Biased precursor ingression underlies the center-to-pole pattern of male sex determination in mouse

The gonad is a model for fate commitment at the level of individual cells and at the level of the whole organ. Although single cell transcriptome analysis has advanced our understanding of transcriptional networks within individual cells (Stévant and Nef, 2018), spatial features of organ development are less well understood. The cells of the gonad are initially bipotential: able to commit to male or female fate. The choice is determined by the expression and opposition of pro-ovary and pro-testis gene regulatory networks (GRNs) that are initiated in the supporting cell lineage (Fig. 1A) (reviewed by Nef et al., 2019). After supporting cells differentiate into granulosa or Sertoli cells, other cell types in the gonad differentiate to the corresponding female or male fate, leading to unified development of the entire organ into an ovary or a testis. However, in some circumstances, XY gonads develop into ovotestes, containing both ovarian and testicular tissue. For example, BALB/cWt strain mice produce spontaneous hermaphrodites that harbor ovotestes (Whitten et al., 1979), as do C57BL6/J mice carrying the Y^POS chromosome, which carries a ‘weak’ allele of Sry leading to a reduction or delay in Sry expression (Bullejos and Koopman, 2005; Eicher et al., 1982; Taketo et al., 1991; Wilhelm et al., 2009). In addition, XY gonads in mutants of sex-determining genes can develop into ovotestes: e.g. ovotestes form in Fgfr2 heterozygous mutants (Bagheri-Fam et al., 2008; Kim et al., 2007), Znfr3 homozygous null (Harris et al., 2018) and Map2k3;Map2k6 compound mutants (Warr et al., 2016). Interestingly, in nearly all cases of ovotestis development, the central domain of the gonad retains testis fate while the poles of the gonad commit to ovary fate. This pattern indicates two things: (1) there is a limited amount of fate coordination in the developing gonad; and (2) a spatial bias exists across the bipotential gonad wherein the central domain is more disposed to Sertoli cell fate.

Before gonadal sex determination, transcriptomes of XX and XY cells are indistinguishable (Munger et al., 2013; Nef et al., 2005). At the level of cell fate commitment, expression of Sry in XY supporting cells initiates the transcriptional cascade that drives Sertoli cell differentiation around embryonic day 11.0 (E11.0). SRY directly activates Sox9 (Sry box 9) (Sekido et al., 2004) and Fgf9 (fibroblast growth factor 9) (Li et al., 2020), which drive expression of Amh (anti-Müllerian hormone) and other Sertoli-specific factors (reviewed by Nef et al., 2019). Mutations in Sry, Sox9 or Fgf9 lead to loss of Sertoli cell development, commitment to granulosa cell fate and development of an ovary (Biason-Lauber and Chaboissier, 2015). The competition between the testis and ovary pathways is heavily influenced by the mutual antagonism between Fgf9 and Wnt4 (Kim et al., 2006), which act at the interface between single cell differentiation and the promotion of organ fate across the gonad field.

In the mouse, the gonad first forms on embryonic day 9 (E9) as a long and thin layer of cells that accumulates between the mesonephric ducts and the coelomic cavity (Brambell, 1927; reviewed by Windley and Wilhelm, 2015). The first steps of gonad development are reported to occur in an anterior-to-posterior direction (Brambell, 1927; Sasaki et al., 2021), including upregulation of consecutive early gonadal genes, including Gata4 and Sf1 (steroidogenic factor 1, Nr5a1) (Bunce et al., 2021; Hu et al., 2013). The anterior-to-posterior pattern is maintained during ovary differentiation (Suzuki et al., 2015). In contrast, testis differentiation, including upregulation of Sry and Sox9 is reported to occur first in the center of the gonad (Bullejos and Koopman, 2001, 2005), followed by expansion to the gonadal poles believed to be mediated by Fgf9 (Hiramatsu et al., 2010). These patterns of gonad specification make sense of the central domain of testis tissue seen in ovotestes, as well as the anterior domain of ovarian tissue, but do not readily explain why ovarian tissue would be found at the posterior pole of the gonad. Although several genes expressed in the early genital ridge have been reported to regulate Sry (Kashimada and Koopman, 2010), an inducer of Sry that is expressed in the center of the gonad before the poles has not been found.

Analysis of the spatial patterns of SRY and SOX9 in the early gonad showed no evidence of center-first SRY and SOX9 expression or of a cellular relay pattern. Instead, we discovered a center bias in supporting cell precursor ingression. We hypothesize that initiation of the Sertoli pathway occurs in an anterior-to-posterior direction, but stabilization of the Sertoli pathway is density dependent and likely mediated through Fgf9 and other paracrine factors. We suggest that this morphogenetic feature underlies the center-biased male pattern.

Sry transcription begins in the XY gonad around E10.5 (Hacker et al., 1995; Koopman et al., 1990). SRY directly activates SOX9 (Sekido and Lovell-Badge, 2008) and SOX9 directly activates AMH (Fig. 1A) (Arango et al., 1999). We expected that SRY, SOX9 and AMH proteins would be enriched in the center of the gonad, following the in situ hybridization pattern seen for the Sry transcript at the level of the whole gonad. In those studies, Sry expression appeared to be center initiated, with complete expansion to the gonadal poles around E11.5 (Bullejos and Koopman, 2001). Using immunofluorescence, we compared the anteroposterior (AP) distribution of SRY with SOX9, and SOX9 with AMH (see Materials and Methods section ‘Pixel-based analysis’). At E11.0, SRY and SOX9 did exhibit central biases, with SRY having a broader distribution and SOX9 having a stronger peak located halfway between the anterior and posterior poles of the gonad (Fig. 1B). The SOX9 peak flattened by E11.25 and became nearly evenly distributed across the gonad by E11.75, when AMH expression began with reduced expression in the poles and a peak within the anterior half of the gonad (Fig. 1C, Fig. S1A). The AMH peak at E11.75 was shifted anteriorly compared to the SOX9 peak at E11.0, although the overall pattern remained center biased when compared with the anterior-biased pattern of the ovarian pathway gene Foxl2 (forkhead box L2) in XX gonads at E11.75 (Fig. S1B). These data confirmed that immunofluorescent signal quantification captures the previously reported patterns of expression during gonad differentiation. To determine the earlier spatial dynamics of central accumulation and expansion to the poles at the cellular level, we directed our attention to the spatial organization of the first cells to express SRY and SOX9.

SRY and SOX9 are first expressed between E10.5 and E11.0 when the gonad remains thin and very elongated (Hacker et al., 1995; Munger et al., 2013). We quantified the AP position of individual SRY⁺ and SOX9⁺ cells with immunofluorescence in ventral views of XY genital ridges that contained fewer than 250 positive cells (see Materials and Methods section ‘Cell-based analysis’), which were collected at E10.75 or E11.0. The AP centerline of the gonadal region was determined by the presence of germ cells (PECAM) or gonad-specific labels (GATA4 or SF1) in the coelomic epithelium (CE) (Fig. 2A,B). Predictably, we found some variation in the number of SRY⁺ and SOX9⁺ cells within individual gonads, even though they were collected at the same somite stage. We used the total number of SRY⁺ or SOX9⁺ cells per sample to represent a pseudo-timeline for graphical assessment of development of the pattern (Fig. 2C,D). The earliest population of SRY⁺ cells exhibited considerable distance between cells and was widely spread across the gonad by the time there were 20-30 SRY⁺ cells (Fig. 2C). The initial population of SOX9⁺ cells had a similar distribution and appeared across more than half of the gonad length by the time there were 10 SOX9⁺ cells (Fig. 2D). Both SRY⁺ and SOX9⁺ cell populations stretched across the full AP length of the gonad before the accumulation of 200 positive cells for each marker. Neither SRY nor SOX9 revealed a single initial cluster of positive cells, although both appeared in the anterior region of the gonad earlier than the posterior. When the position of SOX9⁺ cells was compared with the SRY⁺ region, SOX9⁺ cells displayed a distribution and range similar to either factor alone (Fig. S1C). The distributions demonstrate that there is no single AP position where SRY or SOX9 is first expressed, and expression reaches the full length of the AP axis quickly within individual genital ridges. These results indicate that, in contrast to the previously proposed models, the center bias in Sertoli cell fate in XY gonads is not explained by center-limited initial activation of Sry. Between-sample variability in the peak of pro-testis factor distribution may indicate stochasticity in the initial distribution, additional influences on Sertoli cell fate commitment or regional differences in proliferation.

Sex determination is associated with changes in cell cycle status (Gustin et al., 2016; Nef et al., 2005). Although supporting cell precursors in the CE are actively proliferating at E11.5 (Schmahl et al., 2000), cells that have ingressed into the XX and XY gonad express the cell cycle inhibitor p27^kip1 (p27) and exhibit low rates of proliferation (Gustin et al., 2016). Subsequently, cells in the XX gonad remain enriched in p27 and p21^cip1 (p21), and remain in cell cycle arrest until after birth (Mork et al., 2012b) whereas somatic cells in the XY gonad downregulate cell cycle inhibitors and re-enter active mitotic cycles (Fig. S2A,B, summarized in Fig. 3A). At E12.5, SOX9⁺ cells in the XY gonad exhibit considerably higher proliferation rates compared with E11.5 (Gustin et al., 2016).

To determine whether cell cycle re-entry in the XY gonad occurs in a center-to-pole pattern, which could indicate a molecular connection between cell cycle and Sertoli cell fate disposition, we performed p27 immunolabeling in E11.75 XY gonads. At this stage, p27 was extensively expressed in the poles, but almost entirely absent from the center of the gonad (Fig. 3B, Fig. S2C), suggesting that cell cycle re-entry in the XY gonad occurs first in the center of the gonad. Labeling with 5-ethynyl-2′-deoxyuridine (EdU, a nucleoside analog that is incorporated into DNA during S-phase) confirmed that p27⁺ cells were in cell cycle arrest (EdU⁻) (Fig. S3A). At E11.75, EdU overlapped with SOX9 in several cells scattered across the XY gonad (Fig. S3B).

To determine whether Sertoli cell fate specification required re-entry into the cell cycle, we labeled XY cells with antibodies against p27, SRY, SOX9 and AMH. Segmentation and labeling of XY gonads at E11.0 revealed cells that were SRY-only, p27-only and SRY⁺/p27⁺ (see Materials and Methods section ‘Segmentation and labeling’). In the central region of the gonad, single-positive cells were found in similar numbers (Fig. 3C, E11.0), demonstrating that initiation of SRY expression does not require a cell to be actively proliferating or in cell cycle arrest (see Fig. S4A for labeling of the full gonad). Co-labeling with SOX9 and p27 at E11.25 revealed that pre-Sertoli cells can activate SOX9 in a p27⁺ or p27⁻ state (Fig. 3C, E11.25). The higher quantity of p27-only cells compared with SOX9-only may arise from initial stochastic activation of SRY and p27 followed by SOX9 upregulation several hours later in exclusively SRY⁺ cells (see Fig. S4B for labeling of the full gonad). At E11.75, co-labeling with p27, SOX9 and AMH revealed that SOX9⁺ Sertoli cells do not need to re-enter the cell cycle (lose p27) to express AMH, as SOX9⁺/AMH⁺/p27⁺ cells were found (Fig. 3D). In addition, Sertoli cells may lose p27 (SOX9-only) before AMH upregulation. Contrary to our expectations, these data suggest that initial steps of Sertoli differentiation are not constrained by cell cycle dynamics (Fig. S4C). These results imply that cell cycle regulation is unlikely to underlie the center bias in Sertoli cell fate disposition.

Two phases of Sertoli cell fate commitment, initiation and stabilization, have been distinguished through analysis of XY sex-reversal mutants. The initiation phase involves the activation of Sertoli-specific transcription factors, Sry and Sox9, within the critical window for sex determination (Hiramatsu et al., 2009). The pro-testis GRN is subsequently stabilized against sex reversal, possibly through FGF9 signaling pathways. Although supporting cells in Fgf9 mutant XY gonads initially express Sry, Sox9 and Amh, in the absence of Fgf9, they revert to granulosa fate (Kim et al., 2006). FGF9 antagonizes the granulosa-associated Wnt pathway (Jameson et al., 2012b; Kim et al., 2006) and promotes proliferation in the XY CE, which may be important to establish the threshold number of Sertoli cells required to commit to the testis pathway (Schmahl and Capel, 2003).

Previous studies have demonstrated that E12.5 XX gonads cultured with FGF9-soaked agarose beads on their surface express SOX9 in naive CE cells (Kim et al., 2006). To investigate the dynamics of transdifferentiation in pre-granulosa cells already expressing FOXL2 and p27, we devised a system to place the bead within the gonad (Fig. S5A,B). FGF9-coated beads induced SOX9 expression in cells within the XX gonad after 24 h in culture. Furthermore, XX supporting cells remain sensitive to FGF9 until E13.5 (Fig. S5C,D). To determine whether FGF9 permanently impacts the Sertoli GRN and cell cycle pathway, gonads were cultured for 3-96 h after FGF9-soaked bead application, followed by immunostaining for SOX9, FOXL2, p27 and MKi67 (Ki67, a marker of active cell cycle). Within 3 h, XX cells rapidly activated SOX9 (Fig. 4A). FOXL2 and p27 were absent from SOX9⁺ cells around the bead within 6 h with no consistent order, suggesting that either FOXL2 or p27 could be downregulated first (Fig. 4A). Cells expressing SOX9 were positive for the proliferation marker MKi67 within 24 h (Fig. 4B, left column). As a measure of full Sertoli specification, we investigated whether AMH was activated by FGF9-bead induction. After 48 h, AMH was not detected, despite broad areas of SOX9 activation and FOXL2 loss (Fig. 4C, center column). After 96 h in culture, cells around the bead lacked SOX9 and resumed expression of p27, indicating instability of the Sertoli pathway and a return to granulosa cell fate (Fig. 4D, right column). As FGF9 application initially induced p27 loss, FGF9 may contribute to the center-to-pole loss of p27 during testis development.

Fgf9 previously was purported to underlie center-to-pole expansion of Sertoli cell fate initiation (Harikae et al., 2013). When XY gonads are trisected in culture, the poles are prone to sex reversal (Hiramatsu et al., 2003), but culture with exogenous FGF9 rescued testis fate in the gonad poles (Hiramatsu et al., 2010). To determine the role of Fgf9 in the center-biased testis patterns, we analyzed Fgf9 mutants. XY Fgf9 mutant gonads express SOX9 across the full AP axis of the gonad before the initiation of transdifferentiation at E12.0 (Colvin et al., 2001). To determine whether FGF9 is required for XY-specific supporting cell cycle re-entry, we immunolabeled Fgf9 null (Fgf9^lacZ/lacZ) XY gonads with antibodies against SOX9, FOXL2 and p27 at E12.5, and quantified the marker-positive area in sagittal planes midway through the gonad. Cells expressing only SOX9 could be found in each region of the gonad (designated by anterior, central and posterior third) (Fig. 5A,B, orange arrowhead). Absence of p27 in SOX9⁺ cells indicates that p27 loss occurred despite the absence of Fgf9, but cell cycle arrest resumed with sex reversal. The greater proportion of p27-only (Fig. 5A,B, purple arrowhead) area compared with FOXL2-only (Fig. 5A,B yellow arrowhead) and FOXL2⁺/p27⁺ area suggests that the more common state path for supporting cell sex reversal is p27 gain before FOXL2 gain. However, because areas containing each marker combination were found, no state path could be eliminated from consideration (Fig. 5B).

We reasoned that if Fgf9 were not required for the center bias in Sertoli cell fate, Fgf9 mutants would retain a center bias and sex-reversing gonads would exhibit a transient ovotestis, as seen in other cases of XY sex reversal. In E12.5 XY mutant gonads, SOX9 exhibited an anterior-to-posterior increase in signal density up to the posterior third (AP position 0.0 to 0.7), with a sharp decrease across the posterior third (AP position 0.7 to 1.0). (Fig. 5C). This resulted in a SOX9 pattern that was center biased with a posterior skew. FOXL2 was enriched in the anterior pole with no overt gradient across the AP axis, whereas p27 exhibited a reduction in signal area from the anterior to posterior pole. These patterns are consistent with FOXL2 and p27 upregulation occurring in an anterior-to-posterior pattern while SOX9 undergoes pole-to-center downregulation with faster loss in the anterior. As p27 is lost in a center-to-pole pattern in wild-type XY gonads (Fig. 3B, Fig. S2C), the lack of a center bias in p27 dynamics in the Fgf9 mutant supports the idea that Fgf9 mediates the typical pattern of cell cycle re-entry.

Paracrine relay between pre-Sertoli cells could explain center-to-pole testis patterning, as proposed by Kanai and colleagues (Hiramatsu et al., 2010; reviewed by Harikae et al., 2013). Evidence for this first came from XX↔XY chimeras, where XX pre-supporting cells, which lack Sry, could become Sertoli cells when incorporated into testis cords with XY cells (Palmer and Burgoyne, 1991). More recently, cultured SOX9⁺ gonadal cells could induce SOX9 expression in neighboring cells in vitro, possibly through prostaglandin signaling (Adams and McLaren, 2002; Wilhelm et al., 2005, 2007).

To establish the timeframe for testing paracrine induction of Sertoli cell fate, we revisited XX↔XY chimera experiments. Chimeric embryos were generated by randomly combining wild-type and GFP^+/− XX and XY morulas (Fig. 6A). Embryos were collected at E11.5 and E13.5, and the chromosomal sex of the GFP⁺ and GFP⁻ component was determined (Mork et al., 2012a; Fig. 6A). Once this was established for each chimera, gonads were immunolabeled for GFP and SOX9. At E11.5, only XY cells (GFP⁻ in the image presented) expressed SOX9 (Fig. 6B, arrowheads). At E13.5, similar to previous results (Palmer and Burgoyne, 1991), XX cells (GFP⁻ in the image presented) were incorporated into testis cords with XY cells (GFP⁺) and expressed SOX9 (Fig. 6B, arrows). These results suggest that sometime after E11.5 and before E13.5, XY cells specified with Sertoli cell fate can induce SOX9 in neighboring XX cells.

We then tested the hypothesis that Sertoli cells in the central region of the gonad can induce Sertoli cell fate in adjacent supporting cells. XX and XY gonad/mesonephric complexes were bisected at E11.75, E12.75 and E13.75, and halves of a wild-type complex were juxtaposed with halves from a complex of the opposite sex carrying a genetic label in most somatic cells [SF1:eGFP line, Tg(Nr5a1-GFP)] (Fig. S6A). Reconstructed gonad/mesonephric complexes cultured from E12.5 fuse together within 6 h of assembly in agar blocks (Fig. S6B). After 24 h, immunolabeling revealed that the tissue was cellularly continuous across the border, although a break remained in the basement membrane beneath the coelomic surface (Fig. S6C, arrow). To our surprise, SOX9 expression was only rarely detected in XX (GFP⁺) cells after reconstruction with XY (GFP⁻) tissue at any tested stage (Fig. 6C, arrowheads). Similarly, XY (GFP⁺) cells did not exhibit FOXL2 expression after 24 h in a culture reconstructed with XX (GFP⁻) tissue (Fig. 6D). However, XY cells were found invading XX tissue (Fig. 6D, gray arrowhead). To test whether cell migration occurs in the context of two XY or two XX halves, we reconstructed GFP labeled and unlabeled gonads from the same sex. After 24 h in culture, a distinct GFP border was maintained in same sex XY and XX samples from all stages (Fig. S7A,B). In contrast, at all stages examined, XY (GFP+) cells, including Sertoli cells (SOX9⁺), exhibited a disrupted border when paired with XX (GFP⁻) tissue (Fig. S7C). The difference in migratory behavior may be due to an inhibitory influence within XY gonadal tissue that is absent from XX gonadal tissue. However, the failure of XY cells to migrate into adjacent XY tissue suggests that supporting cell migration along the AP axis is unlikely to occur during typical testis development.

Gonadal supporting cells are primarily derived from the CE, and early evidence suggested that FGF9 promotes proliferation in CE cells downstream of Sry (Schmahl et al., 2004) to establish a threshold number or density of pre-Sertoli cells thought to be required to promote testis development (Burgoyne et al., 1988). Furthermore, inhibition of proliferation around E11.0, soon after initial Sry expression, can lead to XY sex reversal (Schmahl and Capel, 2003). Recent analyses of early gonad morphology demonstrated that the central domain of the gonad increases in size more rapidly than the poles in both sexes (Bunce et al., 2021). We hypothesized that the center bias in stability of Sertoli cell fate commitment follows from a center bias in the density of supporting cell precursors secreting FGF9 or other signals.

To determine whether the CE exhibits a central bias in proliferation, we analyzed EdU labeling in genital ridges at E11.0, E11.5 (when Sertoli cells are still being produced from the CE) and E11.75 (when only interstitial cells are produced) (Karl and Capel, 1998; Schmahl et al., 2000). Whereas EdU was incorporated across the entire AP axis of the CE at E11.5 (Fig. 7A), quantification of the AP position of EdU⁺ CE cells revealed an anterior bias at E11.0 and E11.5 (Fig. 7B). At E11.75, EdU labeling was center biased (Fig. S8A). However, as supporting cells are not produced from the CE at E11.75, this proliferation pattern is unlikely to be responsible for accumulation of XY supporting cells in the center domain.

A center bias in pre-supporting cell population size could also arise from differential rates of supporting cell precursor ingression along the AP axis. To test this possibility, we took advantage of the fact that supporting cell precursors proliferate in the CE, but immediately enter cell cycle arrest once they ingress (see Fig. 3, Fig. S2, Mork et al., 2012b). This makes it possible to use EdU labeling for short-term lineage tracing to analyze the pattern of CE-derived supporting cell precursor ingression (Fig. 7C). Runx1, a transient marker of pre-supporting cells in the bipotential gonad (Fig. S8B), was found to be faintly detectable in cells of the CE and strongly upregulated upon ingression (Fig. 7D, arrowheads). To trace proliferating CE cells, EdU was administered 1-2 or 4-5 h before collection at E11.5 or E11.75. Although the time delay between DNA synthesis and ingression among CE cells is unknown, we expected fewer cells would incorporate EdU and migrate within 1-2 h compared with the longer delay. In addition, the CE stops producing pre-supporting cells at E11.5. Together, these factors lead to enrichment of supporting cell ingression in the 4-5 h E11.5 group compared with the others. The AP distribution of EdU-labeled supporting cells (EdU⁺/RUNX1⁺ or EdU⁺/SOX9⁺ area) is shown in Fig. 7E (E11.5) and Fig. S8C (E11.75). In XY samples collected at E11.5, the distribution of supporting cell EdU 1-2 h after administration was anterior biased. After 4-5 h, XY E11.5 supporting cell EdU was reduced in both poles, producing a center-biased pattern. This difference between 1-2 and 4-5 h collection delays did not appear in samples collected at E11.75, wherein both groups showed a minor anterior bias in supporting cell EdU (Fig. S8C). To determine whether the center-biased pattern of ingression was XY specific, we investigated the ingression pattern among the earliest pre-supporting cells using a transgenic Runx1 reporter (Gong et al., 2003). In this line, GFP is expressed specifically in supporting cells of the gonad in the layer of cells below the CE (Fig. 8A), indicating that the reporter faithfully recapitulates RUNX1 expression. In E10.75 XY genital ridges, GFP was enriched in the central region of the gonad, slightly posterior to the mesonephric tubules (Fig. 8B). At E11.0, both sexes exhibited a greater quantity of GFP in the center compared with the poles (Fig. 8C). These data show that supporting cell precursor ingression is center biased during establishment of the pre-supporting cell population in both sexes.

Two primary spatial patterns of gonad development have informed models of gonadal cell differentiation and sex determination. Initial commitment to gonadal fate, including expression of Gata4 and Nr5a1, occurs from anterior to posterior along the intermediate mesoderm (Brambell, 1927; Hu et al., 2013; Sasaki et al., 2021). Subsequent events of ovary differentiation follow the same pattern (Suzuki et al., 2015). However, events of testis development have been reported to follow a center-to-pole pattern (Harikae et al., 2013) with the central domain of the bipotential gonad showing a greater disposition towards Sertoli cell fate than the poles (Hiramatsu et al., 2003). This pattern was proposed to arise from center-first expression of the Sry transcript (Bullejos and Koopman, 2001) followed by relay of Sertoli cell fate to the poles governed by paracrine signals such as FGF9 (Hiramatsu et al., 2010) or prostaglandins (Rossitto et al., 2015; Wilhelm et al., 2005). Our results contest this model, indicating broad initial distribution of SRY and SOX9 (Fig. 2), and negligible paracrine relay during sex determination (Fig. 6). Instead, our results indicate a center bias in supporting cell precursor ingression in both sexes, resulting in a greater density of pre-supporting cells in the central domain. In this model, reinforcement of Sertoli cell fate through density-dependent mechanisms, such as paracrine FGF or prostaglandin signaling, biases the center of the XY gonad towards testis differentiation compared with the poles (Fig. 9). The resultant spatial bias becomes the dominant developmental pattern in the testis. In contrast, the pattern of early gonad differentiation is not overridden in the XX gonad, leading the ovary to exhibit a primarily anterior-to-posterior pattern of development.

This model changes our understanding of both testis and ovary pathway regulation. It suggests that early granulosa cell commitment is significantly less dependent on density-based signals than testis development, as the central cell density bias does not result in center-biased granulosa fate. For testis development, it focuses attention on the spatial arrangement of gonadal precursor cells and density-dependent signaling. Previous work on ovotestis and chimeric XX↔XY gonads suggests mechanisms for testis-specific coordination of cellular decisions that canalize whole-organ fate commitment (Burgoyne et al., 1988; Palmer and Burgoyne, 1991). These results indicate a role for early gonad morphogenesis in Sertoli cell pathway stabilization. It will be important to determine the relationship between increased accumulation of supporting cell precursors in the central region, and the rate and strength of Sertoli cell commitment. The molecular dynamics of CE cell ingression and the basis for its center bias are unknown. Recent efforts in 3-dimensional in situ imaging of the early genital ridge highlight several morphological AP regional differences that could play a role. These include curving and compaction of the ridge, as well as development of the mesonephric tubules (Bunce et al., 2021).

Paracrine signaling has been implicated in testis commitment by multiple sex-reversal experiments. For example, in gonads composed of XY and XX cells, XX cells can commit to Sertoli cell fate, indicating that non-autonomous induction can occur (Palmer and Burgoyne, 1991). Testis development requires as few as 36% XY cells in XX↔XY chimeras (Burgoyne et al., 1988). In addition, XY gonadal poles can be sex reversed by separation from the center using either physical separation or the insertion of an impermeable barrier (Hiramatsu et al., 2003, 2010).

Multiple secreted signaling factors expressed in Sertoli cells can induce SOX9 in granulosa cells in culture, including FGF9 and PGD2 (Kim et al., 2006; Wilhelm et al., 2005). Unlike previous models that implicated these factors in paracrine relay of Sertoli cell fate, our model suggests that the crucial role of paracrine signaling is density-dependent stabilization of Sertoli cell fate (Fig. 9). We propose that initial regional biases in the density of supporting cells caused by different rates of CE cell ingression lead to differences in speed or strength of commitment to Sertoli cell fate downstream of Sry. In this way, although initial activation of pro-testis genes is widely distributed, the central region of the gonad progresses through Sertoli cell differentiation faster than the poles. As there is a limited time window for activation of the male pathway and repression of the female pathway during commitment to testis development in the XY gonad (Hiramatsu et al., 2009), reduced supporting cell precursor ingression in the poles would lead to the poles exhibiting greater susceptibility to sex reversal. In contrast, commitment to granulosa fate in supporting cells may be primarily autonomous, as the presence of a central Sertoli-enriched region in an ovotestis does not prevent granulosa fate commitment in the posterior pole. Wnt signaling, which is a major component of granulosa fate determination, involves membrane-tethered ligands (Stewart et al., 2019), resulting in short-range cell contact-based signaling in the development of other organs, such as the intestine (Farin et al., 2016). The diffusibility of FGF ligands allows longer-range signaling (Goetz and Mohammadi, 2013). FGF9-soaked beads induce supporting cell fate reversal in cells far from the bead (Fig. 4), but the fate reversal is not stable, possibly because the bead creates a short burst of FGF9 rather than persistent exposure. The difference between Wnt and FGF signaling range suggests the sexes may have opposing schemes of cell fate regulation: an underlying autonomous pro-granulosa network and a non-autonomous pro-Sertoli network. This opposition may be an integral system property for achieving coordination of cell fate and whole-organ bipotentiality. Considerations of evolutionary forces shaping the sex-determination pathway are beyond the scope of this work.

We reconfirm previous results showing that XX cells can commit to Sertoli cell fate in XX↔XY chimeras sometime after E11.5 (Palmer and Burgoyne, 1991). However, surprisingly, we found that the central region of the developing testis does not induce Sertoli cell fate in XX cells in E11.75 reconstructions cultured for 24 h (Fig. 6). As upregulation of Sry and Sox9 across the full AP axis of the wild-type XY gonad occurs in around 6 h (Bullejos and Koopman, 2001), we expected 24 h to be greater than the required time for fate induction, even if interactions were delayed because of fusion and acclimation to culture conditions. Several entangled factors may explain the differences between chimeric gonads and reconstructions, including cell fate stability, signal duration and the extracellular matrix. The paracrine activity of FGF9 is dependent on heparin binding (Harada et al., 2009; Hecht et al., 1995) and may have limited activity in the XX matrix. Gonad sandwich cultures assembled at E11.5 and cultured for 48 h showed no evidence of XX cell conversion to Sertoli cell fate, although XY cells invaded the XX gonad (Yao et al., 2003), similar to reconstruction experiments conducted here. Although granulosa fate in the XX gonad may be stabilized by E11.75, pre-granulosa cells continuously ingress from the XX CE until E14.5 (Mork et al., 2012b), so we expect a population of early-stage granulosa precursor cells in each gonad reconstruction experiment (E11.75-E13.75). Reconstruction experiments of very early-stage gonads (prior to E11.5) might lead to a different outcome.

Cell cycle regulation has been connected to gonad-supporting cell differentiation in several ways. For example, Sry upregulates proliferation in the CE of the XY gonad through a FGF9-dependent mechanism, and disruption of receptors involved with insulin signaling, a pathway linked to proliferation, causes XY sex-reversal phenotypes (Nef et al., 2003; Pitetti et al., 2013; Schmahl et al., 2000, 2004). Blocking proliferation during a narrow window of time around E11.0 leads to a high incidence of sex reversal in XY gonads (Schmahl and Capel, 2003), but the molecular reason for this is unknown. Although we anticipated finding a central bias in CE proliferation in the XY gonad at E11.5, we instead observed a bias in supporting cell precursor ingression (Fig. 7). This conclusion considers cell cycle and ingression as separate behaviors. Recent work on Numb signaling in the bipotential gonad indicates that supporting cell precursor ingression may be tied to asymmetric cell division (Lin et al., 2017). If cell cycle and ingression are tightly coupled in the CE, patterns of cell cycle may be obscured by ingression. At E11.75, when the CE no longer produces supporting cells, it shows central enrichment of proliferation (Fig. S8A). The change between E11.5 and E11.75 indicates supporting cell precursors in the CE may exhibit a different pattern of proliferation from interstitial cell precursors or from the CE overall, which we did not detect using EdU incorporation. It is not known whether the increase in proliferation in the XY CE after Sry expression occurs within all cells of the CE or is specific to supporting cell precursors, or why the CE stops producing supporting cell precursors after E11.5.

Supporting cell precursors enter cell cycle arrest upon ingression from the CE and return to active cycling between E11.75 and E12.0 if committing to Sertoli cell fate, or, if commited to granulosa fate, upon follicle activation after birth (Mork et al., 2012b; Nef et al., 2005). FGF9 may promote pre-Sertoli cell cycle re-entry, as exogenous FGF9 can induce cell cycle reentry in arrested granulosa cells (Fig. 4D). However, p27 analysis in XY Fgf9-KO gonads demonstrates that pre-Sertoli cells transiently re-enter the cell cycle without FGF9 (Fig. 5), indicating that FGF9 is not necessary for this process. The primary means of cell cycle regulation in supporting cells may be granulosa-associated inhibition, which abates when the Sertoli pathway is initiated. Concomitant return to cell cycle arrest and granulosa fate in supporting cells of XY Fgf9-KO gonads indicates that stabilization of Sertoli cell fate may be linked to cell cycle re-entry. Although there was little overlap of SOX9 and FOXL2 during Fgf9-KO sex reversal, analysis of p27 with SOX9 and FOXL2 did not reveal a consistent path from Sertoli to granulosa states (Fig. 5B). Future analysis of additional cell cycle factors may clarify the mechanism of cell cycle regulation during Sertoli cell or granulosa fate commitment.

Although two phases of Sertoli cell fate commitment, initiation and stabilization, have been distinguished through analysis of XY sex-reversal mutants, such as Fgf9-KO, the extent of molecular overlap between the processes remains unclear. Although activation of Sox9 by Sry is considered to be the primary step in the GRN governing Sertoli cell fate, Fgf9 is an integral stabilization factor. Direct binding of SRY to the Fgf9 promoter region was recently demonstrated (Li et al., 2020), indicating that the molecular regulatory networks for initiation and stabilization may separately and simultaneously stem from SRY. The role of FGF9 in promoting SOX9 may be indirect, through downregulation of Wnt signaling. This work suggests that the center bias in Sertoli cell fate arises through density-dependent suppression of the opposing granulosa pathway and is related to the stabilization rather than the initiation of Sertoli cell fate.

In summary, we analyzed spatial patterns of cellular events during gonadogenesis and sex determination to determine the source of the center-biased pattern of testis fate disposition in XY gonad development. Hypotheses of center-first expression and paracrine relay were ruled out by the patterns of SRY⁺ and SOX9⁺ cells, which appear first in a widespread pattern and then exhibit a bias towards the center over time. Analysis of FGF9-soaked bead induction in XX gonads and Fgf9 knockout XY gonads led to the conclusion that FGF9 is not solely responsible for the pattern but is integral to enhancing it. In both sexes, we observed a central bias in supporting cell precursor ingression from the CE. In the XY gonad, subsequent density-dependent reinforcement of Sertoli cell fate explains the center bias in testis fate disposition.

Unless otherwise stated, transgenic mouse lines were maintained on a C57BL/6J (B6) background. CD-1 mice obtained from Jackson Laboratories were used as a wild-type strain. The Fgf9^lacZ mouse line was generated using targeted ES cells from the KOMP repository (project 24486), followed by cre-mediated excision of the neo selection cassette (Huh et al., 2015). Sex reversal in the Fgf9^lacZ line has been previously described (McKey et al., 2019). Mice expressing EGFP from the β-actin promoter and CMV intermediate early enhancer [FVB.Cg-Tg(CAG-EGFP)B5Nagy/J] were maintained as homozygotes on the FVB background. The Tg(Nr5a1-GFP) (SF1:eGFP) mouse has been previously described (Stallings et al., 2002). The Tg(Runx1-EGFP)EI184Gsat/mmucd (Runx1-eGFP) mice were a gift from Humphrey Yao (MMRRC stock number 010771-UCD; Gong et al., 2003). All mice were housed in accordance with National Institutes of Health guidelines, and experiments were conducted with the approval of the Duke University Medical Center Institutional Animal Care and Use Committee.

To obtain embryos at specific stages of development, males were set up in timed matings with several females. Each female was checked daily for the presence of a vaginal plug. The date of plug detection was considered embryonic day 0.5. Female CD-1 mice were used for wild-type and transgenic reporter experiments.

Gonad/mesonephros and urogenital complexes were collected from embryos dissected in PBS. For embryos in the range of E10.5 to E12.5, tail somites were counted to reduce stage variability: eight tail somites corresponds to approximately E10.5, 13 tail somites to E11.0, 18 tail somites to E11.5 and 30 tail somites to E12.5 (Hacker et al., 1995). Tissue samples were fixed in 4% PFA/PBS overnight at 4°C, rinsed three times for 20 min with PBS with 0.1% Triton X-100, and dehydrated stepwise into 100% methanol for storage at −20°C. Tail samples were collected and processed for genotyping. Sex and transgenic genotypes were determined by genotying for Sry (primer sequences 5′→3′: GTGTCTCAAAGCCTGCTCTTC and CATGTACTGCTAGCAGCTATC) or Ddx3X/Y (primer sequences 5′→3′: ATAATCACAGCTGGTAATGTAC and GCTGGACGTTCTAAAAGGTAAG) and the appropriate transgene according to established protocols (Gong et al., 2003; Huh et al., 2015; Stallings et al., 2002). Exceptions were made for determining sex in wild-type embryos after E12.5, in which case the presence of testis cords was used as an indication of the XY genotype. In the case of some fluorescent reporters, mice were not genotyped when the fluorescent signal was strong enough to observe with a fluorescence detection system on the dissection microscope (e.g. NIGHTSEA).

Tissue samples were fixed in 4% PFA and PBS overnight at 4°C, rinsed three times for 20 min with PBS with 0.1% Triton X-100, and dehydrated stepwise into 100% methanol for storage at −20°C. Immunofluorescence was performed as follows: tissues were rehydrated stepwise into PBS, washed three times in PBS with 0.1% Triton X-100, incubated for 1 h at room temperature in blocking solution (10% FBS, 3% BSA and 0.2% Triton X-100 in PBS) and incubated with primary antibodies in blocking solution overnight at 4°C (see Table S1 for details on primary antibodies). The following day, samples were washed three times for 30 min in washing solution (1% FBS, 3% BSA and 0.2% Triton X-100 in PBS) and incubated with 4′,6-diamidino-2-phenylindole (DAPI: D9542; Sigma-Aldrich) and secondary antibodies (1:500) in blocking solution overnight at 4°C (see Table S2 for secondary antibody details). The next day, samples were washed three times in washing solution and mounted in DABCO. Images were taken with an LSM710 confocal microscope using the associated Zen software (Zeiss).

Anteroposterior (AP) axis analysis was conducted using whole-mount samples imaged with confocal microscopy to determine the distribution of fluorescent signal or positively labeled cells along the AP axis of the genital ridge, gonad or gonadal region. Initial steps of image processing, including selection of the gonad area, the anterior point and individual cells, as well as segmentation, filtering and thresholding, were performed using FIJI (ImageJ). Subsequent processing and analysis were performed with JupyterLab (Anaconda distribution, Python) using skimage (van der Walt et al., 2014) and standard scientific Python libraries.

Images were typically scaled down by a factor of 10 to reduce processing time. Images for analysis were acquired with an LSM710 confocal microscope using the associated Zen software (Zeiss). Samples were imaged as ‘tile scan’ z-stacks with 2.5 µm optical sections, 10% tile overlap and 5 µm or 10 µm intervals between sections. Image selection for analysis was dependent on the completeness of the genital ridge and the type of analysis. Samples with a damaged or incomplete gonad were not used for analysis. For analysis of sagittal sections (lateral views), image planes were collected across the majority of the depth (lateral to medial) and the image that contained the largest gonadal area was processed for analysis. For frontal sections (ventral views), samples were imaged across the entire depth of the gonad tissue (CE to mesonephros) with a minimum interval of 5 µm in order to capture every nuclei in at least one plane. Nuclei were found to have a diameter of 5 to 10 µm. In this case, the gonad area used for analysis was the union of the gonad areas in all slices, as determined by maximum intensity projection.

AP axis calculation began with binary images of gonad areas. Gonad areas were hand drawn for sagittal sections as the area between the coelomic surface of the gonad and the bottom of a general gonadal or supporting cell marker. For frontal sections, gonad area was hand drawn as the area around a general gonadal or germ cell marker. Gonad areas from individual slices were merged to make gonad areas for full z-stacks. For each gonad area, a binary axis was calculated by thinning with the ‘skimage.morphology.thin’ function. In order to achieve a skeleton that was 1 pixel in width, a convex hull algorithm was applied to the image by sliding windows with increasing window size until a binary hit-or-miss algorithm detected only two end points. The pixel coordinates of the resultant skeleton were ordered by increasing the distance to the anterior pole of the gonad. The position along the AP axis for each point in the skeleton was calculated as the relative position across the ordered list. The anterior, central and posterior regions of the gonad were calculated by dividing the skeleton into three equal parts. For EdU analyses, the same protocol was applied to a separately drawn CE area to generate a skeleton with a width of a single pixel. All sagittal images processed were analyzed as individual marker-positive pixels.

For pixel-based analysis of sagittal sections, a binary image was generated from confocal optical sections using FIJI by first applying a gaussian filter, followed by masking to isolate the gonad area and applying the Otsu threshold. Pixels with signal above the threshold were considered positive for immunofluorescent signal. The position along the AP axis for each pixel was determined as the AP position of the nearest point of the skeleton generated for the corresponding gonad area. Histograms generated from AP position data were calculated as probability density functions and averaged over multiple samples. As the most biologically relevant discrete regionalization of the gonad remains unknown, histograms were generated with 20 bins to allow visual assessment of the distribution and comparison between experiments with minimum bias. The number of samples analyzed for each plot is indicated by ‘n’. Conclusions are drawn from graphs as presented.

For cell-based analysis of frontal sections, SOX9-positive and SRY-positive cells were identified by eye in the z-stack and each cell was labeled with a single pixel in a binary image, making sure not to duplicate labeling for any cells that appeared in multiple sections. The stack of labels was flattened into a single image where each labeled pixel represented a cell. For each image, labeled pixels were transformed into coordinates and AP positions were assigned as the AP position of the nearest pixel in the skeleton.

For segmentation and labeling of p27, SRY and SOX9 in sagittal sections, binary images were generated from confocal optical sections in FIJI by first applying a gaussian filter, followed by masking to isolate the gonad area and applying the Otsu threshold. Pixels above the threshold were considered positive for immunofluorescence signal. Nuclear shapes were drawn by hand around areas of positive signal that were confirmed by eye to be nuclei. For double-positive cells, which contain positive pixels for two factors, the larger nuclear shape was selected for representation.

Embryonic gonad/mesonephros complexes were dissected from embryos at indicated stages in PBS with 1 mM CaCl₂ and 0.5 mM MgCl₂ and transferred to agar blocks (1.5% agar) by pipetting following the method of Martineau et al. (1997). Where required, genotyping was performed during dissection: embryonic tails were collected and incubated in 200 µl 50 mM NaOH at 95°C for 10 min before addition of 75 µl 40 mM Tris-HCl (pH 5.5) followed by vortexing and centrifugation at 4000 g for 3 min. 2 µl of supernatant was used to genotype for sex using Ddx3X/Y. Organs were cultured at 37°C with 5% CO₂ in Dulbecco's Minimal Eagle Medium (DMEM), supplemented with 10% fetal bovine serum and 1×antibiotic-antimycotic (15240062, Thermo Fisher Scientific). Phase-contrast images of cultured organs were taken with an Axioplan 2 upright fluorescence microscope equipped with an AxioCam MRm camera (Zeiss).

FGF9-soaked beads were prepared as follows: 10 µl of heparin-agarose beads (H3025, Sigma) was added to 10 µl distilled H₂O and 3 µl of 50 µg/ml human recombinant FGF9 (7399-F9, R&D Systems) or 3 µl of 50 mg/ml BSA (15571-020, Invitrogen). Beads were left to incubate for 2 h at room temperature. After incubation, beads were centrifuged five times for 2 min at 3000 g. After each spin, 20 µl of supernatant was replaced with 20 µl DMEM, being careful not to disturb the pellet of beads. FGF9- and BSA-soaked beads were stored at 4°C until use (maximum, 1 month). For experiments at E11.5, embryos were genotyped during dissection. Gonad/mesonephros complexes were incubated in agar blocks for at least 1 h before bead application using the dissection microscope. Beads were applied to gonads using one of three protocols: on the surface, in a wedge or inside. For all protocols, a single bead was initially transferred to the surface of the tissue. For ‘surface’ application, the bead was moved to the center region of the gonad away from the mesonephros, typically resulting in the bead laying beside the gonad on the agar (as in Kim et al., 2006). For ‘wedge’ application, a small wedge of gonad tissue was cut away from the central region of the gonad adjacent to the agar using dissecting needles. The bead was then placed in the wedge, again typically resting on the agar. For ‘inside’ application, the gonad was pierced with a drawn glass needle to create a hole. The glass needle was used to push the bead into the hole far enough that it would not be instantly expelled. To prevent later expulsion, melted agar (1.5% in DMEM) was pipetted over the gonad. The layer of agar over the gonad was removed after 12-14 h, after which beads were typically retained for the duration of the experiment. For all three protocols, upon collection, samples where the bead was not retained in the appropriate position were discarded. During collection, beads were always lost from ‘surface’ application samples, and sometimes lost when application was by ‘wedge’ or ‘inside’ protocol. Samples were collected, rinsed and fixed overnight in 4% PFA at 4°C.

For tissue reconstruction experiments, embryos were collected at indicated stages from CD-1 dams mated with SF1:eGFP^+/− studs. Cultured organs were incubated in agar blocks for 1 h after dissection before further manipulation using a Leica stereomicroscope. Without displacing from troughs in agar pads, gonad and mesonephros complexes were cut in half transversely with needles to create anterior and posterior halves of equivalent length. Using forceps to move organs, the halves were either paired with halves of other organs or kept together and moved to a new location in the dish as controls. Once new pairings were established, troughs were aspirated, gonad halves were pushed against each other and cultures were returned to incubation at 37°C with 5% CO₂. All reconstructed organs presented consisted of an anterior half paired with a posterior half. For experiments lasting longer than 24 h, culture media was replaced and organ samples were briefly floated within the trough every 24 h. Upon collection, reconstructed organs were transferred to PBS and organ halves that did not remain together were discarded. Samples that remained fused were fixed overnight in 4% PFA at 4°C.

Chimera production was carried out as described by Mork et al. (2012a). Chimeras were produced between embryos derived from wild-type CD1×CD1 matings and embryos derived from CD1×EGFP (GFP^+/?) matings. Embryos collected from pregnant females at E2.5 were assembled in pairs (one from each mating) in shallow wells, cultured overnight to the blastocyst stage, then transferred to E2.5 pseudopregnant female hosts (Nagy et al., 2003). Because the sex of embryos was unknown at the time of assembly, chimeras could be XX↔XX, XY↔XY or XX↔XY. To assign sex to the GFP⁺ and GFP⁻ components, the fetal liver was collected from each chimeric embryo at the time of dissection and used for both immunohistochemistry and FACS. Cells from half the liver were dissociated, and GFP⁺ and GFP⁻ cells were separated by FACS and PCR genotyped for the presence or absence of the Y chromosome. To confirm the PCR results, a fragment of liver tissue from the same embryo was labeled with antibodies against H3K27me3, which strongly label the inactive X chromosome (Barr body) present exclusively in XX cells (Mork et al., 2012a).

For EdU experiments, pregnant females were injected intraperitoneally with 25 or 50 mg/kg 5-ethynyl-2′-deoxyuridine (EdU; 20540, Lumiprobe) dissolved in PBS at indicated stages. Embryonic samples were collected at a minimum of 1 h after EdU injection. As EdU is eliminated within 1 h after injection (Cheraghali et al., 1995), this method established consistent duration of EdU exposure across all experiments. Upon collection, samples were fixed overnight at 4°C. EdU-treated samples were imaged using the whole-mount immunofluorescence protocol with the addition of a click reaction step between rehydration and blocking. Samples were incubated in click reaction solution [20 mg/ml ascorbic acid, 2 mM cupric sulfate and 4 µM sulfo-Cy3 azide dye (A3330, Lumiprobe) in PBS] for 1 h rocking at room temperature.

We thank members of the Capel laboratory for comments and advice on the manuscript. We also thank Dagmar Wilhelm for providing the SRY antibody. All confocal imaging was performed in the Duke Light Microscopy Core Facility at Duke University.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201060.reviewer-comments.pdf