Canonical Wnt/β-catenin activity and differential epigenetic marks direct sexually dimorphic regulation of Irx3 and Irx5 in developing mouse gonads

Early in development, the bipotential mammalian gonad can transform into a testis or an ovary depending on the activation or repression of signaling cascades in the somatic cell lineage (reviewed by Svingen and Koopman, 2013). In the ovary, the canonical Wnt4/Rspo1/β-catenin pathway plays a crucial role for proper differentiation and development (reviewed by Nicol and Yao, 2014). In XX mouse embryos, knockouts of Wnt4 (Vainio et al., 1999; Jeays-Ward et al., 2004) and Rspo1 (Chassot et al., 2008; Tomizuka et al., 2008), or somatic cell loss of their downstream mediator, β-catenin (Manuylov et al., 2008; Liu et al., 2009), results in a partial ovary to testis sex reversal and subsequent loss of 90% of the germ cell population by birth. Conversely, stabilization of β-catenin in the somatic cell population of the XY gonad leads to male-to-female sex reversal, suggesting that β-catenin is a crucial regulator of the sex identity of the somatic cell lineage (Maatouk et al., 2008). Multiple ovarian factors are thought to be regulated by β-catenin and its cognate DNA-binding partners TCF/LEF, but a direct relationship in the ovary has yet to be elucidated.

Previously, we reported that two Iroquois homeobox transcription factors Irx3 and Irx5 are expressed in the ovary, beginning shortly after sex differentiation. Each exhibits a dynamic profile during the course of germline nest establishment and breakdown through primordial follicle formation, suggesting they play important roles in ovarian development (Kim et al., 2011; Fu et al., 2018). Iroquois factors are highly conserved and are known for their roles in patterning and embryogenesis, along with organization of the spinal cord, limb, bone and heart (Bruneau et al., 2001; Diez del Corral et al., 1999; Gómez-Skarmeta et al., 2001; Gómez-Skarmeta and Modolell, 2002; Lovrics et al., 2014). Developmental regulation of these factors within these systems is context specific, as a number of signaling pathways have been described. Recently, we showed that null mutation of both Irx3 and Irx5 resulted in improper somatic-germ cell connections within follicles, which culminated in oocyte death (Fu et al., 2018). Notably, it has previously been reported that the Wnt4 knockout mouse also exhibited physical gaps between germ and somatic cells within follicles (Vainio et al., 1999), suggesting that Wnt and Iroquois factors may lie in the same pathway. Irx3 and Irx5 expression have been attributed to the canonical Wnt signaling pathway in other tissues, including the developing mouse ovary (Naillat et al., 2010, 2015), but a direct link to β-catenin/TCF/LEF transcriptional regulation has not been made.

Based on results from our and other studies, we hypothesized that Irx3 and Irx5 are direct transcriptional targets of the canonical Wnt/β-catenin pathway in the developing ovary. We detected no sex-specific regulatory activity within the proximal promoter regions using ex vivo gonad transfection assays. Instead, we uncovered two distant regulatory sequences within the IrxB locus that promote sexually dimorphic expression during crucial stages of gonad differentiation. Herein, we report that active histone marks work together with β-catenin/TCF/LEF to bind and activate at least two enhancer regions within the IrxB locus to stimulate Irx3 and Irx5 transcription in the ovary. Meanwhile, these same sites were enriched for repressor H3K27me3 chromatin marks that actively repressed their transcription in developing testes. Together, these findings increase our perspective of the complex networks that are in place to ensure appropriate sex differentiation of gonads that include cooperation between epigenetic marks and transcription factors on promoter and distant regulatory sequences. In addition, this report uncovers mechanisms by which bipotential regulation can be achieved on the IrxB locus. These data provide a foundation for new discoveries of mechanisms by which canonical Wnt and other regulatory pathways work together to promote IRX3 and IRX5 function in a spatiotemporal manner within the developing ovary and during organogenesis of other systems, including the heart, limb, bone and spinal cord.

Wnt4/Rspo1/β-catenin regulated transcription plays an essential role in ovarian development in somatic cells during sex differentiation. Our lab has previously reported that Irx3 and Irx5 expression increases upon the onset of sex differentiation in the ovary (Fu et al., 2018) and these factors have been linked to canonical Wnt/β-catenin signaling in other tissues (Janssens et al., 2010; Naillat et al., 2010, 2015). Therefore, we hypothesized that canonical β-catenin regulates Irx3 and Irx5 in the somatic cells of the ovary at this time. To test this hypothesis, ex vivo and in vivo approaches were used to manipulate β-catenin activity to cause loss- and gain-of-function in the developing mouse ovary and testis, respectively. Embryonic day 11.5 (E11.5) wild-type ovaries were dissected and then cultured for 24 h with two different doses of iCRT14, a small molecule that inhibits the interaction between β-catenin and TCF/LEF family members to block β-catenin-mediated gene transcription (Fig. 1A) (Yan et al., 2017; Gonsalves et al., 2011). As expected, treatment did not change the expression of Rps29, a ribosomal protein used as a negative control, but inhibited β-catenin-responsive transcription in a dose-responsive manner. The 50 µM dose decreased expression of known β-catenin target genes Axin2 (77% decrease) and Fst (87% decrease), and caused a significant decrease in Irx3 (73% decrease) and Irx5 (76% decrease) transcripts (Fig. 1B). Next, we evaluated Irx3 and Irx5 transcript accumulation in embryonic ovaries lacking somatic cell β-catenin activity that were generated by crossing Sf1Cre to Ctnnb1^F/F mice (Fig. S1). Rps29 transcripts from E14.5 control (Sf1Cre; Ctnnb1^F/+) and mutant (Sf1Cre; Ctnnb1^F/F) ovaries were not changed, whereas Axin2 and Fst were significantly decreased in mutant ovaries (76% and 96% decreased, respectively). In support of the ex vivo culture findings, Irx3 and Irx5 transcripts were also significantly decreased (65 and 60% decreased, respectively) in the mutant ovaries compared with the controls (Fig. 1C). Together, ex vivo and in vivo results showed that the loss of β-catenin and its transcriptional activity in the developing ovary significantly diminished Irx3 and Irx5 expression.

Canonical Wnt/β-catenin is actively repressed in the developing testis (Kim et al., 2006; Uhlenhaut et al., 2009). Indeed, it has been shown that stabilization of β-catenin within the somatic cell population was sufficient to cause male-to-female sex reversal (Maatouk et al., 2008). Therefore, we evaluated whether β-catenin stabilization in the developing testis influenced Irx3 and Irx5 expression. Wild-type E11.5 testes were cultured ex vivo for 24 h with lithium chloride (LiCl) to stabilize β-catenin (Fig. 1A). Results from treated testes showed no change for Rps29 and significantly increased expression of positive controls Axin2 (fourfold) and Fst (10-fold). Irx3 and Irx5 transcripts also increased nine- and fivefold, respectively, compared to vehicle control (Fig. 1D). Previously it has been reported that stabilized β-catenin activity in somatic cells of developing testes from Sf1Cre;Ctnnb1^Δex3/+ (Harada et al., 1999) embryos caused sex reversal (Maatouk et al., 2008). Transcripts from control (No Cre;Ctnnb1^Δex3/+) and mutant (Sf1Cre;Ctnnb1^Δex3/+) testes (Fig. S1) at E14.5 displayed no significant change in Rps29 transcript levels but exhibited significantly increased expression of Axin2 (11-fold), Fst (7-fold), Irx3 (16-fold) and Irx5 (20-fold) (Fig. 1E). Bmp2 was also used to test for Wnt/β-catenin specificity due to its role as a pro-ovarian gene that is not regulated by Wnt signaling.

Later in ovarian development, upon germline nest breakdown, Irx3 expression expands to include both somatic cells and oocytes (Fu et al., 2018). β-Catenin is also present in oocytes at this stage as shown by our immunohistochemistry results from ovaries at E14.5 and P7 and supported by previous reports (Figs S1 and S2) (Yan et al., 2019; Bothun and Woods, 2019; Kumar et al., 2016; Usongo et al., 2012; Chassot et al., 2011; Jameson et al., 2012). To test whether β-catenin activity regulates expression of Irx3 within oocytes post germline nest breakdown, we targeted loss of Ctnnb1 in oocytes using FiglaCre (Lin et al., 2014) and evaluated ovaries at P0 and P7. Germ-cell-specific loss of β-catenin using FiglaCre was confirmed (Fig. S2A); however, immunohistochemistry analysis indicated no obvious change in IRX3 within oocytes of mutant compared with control mice (Fig. S2B). Altogether, these data suggest that canonical β-catenin transcriptional activity promotes Irx3 and Irx5 expression within somatic cells of the germline nest but does not regulate their transcription within oocytes upon their appearance during germline nest breakdown.

Irx3 and Irx5 are on opposing strands of DNA located 550 kb apart within the IrxB cluster on chromosome 8 in the mouse (Cavodeassi et al., 2001; Peters et al., 2000). Given their proximity, we set out to identify accessible regions of chromatin within the IrxB locus. Previously, we performed DNaseI- and ATAC-seq on XY and XX somatic cell populations sorted from embryonic gonads at E10.5 (pre-sex determination) and E13.5 (post-sex determination) (Maatouk et al., 2017; Garcia-Moreno et al., 2019a). These datasets were used to interrogate chromosome 8 spanning 600 kb on either side of the Irx3 transcription start site (TSS) to search for areas of open chromatin that also included the consensus motif for β-catenin-binding partners TCF/LEF (TCAAAG) (van de Wetering et al., 1997). The sites that were used for further evaluation included those that were either resolved (R) by or derived de novo (D) by E13.5 in the ovary (Garcia-Moreno et al., 2019a). Five sites of interest were identified and named based on their distances from the Irx3 TSS: +205 kb, +86 kb, −305 kb (A,B) and −580 kb (see boxed peaks in Fig. 2). The site at −305 kb contained two separate TCF/LEF binding motifs, labeled ‘A’ and ‘B’; all others harbored a single consensus element. A map detailing the approximate location of each site relative to Irx3 and Irx5 is outlined in Fig. 2.

To evaluate these open chromatin sites, we harvested ovaries and testes from E13.5-E14.5 embryos to perform chromatin immunoprecipitation (ChIP) qPCR using antibodies for H3K27ac to mark active enhancer sites, and TCF7L2 to identify TCF/LEF binding sites relevant to the developing ovary. Of TCF/LEF factors, TCF7L2 was chosen for the following reasons: robust ChIP-seq data are available on the ENCODE database; microarray data indicating TCF7L2 is expressed predominantly in the somatic cells of the gonad (Jameson et al., 2012); and the GUDMAP database shows that TCF7L2 expression is detected in the ovary via in situ hybridization, whereas other TCF/LEF factors are negative (Harding et al., 2011; McMahon et al., 2008). For each replicate, whole-gonad ChIP was first validated by showing RNA polymerase II enrichment at the GAPDH promoter and TCF7L2 presence at a known β-catenin/TCF complex target, the SP5 promoter (Kennedy et al., 2016), in ovaries and testes (Fig. S3). TCF7L2 and H3K27ac are present in other cells besides pre-granulosa cells; therefore, we anticipated variability in ChIP-PCR data from replicates sourced from whole-gonad tissue. Despite this potential barrier, our ChIP-PCR results showed substantial enrichment of H3K27ac and TCF7L2 binding on putative enhancer sequences in ovary (XX) compared with testis (XY) tissue (Fig. 3). Combined evaluation of H3K27ac and TCF7L2 results from ovary tissue suggest that β-catenin/TCF/LEF transcription factors bind and act on enhancer sequences at the +86 kb (H3K27ac 20.1-fold; TCF7L2 2.2-fold enrichment) and −580 kb (H3K27ac 11.1-fold; TCF7L2 2.7-fold enrichment) sites to regulate Irx3 and Irx5 expression within developing ovaries. These results also suggest the potential for sex-specific regulation.

To test β-catenin responsive enhancer activity, each potential regulatory site was cloned into a luciferase reporter vector containing a minimal E1b promoter (Huang et al., 2006). In addition, each reporter vector was altered to include a single point mutation (in bold) of the TCF/LEF-binding motif (TCAAAG to CCAAAG), which is the same mutation that differentiates the TOPflash (active) versus FOPflash (inactive) β-catenin reporter plasmids (Korinek et al., 1997) (Table S1). Reporter plasmids were transfected into HEK293 cells along with a constitutively active β-catenin expression vector, CMV-S37A (Jordan et al., 2003). Specific β-catenin activity of the CMV-S37A expression vector was confirmed using co-transfection with positive and negative control reporter vectors, TOPflash and FOPflash, respectively (Fig. S4). Among all reporter vectors, including +250 kb, +86 kb, −305 kb (A,B) and −580 kb, only the +86 kb and −580 kb plasmids exhibited a significant increase in reporter activity that was specific to the putative TCF/LEF-binding site. Of note, the larger plasmid containing wild type −305 kb sequences, which includes A and B TCF/LEF binding sites, was not responsive to CMV-S37A, and the double mutation of A and B had no effect (Fig. 4A). To test whether the +86 kb and −580 kb DNA enhancers (together equals 209 bp) could stimulate promoter activity, both were cloned into the pGL3 basic luciferase reporter in front of 2080 bp of the mouse Irx3 promoter (+86 kb; −580 kb; −1634/+446 bp mIrx3 pGL3). Constitutively active CMV-S37A co-transfected with the enhancer plus promoter reporter stimulated a threefold increase in activity compared with promoter alone. In addition, single base pair point mutations of the TCF/LEF-binding site in each enhancer sequence completely disrupted enhancer activity (Fig. 4B). Together, these data suggest that the +86 kb and −580 kb enhancer sequences confer β-catenin-specific regulatory activity within the context of the Irx3 promoter.

Based on ovary-specific expression of Irx3, we reasoned that sequences within the Irx3 promoter would confer ovary-specific expression. To test this hypothesis, three different sized segments of the mouse Irx3 promoter (−351/+446 bp, −603/+446 bp and −1634/+446 bp) were cloned into a luciferase reporter plasmid and transfected into ovaries and testes from E13.5-E14.5 embryos (Fig. 5A). Although reporter activity increased along with longer promoter sequences, none of the promoters exhibited a significant difference when testis and ovary reporter activities were compared (Fig. 5B). Next, we tested whether β-catenin-specific activity within the +86 kb and −580 kb enhancer sequences would promote ovary-specific expression. Both enhancers and their mutated counterparts were cloned in front of the most active promoter (−1634/+446 bp mIrx3pGL3) reporter vector and transfected into E14.5 gonads. Although the enhancer plus mIrx3 promoter was equally expressed in both ovary and testes, only ovary expression was disrupted upon single basepair point mutations of the TCF/LEF-binding sites (60% decrease from wild-type enhancers) (Fig. 5C). Based on these results, we conclude that the +86 kb and −580 kb enhancer sequences promote β-catenin responsive activity only within the ovary.

Plasmid vectors containing the enhancer sequences linked to the mouse Irx3 promoter did not confer ovary- versus testis-specific reporter activity, as expected; however, one limitation to this analysis is that plasmid reporter vectors lack epigenetic decorations that may have a profound impact on enhancer and/or promoter activity. Thus, we hypothesized that repressor histones suppress the +86 kb and −580 kb enhancer sequences within the developing testis. To test this hypothesis, we performed ChIP-Seq for the repressive histone modification H3K27me3 on FACS-purified XX and XY supporting cells from E13.5 gonads of TESMS-CFP (Gonen et al., 2018) and Sox9-CFP (Sekido and Lovell-Badge, 2008) transgenic mice, which fluorescently label granulosa (ovary) and Sertoli (testis) cells, respectively. Each ChIP-seq experiment was performed on two biological replicates containing pooled cells from multiple gonads. To maintain consistency, we performed ChIP-seq on the same somatic cell populations used for ATAC-seq (Fig. 2). To validate our datasets, we compared our results with previously published H3K27me3 ChIP-seq (performed on the same somatic populations) and found high correlation among all four biological replicates (Fig. 6A). ChIP-seq data had previously been validated on promoters from genes known to drive sex determination and differentiation (Garcia-Moreno et al., 2019b). Results from ovary and testis H3K27me3 ChIP-seq are presented as peaks from individual replicates and include a solid horizontal bar above each set to illustrate the statistically positive sites, as determined by HOMER analysis. Black vertical bars are included above these sites to illustrate positive ATAC-seq data. Results within the Irx3/Irx5 locus show that H3K27me3 marks are enriched at each of the four selected sites in the Sertoli cells but are absent in granulosa cells. In contrast, ATAC-seq peaks are stronger in granulosa cells when compared with Sertoli cells (Fig. 6A). To illustrate the dynamic nature of epigenetic regulation that occurs during sex determination, we present a magnified view of the select sequences with a representative H3K27me3- and ATAC-seq replicates at E10.5 (pre-sex determination) and E13.5 (post-sex determination) in XX and XY supporting cells (Fig. 6B). Stage-dependent epigenetic control is evident at each enhancer site. At +205 kb, chromatin, which is initially accessible pre-sex determination, remains open in granulosa cells, whereas it transitions to a repressed and closed state in Sertoli cells at E13.5. At +86 kb, chromatin is initially open in both XX and XY at E10.5. At E13.5, accessibility increases and H3K27me3 decreases in granulosa cells, whereas H3K27me3 levels increase in the Sertoli cells. Finally, closed chromatin at −306 kb and −580 kb, becomes accessible in granulosa cells at E13.5, whereas this site accumulates H3K27me3 in Sertoli cells.

Taken together, our data uncovers two specific enhancers within the IrxB locus that confer ovary versus testis specific promoter activity. In the ovary, canonical β-catenin activity cooperates with active epigenetic marks on open chromatin to stimulate the +86 kb and −580 kb enhancers while expression is silenced in the testis due to the combined effects of histone methylation repression and the lack of functional β-catenin activity.

Canonical Wnt/β-catenin signaling has been reported to promote Irx3 and Irx5 expression in the ovary (Naillat et al., 2010, 2015) and other tissues such as the brain (Braun et al., 2003), lung (Bell et al., 2008), neural axis (Janssens et al., 2010), kidney (Holmquist Mengelbier et al., 2019) and in colon cancer (Hovanes et al., 2001), but evidence for direct trans-acting regulation via DNA-binding partners has not been elucidated. Here, we used in vitro, ex vivo and in vivo approaches to provide a direct link between canonical Wnt/β-catenin signaling and Irx3 and Irx5 expression within the somatic cell population of the developing ovary. In addition, we previously showed that both Iroquois factors emerge in germ cells in late stages of ovarian development (Fu et al., 2018), but here we report that their regulation in this cell type is independent of β-catenin. Based on these data, we developed a model to describe regulation of Irx3 and Irx5 expression within the somatic cell population (Fig. 7). We uncovered two enhancer sequences that, although distant from the transcription start site, provide the focus for regulation within the ovary and testis. Our data indicate that chromatin enhancer marks work in conjunction with β-catenin/TCF/LEF at these sites to stimulate the IrxB locus in the ovary, while the absence of activated β-catenin in somatic cells along with repressive histone marks enriched at these same sites functionally antagonize expression of Irx3 and Irx5 in the testis. Together, these findings highlight interactions between signaling pathways and epigenetic marks that regulate Irx3 and Irx5 to ensure appropriate expression based on time, sex and cellular environments within developing gonads.

We examined ∼1200 kb of chromosome 8, which included the IrxB and Fto loci, for female-specific open chromatin sites that could also mediate canonical β-catenin regulation within somatic cells after sex differentiation. Altogether, our DNAseI-Seq (Maatouk et al., 2017) and ATAC-Seq (Garcia-Moreno et al., 2019a) data, along with results from ENCODE ChIP-Seq derived TCF/LEF enrichment in human cell lines (ENCODE Project Consortium, 2012) illuminated five putative sites that met these criteria. Notably, none was identified within the proximal promoters of either Irx3 or Irx5. One limitation of this study is that only perfect matches to the TCF/LEF-binding domains were explored. HMG box transcription factors, such as TCF/LEF, can also bind to DNA motifs that do not match the perfect consensus sequence; therefore, all potential binding regions were not explored. There were also several sex-specific sites of open chromatin within the IrxB locus that were not considered. Regarding the nucleosome-depleted regions that were evaluated in this study, JASPAR database interrogations uncovered a variety of other putative binding sites (Khan et al., 2018). These in silico analyses did not distinguish any common suite of transcription factor-binding sites within either the resolved or de novo open chromatin regions. In particular, the two sites that we identified as the most promising β-catenin-responsive enhancers in the ovary, +86 kb and −580 kb, were characterized as resolved and de novo sites, respectively. Besides TCF7L2, the +86 kb site harbors sequences that also bind CTCF, p300 and YY1, factors that are important for facilitating higher order chromatin structures (Ghirlando and Felsenfeld, 2016; Chan and La Thangue, 2001; Deng et al., 2010). These findings, in combination with the significant distance between enhancer sites and our results showing that 20-40% of Irx3 and Irx5 transcripts remain after elimination of β-catenin, suggest that chromatin encompassing the IrxB locus loops and undergoes extensive remodeling in response to sex-specific signals and developmental time.

An important goal of our study was to determine the relevance of putative enhancer sequences in promoting ovary-specific expression of Irx3 and Irx5. To that end, we combined traditional cell-based transfection assays with our previously described microinjection and electroporation technique to transfect reporter plasmids into embryonic gonads (Gao et al., 2011) with a specific focus on the mouse Irx3 promoter (mIrx3-pGL3). We were surprised to find that the enhancer/Irx3 promoter reporters were equally active in ovaries and testes. There are a number of potential reasons for this result. For example, it is recognized that reporter plasmid DNA transfection assays are used to focus attention to specific sequences, which are not in their normal context and, therefore, must be interpreted as such. In addition, plasmid DNA is devoid of epigenetic information, which has important implications on regulation. Indeed, our H3K27me3-Seq data show substantial enrichment at the proximal promoters of Irx3 and Irx5 only in XY cells (Fig. S5). Furthermore, our results showed that the response to mutation of the canonical TCF/LEF-binding site was not present in transfected testes, but was sensitive in cell and ovary transfection, both of which are β-catenin-responsive environments. Additional JASPAR analysis of the enhancer and promoter sequences uncovered several putative binding sites, including SOX, GATA, EZH2, CEBP and SP1 factors, among others. Together, these findings support the hypothesis that ovary-specific regulation for Irx3 and Irx5 is linked to canonical β-catenin signaling and opens the door for other means of regulation in the context of the loss of epigenetic marks that might explain high levels of testis reporter activity.

Because we expected reporter activity to be lower in transfected testes compared with ovaries, we evaluated the enhancer sites for histone repressor marks and found H3K27me3 marks were specifically enriched at +86 kb and −580 kb enhancer sites in addition to the proximal promoters of Irx3 and Irx5 in somatic cells of testes, but not ovaries. Thus, taken together, we conclude that both +86 kb and −560 kb open sites are subject to changing epigenetic landscapes. In newly differentiated ovarian somatic cells, active β-catenin/TCF complexes accumulate on the IrxB locus to stimulate Irx3 and Irx5 transcription. In contrast, somatic cells that are destined for the testis phenotype lack β-catenin and, instead, recruit epigenetic decorations consistent with transcriptional repression.

Irx3 and Irx5 show dynamic ovary-specific expression profiles (Fu et al., 2018). Besides canonical Wnt/β-catenin signaling, Irx3 has also been shown to be controlled by other pathways, including TGFβ (Cavodeassi et al., 2001; Gómez-Skarmeta et al., 1998), SHH (Briscoe et al., 2000; Kobayashi et al., 2002), FGF (Kobayashi et al., 2002), and retinoic acid (Sirbu et al., 2005). Notably, many of these ligands have been established as sex-specific signals that also depend on a cadre of active transcription factors and epigenetic marks within the somatic cell population during the sex differentiation window (Garcia-Moreno et al., 2018, 2019a; Katoh-Fukui et al., 2012; Hiramatsu et al., 2009; Morais da Silva et al., 1996). But important questions remain: which comes first and which regulates which? An interesting conundrum related to this question was illustrated in the evolving story of CBX2-mediated regulation of Sry. Originally, it was proposed that CBX2, a subunit of the canonical polycomb repressive complex 1 (PRC1) acted as a direct activator of Sry (Katoh-Fukui et al., 2012). New studies have since refined this discovery and now show that, within the developing testis, CBX2 and PRC1 establish repressor H3K27me3 marks to extinguish the rising profile of ovary pathway genes, which allows for accumulation of Sry (Garcia-Moreno et al., 2019b). Other chromatin modifiers, including GLP-G9a/JMJD1 and CBP/p300, contribute to Sry expression by modulating H3K9Me2 repressor and H3K27ac marks (Kuroki et al., 2013, 2017; Carré et al., 2018). These new data illustrate how epigenetic writers and readers can play a crucial role in sex determination and differentiation. Important insight can also be learned from species where sex determination is influenced by both genes and environmental cues. Indeed, there is a growing field in developmental epigenetics that increasingly recognizes that environmental cues are translated into specific sex phenotypes via epigenetic manipulations of sex determining genes (Navarro-Martín et al., 2011; Matsumoto et al., 2013; Parrott et al., 2014; Piferrer, 2013). Ultimately, plastic epigenetic marks provide flexibility and a means of preserving survival of sexually reproductive species.

The results of this study highlight the importance of transcription factor binding and local epigenetic landscape in illuminating cell and sex-defining fates during gonadogenesis. Distant enhancer sites have long been implicated in gene control in the gonad and new technologies are improving our capacity to identify and validate their importance (Sekido and Lovell-Badge, 2008; Gonen et al., 2018, 2017). Here, we describe two distal enhancer sites on the IrxB locus that are actively repressed in developing testes, while at the same time being engaged with active chromatin marks and β-catenin/TCF to stimulate Irx3 and Irx5 expression within the developing ovary. Thus, Irx3 and Irx5 are bona fide downstream targets of Wnt/Rspo1/β-catenin that, together, are crucial mediators of ovary development and oocyte survival. These findings allow us to begin to unravel the means by which specific cell environments control Irx3 and Irx5 expression within the fetal ovary. We suggest that these same principles could be applied to the developing brain, spinal cord, lung and kidney, or to abnormal cellular activity in Iroquois-positive cancers.

Mouse strains included CD1 outbred mice [Crl:CD1(ICR), Charles River]; Sf1Cre mice (C57BL/6), originally obtained from the Keith Parker Lab (University of Texas Southwestern Medical Center) (Bingham et al., 2006); Ctnnb1 conditional loss-of-function (LOF) mice (B6.129-Ctnnb1^tm2Kem/KnwJ, Jackson Labs); and Ctnnb1 conditional gain-of-function (GOF) mice (C57BL/6; β-cat^fl.ex3), obtained from Dr Makoto Mark Taketo (Kyoto University, Kyoto, Japan) (Harada et al., 1999), Sox9-CFP (Sekido and Lovell-Badge, 2008) and TESMS-CFP (Gonen et al., 2018). Timed mating was identified by the presence of a vaginal plug, which was designated as embryonic day 0.5 (E0.5). Animals were dissected at the appropriate time and genomic DNA was isolated from tails or ear notches and subjected to PCR using gene-specific primers: Sf1Cre, 5′-GAGTGAACGAACCTGGTCGAAATCAGTGCG-3′ and 5′-GCATTACCGGTCGATGCAACGAGTGATGAG-3′; Ctnnb1 wild-type and floxed (LOF) allele, 5′-AAGGTAGAGTGATGAAAGTTGTT-3′ and 5′-CACCATTGTCCTCTGTCTATTC-3′; and Ctnnb1 wild-type and β-cat^fl.ex3 (GOF) allele, 5′-GCTGCGTGGACAATGGCTACTCAA-3′ and 5′-GCCATGTCCAACTCCATCAGGTCA-3′. In the case where sex could not be determined visually, PCR for SRY was performed using 5′-TGCAGCTCTACTCCAGTCTTG-3′ and 5′-GATCTTGATTTTTAGTGTTC-3′.

Animal housing and all procedures described were reviewed and approved by the Institutional Animal Care and Use Committee at the University of Wisconsin-Madison and were performed in accordance with the National Institutes of Health Guiding Principles for the Care and Use of Laboratory Animals. Mice were housed in disposable, ventilated cages (Innovive). Rooms were maintained at 22±2°C and 30-70% humidity on a 12 h light/dark cycle.

Gonad cultures were performed using a modified version of previously described protocols (Martineau et al., 1997; Maatouk et al., 2008). Briefly, E11.5 gonad/mesonephros complexes were cultured at 37°C with 5% CO₂/95% air in ∼20 μl droplet of culture media [DMEM F-12 (Fisher, SH3002301)] supplemented with 10% fetal bovine serum (Fisher, SH3091003) and 1% penicillin-streptomycin (Fisher, ICN1670249). The sex of the gonads was determined by genotyping PCR for SRY (see above). Gonad/mesonephros complexes were placed in round droplets of media on an inverted lid of a 100 mm Petri dish within a humidified chamber. Gonads (XX and XY) were cultured in a droplet supplemented with either vehicle control, the indicated concentrations of iCRT14 (XX gonads, Sigma SML0203) or LiCl (XY gonads, Fisher L121-100) for 24 h, rinsed with PBS and then harvested for RNA extraction and qPCR analysis.

RNA was extracted using Trizol (Invitrogen, 15596026) according to the manufacturer's instructions and quantified using a NanoDrop 2000. RNA from each sample (500 ng) was used for First-Strand cDNA synthesis by SuperScriptII-RT (Invitrogen, AM9515). cDNA was diluted 1:5 and then 2 μl was added to 5 μl SYBR green PCR mixture (Applied Biosystems), 2.4 μl water and 1.25 pmol primer mix. PCR reactions were carried out using the ABI Prism 7000 Sequence Detection System (Applied Biosystems). RNA transcripts were quantified using the ΔΔCt method (Livak and Schmittgen, 2001). Briefly, to control for overall gene expression in each time point, the average cycle threshold (aveCt) for 36B4 was subtracted from the aveCt value for each gene to generate ΔCt. Next, ΔCt for each gene was compared with ΔCt of that same gene for the mutant genotype (e.g. ΔCt Irx3_{female control}−ΔCt Irx3_{female mutant}) to generate ΔΔCt. Finally, fold-change was calculated as 2 to the −ΔΔCt power (2^−ΔΔCt). Primers are listed in Table S2.

DNAse-I, ATAC-seq and H3K27me3 data were mined from previous studies (Maatouk et al., 2017; Garcia-Moreno et al., 2019a,b). These data were analyzed for open chromatin regions within 600 kb on either side of the Irx3 transcription start site (TSS) of chromosome 8 in the mouse. Open chromatin regions that were specific to the granulosa cells after sex determination (E13.5) were explored for TCF/LEF-binding motifs using the JASPAR database (jaspar.genereg.net). The sequences containing the highest scores for binding potential were chosen for further investigation.

E13.5-E14.5 CD1 ovaries and testes without mesonephros were harvested, snap frozen and stored at −80°C. 100-150 pairs of snap-frozen gonads were thawed and fixed in 1% formaldehyde for 15 min at room temperature with gentle shaking. The reaction was quenched with 160 μl of 1.25 M glycine for 5 min at room temperature with gentle shaking. Samples were washed twice with PBS and cOmplete protease inhibitor (CPI) tablets (Roche, 04693116001) then resuspended in 400 μl RIPA lysis buffer+CPI tablets. Samples were homogenized with a pestle, then chromatin shearing was performed by lightly sonicating via probe-based sonication to fully lyse cells and the nuclear envelope followed by 1 min incubation at 37°C using 1000 gel units of Micrococcal Nuclease (MNase) (New England Biolabs, M0247S). A separate 5 μl sample was incubated with Proteinase K to validate efficient shearing of DNA (between 300 and 900 bp). The MNase reaction was stopped with 1.25 μmol EGTA. Debris was removed by centrifugation at 15,000 g for 10 min at 4°C and then 100 μl of each lysate was diluted in 200 μl IP buffer (PBS+0.05% Triton X-100) and incubated overnight with 2 μg of antibody. After overnight antibody incubation, 25 μl of Dynabeads protein G magnetic beads (Life Technologies, 10004D) was added and mixed for 2 h at 4°C with gentle rocking. Samples were washed sequentially with 500 μl low salt, 500 μl high salt and 500 μl TE buffers, then resuspended in digestion buffer [50 mM Tris, 10 mM EDTA, 0.5% SDS (pH 8.0)] and proteinase K for 2 h at 62°C. DNA was isolated via ethanol precipitation. qPCR analysis was performed to quantify relative amounts of DNA enrichment; immunoprecipitated (IP) samples were normalized to input and IgG. Antibodies used were anti-phospho RNA PolII (Ser2) clone 3 (Millipore, MABE954), normal mouse IgG (Sigma, M8695), TCF4 (C4H811) (Cell Signaling Technology, 2569S) and Histone H3K27ac (Active Motif, 39133).

ChIP-seq was performed and analyzed as described by Garcia-Moreno et al. (2019b). Briefly, XX and XY supporting cells were FACS purified from E13.5 XY SOX9-CFP gonads and E13.5 XX TESMS-CFP gonads on the same day, and immediately processed for ChIP-seq. ChIP-seq was performed with no modifications (Van Galen et al., 2016) on two biological replicates, each containing ∼150K FACS-purified supporting cells from pooled gonads. 400K Drosophila S2 cells were added per IP as carrier chromatin. ChIP-seq was performed using 3 µl of H3 antibody (Active Motif, 39763) (used as input) or 5 µl H3K27me3 antibody (CST, 9733S).

Sequence alignment to the mm9 mouse genome was performed using Bowtie. H3 ChIP-seq was used as input. To identify regions significantly enriched for H3K27me3 compared with flanking regions (peaks), HOMER was used for each independent replicate using the findPeaks function and settings ‘—style histone’ and ‘-C 0’, with a size of 5000. BigWig files were created using bedGraphToBigWig for visualization on the UCSC genome browser.

Luciferase reporters were generated from mouse genomic sequences of the enhancers at +205 kb, +86 kb, −305 kb and −580 kb from the Irx3 transcription start site (TSS) specific to the region containing the TCF/LEF-binding motif via PCR with the addition of the KpnI and XhoI restriction enzyme sequences (Table S1). Each sequence was digested and inserted into the pGL3 basic vector containing a minimal E1b promoter (Huang et al., 2006) digested at the KpnI and XhoI sites. The QuikChangeII site-directed mutagenesis kit was used to make a single base pair mutation for each TCF/LEF-binding site, as directed in the manufacturer's protocol (Stratagene) (Table S1). The mouse Irx3 promoter construct was generated from mouse genomic sequence using primers specific to 1634 bp upstream and 446 bp downstream of the Irx3 TSS and placed into the pGL3 basic vector. The +86 kb and −580 kb sequences were inserted in front of the mouse Irx3 promoter using the NEBuilder HiFi DNA Assembly Cloning Kit according to the manufacturer's instructions (New England BioLabs, E5520S). Each reporter construct was sequenced for accuracy after initial construction and proper mutation following mutagenesis (Sanger sequencing, UW Madison Biotech Center).

Plasmids containing promoter regions of Irx3 were constructed via the Ensembl Irx3 gene sequence and primer design software (Primer Designer version 1.01). PCR primers targeted the promoter region 5′ of the Irx3 TSS. Genomic DNA was amplified, and inserts were blunt-end ligated into the pST-blue Accepter vector (Novagen). Sequencing was then performed (Keck Center, University of Illinois); the inserted sequence was compared with the archived DNA sequence (NT_078586.1) and validated for accuracy.

80,000 HEK293 cells (purchased from ATCC, CRL-1573, validated before shipment) were plated in 24-well plates (Thermo Scientific, 12565163) for transfection assays. Plasmids were prepped using column-based mini or midi prep kits (Qiagen, 27104, 12143) and quantified using a NanoDrop 2000. Cells were transfected using Lipofectamine 2000 (Invitrogen, 11668019) with plasmid DNA diluted in OPTI-MEM media (Fisher, 31985070) according to the manufacturer's instructions. Luciferase reporter vectors were transfected at 0.8 μg/well along with 50 ng/well co-expression vector CMV-EGFP (Addgene, 11153) or CMV-S37A-β-catenin (Jordan et al., 2003) (kindly provided by Dr Vincent Harley, Hudson Institute for Medical Research, Monash University, Melbourne, Australia) for normalization or treatment, respectively. The Lipofectamine 2000 mixture was incubated with the cells for 16-18 h followed by a media change. After 24 h, the cells were lysed using 1× passive lysis buffer and read using the Dual Luciferase Reporter Assay (Promega, E1910). Treatment groups were plated in triplicate and experiments were repeated at least three times. Luciferase values from the treatment group were normalized to the non-treatment group and also normalized to the empty vector control.

Transient transfection assays in urogenital ridge explant cultures were based on previously reported methods of the explant culture system (Jorgensen and Gao, 2005). The sex of the gonad tissue was determined by characteristic findings of a coelomic vessel and testicular cords in the male and the lack of these in the female. Urogenital ridges were harvested from embryos at E14.5 and injected with ∼0.5 µl of a DNA cocktail containing 4 μg/μl pGL3, wild-type +86 kb/−580 kb/mIrx3 promoter pGL3 or mutated +86 kb/−580 kb/mIrx3 promoter pGL3 plus 2 μg/μl SV40-Renilla luciferase in Dulbecco phosphate-buffered saline (PBS; Sigma D8537). An additional aliquot of 25 μl of sterile PBS was placed on the gonad for electroporation. Immediately thereafter, five square electrical pulses of 65 V, 50 ms each at 100 ms intervals, were delivered through platinum electrodes from an electroporator. After electroporation, urogenital ridges were placed back into the culture for 24 h. Explant cultures were maintained at 37°C with 5% CO₂/95% air in 50 μl of Dulbecco minimal Eagle medium (DMEM) supplemented with 10% FBS (fetal bovine serum) and 1% penicillin/streptomycin. Transfected gonad explants were harvested in 50 μl passive lysis buffer, snap frozen, subjected to three freeze-thaw cycles and then processed for dual luciferase assays. Data were calculated by taking the ratio of luciferase to renilla expression with at least three biological replicates for each injected plasmid.

Statistics between groups were carried out using a two-tailed t-test assuming unequal variances. Results were considered statistically significant if P≤0.05. One-way ANOVA with a post-hoc Tukey's comparison was performed where appropriate.

Special thanks to all members of the Jorgensen lab for ideas and support throughout this work, and to Drs John Svaren and Linda Schuler for extraordinary consultation. We are also extremely grateful for the technical support from Dr Camila Lopez-Anido and Dr John Svaren with our ChIP experiments. We also thank Janelle Ryan and Dr Vincent Harley for the CMV-S37A active β-catenin plasmid. We have an immense amount of gratitude to the cohort of undergraduates that helped with immunofluorescence, genotyping and sectioning, especially Linda Wang, Ann Turcotte and Rebekah Schroeder. In addition, we thank the University of Wisconsin Biotron for the exemplary mouse husbandry help. Dedicated to Danielle M. Maatouk, a wonderful colleague, scientist and friend.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.183814.reviewer-comments.pdf