Divergent Expression Regulation of Gonad Development Genes in Medaka Shows Incomplete Conservation of the Downstream Regulatory Network of Vertebrate Sex Determination

Genetic control of male or female gonad development displays between different groups of organisms a remarkable diversity of “master sex-determining genes” at the top of the genetic hierarchies, whereas downstream components surprisingly appear to be evolutionarily more conserved. Without much further studies, conservation of sequence has been equalized to conservation of function. We have used the medaka fish to investigate the generality of this paradigm. In medaka, the master male sex-determining gene is dmrt1bY, a highly conserved downstream regulator of sex determination in vertebrates. To understand its function in orchestrating the complex gene regulatory network, we have identified targets genes and regulated pathways of Dmrt1bY. Monitoring gene expression and interactions by transgenic fluorescent reporter fish lines, in vivo tissue-chromatin immunoprecipitation and in vitro gene regulation assays revealed concordance but also major discrepancies between mammals and medaka, notably amongst spatial, temporal expression patterns and regulations of the canonical Hedgehog and R-spondin/Wnt/Follistatin signaling pathways. Examination of Foxl2 protein distribution in the medaka ovary defined a new subpopulation of theca cells, where ovarian-type aromatase transcriptional regulation appears to be independent of Foxl2. In summary, these data show that the regulation of the downstream regulatory network of sex determination is less conserved than previously thought.


Introduction
Sex determination, the decision whether the bipotential gonad anlage will become a testis or an ovary, is a complex and tightly controlled developmental process. The fate determination and cell differentiation programs are regulated and tuned by cascades or networks of genes. Comparative studies on sex determination cascades of different organisms revealed a remarkable diversity of "master sex-determining genes" at the top of the genetic hierarchies, whereas downstream components surprisingly appeared to be evolutionarily more conserved and tend to converge upon the regulation of common effectors. Hence, a comparative view on genetic sex determination mechanisms led to the paradigm that "masters change, slaves remain" (Graham et al. 2003). A well-known example illustrating this paradigm is the SRY gene, the master sex-determining gene of mammals, which has not been detected outside of the therian mammals. However, its subordinated genes (SOX9, WT1, DMRT1, AMH, SF1, FOXL2) or signaling pathways (TGF-beta, WNT4/ beta-catenin, Hedgehog) have homologs in a much broader spectrum of species, including nonvertebrates, where they apparently are also involved in sex determination. These observations led to the emergence of the subversive stereotype that master sex-determining genes individually spark and orchestrate the irreversible action of uniform and integrated gender-specific pathways. Conservation at the bottom and diversity at the top could be convincingly explained by an evolutionary scenario in which these hierarchies evolve from common core downstream components that acquire new upstream regulators (Wilkins 2007). Although the global rule of sex determination evolution is intuitively appealing and well accepted, only the variety at the top is well supported by comparative experimental data (Graham et al. 2003;Haag and Doty 2005;Herpin and Schartl 2008). The downstream conservation is less studied and relies only on a few gene expression studies. We have used the medaka as a versatile model system to study gene regulatory interactions and their evolutionary conservation (Wittbrodt et al. 2002;. In the medaka fish, which has XY-XX sex determination, dmrt1bY, the duplicated copy on the Y-chromosome of dmrt1a, was shown to be the dominant master regulator of male development (Matsuda et al. 2002;Nanda et al. 2002), similar to Sry in mammals. Interestingly, dmrt1, the ancestor of dmrt1bY, is one of the downstream effectors of SRY in the mammalian male pathway. The duplicated copy of dmrt1 on the Y-chromosome has acquired an upstream position in the sex-determining cascade. Remarkably this evolutionary novelty, requiring a rewiring of the regulatory network, was brought about by co-optation of "ready-to-use" pre-existing cis-regulatory elements contributed by transposable elements (Herpin et al. 2010). With respect to their biochemical functions, both medaka dmrt1 paralogs act as transcriptional regulators (Herpin et al. 2010). Dmrt1bY was shown to be responsible for male-speciEc primordial germ cell mitotic arrest in the developing gonad at the sex-determination stage . In contrast, the autosomal dmrt1a medaka gene is essential for testis maintenance (Masuyama et al. 2012).
The medaka gonad is formed by the coordinated development of two different cell lineages: the germ cells and the somatic gonadal mesoderm surrounding the germ cells. Shortly before hatching, at the time of expression of dmrt1bY in the male gonad primordium, the germ cells in the female gonad actively proliferate and undergo meiosis, whereas this is not observed in male gonads (Kobayashi et al. 2004;Herpin et al. 2007). It is only 10 days later that the first somatic gonadal dimorphisms are apparent with the formation of the acinus (the seminiferous tubule precursor) and the follicles in gonads of male and female, respectively. Interestingly, ovarian cords within the germinal epithelia of medaka ovaries have been recently characterized . These cords composed of somatic sox9b-expressing cells and mitotic nos2-expressing oogonia continually give rise to germ cells and form a stem cell niche referred to as the germinal cradle ). These cradles containing germline stem cells contribute to the production of fertile eggs during the life cycle of the adult ovary.
Like in other vertebrates studied so far, many components of the classical repertoire of mammalian sex-determining genes could be inventoried in medaka as well (Matsuda 2005;Siegfried 2010). To elucidate the gene regulatory network that controls specification and patterning of the gonads in the medaka fish, in this study we report the in vivo expression dynamics of classical mammalian markers within the forming and the adult gonads in medaka.
Because of their important roles in initiating male or female gonadal development in mammals, the respective implications of either Dmrt1 or Foxl2 transcription factors were examined. Further on, two of the major signaling pathways central for early gonadal induction and maintenance in mammals, namely the canonical Hedgehog and Wnt4/b-catenin signaling pathways (Wilhelm et al. 2007;Liu et al. 2010;Franco and Yao 2012), were investigated.
The Hedgehog (HH) signaling pathway plays an essential role in a wide variety of developmental processes (Ingham and McMahon 2001). Three HH proteins have been identified in mammals; SONIC (SHH), INDIAN (IHH), and DESERT (DHH) (Hooper and Scott 2005). Dhh and Ihh are coexpressed in the adult ovary where they stimulate proliferation and steroidogenesis of theca cells (Spicer et al. 2009). Dhh is also required for maintenance of the male germ line and spermatogenesis in mice (Bitgood et al. 1996;Clark et al. 2000). In mammals all three HH ligands signal by binding to one of two homologous transmembrane receptors, PATCHED homolog 1 and 2 (PTCH1 and 2). HH signaling is modulated by HH-induced transcription of the HH antagonistic interacting protein (HHIP) that binds to HH ligands and prevents their interaction with PTCH receptors (Chuang and McMahon 1999).
R-spondin1 (Rspo1), a member of a small family of secreted growth factors, is a key female-determining factor. RSPO protein operates through the canonical WNT signaling pathway (Tomizuka et al. 2008) to activate the b-catenin pathway as well as via upregulation of Follistatin (Fst) through WNT4 (Yao et al. 2004). It is well established that mammalian R-SPO-1, WNT4, b-catenin, and FST are components of a single pathway that promotes ovarian development and suppresses the formation of testis cord (Chassot et al. 2008).
The winged helix/forkhead transcription factor FOXL2 is mainly expressed in the somatic cells of the female gonad (Crisponi et al. 2001). The major role of Foxl2 during gonadal differentiation and maintenance has recently been shown via the mutual antagonistic relationship of Foxl2 and Dmrt1. FOXL2 suppresses expression of Dmrt1 and vice versa for maintaining female or male gonadal fate, respectively (Uhlenhaut et al. 2009;Matson et al. 2011). Additionally it has been reported that FOXL2 and WNT4 (Yao 2005;Ottolenghi et al. 2007;Garcia-Ortiz et al. 2009) cooperate in regulating FST expression during ovarian development. Interestingly in the ovary the expression profiles of Foxl2 highly correlate with that of Aromatase (Cyp19), suggesting that Foxl2 is involved in the regulation of estrogen synthesis via direct transcriptional upregulation of Aromatase (Pannetier et al. 2006). On the other hand, several other factors (e.g., testosterone, TGF-b1, TNF-, and glucocorticoids) have been shown to direct the expression of the aromatase gene in Sertoli, Leydig, and germ cells of rat testis (see Bourguiba et al. 2003 for review).
Analyzing several fluorescent reporter lines established from a bacterial artificial chromosome recombination (BAC recombination) method resulting in optimal spatial resolution and high reliability of gene expression (Giraldo and Montoliu 2001;Nakamura et al. 2008b;Suster et al. 2011), we find major discrepancies between mammals and medaka, notably amongst spatial and temporal expression patterns of the canonical signaling pathways. Using in vivo whole tissues chromatin immunoprecipitation and in vitro gene regulation assays, we can reveal possible interactions between these pathways that emphasize the importance of the cellular context on modulating these regulations and call into question a strict conservation of regulatory and functional interactions of sexual development genes in vertebrates.

Patched-2 Expression and Hedgehog Pathway Regulation
To determine the role of the hedgehog signaling pathway in gonadal development of medaka, we recorded the temporal and spatial expression patterns of the key receptor, Ptch2.
2329 Downstream Regulatory Network of Vertebrate Sex Determination . doi:10.1093/molbev/mst130 MBE Although expressed quite early on during somatogenesis (data not shown), patched-2 was surprisingly neither detected throughout the early phase of gonadogenesis when the undifferentiated gonad anlage grows up to hatching stage ( fig. 1A) nor further on up to the stage when the dimorphic gonad develops 10 days after hatching ( fig. 1B and C). The first specific gonadal ptch-2 expression was detected in the young ovary ( fig. 1D-H). Here expression is restricted to the somatic cells that express sox9b and surround the germline stem cells of the germinal cradle within the ovarian cord ) ( fig. 1G and H). Interestingly, r-spo-1 is also co-expressed with sox9b in these cells (figs. 1I-K and 2A-D compared with fig. 8A-F). In ovaries of fish of later reproductive phase, when the cradle number has declined, ptch-2 expression could be only noticed in the interstitium ( fig. 2E-G). Unexpectedly and in contrast to patched expression in mammals, ptch-2 expression was only detected at background levels throughout stages of testis development (data not shown). This inconspicuous role for the Hedgehog signaling in gonad development and maintenance was also apparent in real-time polymerase chain reaction (PCR) analysis. Although similarly expressed in different tissues, both patched receptor transcripts (1 and 2) are only expressed at background levels in adult gonads of both sexes ( fig. 3A).

In vivo Dmrt1bY Binding to hhip Promoter Region
Given the pivotal role of Dmrt1 transcription factors during medaka sex determination, patched-2 and the antagonistic regulatory HH interacting protein (hhip) promoter regions were scanned for putative Dmrt1bY and Dmrt1a target sites. Although no Dmrt1-binding sites could be identified in the patched-2 promoter region (10 kb upstream scanned), two sites were predicted with high fidelity in the hhip 5 0 region ( fig. 4B). To assess the in vivo relevance of the predicted Dmrt1 interaction, two stable transgenic lines expressing either the full-length Dmrt1bY protein (Dmrt1::GFP) or a truncated form lacking the DNA-binding domain (ÁDmrt1::GFP), both fused to GFP, were utilized (Herpin et al. 2010). These two lines were used for in vivo tissue chromatin immunoprecipitation (in vivo tissue ChIP) on testis tissue using GFP antibody for immunoprecipitation ( fig. 4A). For the predicted two proximal Dmrt1-binding sites in the hhip promoter, a more than 2.3-fold enrichment after immunoprecipitation validated Dmrt1 binding ( fig. 4B). Under the same conditions, no binding was detected in ovary (data not shown). This indicates that, in vivo, Dmrt1bY and/ or Dmrt1a are potentially regulating the gonadal HH signaling through direct transcriptional regulation of the antagonist hedgehog interacting protein, hhip. In line with the in vitro regulation data, patched-1/2 expression in gonads of both sexes was not above background ( fig. 3A). Due to the absence of the dmrt1bY gene in females, and since a background level of dmrt1a expression is detected in ovary , the high expression of hhip and the suppression of ptch-2 must be exclusively regulated by the autosomal dmrt1a ortholog in the ovary of medaka.

R-spondin 1 and Follistatin Expression and Wnt Pathway Regulation
Although ubiquitously expressed at early stages of development, neither medaka r-spo-1 nor fst expressions could be detected in the presumptive gonadal mesoderm before the PGCs reach the undifferentiated gonadal primordium at stage 30 ( Interestingly, this very brief pulse of r-spo-1and fst expression occurs shortly before the rise of dmrt1bY expression at that stage (Kobayashi et al. 2004;Hornung et al. 2007). At hatching stage when sex determination occurs, no r-spo-1or fst expression could be detected in the gonadal primordia of males and females ( fig. 7G-I). This lack of later gonadal expression is in line with similar findings in zebrafish (Zhang et al. 2011), turtle (Smith et al. 2008), chicken (Smith et al. 2008), and mice (Yao et al. 2004;Parma et al. 2006). Further on, while the ovary develops in juvenile females, r-spo-1 expression is restricted to few so-  (Zhang et al. 2011) and mouse (Smith et al. 2008), expression of medaka r-spo-1 was also detected in granulosa cells around young oocytes (fig. 8G-L and table 1). Granulosa expression is then progressively lost while the oocytes are growing. In male gonads, only a low expression of r-spo-1 is detected ( fig. 3B-D). Medaka fst could not be detected in the ovary at this stage. In contrast to germ cell expression of r-spo-1 in chicken (Smith et al. 2008), zebrafish (Zhang et al. 2011) and mouse (Smith et al. 2008) ovaries or zebrafish testis (Zhang et al. 2011) or fst in rat testis (Meinhardt et al. 1998), neither r-spo-1 nor fst could be detected in gonadal germ cells or spermatogonia in medaka ( fig. 8 and table 1). In older ovaries, r-spo-1 and fst are also expressed in the interstitium as well as in the ovarian epithelium ( fig. 8M, N, and Q-S). Of note, sparse clusters of fst expressing interstitial somatic cells were detected in testes ( fig. 8O and P).
In vivo Dmrt1bY Binding to the r-spo-1 and dkk1 Promoter Regions and Regulation of the r-spo-1 Pathway We next analyzed the capacity of Dmrt1 to transcriptionally regulate positive (r-spo-1 and fst) and negative (dkk1) 2333 Downstream Regulatory Network of Vertebrate Sex Determination . doi:10.1093/molbev/mst130 MBE effectors of the Wnt/Rspo1/Fst pathway. Hence, the ability of Dmrt1 to directly bind in vivo to the fst, r-spo-1, and dkk1 promoters was investigated. Out of several putative Dmrt1-binding sites within the r-spo-1 and dkk1 promoter regions, in vivo ChIP revealed direct Dmrt1bY interaction with both promoters ( fig. 4C and D). Of note, robust binding of Dmrt1bY to dkk1 promoter region was seen in testes, but similar interactions did not occur in ovaries ( fig. 4D). Although not highly significant, an analogous trend is observed for Dmrt1bY binding to r-spo-1 promoter ( fig. 4C).

Foxl2 Expression in the Adult Gonad
To investigate the role of Foxl2 in ovarian differentiation, we analyzed Foxl2 protein distribution in the medaka ovary (figs. 9 and 10). During the transition process of germ line stem cells to oocytes within the germinal cradle, medaka  Particularly, Foxl2 is present in the nuclei of all granulosa cells ( fig. 10B-D). Interestingly, while Foxl2 has been reported to be a strong inducer of the steroidogenic activity of granulosa cells via upregulation of the aromatase gene (the ovarian-type Cyp19a1) (Hudson et al. 2005;Wang et al. 2007;Guiguen et al. 2010), unexpectedly, and in contrast to mammals, a minority of theca cells do also express Foxl2 in medaka ( fig. 10E-G). Our results reveal two subpopulations of cyp19a1-positive theca cells, which are either Foxl2 positive or do not express the transcription factor ( fig. 10H-K).

Discussion
Comparative studies on sex determination cascades of different organisms revealed that the genetic control of male or female gonad development displays between different groups of organisms a remarkable diversity of "master sexdetermining genes" at the top of the genetic hierarchies, whereas downstream components surprisingly were found more widespread and are evolutionarily more conserved. Without much further studies, these observations led to the reasoning that conservation of sequence equalizes to conservation of function. While vertebrates have at least a common set of transcriptional regulators, including DMRT1 and FOXL2, as well as some signaling molecules and pathways such as the Hedgehog and R-spo-1/Wnt4 pathways, their molecular interplay and epistatic relationships are nevertheless far from being understood. The purpose of our work was to examine this molecular interplay in fish.
Absence of R-spo-1, fst, and ptch-2 in Medaka Germ Cells In the mouse it was proposed that activation of the R-spo-1/ Wnt/Fst signaling pathway in both somatic and germ cells, besides triggering meiosis in fetal germ cells, is required for ovarian differentiation and maintenance of ovarian cell identity (Chassot et al. 2008). Indeed inherent to their inductive epigenetic mode of specification, the germline sex is likely to be determined early on by Sry acting in the somatic cells (McLaren and Southee 1997;Sekido and Lovell-Badge 2008;Bowles et al. 2010) in a noncell-autonomous manner (see De Felici 2009 for review). Sry controls whether bipotential precursor cells differentiate into testicular Sertoli cells or ovarian granulosa cells (Koopman et al. 1991). This pivotal decision in a single gonadal cell type ultimately controls sexual differentiation throughout the body. Sex determination can be viewed as a battle for primacy in the fetal gonad between a male regulatory gene network in which Sry activates  In contrast, the sexual plasticity of medaka germ cells seems to be retained much longer than in mammals as illustrated by XX/XY transplantation chimeras. Although XY somatic cells differentiate into male cells according to their sex chromosome composition, in this environment XX germ cells differentiate into male cells regardless of their sex chromosome composition . Hence, while unlike their mammalian counterparts none of the different medaka marker genes analyzed (r-spo-1, fst or ptch-2) were detected in germ cells at any time of gonadal development, this major inconsistency between mammals and fish certainly reflects intrinsically divergent modes of germ cell commitment and interaction between germ and somatic cells possibly accounting for a higher sexual plasticity of germ cells in fish.

Role of the Rspo1/Wnt/Fst Pathway during Gonad Development and Maintenance
In line with observations made in zebrafish (Zhang et al. 2011), turtle (Smith et al. 2008), chicken (Smith et al. 2008), and mice (Yao et al. 2004;Parma et al. 2006), the absence in medaka of r-spo-1 and fst dimorphic expression during sex determination stages does not support a role during gonad induction. Of particular interest and unlike in mammals, the lack of medaka ptch-2 expression during the same MBE period additionally rules out any involvement of the Hedgehog pathway for gonad induction. Further on, while the ovary develops in the juvenile female, the strict co-expression of r-spo-1 and ptch-2, together with sox9b in the somatic cells of the germinal cradles, likely indicates a role in differentiating and specifying the somatic supporting lineage of the ovary. Of note the early decoupling of r-spo-1 and fst expression patterns, although expected to be involved in the same signaling pathway in mammals (Yao et al. 2004), de facto invalidates such an interplay in the medaka ovarian cradle.
Our results support the view that r-spo-1 has a globally conserved female-specific expression profile in vertebrate gonads despite some slight but intriguing divergences. In all analyzed vertebrate species it is apparent that r-spo-1 expression goes along with the process of somatic cell organization within the young ovary (table 1). In mouse, chicken, and medaka, r-spo-1 expression at that time, predominantly or exclusively in the somatic cells, suggests its implication in a conserved pathway leading to folliculogenesis. On the other hand, the absence of medaka r-spo-1 expression in gonad embedded germ cells, unlike in zebrafish, mice, chicken, and turtle, reveals that a role for germ cell development might not be accordingly conserved in the adult gonad of medaka.
In mammals, R-spo-1 engages the effector pathway of Wnt signaling and b-catenin and thereby activates Fst expression (Carmon et al. 2011). Hence, also medaka fst spatial and temporal expression pattern was expected to overlap with r-spo-1. While medaka r-spo-1 and fst indeed display similar expression in adult tissues ( fig. 3B), medaka fst was, however, absent during the early and late stages of ovarian induction and development. Fst expression was perceptible only in the interstitium and epithelium of old ovaries together with r-spo-1 ( fig. 8M, N, and Q-S). Unexpectedly, fst was also FIG. 7. R-spondin-1 and follistatin spatial and temporal expression patterns during medaka gonad primordium formation. Two BAC reporter fish transgenic lines expressing either GFP or mCherry were established to follow r-spo-1 and fst expression dynamics during gonad formation in vivo. Neither r-spo-1 nor fst expression was detected in the presumptive gonadal mesoderm at early stages of development (A-C). Between stages 33 and 35 in the dorsal region of the hindgut r-spo-1 and fst are co-expressed (D-F). At hatching stage, although ubiquitously expressed, neither r-spo-1 nor fst were detected in the gonadal primordium (G-I).
FIG. 8. R-spondin-1 and follistatin expression in the adults gonads. In the ovary of juvenile females, r-spondin expression is restricted to somatic cells surrounding the germ line stem cells of the germinal cradle within the ovarian cord (A-F). Co-expression with sox9b (A to F) and patched-2 (see fig. 2A-D) is observed. R-spondin-1 expression in granulosa cells is also detected around young oocytes (G-L). In the ovary, follistatin expression was only detected in the interstitium (M and N) and in the ovarian epithelium together with r-spondin-1 (Q-S). Sparse clusters of follistatin expressing interstitial somatic cells are also detected in adult testes (O and P).
2338 Herpin et al. . doi:10.1093/molbev/mst130 MBE detected in testes ( fig. 8O and P). This expression pattern indicates that in medaka and unlike in mammals, Fst is probably neither acting throughout early female gonad patterning nor during maintenance of cell identity in the adult ovary. Instead medaka fst expression appears to be more a marker of the aging ovary, probably acting during follicular atresia. Of interest, while exclusive ovarian expression is generally described for R-spo-1 and Fst in vertebrates, such strict female dimorphism was not observed for r-spo-1 and fst in zebrafish, rat, and medaka (table 1 and fig. 3).

Role of the Hedgehog Pathway during Gonad Development and Maintenance
The gonadal expression pattern of two components of the Hedgehog pathway in medaka is peculiar and different from what has so far been reported for other vertebrates including mammals (table 1). Unlike in mammals it appears that medaka gonadal HH signaling through the patched-2 receptor is not involved in inducing and specifying the gonad primordia. It would rather act late exclusively in the process of somatic cell differentiation in the ovarian cradle. In strict contrast to its mammalian counterparts, the quasi-absence of medaka patched-2 expression during testis formation in larvae and for testis cell identity maintenance in the adult rules out any functions during these processes (table 1). Certainly, the low testicular expression of patched-2 together with Dmrt1a/1bY-induced transcriptional upregulation of the HH antagonist hhip indicates a general function of Dmrt1 in actively downregulating the Hedgehog pathway in medaka testes. Taken together, we can conclude that although apparently downregulated at the transcriptional level, a background expression of patched-2 remains. This phenomenon is known as illegitimate transcription (Chelly et al. 1989;McLeod and Cooke 1989). We speculate that the high expression of the hhip hedgehog pathway antagonist in gonads (about 5 to 10 times higher than patched1/2 (E) Predominantly in the somatic cells of the developing ovary (Smith et al. 2008). (E) Germ cells (during meiosis at low levels) (Smith et al. 2008).
(E) Not detected in embryonic testes (Menke and Page 2002).
Rat: (A) Sertoli and endothelial cells, germ cells, spermatogonia, spermatocytes, and round spermatids (Meinhardt et al. 1998). A Newly Identified Subpopulation of Theca Cells Expressing Aromatase but Not Foxl2 Examination of Foxl2 protein distribution in the medaka ovary allowed us to define a new subpopulation of theca cells expressing Foxl2. Also expressing cyp19a1 (aromatase) these cells are then suspected of having a steroidogenic activity. In contrast to mammals where ovarian-type aromatase is only produced by granulosa cells, the biological significance of such cells expressing cyp19a1 and Foxl2 within the thecal layer remains unclear. Of note, in ovaries of cyp19a1/p450c17I double transgenic medaka reporter lines, two subpopulations of theca cells were previously identified, being either cyp19a1 or p450c17I positive in a mutually exclusive manner ). The cyp19a1 expressing subpopulation of theca cells was already considered as the precursors of the theca lineage ). The strict co-expression of Foxl2 and Aromatase (Cyp19) in the mammalian ovary led to the further demonstration that Foxl2 is involved in the regulation of estrogen synthesis via direct transcriptional upregulation of ovarian-type Aromatase (see Pannetier et al. 2006 for review). Surprisingly in medaka we found, within the thecal layer, aromatase-only positive theca cells that remained Foxl2-negative. In that perspective it is interesting to note that birds also have multiple populations of theca cells some of which are also steroidogenic (Nitta et al. 1991). In contrast to the main consensus, the discordance of spatial expression patterns of Foxl2 and ovarian-type aromatase (cyp19a1) calls into question an exclusive transcriptional regulation of cyp19a1 by Foxl2 in the ovary of medaka ( fig. 11). Although we cannot exclude that aromatase-only positive cells have not been previously also positives for Foxl2, implying the requirement of Foxl2 for the induction of the aromatase expression, our results indicate that foxl2 is nevertheless not required for the maintenance of the aromatase expression.

Variable Molecular Interplay among the Repertoire of Gonadal Markers during Medaka Gonad Formation and Maintenance
We could show a direct regulation of the Hedgehog and R-spo-1 pathways by Dmrt1bY ( fig. 12). It is thus becoming apparent that despite its tangible requirement for mammalian testis formation and later on in regulating

2341
Downstream Regulatory Network of Vertebrate Sex Determination . doi:10.1093/molbev/mst130 MBE Leydig and myoid cell function (Clark et al. 2000;Pierucci-Alves et al. 2001;Canto et al. 2004), the Hedgehog pathway might not only be dispensable during medaka male gonadogenesis and maintenance but even needs to be suppressed as it is actively repressed by Dmrt1 genes. This would also explain the specific lack of ptch-2 expression in medaka testes shown to be indirectly downregulated by Dmrt1bY (see fig. 12 for summary). In contrast the female-specific, Dmrt1a-triggerd upregulation of follistatin transcription might nevertheless point out the importance of the R-spo-1 pathway during female gonad differentiation although the upstream components of this same pathway are tightly regulated by the Dmrt1 co-orthologs ( fig. 12).
Importantly, showing that under certain conditions Dmrt1 paralogs are able to strongly upregulate the female-specific r-spondin-1 gene expression, we could also demonstrate that Dmrt1bY/Dmrt1a-triggered regulations are highly dependent of the cellular context and might suggest requirement of co-factors ( fig. 12). These findings are reminiscent of observations showing Stra8 transcription to be directly repressed by DMRT1 in mouse testes while activated in the fetal ovary (Krentz et al. 2011).
Interestingly, the canonical Wnt/ b-catenin (R-spo-1 and Fst) pathway strongly antagonizes FGF9, a robust component of the male sex-determining cascade in mammals (Kim et al. 2006;Matson et al. 2011). However, FGF9 is absent in the fish lineage and no indication of a redundant action of related FGF has been obtained, questioning the importance of FGF signaling in fish sex determination (Forconi et al. 2013). Into that direction the importance of the FGF9 signaling in vertebrates is actually also challenged by the finding that in chicken embryos Fgf9 does not show any sexually dimorphic expression pattern during gonadal differentiation (Cutting et al. 2013). Similarly, although phylogenetically preserved, the Sox9 gene, a direct target of Sry in mammals, has been shown to be functionally dispensable for medaka testis determination (Nakamura et al. 2008a(Nakamura et al. , 2012. In this context our data might reflect a profound reorganization of that part of the fish gonadal regulatory network compared with mammals. While some components such ase DMRT1, SOX9, FOXL2 and pathways such as Hedgehog or R-spo1/Wnt/Fst of the gonadal gene regulatory network are conserved on the DNA sequence level across phyla, their functions, regulation, and interplays might be considerably different.

Generation of BAC Transgenic Medaka Lines and Imaging Analyses
The Carbio (WLC# 2674) strain of medaka (Oryzias latipes) was used for establishment of the transgenic lines. Microinjection of DNA was performed as described previously ) using BAC clone DNA at a concentration of 50-100 ng/mL. Adult G0 fish were then screened for fluorescence, and positive individuals were raised to adulthood. Siblings from positive G0 fish were mated to each other and the offspring were again sorted for fluorescence. Sox9b and cyp19a1 (aromatase) transgenic lines were described earlier (Nakamura et al. 2008a. For imaging embryos, hatchlings or tissues were mounted with 1.2% low melting temperature agarose. Confocal pictures and image stacks were acquired using a Nikon C1 (eclipse Ti) confocal laser scanning microscope and the NIS elements AR software.

Immunochemistry
Ovaries or testes from juvenile and adult fish were fixed with 4% paraformaldehyde/balanced salt solution (111 mM NaCl, 5.37 mM KCl, 1 mM CaCl 2 ÁH 2 O, 0.6 mM MgSO 4 Á7H 2 O, 5 mM Hepes, pH 7.3) for 30 min on ice. After fixation samples were washed three times for 10 min with MABT buffer (100 mM maleic acid, 150 mM NaCl, pH 7.5, 0.1% Triton X-100) and subsequently twice for 30 min with MABDT buffer (MABT buffer complemented with 1% BSA and 1% DMSO). After blocking in MABDT-blocking buffer (MABDT buffer supplemented with 2% lamb or sheep serum), the tissues were incubated in MABDT-blocking buffer together with the primary antibody (1:150 dilution) overnight at 4 C. Samples were then washed three times 5 min in MABDT buffer and 2343 Downstream Regulatory Network of Vertebrate Sex Determination . doi:10.1093/molbev/mst130 MBE washed again four times for 30 min in MABDT-blocking buffer on ice. Thereafter samples were incubated overnight at 4 C with the secondary antibody diluted at 1:600 in MABDT-blocking buffer. Finally the tissues were washed in PBS, stained with Hoechst solution for 3 h at 4 C, mounted, and imaged with a confocal microscope (Nikon C1 confocal microscope).

In vivo Chromatin Immunoprecipitation
For in vivo chromatin immunoprecipitation, the EpiQuik Tissue Chromatin Immunoprecipitation kit (Epigentek) was used according to the manufacturer's instructions, using 20 mg of testis tissue samples either from dmrt1bY::GFP or deltadmrt1bY::GFP transgenic fish Herpin et al. 2010) (20 testes for each) and GFP antibody (3 mg, Upstate) for immunoprecipitation. After tissue disaggregation and cell re-suspension, DNA was sheared by sonication (9 pulses of 10 s with an amplitude of 10%). After immunoprecipitation (

Bioinformatic Analyses
Binding sites for Dmrt1bY were identified using the matrix provided by (Murphy et al. 2007) together with the Regulatory Sequence Analysis Tools portal; RSat (http://rsat. ulb.ac.be/rsat/, last accessed August 12, 2013).

In Vitro Expression Regulation Analyses and Real-Time PCR
Medaka spermatogonial (SG3) and fibroblast-like (OLF) cell lines were cultured as described (Etoh 1988;Hong et al. 2004). For transfection cells were grown to 80% confluency in 6-well plates and transfected with 5 mg expression vector using FuGene (Roche) reagent as described by the manufacturer.
Total RNA was extracted from fish tissues or transfected cells using the TRIZOL reagent (Invitrogen) according to the supplier's recommendation. After DNase treatment, reverse transcription was done with 2 mg total RNA using RevertAid First Strand Synthesis kit (Fermentas) and random primers. Real-time quantitative PCR was carried out with SYBR Green reagents and amplifications were detected with an i-Cycler (Biorad). All results are averages of at least two independent reverse transcription reactions. Error bars represent the standard deviation of the mean. Relative expression levels (according to the equation 2-DeltaCT) were calculated after correction of expression of elongation factor 1 alpha (ef1alpha) and brain expression was set to 1 as a reference.