (green) and Orb (red) staining. Orb related phenotypes enhance with age, with 18 day old flies having a high proportion of ovarioles that have fewer (C) or no oocytes (Suppl. Figure 1). DAPI (blue), F-actin (green) and Orb (red) are used to mark ovarioles. Insets are a 2X Zoom of the images. (D-E) Quantitative data representing the percentage of ovarioles that show the phenotype on third, tenth and eighteenth day after fly eclosion. Ovaries from Canton S 10 day old brwl MIO /brwl KGO animals were used as a control for statistical analyses. N = 3-15, n = 30-50.
four ring canals. One of these two cells eventually would assume the oocyte identity while forcing the other to acquire a nurse cells fate like the remaining cells with fewer number of ring canals. Nurse cells are responsible for the production of cytoplasmic components such as proteins and RNA that are trafficked to the oocyte. As ring canals are the structures through which nurse cells pass the nutrients to the oocyte, reduced number of ring canals raised obvious questions about the oocyte identity in brwl KG0 / brwl MI0 egg chambers.
brwl mutants affect oocyte specification: Proteins like Orb, Egl and BicD (Huynh and St Johnston 2000) that accumulate in the prospective oocyte are essential for oocyte specification and axis determination. orb encodes for Drosophila homolog of Cytoplasmic Polyadenylation Element Binding protein, (CPEB) and has been shown to be required at multiple levels during oogenesis including formation of a 16-cell cyst, oocyte specification, and polarity establishment and/or maintenance (Lantz et al. 1994). Orb accumulation is progressively restricted to the presumptive oocyte during early stages of oogenesis. Furthermore, Orb is localized in a crescent shaped pattern at the posterior end of the developing oocyte. To assess if loss of brwl influences oocyte specification, brwl KG0 / brwl MI0 ovaries were stained with anti-Orb antibodies ( Figure 3A-C). The staining pattern of Orb protein in mutant egg chambers suggested age dependent loss of oocyte specification. Other phenotypes included mis-positioned oocytes or complete loss of oocyte identity. On several occasions although present, Orb protein was not restricted to a single cell and was instead detected in a group of cells. Consequently, by 10 th day, 77.6% of the total ovarioles (n = 45) showed either mis-positioned oocyte and/or several mis-specified i.e., Orb positive cells ( Figure 3B). Moreover, 20.83% egg chambers showed absence of a presumptive oocyte as assessed by lack of Orb accumulation in any cell (Suppl. Figure 1D). By the 18 th day 63.3% of the ovarioles (n = 32) showed egg chambers with a complete absence of oocyte, while around 36.7% showed mispositioned oocytes ( Figure 3D,E). This demonstrates progressive worsening of the phenotype and deterioration of the tissue from day 10 to day 18.
brwl mutants show age dependent Caspase activation: DAPI staining of 10 day brwl KG0 /brwl MI0 mutants showed pycnotic nuclei indicative of apoptotic cell death. To test whether this is indeed the case, brwl KG0 /brwl MI0 ovaries were stained with cleaved Caspase-3 antibody (Sarkissian et al. 2014), which is a marker for cells undergoing apoptosis. brwl mutant ovaries undergo cell death as shown ( Figure 4A-D). 88.88% (n = 39) of ovarioles show extensive apoptosis as assessed by the presence of cleaved Caspase on the 10 th day after fly eclosion which deteriorates further by 18 th day post-eclosion as all the ovarioles positive for caspase activity (n = 28). This drastic increase in cell death likely results in oogenesis arrest in these mutants (Buszczak and Cooley 2000).
Mitosis and Meiosis is affected in brwl mutants. Mitosis in Drosophila ovaries is tightly regulated at multiple levels by different pathways (Deng et al. 2001). Wild type ovaries when stained for phospho-histone-3 S10 (PH3) show mitotically active cells until stage 6 ( Figure 5A-C) after which they enter an endoreplication phase. Interestingly, in brwl KG0 /brwl MI0 ovaries this pattern of mitosis is lost. It is difficult to stage the mutant ovary samples owing to degeneration, but it appears that the total number of mitotic nuclei is considerably reduced in brwl egg chambers. In 10 days old ovaries very few, and widely distributed mitotically active cells are observed ( Figure 5B) whereas by 18 th day old ( Figure 5C) unevenly distributed mitotically active cells are seen throughout the ovariole. We confirmed this by staining the ovaries with anti-cyclin E antibody, which also shows an aberrant pattern ( Figure 5E-F) and is indicative of cell division defects in brwl KG0 /brwl MI0 ovaries.
Meiosis initiated in the cysts after 4 th mitotic division is essential for oocyte specification in Drosophila ovary. It has been shown that onset of meiosis occurs in 2 pro-oocytes with 4 prospective ring canals after the meiotic markers accumulate in these 2 cells (Carpenter 1975) (Jeffress et al. 2007;Lake and Hawley 2012). Failure of proper cyst formation and accumulation of meiotic markers hinders oocyte specification (Huynh and St Johnston 2000). Since brwl mutants showed defects in oocyte specification we stained the brwl mutants with a meiosis marker C(3)G which marks the synaptonemal complex in a meiotic cell (Jeffress et al. 2007). brwl mutants show complete loss C(3) G staining on 10 th day post eclosion (Figure 5G-I). This indicates that none of the cells in cysts form the synaptonemal complex. This is similar to that of BicD null mutants where oocyte specification is not achieved (Huynh and St Johnston 2000).
brwl mutant phenotype is rescued by over-expression of stwl in germ stem cells We are interested in the sequence homology and its relationship with the functional redundancy among the members of MADF-BESS family of proteins. Detailed phenotypic characterization of brwl mutants revealed remarkable overlap with stwl loss of function. Both the proteins seem to influence similar phenotypic traits including the establishment and/or maintenance of GSC fate, proper cystoblast differentiation and oocyte specification. We thus wondered if the two proteins function in a redundant manner. To directly test this possibility, we sought to assess if brwl mutant phenotypes could be rescued by expression of stwl. As a control, we also decided to test other MADF-BESS domain proteins in an identical experimental set-up. Among the different lines tested UAS-CG13204, UAS-CG3838, UAS-stwl, UAS-CG4404, UAS-CG8359, UAS-Coop and UAS-CG3919 showed expression in the ovary, when driven by nos-Gal4 and as monitored by HA staining. Unfortunately however, flies of only two genotypes survived to adulthood and could be analyzed further: brwl MIO /Brwl KGO ; nos-Gal4/UAS-stwl and brwl MIO /brwl KGO ; nos-Gal4/UAS-CG3919. All the other genetic combinations failed to eclose making further analysis untenable. Figure 5 brwl mutants affect mitotis and meiosis. (A-C) brwl KG0 /brwl MI0 mutants show decrease in the number of mitotically active cells, as stained by a phospho-histone3 S10 (PH3) antibody (red) and counterstained by phalloidin (green), in 10 day old ovaries. This pattern appears abrupt and random by 18 th day where mitotic cells are seen randomly situated in the ovariole. N = 5-7, n = 20 = 25. (D-F) brwl KG0 /brwl MI0 mutants show decrease in the number of mitotically active cells, as stained by the Cyclin E antibody (red) in 10 day old ovaries. This pattern is further disrupted by 18 th day (C) where randomly scattered mitotic cells are seen in the ovariole. N = 5, n = 15-35. (G-I) brwl MI0 /brwl MI0 mutants show complete loss of the formation of synaptonemal complex, as marked by the anti-C(3)G antibody, on the 10 th day post eclosion. 100% of ovarioles show this phenotype. N = 5, n = 15-25.
Interestingly we observed a substantial rescue of several phenotypes associated with loss of brwl upon germline specific stwl expression whereas CG3919 expression was unable to achieve rescue. Overexpression of stwl with nos-Gal4 in brwl MI0 / brwl MI0 background led to an amelioration of the ovarian phenotype induced by loss of brwl. For instance, ovarioles from 10 day old female ovaries consisted of normal egg chambers with 15 nurse cells and a properly developing oocyte ( Figure 6A-B, E-F). Only 32.8% of total brwl MI0 /brwl MI0 ; nanos-Gal4/ UAS-stwl-HA ovarioles (n = 56) show ovarian defects which is substantially lower as compared to brwl MI0 / brwl MI0 ovarioles. Moreover, considerable morphological improvement was also observed as brwl MI0 / brwl MI0 ; nos-Gal4/UAS-stwl-HA ovarioles display regularly spaced ring canals with in the egg chambers ( Figure 6B). Notably however, overexpression of stwl is unable to sustain the rescue levels for the 18-day old ovarioles of brwl mutant females (n = 25) ( Figure 6C).
Overexpression of stwl in brwl MI0 / brwl MI0 animals can also partially rescue the oocyte specification defects as assessed by Orb staining ( Figure  6F,H). In brwl mutants the percentage of ovarioles with defective oocytes (no oocytes + mis-patterned oocytes) on the 10 th day is $95%, with ,5% normal oocytes. Expression of stwl in brwl MI0 / brwl MI0 leads to a dramatic rescue with .50% of the ovarioles (n = 25) showing normal patterning as compared to ,5% in the mutant. The rescue at day 18 is also significant ( Figure 6G,H) with egg chambers without Orb staining decreasing from 80% (in brwl MI0 / brwl MI0 ) to 40% after Stwl overexpression. This clearly indicates that overexpression of stwl in brwl mutants delays the accumulation of brwl associated defects.
Lastly, we examined if the germ cell division pattern is restored by overexpression of stwl in brwl loss of function background. As judged by anti-pH 3 S10 (Suppl. Figure 2A) and anti-CyclinE (Suppl. Figure 2B) stainings the 10-day old brwl MI0 / brwl MI0 ; nanos-Gal4/UAS-stwl-HA ovarioles show marked improvement. However, consistent with earlier observations mitotic defects persist in the 18 th day brwl MI0 / brwl MI0 ; nanos-Gal4/ UAS-stwl-HA ovarioles. Taken together these data argue that bwl activity is continuously required during germline development and Stwl protein can partially substitute loss of brwl albeit in a temporally restricted manner.
Consistent with this limited ability to rescue overexpression of stwl in brwl MI0 / brwl MI0 animals did not decrease the apoptotic cell death that brwl MI0 / brwl MI0 ovarioles show on 10 th (n = 20) and 18 th day respectively (n = 15) ( Figure 6I-L). The temporal rescue of Brwl by Stwl does not appear to extend to blocking the apoptotic program that gets initiated in the absence of brwl function. It is however possible that higher dose of Stwl protein may be sufficient to rescue all the phenotypes due to loss of brwl effectively and additional experiments involving better reagents will be required to examine this possibility.

Negative purifying selection for stwl and brwl
Taken together our analysis of Brwl and Stwl suggests that these two proteins are similar enough but they may have retained independent functions. We decided to explore this possibility using phylogenetic analysis. In an earlier study (Shukla et al. 2014), we found evidence for gene duplication and lineage specific expansion of the entire MADF-BESS family. Since our experimental data suggests that stwl and brwl are functional duplicates in the MADF-BESS family, we analyzed Figure 6 Ovarian defects due to loss of brwl are ameliorated by over-expression of stwl. (A-C) brwl MI0 /brwl MI0 mutants show ovary defects, as measured by DAPI staining (blue) and phalloidin (green) that are rescued in brwl MI0 / brwl MI0 ; nos-Gal4/UAS-stwl-HA on the 10 th (B) and 18 th (C) day PE. The rescue is however not as strong on the 18 th day PE. (D) Quantitative data representing the number of ovarioles showing the phenotype on 3 rd (n = 10), 10 th (n = 56) and 18 th (n = 25) day after fly eclosion. Ovaries from three day old brwl MIO /brwl KGO animals were used as a control for statistical analyses. (E-G) brwl MI0 /brwl MI0 ; nosGal4/UAS-stwl-HA ovarioles showing reduced oocyte specification defects. Oocyte defects are significantly rescued in brwl MI0 / brwl MI0 ; nosGal4/ UAS-stwl-HA ovarioles on the 10 th day (n = 25). Mild rescue is seen on the 18 th day (n = 15) with a decrease in ovarioles without orb staining. (H) Quantitative data representing the percentage of ovarioles that show the phenotype on third, tenth and eighteenth day after fly eclosion. Ovaries from three day old brwl MIO /brwl KGO animals were used as a control for statistical analyses. N = 3-15; n = 25-60. (I-K) brwl MI0 / brwl MI0 mutants show apoptotic egg chambers which is not rescued in 10 day old brwl MI0 / brwl MI0 ; nanosGal4/ UAS-stwl-HA ovarioles. (L) Quantitative data representing the percentage of ovarioles that show the phenotype on third, tenth (n = 20) and eighteenth (n = 15) day after fly eclosion. Ovaries from three day old brwl MIO /brwl KGO animals were used as a control for statistical analyses. N = 3-10; n = 15-20. codon substitution patterns for the family with a specific focus on stwl and brwl (See Materials & Methods). We find that the mean pairwise raw genetic distance was two times higher in MADF of stwl than brwl and 1.4 times higher in BESS of stwl than brwl ( Figure 7A). This indicates that the rate of evolution of both MADF and BESS domains of stwl was significantly higher than that for brwl. Model selection for nucleotide substitutions suggests that the first two codon positions are evolving by either simple one parameter (Jukes and Cantor 1969) model with equal rates for substitutions of the four nucleotides or with a two parameter (Kimura 1980) Figure 7B) indicates that MADF and BESS domains of stwl accumulated higher number of non-synonymous substitutions as compared to the respective domains of brwl, probably because of their higher rate of mutation. This suggests that brwl is under more stringent-evolutionary control in terms of codon substitutions than stwl. The dN/dS ratio for pairwise comparison between species showed that for both MADF and BESS domains of stwl and brwl was much less than one ( Figure 7C), indicating that there is purifying or negative selection. This was also evident from negative dN-dS values for codon-by-codon analysis considering best fit codon substitution model ( Figure 7C). Negative or purifying selection indicates selective removal of deleterious alleles, suggesting vital role of the domains and by association for both Brwl and Stwl in cellular functions. Stwl and Brwl have thus maintained their amino acid sequences in spite of changes to the DNA coding sequences, over time. The above analysis thus supports that Brwl and Stwl perform overlapping but distinct functions. (Maheshwari et al. 2008) analyzed the charge and isoelectric point (IP) based on amino-acid sequence of MADF domains in Drosophila and discovered that proteins with multiple MADF domains in a single sequence, for example hybrid male resce (Hmr) and lethal hybrid rescue (Lhr) have distinct Charge and Isoelectric points. They reasoned that MADF domains with a strong positive charge were DNA binding proteins and those with negative charge were primarily binding basic proteins like histones. We find that Stwl and Brwl, having positive Charge of 6.1 and 8.8 respectively ( Figure 7D) for their respective MADF domains fall in the category of transcription factors of the type 'MADF1', geared for DNA binding (Compare to Table1, (Maheshwari et al. 2008)). In contrast, the BESS domain appears to differ significantly in its charge and IP for Brwl as compared to Stwl, suggesting a possible difference in interactors between the two proteins.

DISCUSSION
The continuity of a species is almost exclusively dependent upon the transmission of genetic information across generations. Multicellular organisms accomplish this task via sexual reproduction, which typically involves fusion of the male and female gametes resulting in the development of a new organism post-fertilization. The progenitors of the male and female gametes, known as GSCs, are thus under constant scrutiny. Consequently, a wealth of information is available regarding the mechanisms underlying the specification and maintenance of the GSC fate. How the specification process follows a defined route to eventually accomplish a reliable outcome namely continuous production of healthy gametes is less clear, however. This is of special significance as GSCs not only self-renew but are also pluripotent and are competent to give rise to daughter progeny that subsequently differentiates into a variety of cell types. Thus, it is important to elucidate how GSCs maintain 'stemness' by employing symmetric cell division and also have the capability to generate differentiated daughter cells via asymmetric cell division program.
It is proposed that the process of GSC specification relies upon mechanisms that would ensure both qualitatively and quantitatively robust outcomes (Rudel and Sommer 2003;Félix and Barkoulas 2015;Kitano 2004). To achieve dependable and reproducible outcomes, developmental processes are often reinforced by redundant mechanisms that effectively allow dampening of potential disruptions in the functional gene regulatory network. 'Gene duplication and subfunctionalization' is a prominent strategy among the different safeguarding means that have been employed to generate redundancy (Gu 2003;Kitano 2004). Duplicated genes ensure robustness in a system as deleterious mutations on one gene are readily compensated by the activity of the duplicated member. Gene duplicates thus provide functional redundancy, buffering detrimental phenotypes that arise from malfunction of one copy (Gout and Lynch 2015;Shukla et al. 2014). This contributes significantly to the developmental transitions occurring in a sequential and seamless manner from one stage to the next.
In this study, we have presented experimental analysis of two paralogs stwl and brwl, which appear to be partially redundant molecular players that perform similar functions and thus contribute to the robustness to the processes underlying GSC development in Drosophila ovaries. brwl appears to function in the same cell lineage and spatiotemporal domain as stwl, and stwl expression in the germ cell lineage i.e., nos domain can partially rescue/delay brwl phenotypes. Our study finds that similar to stwl, brwl is also required for stem cell maintenance, proper cyst formation and oocyte specification.
First, we examine the implications of the evolutionary data analyzed. Gene duplicates are believed to evolve differently than singletons (Jordan et al. 2004). Post-duplication, the genes could show changes in their rates of evolution with consequences on purifying (negative or positive) selection. Our data, based on the analysis of the MADF and BESS domains suggest that a) stwl and brwl have undergone negative selection, indicating that their ancestral functional state, as defined by the linear sequence of amino acids was critical and had to be maintained and that b) The rate of evolution appears asymmetric with MADF and BESS domains coding regions of stwl and Brwl sequences changing at different rates.
We have compared the stwl phenotypes described in literature with our data as we were unable to collect data for a direct comparison due to unavailability of published Stwl reagents (Clark and McKearin 1996;Akiyama 2002), such as mutant alleles and antibodies. brwl loss of function shows strong age dependent phenotypes that were not reported in the analysis of any stwl mutants. Both stwl and brwl show a range of similar phenotypes, specific to the female germline. The ability of stwl to partially compensate for brwl function indicates that stwl may be a genetic interactor of brwl and that Stwl has function similar to Brwl and can replace, at least partially, Brwl, suggesting redundancy in function. Since both proteins are MADF-BESS domain proteins and predicted paralogs, this leads to a strong possibility that the common function shared by both is an ancestral function retained by both orthologs.
This thesis is further supported by available evidence as summarized below. A comparison of modENCODE data indicates that both brwl and stwl are expressed in the early embryo likely due to maternal deposition. Transcript levels of both the genes are higher in the ovaries as compared to testis ( Figure 8A) consistent with the elevated levels in adult females. Some MADF-BESS genes (Hng1, Hng2) show similar expression patterns with others showing varying levels of expression in different tissues. In contrast to the data that shows equivalence for Brwl and Stwl function, there is also evidence for differences between the two indicating gain/loss of function for one member. For example, Loss of stwl in the wing by RNAi shows wing/wing-hinge phenotypes whereas brwl RNAi has no effect (Shukla et al. 2014). Overexpression of stwl perturbs development, both in wings and in the ovarian stem cell lineage. brwl overexpression on the other hand has no such effect. Loss of brwl leads to a decrease in GSC number, with survivability being one of the possible means by which Brwl regulates GSC number.
A genomic map of distribution of MADF-BESS genes reveals interesting features ( Figure 8B). The MADF-BESS genes (green circles) are distributed all over the genome, suggesting that the duplication event ($40 My ago) was transposon based. Local duplication, possibly and BESS (without MADF, blue circles) domain coding genes are distributed all over the genome. stwl and brwl are on different chromosomes and CG3919, a local duplicate of stwl does not appear to have functional roles in the GSC lineage. The map distances are in centimorgans and genetic distances are based on chromosomal maps, as collated by Flybase. (C) brwl and stwl are expressed in overlapping domains (driven by enhancer, filled triangle) in the GLC lineage and we hypothesize that the functional entity for this overlapping function, which is critical for robust function, is a Stwl/Brwl heterodimer and absence of the Brwl homodimer can be functionally replaced, to a limited extent by a Stwl Homo-dimer.
succeding a transposon based duplication event may also have occurred, for example for MADF genes set CG1602, CG1603 and az2 and possibly for stwl, CG3919 pair are possible local duplicates. Interestingly, CG3919, as tested by RNAi (Table 1), does not appear to retain germ line Cell (GLC) lineage function, nor does it rescue the brwl mutant phenotype. In agreement, ModENCODE data also suggest low expression of CG3919 transcripts in the ovary.
What are the possible mechanisms for redundancy after gene duplication? A possible scenario is a model where MADF-BESS domain proteins are competent fo function as homo as well as heterodimers and and the Stwl-Brwl heterodimer can substitute the Brwl homodimer ( Figure 8C). In the female germline, both Stwl and Brwl are expressed, thus retaining ancestral expression and functionality. Although the BESS domains have been suggested to be protein:protein intercation surfaces, there is no experimental evidence for direct molecular interaction between Stwl and Brwl in current interaction databases (Hu et al. 2018).
An open question in the field is the molecular function of MADF and MADF-BESS domain containing proteins. Members appear to modulate diverse processes Song et al. 2010;Cutler et al. 1998;Duong et al. 2008;Shukla et al. 2014;Zimmermann et al. 2006;Timmerman et al. 2013;Casola et al. 2007;Barbash et al. 2000;Brideau et al. 2006) with no apparent common theme. The MADF-BESS family members may either act as transcription activators/repressors or as histone modifiers, they may also associate with different subsets of co-factors for diverse functions. Brwl, based on its functional similarity to Stwl may be a DNA binding protein and a putative chromatin regulator (Yi et al. 2009).
In summary, our study uncovers brwl as a novel player in the maintanence of GSCs in the female germline. While our data do not rule out a somatic function conlusively, our phenotypic analysis and expression data strongly suggest a possible germ cell autonomous function of the gene. brwl phenotypes are pleiotropic suggesting that either the protein has an early function with compounding downstream effects or it has multiple roles in the development of the cyst. This pleiotropy is reminiscent of stwl mutants.
Furthermore, our data point to partially redundant functions shared between brwl and stwl in promoting oocyte development in the female germline. We hypothesize that brwl and stwl likely arose as a consequence of a gene duplication event 40 My ago. We further speculate that ancestral functions are retained by both the gene products and are utilized for robust patterning of the germline cells in the developing ovaries.