Lats1 and Lats2 are required for the maintenance of multipotency in the Müllerian duct mesenchyme

The Müllerian ducts (or paramesonephric ducts) are paired embryonic structures that give rise to most of the female reproductive tract. They comprise a luminal epithelial cell population, mesenchymal cells that surround the epithelium, and an outer coelomic epithelium. The ducts form when specialized cells from the rostral mesonephros invaginate and elongate caudally alongside the Wolffian ducts, until they join at the urogenital sinus. In male embryos, the production of anti-Müllerian hormone (AMH) causes the Müllerian ducts to degenerate, and testosterone spurs the development of the Wolffian duct into the male reproductive tract. In the female, absence of AMH allows the Müllerian ducts to further develop, fuse and differentiate into the oviducts, uterus and anterior vagina (Roly et al., 2018).

All steps of the initiation, invagination, elongation and differentiation (or degradation) of the Müllerian ducts is tightly coordinated by a number of signaling effectors and pathways. Among these, members of the WNT family of secreted glycoprotein signaling molecules play prominent roles. Using cell-fate mapping and neutralizing antibodies, WNT4 was shown to be required for Müllerian duct invagination and elongation (Prunskaite-Hyyrylainen et al., 2016), and Wnt4 null mice are born without Müllerian structures (Vainio et al., 1999). Additional studies in mice using hypomorphic alleles and conditional targeting have shown that Wnt4 is also involved in the development of the oviduct and the myometrium (Prunskaite-Hyyrylainen et al., 2016). Wnt5a null mice lack both cervix and vagina, and have short, coiled uterine horns (Mericskay et al., 2004). Moreover, deletion of the Wnt7a gene results in a partially posteriorized female reproductive tract, with uterus-like oviducts and a uterus with histological characteristics of the vagina (Dunlap et al., 2011). Wnt4, Wnt5a and Wnt7a have also all been implicated in uterine gland formation (Dunlap et al., 2011; Farah et al., 2017; Franco et al., 2011; Mericskay et al., 2004; St-Jean et al., 2019). Although the signaling processes whereby WNT ligands direct the development of the female reproductive tract are poorly understood, the canonical WNT signaling effector β-catenin (CTNNB1) is believed to play an important role, largely based on phenotypic abnormalities observed in Ctnnb1 conditional knockout models that mimic to some extent those observed in knockout models of specific WNT genes (Deutscher and Hung-Chang Yao, 2007; Hernandez Gifford et al., 2009). How WNT signaling itself is regulated in the Müllerian duct and during reproductive tract development remains essentially unknown.

The Hippo signaling pathway plays important roles in the embryonic development of many tissues such as the heart (Heallen et al., 2011), nervous system (Bao et al., 2017) and kidneys (McNeill and Reginensi, 2017; Reginensi et al., 2016) by regulating cellular processes, including proliferation, apoptosis and differentiation. Hippo is not activated by a specific ligand, but rather responds to a multitude of extra- and intracellular cues such as cell-cell contact, cell polarity, mechanical forces and several G protein-coupled receptor ligands (Lv et al., 2015; Yu et al., 2012; Zhao et al., 2007). The pathway consists of a core kinase cascade that culminates in the activation of the kinases large tumor suppressor (LATS) 1 and 2. Once activated, LATS1/2 can phosphorylate the transcriptional co-regulatory molecules YAP (YAP1) and TAZ, resulting in their cytoplasmic sequestration and degradation. Inactivation of the pathway results in the translocation of unphosphorylated YAP and TAZ to the nucleus, which bind to various transcription factors (including those of the TEAD family) to modulate the expression of specific target genes (Tsutsumi et al., 2013; Zhao et al., 2008). Importantly, recent studies have established functional interactions between Hippo and the canonical WNT signaling pathway. For instance, TAZ was found to inhibit WNT signaling by interacting with the WNT signaling effector Disheveled (Varelas et al., 2010). YAP/TAZ can also recruit the ubiquitin ligase β-TRCP (BTRC) to a multiprotein complex that directs the ubiquitylation and degradation of CTNNB1 (Azzolin et al., 2014). Furthermore, YAP and CTNNB1 can physically interact when in the nucleus and bound to their respective transcription factor partners, and this interaction enhances target gene transcriptional activity (Heallen et al., 2011). Although this mutual interdependence between Hippo and canonical WNT signaling has been shown in several cell types, whether or not this occurs within the Müllerian duct during reproductive tract development has never been reported.

A few recent studies have suggested that the Hippo pathway may also play important roles in postnatal uterine physiology and disease processes. For instance, Strakova et al. reported decreased expression of TAZ in the nuclei of human uterine fibroblasts during in vitro decidualization, as well as decreased TAZ expression in the uterine stroma during the secretory phase of the menstrual cycle, suggesting a potential role in decidualization and implantation (Strakova et al., 2010). Increased expression of both YAP (YAP1) and TEAD1 mRNA levels has also been observed during decidualization in cultured human endometrial stromal cells. Knockdown of YAP in these cells delayed the morphological and gene expression changes associated with decidualization (Chen et al., 2017). Further studies have suggested roles for dysregulated YAP and/or TAZ in the pathogenesis of endometriosis (Song et al., 2016) and uterine cancers (Liu et al., 2013a; Zhan et al., 2016).

To determine the roles of Hippo signaling in the development and function of the female reproductive tract, we generated mouse models to inactivate Lats1 and/or Lats2 in Müllerian mesenchyme cells. Unexpectedly, loss of Hippo signaling caused the mesenchyme cells to lose their ability to differentiate into uterine and oviduct stromal and smooth cells. Rather, loss of Lats1/2 directed the mesenchymal cells to adopt the myofibroblast fate, resulting in severe developmental defects of the uterus and absent oviducts. These effects were attributed to increased YAP and TAZ (but not CTNNB1) expression and transcriptional regulatory activity, leading to the activation of the myofibroblast genetic program. Together, our findings indicate a previously unsuspected role for the Hippo pathway in the maintenance of multipotency in Müllerian mesenchyme cells.

To study the roles of the Hippo pathway in the development and physiology of the female reproductive tract, we first determined the expression of its key components. Immunohistochemical analyses showed weak TAZ expression throughout all stages of development compared with the other Hippo effectors. Staining was mostly absent in embryonic day (E) 17.5 and newborn mice, whereas weak cytoplasmic staining appeared in the endometrium and myometrium of adult mice (Fig. S1). Conversely, high levels of LATS1, LATS2 and YAP expression were found at all stages of development. Their expression localized mostly to the nucleus during early stages of development (E17.5 and newborn) in the Müllerian mesenchyme/uterine stroma, along with scant cytoplasmic signal in the luminal epithelium. In adult mice, LATS1, LATS2 and YAP nuclear and cytoplasmic expression was observed in all parts of the uterus (Fig. S1). PhosphoYAP expression mirrored that of YAP at E17.5, but was absent in the uteri of newborn mice. Cytoplasmic phosphoYAP expression then reappeared throughout the uterus in adult mice. (Fig. S1).

To inactivate Hippo signaling in the female reproductive tract, mice bearing floxed alleles for Lats1 and/or Lats2 were crossed with the Amhr2^cre strain, which targets the mesenchyme cells of the Müllerian duct, starting at the rostral end as early as E12.5 in mice of both sexes (Arango et al., 2008; Jamin et al., 2002). Whereas Lats1^flox/flox; Amhr2^cre/+ and Lats2^flox/flox; Amhr2^cre/+ females were fertile and had reproductive tracts that appeared normal at the gross and histological levels (not shown), Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ females were sterile (Table S1). Developmental time-course analyses of the reproductive tracts of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice revealed numerous anomalies. Although no changes were apparent at E14.5 (not shown), by E17.5 the Müllerian duct was markedly thickened, with the presence of an abnormal cell population with an elongated, abundant, slightly eosinophilic cytoplasm and a plump nucleus. These appeared to replace most of the antimesometrial mesenchyme cell population (Fig. 1A). RT-qPCR analyses confirmed the loss of Lats1 and Lats2 expression in the Müllerian ducts of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice, and immunohistochemical analyses showed that this loss of expression localized specifically to the ectopic cell population (Fig. S2). The changes in the Müllerian duct appeared to prevent it from elongating normally, resulting in coiling that could be seen macroscopically by E19.5 (Fig. 1B). Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ females were born with plump and shortened uterine horns (Fig. 1B), with the ectopic cell population now appearing more differentiated and occupying most of the axial and antimesometrial regions of the uterine horns (Fig. 1A). In juvenile and adult mice, there was a severe architectural disorganization of the uterine layers, with an absence of demarcation between the endometrium and myometrium. As a result of the coiling, multiple lumens were often visible in transverse sections, and glands were sparse or absent. Cystic structures were often present adjacent to the ovary in juvenile mice (Fig. 1B). These were lined with ciliated epithelial cells (Fig. 1B, ‘P21’ inset), and presumed to be severely dilated oviduct. These structures typically ruptured by adulthood, leaving tattered remains and no recognizable oviduct. The ovaries were connected to the uterine walls by bands of tissue consisting of disorganized fibroblasts and smooth muscle cells (Fig. 1B, ‘Adult’ inset).

To characterize the cellular changes observed in the Müllerian ducts of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice, we assessed proliferation by 5-bromo-2′-deoxyuridine (BrdU) incorporation at E17.5. Interestingly, the ectopic cell population was found to be non-proliferative, in contrast to the adjacent mesenchyme cells and Müllerian and coelomic epithelial cells (Fig. 2A). RT-qPCR and immunohistochemical analyses also showed a reduction in expression of the pluripotency markers Sox2, Nanog and Oct4 (Pou5f1) in the Müllerian ducts, which was due to their absence in the ectopic cell population (Fig. 2B). These cells also stained positive for vimentin and smooth muscle actin (α-SMA), and Picrosirius Red stain showed abundant collagen deposition in proximity to the ectopic cells, together strongly suggesting that they were myofibroblasts (Fig. 2C). This hypothesis was confirmed by transmission electron microscopy, which showed the presence of ultrastructural features, including thick intracellular actin belts and fibronexus junctions (Fig. 2D), that distinguish myofibroblasts from fibroblasts or smooth muscle cells (Eyden, 2008). Together, these findings indicate that loss of Hippo signaling in Müllerian mesenchyme cells causes them to lose their multipotency, while directing them to adopt the myofibroblast cell fate. In the uterus of adult Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice, the myofibroblasts were found interspersed between haphazard clusters of endometrial stromal cells and myometrial smooth muscle cells (Figs 1A and 2E).

Ovaries from Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice were examined, as the Amhr2^cre allele also drives Cre expression in follicular granulosa cells and (to some extent) the ovarian stroma and surface epithelium (Arango et al., 2008; Baarends et al., 1994; Teixeira et al., 1996). In adult mice, a general reduction in the abundance of follicles was noted (Fig. 3). Corpora lutea were usually present, indicating that at least some follicles were able to develop fully, ovulate and luteinize. Some areas of cellular changes consistent with loss of granulosa cell differentiation were observed, as had been previously observed in the Lats1^flox/flox; Lats2^flox/flox; CYP19-cre model (Tsoi et al., 2019). In these areas, granulosa cells appeared to be replaced by large, round or polygonal cells with abundant, clear and slightly eosinophilic cytoplasm (Fig. 3, inset in image of mutant adult). The most striking change, however, was that most ovaries were partially covered by a thick layer of haphazardly organized, poorly differentiated elongated cells, suggestive of a neoplasm originating from the surface epithelium (Fig. 3). However, immunohistochemical analyses of ovarian cancer markers showed that these cells did not correspond to any known ovarian tumor type (Table S2, Fig. S3). Furthermore, they did not increase in number nor did they become invasive in older animals. A subsequent developmental time-course analysis showed that these cells in fact appeared to originate from the ovarian stroma. Morphological changes in the stroma were evident as early as E17.5 (Fig. 3), and seemed to develop into areas of exuberant overgrowth that extended past the ovarian surface, the growth of which appeared to be self-limiting.

Developmental anomalies of the reproductive tract were also observed in male Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice. Rather than undergoing apoptosis during Müllerian duct regression (which is normally mostly completed around E15.5; Arango et al., 2008), Müllerian mesenchyme cells persisted in Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ males and underwent changes similar to those observed in females. Histological analyses of E17.5 embryos showed the presence of Müllerian duct-derived myofibroblasts adjacent to the Wolffian ducts (Fig. 4A). These subsequently interfered with the elongation of the Wolffian ducts in a manner analogous to the Müllerian duct elongation defects observed in Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ females. As a result, newborn mice had thick and abnormally coiled epididymides and deferent ducts, with the head and body of the epididymis ectopically positioned at the cranial pole of the testis alongside the tail (Fig. 4A). The testes failed to descend in the mutant mice, resulting in pathological changes associated with cryptorchidism (e.g. testicular hypoplasia, atrophy, fibrosis, mineralization) in adults (Fig. 4B). Dilation of seminiferous tubules and of the epididymis (Fig. 4B) was also frequently observed in adult mice, suggesting that the developmental defects resulted in mechanical obstruction. Spermatozoa were rarely observed in the epididymis, and most Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ males failed to sire litters in breeding trials.

As previous studies have suggested a role for the Hippo pathway in decidualization (Chen et al., 2017; Strakova et al., 2010), we investigated whether loss of Lats1 and Lats2 affects the ability of endometrial stromal cells to decidualize. As the uterine developmental defects in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model precluded its use in conventional, in vivo decidualization assays, endometrial stromal cells from Lats1^flox/flox; Lats2^flox/flox mice were placed in culture and infected with an adenovirus to drive Cre expression (Ad-Cre, to inactivate the floxed alleles) or an adenovirus that drives expression of eGFP (Ad-eGFP, control) for 24 h. Ad-Cre treatment resulted in ∼99% and ∼90% reduction of Lats1 and Lats2 mRNA levels, respectively (Fig. 5A). Following removal of the adenovirus, stromal cells were co-cultured with endometrial epithelial cells (separated by a semi-permeable membrane) and treated or not with medroxyprogesterone and cAMP to induce decidualization over a period of 48-72 h. Both the Ad-Cre- and Ad-eGFP-treated stromal cells responded to medroxyprogesterone+cAMP with a dramatic increase in decidual prolactin (Prl8a2) mRNA levels, indicating that decidualization occurred in both groups (Fig. 5A). Ad-Cre treatment did not change the mRNA levels of myofibroblast markers (Fig. 5B), indicating that loss of Lats1/2 did not cause differentiated endometrial stromal cells to adopt myofibroblast characteristics. These results therefore suggest that Hippo signaling is dispensable for decidualization and cell fate maintenance in the uterine stroma.

To elucidate the mechanism whereby loss of Hippo signaling causes Müllerian mesenchyme cells to commit to the myofibroblast cell fate, we first examined CTNNB1 expression in the uteri of newborn Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice. Although loss of Lats1/2 was predicted to result in the hypophosphorylation and stabilization of CTNNB1 (Varelas et al., 2010), both activated (i.e. unphosphorylated) and total CTNNB1 levels were in fact decreased in Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ myofibroblasts relative to control Müllerian mesenchyme and endometrial stromal cells (Fig. 6A). Conversely, YAP and TAZ levels were markedly increased in the mutant myofibroblasts. RT-qPCR analyses of uteri from Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice showed a strong and significant increase in the mRNA levels of the well-established YAP/TAZ target genes Ctgf (Ccn2; ∼6-fold) and Cyr61 (Ccn1; ∼12-fold) relative to controls (Fig. 6B). Together, these results indicate that the loss of Lats1/2 resulted in inactivation of the canonical Hippo pathway in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model, but did not induce CTNNB1 signaling.

To further determine how the loss of Hippo signaling affected the transcriptome in the Müllerian mesenchyme, ducts from E17.5 Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ and control mice were analyzed by microarray. These analyses identified 716 genes that were differentially expressed between genotypes by 2-fold or more. Of these, genes that were upregulated in the Müllerian ducts of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice (n=547) were subjected to gene ontology analysis using the Metascape gene annotation and analysis resource. This showed that the upregulated genes regulate biological processes including blood vessel morphogenesis, the inflammatory response, extracellular matrix organization, wound healing and actin filament-based processes (Fig. S3), all of which are associated with myofibroblast functions (Klingberg et al., 2013; Gerarduzzi and Di Battista, 2017). Furthermore, six out of the 20 most highly upregulated genes have been associated with myofibroblast activation, differentiation or function (Table 1, genes in bold) (Elbjeirami et al., 2010; Garrett et al., 2004; Gomez et al., 2016; Lenga et al., 2008; Lipson et al., 2012; Mifflin et al., 2002; Nightingale et al., 2004; Sobral et al., 2011). RT-qPCR analyses were performed for selected genes to confirm the microarray analyses (Fig. 6B), and the most highly upregulated gene (Ptgs2) was evaluated by immunohistochemistry, which localized its expression to the myofibroblast cell population of newborn Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice, and confirmed its absence in controls (Fig. 6C). Genes that were downregulated in the Müllerian ducts of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice were fewer in number (n=169) and were associated with biological processes including reproductive structure development and developmental growth (Fig. S4). The most downregulated gene identified in the microarray analysis was Amhr2 (Table 2), the expression of which is normally restricted within the Müllerian duct to the mesenchyme cells (Arango et al., 2008). These results support the notion that loss of Lats1/2 results in activation of the myofibroblast genetic program, while shutting off that of the Müllerian mesenchyme cells.

To determine how YAP and TAZ may activate the myofibroblast genetic program, we examined their physical association with the promoter of Ctgf, a gene with established roles in myofibroblast differentiation and activation, and which is overexpressed in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model (Fig. 6B, Table 1). Chromatin immunoprecipitation analyses showed enhanced binding of YAP to the Ctgf promoter in the uteri of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice relative to controls, within ∼100 bp from the transcriptional start site, in an area that encompasses three consensus TEAD binding motifs (Fig. 7, top). A much more marked enhancement (∼10-fold) was obtained for TAZ in the same region of the promoter (Fig. 7, bottom). These data indicate that YAP/TAZ induce Ctgf transcription through direct binding of its promoter, and that TAZ may play a preponderant role in the induction of Ctgf expression in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model.

Several reports have shown that the Hippo pathway can act in pluripotent cell types to direct cell fate specification. For instance, during early embryonic development, Hippo signaling must be differentially regulated in the inner and outer cells of the morula to permit proper adoption of either the inner cell mass or trophectoderm cell fate (Sasaki, 2017). In the optic vesicle, YAP and TAZ cause pluripotent progenitor cells to adopt the retinal pigment epithelial cell fate (Miesfeld et al., 2015). In the liver, YAP/TAZ expression and activity determine whether hepatoblasts/progenitor cells become either hepatocytes or cholangiocytes (Lee et al., 2016; Nguyen et al., 2015). Likewise, Hippo signaling also appears to direct progenitor cells to give rise either to nephron epithelial cells or myofibroblasts in the developing kidney (McNeill and Reginensi, 2017). In this article, we show that Hippo signaling plays a similar role in Müllerian mesenchyme cells. Indeed, loss of Hippo activity in Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice caused Müllerian mesenchyme cells to cease proliferating, and adopt morphological, functional and gene expression characteristics of myofibroblasts. Simultaneously, these cells lost the expression of markers of pluripotency, apparently along with the capacity to become endometrial stromal and myometrial smooth muscle cells. That the inactivation of Lats1 or Lats2 alone in Müllerian mesenchyme cells has no discernable effect shows that Lats1 and Lats2 are entirely functionally redundant in this context. Together, these results define a key function of Hippo signaling in the developing female reproductive tract, and provide novel insight into the biology of Müllerian mesenchyme cells.

As Hippo is a known regulator of WNT signaling, we originally expected that the inactivation of Hippo signaling in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model would result in the stabilization of CTNNB1 in Müllerian mesenchyme cells. However, we were unable to find any evidence of enhanced CTNNB1 expression. Consistent with this, the phenotypic changes observed in the reproductive tracts of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice appeared completely unrelated to those in any previously reported mouse models in which WNT signaling was altered in the Müllerian duct (Dunlap et al., 2011; Farah et al., 2017; Franco et al., 2011; Mericskay et al., 2004; Prunskaite-Hyyrylainen et al., 2016; St-Jean et al., 2019; Vainio et al., 1999). Inactivation of Hippo in Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice did, however, clearly result in the activation of the canonical Hippo effectors YAP and TAZ, as shown by their increased expression, along with that of their transcriptional target genes. WNT and Hippo signaling may therefore work independently of each other in Müllerian mesenchyme cells. Moreover, YAP/TAZ can even antagonize CTNNB1 signaling in certain contexts (Park et al., 2015), although we were unable to determine whether this occurs in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model.

CTGF has been previously described as a key driver of myofibroblast activation and differentiation (Garrett et al., 2004; Grotendorst et al., 2004; Klingberg et al., 2013; Lipson et al., 2012; Sakai et al., 2017; Yang et al., 2015). Furthermore, it is thought to be a direct transcriptional target of the Hippo pathway (Di Benedetto et al., 2016; Zhao et al., 2008), and is among the most highly overexpressed genes in the Müllerian ducts of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice. Together, these observations suggested that the induction of Ctgf could be a key mechanism whereby loss of Hippo signaling causes Müllerian mesenchyme cells to commit to the myofibroblast fate in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model, and prompted us to investigate the underlying mechanism. We were able to show an enhanced recruitment of YAP and TAZ to the proximal promoter of Ctgf in a region that includes a cluster of three consensus TEAD binding motifs, strongly suggesting that overexpressed YAP and TAZ induce Ctgf transcription through direct association with its promoter. This does not, however, exclude the possibility that additional mechanism(s) may be involved. For instance, Ctgf expression can also be regulated by TGFβ signaling via the SMAD2/3/4 pathway (Liu et al., 2013b; Parada et al., 2013) in contexts such as tissue repair and fibrosis (Lipson et al., 2012; Rachfal and Brigstock, 2003; Sakai et al., 2017; Tian et al., 2012). Several reports have shown that YAP/TAZ interact with SMAD complexes following TGFβ stimulation, and that this interaction serves to regulate both SMAD2/3/4 transcriptional activity and subcellular localization (Attisano and Wrana, 2013; Grannas et al., 2015; Mauviel et al., 2012; Szeto et al., 2016). It is therefore possible that YAP/TAZ modulation of the TGFβ pathway contributes to driving Ctgf overexpression in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model. Further experiments will be required to test the latter hypothesis, as well as to determine the relative role of CTGF in driving the phenotypic changes observed in Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice.

Male reproductive tract defects were not expected in the Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ model, given the specificity of the Amhr2^cre allele for the Müllerian duct. Although the exact mechanism underlying the development of these defects remains unclear, it can be surmised that loss of Lats1 and Lats2 occurred in the Müllerian ducts of male mice prior to the AMH signal that would have initiated their degradation. Our microarray data showed that Amhr2 expression is dramatically downregulated in the Müllerian ducts in Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ female mice, suggesting that loss of Lats1 and Lats2 could result in rapid loss of AMHR2 and consequential insensitivity to AMH. Once the Müllerian mesenchyme cells were committed to the myofibroblast fate, they appeared to interfere mechanically with the coiling and development of the epididymides and deferent ducts, and caused the testes to fail to descend, leading to pathological secondary changes in the testes as well. Some of the changes observed in the testes of Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice may also have been due to loss of Lats1/2 expression in Leydig and/or Sertoli cells, as the Amhr2^cre allele has previously been shown to drive some level of Cre expression in both cell types (Boyer et al., 2008; Jeyasuria et al., 2004). However, it was impossible to unambiguously attribute phenotypic changes specifically to conditional gene inactivation in testicular cells owing to the confounding factor of cryptorchidism.

A small number of studies have suggested that YAP/TAZ may play a role in the adult uterus in processes such as decidualization and implantation (Chen et al., 2017; Strakova et al., 2010). This prompted us to determine whether the loss of Hippo signaling in endometrial stromal cells would affect their ability to decidualize. We found that loss of Lats1/2 had no effect on decidualization in cultured endometrial stromal cells. Although this suggests that Hippo signaling is dispensable in this context, it does not necessarily indicate that YAP and TAZ do not play a role. At most, it suggests that YAP and TAZ overexpression does not inhibit decidualization in the short term. Further experiments targeting Hippo pathway effectors including Yap and Taz in vivo will be required to make a definitive conclusion. Our findings indicate that loss of Lats1/2 in fully differentiated endometrial stromal cells does not cause any apparent loss of function or cause them to become myofibroblasts, unlike what occurs in their Müllerian mesenchyme precursors. For this reason, inactivation of Hippo effectors in vivo to study their roles in decidualization will likely require a strategy that avoids targeting the uterine stroma at developmental stages prior to the commitment of mesenchyme cells to the stromal cell fate.

In summary, we report here a previously unsuspected role of the Hippo pathway in the development of the female reproductive tract. Lats1/2 activity was found to be necessary to preserve the multipotency of Müllerian mesenchyme cells and prevent them from committing to the myofibroblast cell fate. Further studies will be required to better define how Hippo signaling is regulated and acts throughout reproductive tract development, as well as its potential roles in reproductive physiology.

The Lats1^tm1.1Jfm (referred to herein as Lats1^flox) and Lats2^tm1.1Jfm (referred to herein as Lats2^flox) strains were obtained from Randy L. Johnson (M.D. Anderson Cancer Center, Houston, TX, USA). The Amhr2^tm3(cre)Bhr (referred to herein as Amhr2^cre) strain was obtained from Richard R. Behringer (M.D. Anderson Cancer Center). Lats1^flox, Lats2^flox and Amhr2^cre strains were crossed in order to obtain the Lats1, Lats2 and Lats1/2 conditional knockout genotypes described in the text. All strains were maintained on a C57BL/6J genetic background. Genotyping analyses were performed by PCR on DNA obtained from tail biopsies as previously described (Heallen et al., 2011; Tsoi et al., 2019).

To assess fertility in Lats1^flox/flox; Amhr2^cre/+, Lats2^flox/flox; Amhr2^cre/+ and Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ mice, 8-week-old females were housed with 8-week-old wild-type (C57BL/6J) males and monitored daily for litters. Litter sizes were recorded at birth. The males were removed after 6 months and the experiment was terminated 22 days later to allow for the birth of a final litter. Seven 8-week-old Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ males were similarly housed with females for 2 months. Of these, five failed to sire pups, one sired a litter of four pups, and another sired three litters of one, four and one pup, respectively.

Lats1^flox/flox; Lats2^flox/flox (control) and Lats1^flox/flox; Lats2^flox/flox; Amhr2^cre/+ embryos and mice were obtained at E14.5, E17.5, E19.5, postnatal day (P)1, P21 and 2-4 months (adult) of age. For E14.5, E17.5 and E19.5 embryos, Lats1^flox/flox; Lats2^flox/flox adult females were crossed with either Lats1^flox/+; Lats2^flox/flox; Amhr2^cre/+ or Lats1^flox/flox; Lats2^flox/+; Amhr2^cre/+ males (as conditional knockout males were infertile), and observed every morning for the presence of a vaginal plug (defined as E0.5). Gravid uteri were dissected to collect the embryos, tail biopsies taken from each embryo for genotype analyses, and the whole embryos fixed for 24 h in 4% paraformaldehyde (PFA) prior to paraffin embedding. A subset of E17.5 and E19.5 embryos was dissected to isolate the Müllerian ducts prior to fixing (PFA, 4 h) and embedding. Reproductive tracts from P1, P21 and adult mice uteri were collected, fixed in 10% neutral buffered formalin and paraffin embedded. Sections of whole embryos, Müllerian ducts and reproductive tracts were prepared at a thickness of 3 µm and stained either with Hematoxylin and Eosin-Phloxine (HE), Hematoxylin-Eosin-Phloxine-Saffron (HEPS), or Picrosirius Red. For transmission electron microscopy, uteri from P1 mice were collected and fixed in 2.5% glutaraldehyde in sodium cacodylate buffer. Processing of the tissue and transmission electron microscopy analyses were performed by the Facility for Electron Microscopy Research (FEMR) at McGill University.

Müllerian ducts (E17.5) and uteri (P1) were flash frozen in liquid nitrogen and stored at −70°C prior to RNA extraction. Total RNA was extracted using the Qiagen RNeasy mini kit according to the manufacturer's instructions. RNA quantity and quality was assessed using a NanoDrop-1000 spectrophotometer (Thermo Fisher Scientific). RNA was reverse-transcribed using the SuperScript Vilo cDNA Synthesis Kit (Invitrogen) following the manufacturer's instructions. Real-time PCR reactions were run using a CFX96 Real-Time System/C1000 Touch thermal cycler (Bio-Rad) and Universal SsoAdvanced SYBR Green Supermix (Bio-Rad). PCR reactions consisted of 2.3 µl of H₂O, 6 pmol of each forward and reverse gene-specific primer, and 7.5 µl of SYBR Green Supermix. Standard curves were generated using serial dilutions of cDNA, and amplification efficiency (E) values were obtained using the slope of the log-linear phase derived from the formula E=e(1/slope). Only primer pairs with efficiency values between 1.8 and 2.2 were used. The thermal cycling program to amplify transcripts typically consisted of 3 min at 95°C, followed by 40 cycles of 15 s at 95°C, 30 s at 60°C and 30 s at 72°C. To quantify relative gene expression, the cycle threshold (Ct) values for each transcript were compared with that of Rpl19, according to the ratio R=(E^{Ct Rpl19}/E^{Ct target}). MIQE guidelines were followed throughout. Primer sequences are listed in Table S3.

Total RNA from E17.5 Müllerian ducts (n=3 per group) was submitted to the McGill University and Génome Québec Innovation Centre (Montreal, Quebec, Canada) for microarray analysis using the Affymetrix Clariom S Mouse assay. Results were analyzed using Transcriptome Analysis Console software (TAC 4.0) (Thermo Fisher Scientific). Thresholds used for P-value and fold change were 0.05 and ±2, respectively. Further analyses were performed using the Metascape gene annotation and analysis resource (http://metascape.org) (Zhou et al., 2019).

Female Lats1^flox/flox; Lats2^flox/flox mice ranging from 8 to 12 weeks of age were injected daily with estradiol (Sigma-Aldrich) (100 ng, subcutaneously) for 3 days prior to endometrial stromal and epithelial cell isolation and culture. All steps of cell purification, culture and in vitro decidualization were performed essentially as described by De Clercq et al. (2017). Freshly isolated stromal cells were incubated overnight to permit attachment to the cell culture plates, and subsequently infected with Ad-Cre or Ad-eGFP (Vector development laboratory, Baylor College of Medicine, Houston, TX, USA) at a multiplicity of infection of approximately 50 for 24 h. Following washes, stromal cells we co-cultured with epithelial cells grown in ThinCert cell culture inserts (Greiner Bio-One International) such that the stromal and epithelial cells remained separate but were able to communicate via a semi-permeable membrane. Following medroxyprogesterone/cAMP (Sigma-Aldrich) treatment for 48-72 h, Lats1, Lats2 and Prl8a2 mRNA levels were assessed by RT-qPCR as described above. Similar experiments were also performed without co-culturing with epithelial cells and Acta2, Cnn1, Cald1 mRNA levels were assessed by RT-qPCR as described above.

Immunohistochemical (IHC) analyses were performed either on PFA (E17.5) or formalin (P1)-fixed, paraffin-embedded Müllerian ducts and uteri were sectioned at a thickness of 3 µm. Following deparaffinization, rehydration, sodium citrate heat-mediated antigen retrieval and peroxidase block, primary antibodies were diluted in Signal Stain Antibody Diluent (8112; Cell Signaling Technology) and incubated overnight at 4°C. Antibodies used include LATS1 at 1:100-1:500 (SC-28223; Santa Cruz Biotechnology), LATS2 at 1:600 (orb6306; Biorbyt), YAP at 1:200 (4912S; Cell Signaling Technology), TAZ at 1:500 (4883S; Cell Signaling Technology), pYAP at 1:100 (13008S; Cell Signaling Technology), vimentin at 1:100 (5741S; Cell Signaling Technology), PTGS2 at 1:300 (12282S; Cell Signaling Technology), total CTNNB1 at 1:100 (D10A8; Cell Signaling Technology), active CTNNB1 at 1:600 (D13A1; Cell Signaling Technology), pre-diluted CK19 (ab961; Abcam), PAX8 at 1:1000 (10336-1-AP; Proteintech), inhibin at 1:100 (M3609; Dako), calretinin at 1:50 (MA5-14540; Thermo Fisher Scientific), WT1 at 1:100 (M3561; Dako), CK5 at 1:200 (MA5-17057; Thermo Fisher Scientific), PAX2 at 1:100 (ab79389; Abcam) and Ki67 at 1:250 (ab16667, Abcam). Detection was performed with the Vectastain Elite ABC HRP Kit (Vector Laboratories, PK-6101) and the 3,3′-diaminobenzidine peroxidase substrate kit (Vector Laboratories, SK-4100). Slides were counterstained with Hematoxylin prior to mounting. IHC for smooth muscle actin was carried out using a LabVision autostainer (Thermo Fisher Scientific), with the primary antibody (MU128-UC; Biogenex) diluted at 1:100 and incubated for 1 h at room temperature. Detection was performed using Kit Envision System AEC Mouse (K4005; Agilent Technologies). To assay proliferation, BrdU (B5002; Sigma-Aldrich) was dissolved as per manufacturer instructions (10 mg/ml in 0.9% saline) and injected in gravid mice at E17.5 (100 mg/kg, intraperitoneally). Mice were sacrificed 3 h post-injection. Embryos were collected, fixed for 24 h in PFA, paraffin embedded and 3-µm-thick sections prepared. Immunohistochemistry was then performed using the M.O.M. Kit (PK-2200; Vector Laboratories) with the BrdU primary antibody (M0744; Agilent Technologies) incubated at the manufacturer's suggested dilution for 1 h at room temperature. Detection was then performed using the Vectastain Elite ABC kit as described above.

Chromatin immunoprecipitation (ChIP) experiments were performed using frozen uteri (P1) as previously described (Svotelis et al., 2009), with the exception of using magnetic dynabeads (Thermo Fisher Scientific, 10002D) for immunoprecipitation. Briefly, samples (n=4 biological replicates per group, 3 pooled uteri per replicate) were crosslinked with 1.1% formaldehyde for 10 min at 37°C with agitation. The crosslinking reaction was stopped by addition of glycine (0.125 M for 10 min at 37°C with agitation) followed by two washes with cold PBS. Crosslinked uteri were ground in SDS lysis buffer using a mortar and pestle prior to sonication. Immunoprecipitation was carried out using 1 µg of TAZ antibody (4883S; Cell Signaling Technology), 1 µg of YAP antibody (NB110-58358; Novus Biologicals), or no antibody as background control. qPCR assays were performed on immunoprecipitated DNA and quantified using a standard curve derived from total (i.e. input) DNA. Background signal was removed by subtracting the signal obtained in ‘no antibody’ controls from that obtained in samples immunoprecipitated with the YAP or TAZ antibodies. Primers used for the amplification of specific regions of the Ctgf promoter are listed in Table S4. Background results were brought to 0.1% of input for TAZ enrichment and 0.1% of input for YAP enrichment.

Analyses were performed using GraphPad Prism 6 software (GraphPad Software). All data were subjected to the F test to determine the equality of variances. Data were transformed to logarithms if they were not normally distributed. Statistical significance was determined using unpaired Student's t-test or ANOVA, with P<0.05 considered significant. Data are presented as means±s.e.m.

The authors thank Ms Nancy Veilleux and Ms Jacinthe Cardin for assistance with sectioning of processed tissues and Ms Jeannie Mui for assistance during the preparation and evaluation of tissue for transmission electron microscopy.