Lmx1a is required for the development of the ovarian stem cell niche in Drosophila

Drosophila ovaries are composed of about 15-20 functional units called ovarioles that produce eggs. In each ovariole, somatic and germline stem cells are maintained by a group of specialized cells that form the niche (Eliazer and Buszczak, 2011; Kirilly and Xie, 2007). These niches are composed of terminal filaments (TFs) and cap cells, which provide signals required for the maintenance and function of the stem cell populations in adult ovaries, including BMPs and Hh (Eliazer and Buszczak, 2011; Kirilly and Xie, 2007). Although the origin of the somatic lineage in the female gonad has been established, the genetic factors underlying the morphogenetic processes required for the specification and establishment of TF-cap structures are largely unknown. Ovarian somatic cells originate from three mesoderm cell clusters on each side of the embryo (Boyle and DiNardo, 1995). During the mid-larval third instar (ML3), TF cells start to specify, expressing the TF marker Engrailed (En) (Bolívar et al., 2006; Gilboa, 2015). By late third larval instar (LL3), and continuing in white pre-pupae (WPP), TFs form well-organized stacks and promote the proliferation and migration of muscle precursors that will develop into the muscular sheath and delimit individual ovarioles (Fig. S1A) (Irizarry and Stathopoulos, 2015). This process ensures that a single TF stack and a defined population of germline and somatic cells are incorporated in each unit.

The number of ovarioles and the developmental timing of the niche have been shown to be controlled by several signaling pathways, including Insulin, Hippo, Notch, Activin and Ecdysone pathways (Gancz and Gilboa, 2013; Gancz et al., 2011; Green and Extavour, 2014; Hodin and Riddiford, 1998; Lengil et al., 2015; Sarikaya and Extavour, 2015; Song et al., 2007). In addition, the two transcription factors Bab1 and Bab2, encoded by the bric-à-brac (bab) locus, are expressed in TFs and are essential for the formation of these structures (Couderc et al., 2002; Sahut-Barnola et al., 1995). Other BTB transcription factors, Pipsqueak, Trithorax-like and Batman (Lola like), also control TFs and ovariole numbers (Bartoletti et al., 2012). Finally, loss of the Engrailed transcription factor causes abnormal TF cell stacking (Bolívar et al., 2006). Although these signaling pathways and transcriptional regulators are essential for development of stem cell niches, the early genetic events coordinating TF-cap cell specification, formation and function remain largely unknown.

Here, we describe the expression and function of the LIM-homeodomain transcription factor Lmx1a in developing Drosophila ovaries. LIM-homeodomain proteins have essential roles during tissue patterning and cell differentiation in metazoans, from nematodes to vertebrates (Hobert and Westphal, 2000). Mammalian Lmx1a and Lmx1b are pleiotropic regulators of cell differentiation and tissue development in many organs, as shown by the defects and syndromes caused by loss-of-function mutations in these genes (Doucet-Beaupré et al., 2015). For instance, haploinsufficiency of Lmx1b causes nail-patella syndrome (NPS), a disorder affecting dorsal limb structures, kidneys, anterior eye components and the nervous system (Bongers et al., 2008; Dreyer et al., 1998; McIntosh et al., 1998; Vollrath et al., 1998). Studies in genetic mouse models have confirmed a regulatory function of Lmx1b in the development of these organs and the behavior of associated cell types, including serotonergic neurons, podocytes and several eye tissues (Burghardt et al., 2013; Chen et al., 1998; Ding et al., 2003; Miner et al., 2002; Morello et al., 2001; Pressman et al., 2000; Rohr et al., 2002). Similarly, studies of the mouse Lmx1a factor have revealed diverse developmental functions, including brain patterning and cell fate decision, dopaminergic neuron differentiation and insulin expression (Andersson et al., 2006; German et al., 1992). Finally, previous work has implicated Lmx1a and Lmx1b in a variety of other pathologies, including non-NPS renal disease (Edwards et al., 2015), Parkinson's disease (Laguna et al., 2015), and various forms of cancer, including ovarian epithelial carcinoma (He et al., 2014; Lin et al., 2012; Tsai et al., 2012; Tsai et al., 2013), highlighting the complexity of Lmx1a/b biology, as well as the urgent need to understand the function of these factors better.

Two LIM-homeodomain proteins of the LMX subgroup are encoded in the Drosophila genome: Lmx1a/CG32105 and Lmx1b/CG4328. Lmx1a is expressed in the LL3 eye imaginal disc and its overexpression in this tissue causes eye defects (Roignant et al., 2010; Wang et al., 2016). However, the molecular, cell biological and developmental functions of this transcription factor have yet to be described. Here, we show that Lmx1a is expressed in somatic lineages in the developing ovary. We found that its expression becomes restricted to TFs and cap cells by LL3/P0 (freshly pupariated) and is maintained in the adult stem cell niche. Analyzing a CRISPR-generated Lmx1a knockout allele as well as cell type-specific and stage-specific knockdown of Lmx1a, we have determined that Lmx1a is required for ovary development specifically in the TF-cap cell niche at the time at which it forms. We performed transcriptional profiling of developing ovaries and placed Lmx1a downstream of Bab1/2 in the specification of TF cells. We also show that without Lmx1a, several components of signaling pathways crucial to the forming niche are not properly expressed, including Hh, the transcription factors Sox100B, Engrailed and Invected, and confirmed that these genes are required in the Lmx1a lineage. Strikingly, expression of a chicken ortholog of Lmx1a in forming TF-cap cells is sufficient to rescue the Lmx1a null phenotype. We anticipate that these results will further elucidate the genetic and cell biological mechanisms underlying the establishment of the Drosophila ovary stem cell niche and provide insight into the role of LIM-HD factors in tissue development and patterning, homeostasis and disease.

In a search for novel regulators of ovarian development in Drosophila, we found that the transcription factor Lmx1a is expressed in LL3 ovaries. Using RT-qPCR, we detected endogenous Lmx1a mRNA specifically in samples containing ovaries, compared with samples containing only fat body tissues (Fig. 1A). To confirm the specificity of the Lmx1a mRNA detection and to study its function, we generated a null allele, in which the CG32105 open reading frame is replaced by a 3xP3-DsRed cassette (Lmx1aΔ1; Fig. S1B). Genome editing was validated by PCR amplification (Fig. S1C), and RT-qPCR confirmed that the Lmx1a transcript is undetectable in null animals, neither in adults nor in LL3 ovaries (Fig. 1A, Fig. S1D). Next, we took advantage of a reporter line in which the transcription factor Gal4 is expressed under the control of a putative Lmx1a enhancer (Lmx1a-Gal4^GMR42B09; Fig. S1B) to investigate the pattern and developmental history of Lmx1a expression in the ovary. When crossed individually with three RNAi lines against Lmx1a (Fig. S1B), Lmx1a-Gal4 is sufficient to drive a loss of endogenous Lmx1a transcript in LL3/P0 transitioning ovaries comparable to Lmx1a mutant homozygotes, as detected by qPCR of cDNA generated from samples containing fat bodies and developing ovaries (Fig. 1B, Fig. S1E). This demonstrates that, at this stage, Lmx1a is exclusively expressed in all cells, or a subpopulation of cells, in which this driver is active. In P0 pupal ovaries, the Lmx1a-Gal4 is active in apical cells, TFs and cap cells, as determined by morphology, colocalization with the TF marker Engrailed, and localization alongside the germline markers Vasa and α-Spectrin (Fig. 1C,D, Fig. S1F). We also found that the dsRed transgene inserted in place of the Lmx1a coding sequence is strongly expressed in P0 TFs and cap cells and more weakly in apical cells (Fig. 1D, Fig. S1F), a pattern identical to that of Lmx1a-Gal4>UAS-GFP and not observed in any other transgenic lines generated with the same 3xP3-DsRed cassette (data not shown).

We finally combined the Lmx1a-Gal4 driver with the G-TRACE (Gal4 technique for real-time and clonal expression) lineage-tracing system (Evans et al., 2009). GFP-based tracing suggests that Lmx1a-Gal4 is active in the progenitors of all somatic cells (GFP), but its expression becomes restricted to the apical domain, which contains apical cells and TFs, by LL3 (RFP expression) (Fig. 1E). In adult ovaries, RFP expression suggests that this driver remains active only in the germline stem cell niche (TFs and cap cells), and GFP lineage tracing confirms that it is active earlier in the progenitors of follicle cells, follicle stem cells and muscle sheaths (Fig. 1F, GFP).

Together, these data strongly suggest that Lmx1a is expressed early during ovarian development, in the somatic progenitors. Its expression becomes limited to TFs and apical cells by LL3 and is further restricted to the germline stem cell niche in adult germaria.

To begin addressing the function of Lmx1a, we analyzed phenotypes associated with Lmx1a mutation. Surprisingly, a close examination of homozygous null animals revealed no obvious morphological defects. Furthermore, homozygous adults were observed at the expected proportion in crosses between heterozygous animals, indicating no developmental delay or decreased fitness compared with control siblings (data not shown). However, analysis of fertility revealed that homozygous females are sterile and lay no eggs (Fig. 2A), whereas the fertility of males is not affected (Fig. S2A). In adult mutant females, ovaries are generally absent, with no significant amount of tissue attached to the oviducts (Fig. 2B). To confirm Lmx1a mutation as the cause of this phenotype, we generated trans-heterozygous animals carrying the Lmx1a^Δ1 knockout allele and several deletions covering the CG32105 locus. All mutant combinations result in viable adult females with absent ovaries (Fig. 2B). In a limited number of genetic backgrounds or at lower temperatures, residual mutant ovarian tissue can be recovered, presenting abnormal germaria and egg chambers and very few eggs attached to the oviducts or found dispersed in the abdominal cavity (Fig. S2B). The limited ovariole material appears highly disorganized, with egg chambers often fused together rather than separated and packaged by a muscular sheath. In the rare mutant germaria that could be recovered, we observed no morphologically recognizable TFs (Fig. S2B). This suggested to us that the development of muscle and TF-cap structures is aberrant in Lmx1a mutants.

Consistent with this, looking as early as in third instar larva, when TF morphogenesis starts to take place, we found evidence that ovarian development is aberrant. Based on morphology, cell organization and Lmx1a-dsRed expression, we found that although there is partial evidence of TF morphogenesis in Lmx1a^Δ1 homozygotes, these developing ovaries fail to generate the collection of fully formed, linear stacks early during the larval-pupal transition at LL3/P0 (Fig. 2C). By 24 or 48 h after pupariation, TFs are absent in mutant animals, the ovaries show no sign of growth and elongation, and primordial ovarioles do not form (Fig. 2D). To characterize these defects in more detail, we used well-established markers for TF-cap cells, Engrailed protein and Bab1^P-lacZ reporter expression (Fig. 2E,F, Fig. S2C), as well as DE-cadherin (Shotgun in Drosophila) antibody staining to delineate their cellular boundaries (Fig. 2E) and collagen IV deposition around developing TFs (Fig. S2C; Van De Bor et al., 2015), in the ovaries of P0 heterozygous and null animals. Our analysis shows that, although Bab1^P-lacZ-positive cells are detected in Lmx1a^Δ1 homozygotes, these cells fail to organize in individualized stacks with strong DE-cadherin staining (Fig. 2E,F), suggesting that these cells are initially specified but are not capable of proceeding normally through differentiation. Confirming this notion, we found that none of these cells shows the strong Engrailed expression characteristic of TFs and cap cells in P0 ovaries (Fig. 2E). Finally, these aberrant TFs do not support collagen IV deposition like control structures do (Fig. S2C).

Altogether, these data demonstrate that Lmx1a is required for ovarian development and, in particular, the normal differentiation of TFs.

Based on the expression pattern of Lmx1a in the developing ovary and the defect caused by its mutation, we hypothesized that Lmx1a is required in TFs for their differentiation, formation and/or developmental function as organizing centers for niche establishment and ovariole morphogenesis. To test this notion, we first used Lmx1a-Gal4 to rescue Lmx1a expression in Lmx1a null animals. We found that expressing an HA-tagged version of the protein significantly restores overall ovary morphogenesis, including sheath development, TF formation and egg laying (Fig. 3A, Fig. S3A,B). In order to test the functional conservation between Drosophila Lmx1a and Lmx1a/b in vertebrates, we drove expression of chicken Lmx1b with Lmx1a-Gal4 and observed a significant rescue of the null mutant phenotype, comparable to that achieved by the Lmx1a-HA construct in terms of both overall morphology and fertility (Fig. 3A, Fig. S3B; UAS-cLmx1b). Of note, the rescue by both Drosophila Lmx1a and chicken Lmx1b expression in TFs is much more pronounced at a lower temperature, at which development is slower (compare phenotypes at 18°C in Fig. 3A with 25°C in Fig. S3B).

In a set of converse experiments, we used the same driver to express several distinct dsRNA constructs targeting the Lmx1a transcript and knocking down its expression in TFs and apical cells in an otherwise wild-type organism (Fig. 1B, Fig. S4). Consistent with the requirement for Lmx1a in TFs, these manipulations strongly impair ovary development and female fertility, much like null homozygotes (Fig. 3B,D). Interestingly, we identified another Gal4 reporter, driven by a distinct putative Lmx1b/a enhancer (Lmx1b-Gal4^GMR35G09; Fig. S1B), which is also active early in somatic lineages, as shown by lineage tracing, and remains active in P0 apical cells but is not active in P0 TF-cap structures (Fig. S4). Using this driver to knock down the expression of Lmx1a, we did not observe strong ovarian defects: only a mild phenotype with one of the RNAi lines (KK) and no significant phenotype with the other (TRiP) (Fig. 3C). This correlates with a complete loss of egg laying when driving expression of Lmx1a^RNAi(TRiP) with the Lmx1a-Gal4 driver (active in LL3-P0 TF-cap cells) and normal fertility when expression of the same construct is driven by Lmx1b-Gal4^GMR35G09 (not active in LL3-P0 TF-cap cells; Fig. 3D). This strongly suggests that Lmx1a is essential in TFs and cap cells themselves, with a minimal role in apical cells, to support ovarian development.

To confirm these findings, we took advantage of additional drivers active in developing ovaries. The bric-à-brac (bab) locus encodes two transcription factors, Bab1 and Bab2, expressed in apical cells and developing TFs and is essential for ovary development (Couderc et al., 2002; Godt and Laski, 1995; Sahut-Barnola et al., 1995). The bab1-Gal4 drivers are highly active in P0 TFs (Cabrera et al., 2002) (Fig. 3B, Fig. S4). The strong ovarian defects observed in bab1-Gal4>Lmx1a RNAi animals, which recapitulate the Lmx1a mutant phenotype, thus confirm that Lmx1a is required in TFs to support ovarian morphogenesis (Fig. 3B, Fig. S5). In addition, a previous study showed that components of the FGF signaling pathway, thisbe (ths, FGF-like ligand) and heartless (htl, FGF receptor), are expressed in TFs and apical cells, respectively (Irizarry and Stathopoulos, 2015). Therefore, we also used ths-Gal4 and htl-Gal4 to reduce the expression of Lmx1a in these cells (Fig. S5). G-TRACE analysis confirmed that htl-Gal4 activity is restricted to the apical cells in P0 ovaries and showed that ths-Gal4, although active in all TF cells by adulthood (data not shown), drove RFP expression in only a few TF cells at P0 (Fig. S4). Consistent with our model, we found that Lmx1a RNAi expression, using these two drivers, does not affect the development of ovaries. Interestingly, in their study, Irizarry and Stathopoulos demonstrate that TFs are essential for the proliferation of apical cells and proper development of the sheaths (Irizarry and Stathopoulos, 2015). This, combined with our results, strongly suggests that Lmx1a is required for TF differentiation and/or developmental function, and that the absence of muscle sheaths in Lmx1a mutants is a consequence of TF defects.

Finally, to establish the temporal requirement of Lmx1a in TFs, we conducted a series of temperature-sensitive knockdown and rescue experiments using the Lmx1a-Gal4 driver combined with the temperature-sensitive tub-Gal80^ts inhibitor. We found that Lmx1a RNAi expression starting in mid-stage L3 larvae (ML3, around 24 h before pupariation when animals are switched to 29°C) is sufficient to cause severe ovarian defects and recapitulate the ovary phenotype observed when Lmx1a expression is knocked down throughout development (Fig. 4A). In contrast, Lmx1a RNAi expression 24 h after pupariation does not cause any significant defect. Conversely, re-expressing Lmx1a in the null background starting in ML3 rescues the ovarian mutant phenotype, just as much as re-expressing the transgene throughout development, and re-expression starting 24 h after pupariation does not visibly ameliorate ovarian development (Fig. 4B). These results define a developmental window of about 48 h around the larval-pupal transition during which Lmx1a is required for ovary morphogenesis.

Altogether, our data demonstrate that, although Lmx1a is potentially expressed in very early common somatic progenitors and its expression is maintained into adulthood, it is required specifically in TFs during a critical developmental window, the larval-pupal transition, for the proper differentiation of these cells and normal development of the ovary.

The morphogenetic mechanisms that control the establishment of individual ovarioles and overall Drosophila ovary development are beginning to be elucidated. For example, Hedgehog, Notch, FGF, Insulin and Hippo signaling pathways control the specification and number of TF stacks and cap cells that in turn coordinate the development of each ovariole (Forbes et al., 1996; Gancz and Gilboa, 2013; Green and Extavour, 2014; Irizarry and Stathopoulos, 2015; Sarikaya and Extavour, 2015; Song et al., 2007). However, the sequence of signaling events and transcriptional cascade(s) required for the development of these structures are still largely unclear. Because Lmx1a is required specifically in developing TF-cap structures, we hypothesized that Lmx1a is, in turn, essential for many crucial niche-forming developmental pathways, which would explain the near complete loss of ovary material by adulthood in Lmx1a null animals. To test this notion, we analyzed the transcriptomes of LL3/P0 developing ovaries embedded in fat body tissue compared with surrounding fat body alone, and compared Lmx1a null tissues with heterozygous controls.

We first generated a list of genes that are significantly enriched in ovary-containing samples. Among these genes, we identified many transcripts of proteins known to be expressed in developing Drosophila ovaries (e.g. Vasa, Hedgehog, Wnt4, Bab1, Bab2, Engrailed, Invected and Sox100B) (Fig. 5A,C), confirming the specificity of our methodology. Within this list of 1755 genes enriched in ovaries, we then compared Lmx1a^Δ1 homozygotes with heterozygotes in order to identify the genes most highly affected by the loss of Lmx1a. Strikingly, we found components of the Hedgehog, Notch and FGF pathways (hh and ptc, Ser, and stumps, respectively) in addition to engrailed, invected and Sox100B among the 48 reduced transcripts in Lmx1a null developing ovaries (Fig. 5B,C; Table S1), suggesting that these genes are direct or indirect targets of Lmx1a. Conversely, we found that other factors essential for ovary development and mentioned before, such as Bab1, Bab2, Wnt4, Ths and Htl, are not affected in Lmx1a mutants (Fig. 5C), raising the possibility that these factors are regulated upstream of or in parallel to Lmx1a.

To validate these results, we performed qPCR on cDNA generated from the same samples and found no change in vasa, bab1 and bab2 transcript levels and confirmed a significant reduction in hh and Ser transcript levels in Lmx1a homozygous mutants (Fig. 5D).

Prior to this study, the bab1/2 locus was one of the only loci known to be essential cell-autonomously in forming TF-cap niche structures (Bolívar et al., 2006; Couderc et al., 2002). The maintenance of Bab1 and Bab2 expression in Lmx1a^Δ1 homozygotes, as revealed through RNAseq, qPCR (Fig. 5B,C,D) and the Bab1^P-lacZ reporter (Fig. 2F), suggested to us that Lmx1a may function downstream of Bab1/2. Further confirming that Lmx1a is not required for Bab1 expression, we found that knockdown of Lmx1a, although inducing morphologically defective TFs, did not affect Bab1^P-lacZ reporter activity (Fig. 6A). We next investigated whether Bab1/2 and Lmx1a function in the same lineage. Consistent with this notion, we found that expression of RNAi against either bab1 or bab2 in the Lmx1a lineage, using the Lmx1a-Gal4 driver, induces an ovary defect comparable to that previously reported in Bab1/2 mutant homozygotes, with substantially reduced, disorganized and fused ovariole material (Couderc et al., 2002), and is comparable to what we observe in Lmx1a mutants or Lmx1a-Gal4>Lmx1aRNAi animals (Fig. 6B). This raised the possibility that Bab1/2 is essential for Lmx1a expression and function. To test this hypothesis, we analyzed the endogenous level of Lmx1a mRNA in bab1^P/bab1^P homozygous ovaries. It has been previously reported that the Bab1^P-lacZ insertion causes a mutation in Bab1, but not Bab2, and homozygous mutants show a severe ovary defect (Couderc et al., 2002). In P0 ovaries of these mutants, we show that Lmx1a expression is very significantly reduced compared with samples dissected from heterozygous siblings (Fig. 6C), placing Bab1 genetically or temporally upstream of Lmx1a in the development of the TF-cap structure.

We next functionally validated a collection of genes identified as being dysregulated in Lmx1a mutants. To this end, we drove the expression of RNAi constructs directed against Hh, Sox100B, Engrailed and Invected in the Lmx1a-Gal4 lineage in a control background or in a Bab1 heterozygous sensitized background (Bab1^P-lacZ/+). Consistent with the known crucial role for Hh in the function of the ovarian stem cell niche (Forbes et al., 1996), we found that the expression of Hh dsRNA within the Lmx1a-Gal4 lineage induces a strong loss of fertility and a defect in ovary morphogenesis, which is enhanced by Bab1^P-lacZ heterozygosity (Fig. 7). Next, although Sox100B has been reported to be expressed in TFs of developing ovaries, no role for this protein in ovary morphogenesis has been detected, potentially due to the failure of Sox100B mutant flies to fully emerge and produce mature ovaries suitable for proper assessment of adult phenotype (Nanda et al., 2009). Driving the expression of Sox100B dsRNA within the Lmx1a-Gal4 lineage resulted in a mild ovary defect, and reduced fertility when combined with Bab1 heterozygosity (Fig. 7), suggesting a developmental role for Sox100B within the Lmx1a-Gal4 lineage. Finally, Engrailed has been suggested to be required for proper stacking of TF cells (Bolívar et al., 2006), and is essential for decapentaplegic (dpp) expression in cap cells, a crucial niche signal maintaining germline stem cell self-renewal (Eliazer et al., 2014; Luo et al., 2017; Rojas-Ríos et al., 2012; Xie and Spradling, 1998). Because of the possibility of redundancy between Engrailed and Invected, which are both affected in Lmx1a mutants, we simultaneously drove En^RNAi and Inv^RNAi in the Lmx1a-Gal4 lineage. In these strong knockdown conditions (Fig. S6), we found that the combined loss of Engrailed and Invected produces a mild phenotype in young animals comparable to that of Lmx1a-Gal4>Sox100B RNAi animals and slightly reduced fertility in control background (Fig. 7A,C). However, consistent with the known role for Engrailed in adult niche maintenance (Luo et al., 2017), we observed strong ovarian defects in older females (11 days old). Furthermore, we observed a complete loss of ovary morphogenesis and fertility in En/Inv knockdown animals carrying the Bab1^P-lacZ allele (Fig. 7B,C). This confirms that, in addition to its role in the adult niche, Engrailed, along with Invected, plays at least a partial role in the Lmx1a lineage for ovary development.

Based on its amino acid sequence, especially the predicted LIM and homeobox domains, it is very likely that Lmx1a functions as a transcription factor. In order to function, LIM-HD factors often require the co-factor Ldb (LIM domain binding) to mediate either homo- or heterodimer interactions. In support of this, Wang et al. found that Lmx1a requires the co-factor Chip (the known Drosophila Ldb homolog) to mediate eye developmental defects when expressed ectopically in the LL3 eye imaginal disc (Wang et al., 2016). We reasoned that Lmx1a might require Chip to mediate its function in the developing ovary and found that Chip^RNAi expression, using the Lmx1a-Gal4 driver, produces ovary developmental defects similar to Lmx1a mutants (Fig. S7). Previous studies on several LIM-HD factors suggest that Chip mediates the formation of either homodimers or heterodimers with other LIM-HD factors (Hobert and Westphal, 2000; Roignant et al., 2010; Wang et al., 2016). Further studies will be required to determine whether Lmx1a interacts with another LIM-HD factor to regulate TF-cap development and function.

Altogether, our genetic analyses demonstrate that multiple Lmx1a-dependent genes are required in the Lmx1a-Gal4 lineage for ovary morphogenesis. This, combined with our RNAseq analysis, supports a model in which Lmx1a functions upstream of a collection of signaling and transcriptional events essential for the differentiation of TF-cap cells, stem cell niche establishment/maintenance and, consequently, ovary morphogenesis. Additional studies will be required to investigate whether Lmx1a directly and transcriptionally controls the expression of these genes in TF-cap cells.

Despite longstanding interest in the Drosophila ovary as a model for the adult stem cell niche, the genetic mechanisms underlying the initial establishment and development of the niche remain largely unexplored. To date, only two transcription factor loci, bric-à-brac and engrailed/invected, have been identified as required for the development or proper stacking of TFs, and their function remains largely unknown (Bolívar et al., 2006; Godt and Laski, 1995). In this study, we characterized the function of Lmx1a, a highly conserved LIM homeobox transcription factor, in the developing Drosophila ovary. By generating an Lmx1a knockout line as well as taking advantage of a novel collection of Gal4 drivers that produce spatially restricted developing ovary expression patterns, we show that Lmx1a is required in TF-cap structures at the time at which they form and initiate ovarian stem cell niche establishment and ovariole morphogenesis. It remains unclear whether Lmx1a plays a role in TFs, cap cells, or both. Our data also suggest that Lmx1a expression is maintained in the adult niche. Further studies will be required to determine whether it is essential for the function of the adult niche, for example by controlling the expression of self-renewal factors.

Using RNAseq, we have generated a novel list of genes enriched in LL3/P0 ovaries, and from this list have identified several pathways potentially affected by the loss of Lmx1a. Within this list are several transcription factors and components of signaling pathways necessary for ovarian niche development, function and maintenance, including Hh, FGF, Notch, Engrailed and Sox100B. Indeed, RNAi-mediated knockdown of several of these transcripts within the Lmx1a-Gal4 lineage results in ovary developmental defects and reduced fertility. We therefore propose that Lmx1a controls many aspects of the development of this tissue, including proliferation, migration or cell shape changes, by regulating a series of developmental factors and signal pathways. It is important to note that our study does not allow us to distinguish between a direct transcriptional regulation of these genes by Lmx1a and an indirect effect of an impaired development of the TF-cap structure in Lmx1a mutants. Our data analyzed, in young adult females, the consequence of manipulating Hh, Sox100B, Engrailed and Invected in the Lmx1a lineage. Investigating the function of these genes, specifically during the LL3-P0 transition or later during metamorphosis or adulthood, will allow us to test whether their genetic requirement and expression are compatible with a direct regulation by Lmx1a and to understand better the sequence of events required for normal ovarian morphogenesis.

Interestingly, Engrailed has been previously suggested to be required for TF stacking based on clonal analysis of TF cells homozygous mutant for both Engrailed and Invected (Bolívar et al., 2006). Our work suggests that the stacking defects caused by these mutations lead to a mild developmental phenotype in young adults. However, consistent with other work revealing a crucial role for Engrailed in the adult niche (Eliazer et al., 2014; Luo et al., 2017; Rojas-Ríos et al., 2012), we also show that persistent knockdown of Engrailed/Invected well into adulthood leads to an eventual loss of ovaries. We speculate that the progressiveness of this defect is the consequence of a continuous impairment of adult niche function that worsens a mild initial developmental phenotype.

We investigated where Lmx1a functions in relation to the transcription factors Bab1 and Bab2. In support of earlier work, we found that both bab1 and bab2 are required for ovary morphogenesis. In addition, we show a strong reduction in Lmx1a mRNA expression in the ovaries of LL3/P0 bab1 homozygous mutant animals, placing Bab1 genetically and/or temporally upstream of Lmx1a. It is worth noting that, although reduced, the expression of Lmx1a is not eliminated in Bab1 mutants. We speculate that Bab2 activity may be maintaining residual Lmx1a expression in this background. We found that heterozygosity for Bab1 is sufficient to enhance the ovary developmental defects caused by RNAi-mediated knockdown of the Lmx1a-dependent genes tested here, consistent with our model in which the bric-à-brac locus lies genetically upstream of Lmx1a.

Finally, we found that the Lmx1a null homozygous phenotype can be significantly rescued by the expression of chicken Lmx1b, in terms of TF-cap formation, sheath formation and fertility. This reveals striking conservation between the Lmx1a/b factors across the animal kingdom. Even more striking is the conserved relationship between Lmx1a/b and Engrailed. For example, Lmx1b is required for Engrailed expression during both the establishment of the isthmic organizer in the developing mouse midbrain, and the specification and differentiation of dopaminergic neurons (Guo et al., 2007; Sherf et al., 2015). Additionally, simultaneous lentiviral-mediated expression of Lmx1b and Engrailed has been shown to aid in the differentiation of dopaminergic neurons in vitro (Panman et al., 2011). Altogether, these observations point to a likely conserved Lmx1a/b-Engrailed transcriptional module involved in tissue patterning and cell differentiation, across different developing tissues, cell types and species. We therefore anticipate that studies on the function of Lmx1a in the Drosophila ovary will lead to a better understanding of the developmental function of Lmx1a in animals ranging from invertebrates to mammals. Our study also indicates that Drosophila represents a valuable model in which to investigate the mechanisms underlying complex diseases caused by a dysfunction of Lmx1a/b, such as nail-patella syndrome, ovarian carcinoma and Parkinson's disease.

All flies were cultured on standard yeast and molasses-based food at 25°C and 65% humidity, on a 12-h light/dark cycle, unless otherwise indicated. The following stocks were obtained from the Bloomington Drosophila Stock Center: Lmx1a-Gal4 (#41248), Gtrace (#28280), UAS-mCD8-GFP (#32184), TM3-GFP (#4534), Df(3L)ED4475 (#8069), Df(3L)ED4483 (#8070), Df(3L)BSC380 (#24404), Lmx1a^{RNAi TRiP} (#31905), bab-Gal4^Agal4-5 (#6802), bab-Gal4^Pgal4-2 (#6803), Lmx1ab-Gal4^GMR35G09 (#47557), tubGal80^ts (#7108), Invected^{RNAi TRiP} (#41675), Bab1^RNAi (#57410), Bab2^RNAi (#49042), TM3 (#2537), GMR-Gal4 (#1104), Ths-Gal4 (#47051), Htl-Gal4 (#40669), Chip^{RNAi TRiP} (#35435), Hh^{RNAi TRiP} (#25794), Sox100B^{RNAi TRiP} (#57417); the Vienna Drosophila Resource Center: Lmx1a^{RNAi KK} (108747), Lmx1a^{RNAi GD} (#51267), Engrailed^{RNAi GD35697} (#35697), Engrailed^{RNAi GD35698} (#35698), Chip^{RNAi KK} (#107314), Chip^{RNAi GD} (#30454); FlyORF: UAS-Lmx1a-HA (#F000446); Jean Louis Couderc: bab1^P (Sahut-Barnola et al., 1995). Detailed genotypes used in this study are listed by corresponding figure in Table S2.

The Lmx1a^Δ1-dsRed fly stock was generated by a CRISPR-Cas9 homology directed repair (HDR) strategy. Sense and antisense Lmx1a gRNA oligos were annealed and cloned into the pBFv-U6.2B to generate a gRNA-expressing construct targeting the second exon of Lmx1a. 1 kb regions flanking the Lmx1a open reading frame were amplified using the 5′ homology arm and 3′ homology arm forward and reverse primer pairs and cloned on each side of the 3xP3-dsRed cassette in the pHD-DsRed-attP vector. The gRNA and HDR constructs were simultaneously injected into a nos-Cas9/CyO genetic background by Bestgene injection services. Adult F0 males were crossed individually with TM3/TM6 virgins and F1 progeny were then screened for dsRed fluorescent eyes. Lmx1a^Δ1-dsRed genome editing was confirmed by PCR amplification of an amplicon specific to both dsRed and the Lmx1a locus (outside the 3′ homology arm), using the Lmx1a^Δ1-dsRed locus detection forward and reverse primers. Initial analysis of phenotype was performed in this TM3 balanced stock and is reported in Fig. S2B. An isogenic stock was then generated by backcrossing Lmx1a^Δ1-dsRed virgins with w¹¹¹⁸ males for more than five generations and rebalanced over a GFP-marked TM3 balancer. See Table S3 for primer sequences.

The chicken Lmx1b clone (pSKII-cLmx1b, a gift from J. C. Izpisua Belmonte) (Vogel et al., 1995) was digested with EcoRI and NotI and the cLMX1b cDNA was subcloned into the corresponding sites of the pUAST vector to generate pUAS-cLmx1b. The resulting clone was verified by sequencing and injected into yw embryos by standard P-element transformation.

Adult ovaries were scored based on a four-category scale: strong phenotype – little to no ovariole material present in dissected adult females; mild phenotype – reduced ovariole material compared with control with substantial disruption of ovariole architecture (ovarioles not aligned from anterior to posterior or evidence of egg chamber fusion or loss of sheath) and/or evidence of detached eggs in the abdominal cavity during dissection; weak phenotype – similar ovariole material compared with control but with some mild evidence of disruption of ovariole architecture; normal.

Four females were placed with at least six healthy males in a fresh vial containing standard food. The vial was flipped every 24 h and the number of eggs laid on top of the food was counted per day over the course of 5-11 days (Fig. 2A, 5 days; Fig. 3A, 5 days; Fig. 3C, 8 days; Fig. 7C, 6 days; Fig. S3B, 8-9 days). At least four independent crosses were analyzed per genotype. The number of eggs laid/female/day was averaged over all the replicates to generate mean±s.e.m.

Male fertility was determined by placing single males in a vial containing four healthy virgin females. Adult progeny per female per day, rather than eggs laid, was calculated as a measure of male fertility, due to the fact that females are capable of laying unfertilized eggs.

Significance was determined by Student's t-tests across indicated genetic conditions.

The temporal and regional gene expression targeting (TARGET) system was used to temporally restrict the activity of Lmx1a-Gal4. We combined a ubiquitously expressed temperature-sensitive inhibitor of Gal4 (tubGal80^ts) with Lmx1a-Gal4, along with either UAS-Lmx1a-HA (FlyORF) for rescue experiments or UAS-Lmx1a^RNAi for knockdown. Flies were cultured at the Gal80^ts permissive temperature (18°C) to inhibit UAS activation, and placed at 29°C to relieve Gal4 of Gal80^ts repression and allow for UAS activation. Based on our observations of developmental timing at 18°C, we designated 7 days at 18°C as representative of ML3 and 10 days at 18°C as representative of 24 h post-pupariation (P24 h). The four temperature time-course conditions used in Fig. 4 are therefore as follows: 18°C throughout development, 29°C throughout development, 7 days (up until ∼ML3) at 18°C followed by 29°C, and 10 days (up until ∼P24 h) at 18°C followed by 29°C.

LL3 larvae were selected for by their size and ‘wandering’ behavior. P0 animals were selected for as strictly white pupae. Dissection of these early stages was performed in PBS. The entire contents of the LL3/P0 animal was exposed using a pair of fine forceps, and fixed at room temperature for 25 min with 100 mM glutamic acid, 25 mM KCl, 20 mM MgSO₄, 4 mM sodium phosphate, 1 mM MgCl₂, 5% formaldehyde. This was followed by two washes with 1% Triton X-100 in PBS, the first for 5 min and the second for 30 min. All subsequent washes and immunostaining were performed with 0.3% xPBTB (0.3% Triton X-100, 1% bovine serum albumin in PBS). After completion of staining, pieces of fat body containing ovary were mounted in Mowiol-DABCO. All images of LL3 and P0 staged ovaries were taken with a Leica Sp6 confocal microscope. For each image, a single representative z-section was chosen from a collection of z-sections spanning the entire developing ovary. This chosen z-section was consistently selected to reveal both maximal TF architecture as well as the maximum diameter of the ovary, regardless of genotype.

Mature adult ovaries used in this study were from females that were allowed to mate over the course of 2-3 days post-eclosion. The virgin female flies presented in Fig. S2A were selected for as freshly emerged, bloated females lacking pigmentation. Eleven-day-old flies were dissected for some of the experiments reported in Fig. 7. Adult ovaries were dissected under PBS and fixed at room temperature for 45 min with 100 mM glutamic acid, 25 mM KCl, 20 mM MgSO₄, 4 mM sodium phosphate, 1 mM MgCl₂, 4% formaldehyde. All subsequent wash steps and immunostaining were performed with 0.5% bovine serum albumin, 0.1% Triton X-100 in PBS. Ovary pairs were mounted in Mowiol-DABCO. All images of adult ovaries in this study were taken with a Zeiss AxioImager epifluorescence microscope, except for images of ovaries from newly eclosed females in Fig. S2B, which were obtained using a Leica Sp6 confocal microscope.

The primary antibodies used in this study were all from the Developmental Studies Hybridoma Bank: mouse anti-Engrailed/Invected (4D9; 1:20), rat anti-DE-Cadherin (DCAD2; 1:20), mouse anti-Fasciclin II (1D4; 1:10), mouse anti-Fasciclin III (7G10; 1:5), mouse anti-α-Spectrin (3A9; 1:100), mouse anti-β-galactosidase (40-1a; 1:20). The following secondary antibodies were used in this study: Cy5-conjugated goat anti-mouse, Cy5-conjugated goat anti-rat, Alexa Fluor 488-conjugated goat anti-rat (all from Jackson ImmunoResearch; 1:500). Alexa Fluor 633-Phalloidin was used at a concentration of 1:400 to stain for muscular sheath. Hoechst 33342 was used at a concentration of 1:2000 to stain DNA.

All confocal images presented are single z-sections. Images imported in Adobe Photoshop and figures assembled using Adobe Illustrator.

Fat body samples with or without ovary were dissected in PBS as described above. Twenty-four pairs per sample were placed into 300 µl total of RLT buffer (provided by RNeasy Qiagen kit) and frozen immediately at −80°C. Once all samples were prepared, they were thawed, vortexed briefly and RNAs isolated using the RNeasy Qiagen kit. Superscript II enzyme was used to reverse transcribe 680 ng of total RNA per sample. Real-time PCR was performed on a Biorad MyiQ. The primers used to amplify Actin 5C, vasa, bab1, bab2, Fbp2, Hh, Ser and Lmx1a are detailed in Table S3.

RNA (500 ng) was extracted from dissected fat bodies with or without ovaries and submitted for RNA sequencing at the University of Rochester Genomic Research Center. Raw reads generated from the Illumina HiSeq2500 sequencer were demultiplexed using configurebcl2fastq.pl version 1.8.4. Quality filtering and adapter removal are performed using Trimmomatic version 0.36 with the following parameters: ‘SLIDINGWINDOW:4:20 TRAILING:13 LEADING:13 ILLUMINACLIP:adapters.fasta:2:30:10 MINLEN:15’. Processed/cleaned reads were then mapped to the D. melanogaster reference genome (dm6) with STAR_2.5.2b with the following parameters: ‘–twopassMode Basic –runMode alignReads –outSAMtype BAM SortedByCoordinate –outSAMstrandField intronMotif –outFilterIntronMotifs RemoveNoncanonical –outReadsUnmapped Fastx’. Gene-level read quantification was derived using feature counts from the subread-1.5.0p3 package with the dm6 GTF annotation file and the following parameters: ‘-s 0 -t exon -g gene_name’. Differential expression analysis was performed using DESeq2-1.14.1 with an adjusted P-value threshold of 0.05 within R version 3.3.2. Heat maps were generated using ggplot2 after normalizing all values to the median value of Lmx1a^Δ1/+ FatBody+Ovaries samples for each gene. The volcano plot was generated using the same package.

We are grateful to Dirk Bohmann for comments on the manuscript and Dan Bergstralh and members of the Biteau lab for helpful discussions. We would like to thank Victor Zhang for his technical assistance with this project; Jean-Louis Couderc, the Bloomington Drosophila Stock Center, the Vienna Drosophila Resource Center and the Zurich ORFeome Project for fly lines; and the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, for antibodies. RNAseq library construction, sequencing, and primary analysis were performed by the Genomics Research Center at the University of Rochester.