FOXL2 versus SOX9: A lifelong “battle of the sexes”


  • Reiner A. Veitia

    Corresponding author
    1. Institut Jacques Monod, CNRS-UMR 7592, Bâtiment Buffon, Paris Cedex, France
    2. Université Paris Diderot, Paris, France
    • Institut Jacques Monod, CNRS-UMR 7592, Bâtiment Buffon, Paris Cedex, France.
    Search for more papers by this author


Testis determination in most mammals is regulated by a genetic hierarchy initiated by the SRY gene. Early ovarian development has long been thought of as a default pathway switched on passively by the absence of SRY. Recent studies challenge this view and show that the ovary constantly represses male-specific genes, from embryonic stages to adulthood. Notably, the absence of the crucial ovarian transcription factor FOXL2 (alone or in combination with other factors) induces a derepression of male-specific genes during development, postnatally and, even more interestingly, during adulthood. Strikingly, in the adult, targeted ablation of Foxl2 leads to a molecular transdifferentiation of the supporting cells of the ovary, which acquire cytological and transcriptomic characteristics of the supporting cells of the testes. These studies bring many answers to the field of gonadal determination, differentiation and maintenance, but also open many questions.


TES, for testis-specific enhancer of Sox9; TESCO, Tes core sequence; FLRE, FOXL2 response element; SOX9, SRY (sex determining region Y)-box 9; FOXL2, Forkhead box transcription factor L2; WNT, Wingless-type MMTV, Integration Site Family of proteins; Fgf, fibroblast growth factor; BPES, blepharophimosis, ptosis, epicanthus inversus syndrome; Amh, antimullerian hormone; RSPO, member of the R-spondin family; ESR, estrogen receptor


Gonadal sex determination is a process that implies a unique decision to make a testis or an ovary out of a bipotential primordium. In most mammals, sex determination is equated with testis determination, whereas ovarian determination has been considered a default process. That is, absence of the testis-determining factor would automatically lead to the development of an ovary (in appropriate conditions). Recent articles challenge this view and show that in the ovary active mechanisms are required to maintain the differentiated state.1, 2

The bipotential gonad is made up of four presumptive cell lineages: germ cells (which have an extraembryonic origin) and three types of somatic cells. The somatic component involves the supporting cell precursors, which will give rise to Sertoli cells in the male or granulosa cells in the female, the steroidogenic precursors, which differentiate into Leydig cells (male) and theca cells (female), and finally the connective tissue cells yielding other important structures.3

From a genetic perspective, the SRY gene is pivotal in switching on the testis-determining cascade.4 Indeed, murine Sry transcripts appear in the presumptive Sertoli cells from 10.5 days postcoitum and for about 2 days, at the right moment for testis determination.5 In other mammals, including man, SRY is expressed from the period of testis determination until adulthood6). It seems now that the main task of Sry/SRY is to directly activate the gene Sox9, which also encodes a transcription factor. Recently a gonad-specific enhancer that mediates testis Sox9 expression, called TESCO, has been identified.7 Indeed, Sry along with the steroidogenic factor 1 (Sf1) binds to multiple elements within the TESCO. Moreover, mutation, co-transfection, and sex-reversal studies suggest the existence of a positive feedback circuit. First, Sf1 and Sry cooperatively up-regulate Sox9 transcription. Then, Sox9 and Sf1 recognize the enhancer to maintain the expression of the former in the absence of Sry, giving rise to the self-sustained mechanism itself.1, 7

Murine Sox9 is strongly expressed in the male genital ridge. Its expression in the mouse persists during fetal life and in adult testes.8 Consistent with this behavior, it has been shown that Sox9 expression, once triggered, is sustained by another positive feedback loop involving the signaling molecule of the fibroblast growth factor family Fgf9. In short, Sox9 is essential for Fgf9 expression, and Fgf9, in return, maintains Sox9 expression.9 Another positive feedback loop involves the lipocalin prostaglandin D synthase, whose expression, once activated and maintained by Sox9, leads to the accumulation of prostaglandin D2, which in turn activates Sox9 transcription and nuclear translocation. These self-activating mechanisms amplify Sox9 expression and activity during mammalian testicular organogenesis and work in seemingly independent fashions.10 As discussed below, such elementary circuits involving positive feedback loops are sufficient to maintain a differentiated state and in the case of Sox9 the existence of redundant circuits can be interpreted as a fail safe against the effects of (potentially sex-reversing) mutations in one of the branches of the network.

Not surprisingly, overexpression of SOX9 in an XX background has been associated with female-to-male sex reversal, as described in a patient with a duplication of a region containing SOX911 or for the murine Odsex mutation, where a transgene inserted in a regulatory region of Sox9 potentially derepresses its expression.12 This suggests that Sox9/SOX9 is the main effector of testis determination (even though acting downstream of Sry/SRY in most mammals). Accordingly, SOX9 is the common denominator of vertebrate testis determination and might be the highest placed and most ancient known sex-determining factor. Indeed, SRY/Sry has only recently been recruited at the head of the regulatory hierarchy in mammals (13 and references therein).

This behavior of SOX9 is compatible with the “Z-locus” model of sex determination, which postulates that in the male SRY, or SOX9 itself, would repress Z, a repressor of the male pathway. Absence of Z would lead to SOX9 expression, testis development, and XX sex reversal.14 We will come back to this idea below (Fig. 1).

Figure 1.

The “Z locus” model. It postulates that in the male, SRY (or SOX9 itself) would repress Z, a repressor of the male pathway. Absence of Z would lead to SOX9 expression, testis development, and XX sex reversal. Solid arrows represent well-documented interactions. Dashed arrows represent putative or suggested interactions.

FOXL2: A molecular actor in the spotlight

XX sex reversal without SRY is a rare condition in humans. However, in different domestic animals, cases of SRY-negative XX sex reversal have been described. A mammalian locus involved in XX sex reversal was mapped through the study of the goat polled intersex syndrome (PIS), a disorder showing a dominant absence of horns affecting both sexes and autosomal recessive XX sex reversal. Fine-mapping studies suggested that both PIS (on 1q43) and the human blepharophimosis ptosis epicanthus syndrome (BPES) (on 3q23) loci mapped to a 100-kb homologous region.15 The PIS mutation has been characterized and involves a small deletion (probably removing regulatory elements) that leads to a misregulation of several genes in the region.16

More or less at the same time, mutations in the gene FOXL2 were shown to be responsible for BPES.17FOXL2 encodes a forkhead transcription factor whose expression is strongly dysregulated by the PIS mutation16 and thus became a perfect candidate gene to explain sex reversal in the goat model. While mapping data were consistent with this hypothesis, the phenotypes were more difficult to reconcile, since in typical BPES, eyelid abnormalities are associated with ovarian failure in humans and not with sex reversal. Accordingly, we failed to detect any mutation in the coding region of FOXL2 in a series of XX males.18 However, specific regulatory mutations cannot be excluded, and should indeed be looked for.

Rather similar findings have been obtained with two Foxl2 knockout (KO) mouse models.19, 20 XY homozygous animals do not display any defect in testis development and XX Foxl2−/− mice are phenotypically female, but sterile.19, 20 Indeed, ovaries of KO mice are small, severely disorganized, and primordial Follicles either do not form or do not proceed to further maturation stages.19, 20 Interestingly, germ cells of neither Foxl2 KO mice model seem to be affected during early folliculogenesis.19, 20 Moreover, immediately after birth, mutant and wild-type mice display a similar number of oocytes, but soon folliculogenesis stops in the former, leading to massive atresia.19, 20

Other genes, including those encoding the signaling molecules Wnt4 and Rspo1, are also crucial for ovarian development.21, 22Wnt4−/− female mice are masculinized,21 and overexpression of WNT4 might be responsible for sex reversal in a 46,XY patient carrying a duplication of 1p31-p35.23 WNT4 and RSPO1 both induce stabilization of beta-catenin, and it is known that ectopic expression of a stable form of beta-catenin in XY gonads can lead to male-to-female sex reversal.24 The antagonism between FGF and Wnt signals during chondrocyte differentiation is well known. Indeed, Fgf promotes Sox9 expression and Wnt signaling leads to its degradation (25 and references therein). This antagonistic behavior is also conserved in the gonad.9 An integrated picture of the molecular interactions during gonadal determination/differentiation is displayed in Fig. 2.

Figure 2.

An integrated yet highly simplified picture of molecular interactions during gonadal determination/differentiation. Solid arrows represent well-documented interactions. Dashed arrows represent putative or suggested interactions. Is there a yet-to-be-discovered master regulator that orchestrates ovarian development? Known or putative feedback loops are represented.

Recent in-depth analyses of animal models have shown that granulosa cells from Foxl2/Wnt4 (or to a lesser extent Foxl2/c-Kit) double mutants acquire Sertoli-like characteristics, including strong expression of Sox9, Dmrt1, and other genes of testis determination/differentiation even before birth.2, 26 Interestingly, Foxl2 single mutants do not show this molecular phenotype that early. However, it is known that Foxl2 ablation up-regulates Sox9 expression postnatally.27 According to these studies, the transcriptional dysregulation of a critical subset of genes suggests that the female sex determination/differentiation program is impaired in double-mutant embryonic ovaries. This would predispose to a “molecular sex reversal.”

FOXL2: A key factor in maintaining a differentiated state

Uhlenhaut et al.1 showed very recently that Foxl2 is required to prevent transdifferentiation of follicle cells in the adult ovary to “testis-like” cells. Specifically, inducible deletion of Foxl2 in adult ovarian follicles led to histological changes that affected most of the gonad. Notably, the typical follicular structure of the ovary took the appearance of the seminiferous tubules of the testis (3 weeks after induction of Foxl2 deletion). In most tubule-like structures, oocytes seemed to be lost while granulosa cells acquired morphological characteristics of Sertoli cells. For instance, they displayed a thick basal lamina and tripartite nucleoli. In order to gain molecular insights, Uhlenhaut et al. performed transcriptomic analyses of wild-type ovaries, somatically sex-reversed ovaries and testes. These molecular studies showed that a panoply of genes were up-regulated in sex-reversed gonads, including testicular somatic cell markers such as Sox9, Dax1, Dhh, Dmrt1, and Hsd17b3. It is worth recalling that during embryonic development, two factors must be knocked out to robustly induce the bipotential gonad to commit to male rather than female fate. Strikingly, the loss of Foxl2 alone seems to be sufficient to begin converting a fully differentiated ovary to a testis in the adult.

A kinetic study of this transdifferentiation process showed that 2 days after the induction of Foxl2 ablation the corresponding protein was still detectable by immunostaining, a day later Foxl2 was absent from the XX gonads, and Sox9 immunoreactivity appeared 4 days after induction of Foxl2 deletion. The period of 1 day in which both Foxl2 and Sox9 were undetectable by immunohistochemistry suggests that they are mutually exclusive. This is in keeping with the expression of both genes in the gonads of patients with disorders of sex development (DSD). Notably, in some instances, FOXL2 is found within rather well-developed seminiferous tubules, but it is never strongly co-expressed with SOX9 in the same cell.28 The mutual exclusion between SOX9 and FOXL2 is reminiscent of a toggle switch in which two master genes inhibit each other, directly or indirectly, leading to two alternative states/fates (Fig. 3).

Figure 3.

The interactions between Foxl2 and Sox9 represented as a toggle switch. Expression of one of them (upon a triggering signal) leads to silencing of the other. This process is enhanced by positive feedback loops. Direct repression of Sox9 by Foxl2 is now clear. The contrary remains to be shown but is highly predictable. It is possible that within the period when the switch is operating, the levels of both factors fall below the detection threshold, which would be compatible with the observations published.1 Regulatory disruption at several points of the network is likely to lead/sensitize to sex reversal.

The authors also deleted Foxl2 specifically in the oocyte.1 Homozygous mutant mice were fully fertile, demonstrating that Foxl2 does not play a role in oogenesis. Next, they induced diphtheria-toxin-mediated oocyte ablation. Interestingly, ovaries from 8-week-old animals displayed a complete absence of oocytes but Foxl2 (and not Sox9) was still expressed. This and other experiments led the authors to propose that granulosa cells become reprogrammed into Sertoli-like cells in XX Foxl2−/− gonads in a cell-autonomous way (independently of any oocyte influence). However, an interesting feature of the transdifferentiation process is that Sox9 expression started stochastically and was first detected in mural but not in cumulus granulosa cells. This suggests that the latter cells, which are in direct contact with the oocyte, potentially receive signals enhancing their differentiated state. However, these signals seem not to be strong enough in the presence of a massive reprogramming of both granulosa and theca cells (into Sertoli-like and Leydig-like cells, respectively). These results suggest that oocytes are not required to maintain somatic cell identity in the ovary. However, it is clear from a previous body of work that “the mammalian oocyte orchestrates the rate of ovarian follicular development”.29 It is possible to reconcile the finding of Uhlenhaut et al.1 with previous ones (29, 30 and references therein) by proposing that the oocytes might help to determine ovarian somatic cells, but once this step is achieved, the fate of the latter become less dependent on signals from the former. If this is so, the molecule(s) involved in helping in the granulosa cell determination process is (are) yet to be discovered. This is compatible with the phenotype of mouse mutants in which there is no colonization of the gonad by germ cells. This leads to gonads blocked in an apparently undifferentiated state (31 and references therein). Moreover, recall that absence or dysfunction of SRY in an XY background generally leads to a streak gonad or to an impaired ovary,13, 32 which argues for the existence of autonomous signal(s) in the somatic cells but also of signals resulting from a dialog with the gonocytes that would help in the making of a functional ovary.

A snapshot of the molecular events underlying ovarian maintenance

The rapid up-regulation of Sox9 expression in Foxl2-ablated gonads suggests a direct transcriptional repression of Sox9 by Foxl2 in the ovary. An obvious cis-regulatory region to mediate repression is the TESCO. Not surprisingly, TESCO is activated after induced Foxl2 deletion. TESCO direct recognition by Foxl2 was confirmed by chromatin immunoprecipitation.1In vivo evidence was further supported by in vitro results showing that Foxl2 was able to attenuate TESCO activation by Sf1, Sry/Sf1, and Sox9/Sf1. This is in keeping with the fact that Foxl2 overexpression in XY Sertoli cells in transgenic mice leads to seminiferous tubule disorganization and to the development of ovotestis-like gonads.26

Sertoli-like cells expressing Sox9 in postnatal ovaries appear in XX mice double mutant for the estrogen receptor genes (Esr1/Esr233). This, and the fact that FOX factors are able to interact with ESRs,34 prompted the authors to study a potential interaction between Esr (or ER) and Foxl2 in the repression of TESCO. In vitro experiments showed that Esr1 alone was not able to repress the TESCO element but synergized with Foxl2 to induce a significant repression. This is further supported by the fact that Esr1 or Esr2 and Foxl2 are able to interact in vitro and by in vivo data involving conditional double-mutant animals.

Interestingly, FOXL2 recognizes a response element, the FLRE (FoxL2 response element), whose consensus is similar to a half-site for ESR1/2.35, 36 The FLRE seems to be rather specific for FOXL2, and several other FOX factors fail to recognize it.35 Since FOXL2 is essential for ovarian development and testis repression, strong target specificity achieved through the FLRE at least in the context of gonadal determination or differentiation is not unexpected to prevent unscheduled ovarian differentiation in XY individuals (i.e., triggered by another FOX factor). This is in agreement with the observation that deletion of the standard forkhead consensus sites in the TESCO does not abolish Foxl2/Esr1 action. However, mutation of both the classical forkhead and ER sites abolishes repression by Foxl2/Esr1.

Estrogens are key players of ovarian differentiation. They are able to induce male-to-female sex reversal in fish, reptiles, and birds, while counteracting their production (by using, for instance, aromatase inhibitors) leads to female-to-male reversal (37 and references therein). The latter fact might now be explained, at least in part, by a defect in the synergy between FOXL2 and ER to control and maintain granulosa cell identity.

An early estrogenic dependence of ovarian differentiation is not obvious in most eutherian mammals. For instance, in the mouse, there is no ovarian steroidogenesis documented during fetal development. However, the goat is able to produce estrogens from the earliest stages of ovarian development.38 We have shown that aromatase, the key enzyme of estrogen synthesis, is positively regulated by FoxL2 in the fetal goat ovary, which suggests that in PIS animals, estrogens might be low, consequently to the absence/decrease of FoxL2.

According to the results of Uhlenhaut et al.,1 the bottom line is that down-regulation of FoxL2 might be the basis of XX maleness in PIS goats. This lends credence to the equation FOXL2 = Z. However, to further substantiate this idea, it would be interesting to assess whether FoxL2 expression in the testis is repressed directly or indirectly by SRY or SOX9. If this proves to be so, then FoxL2 becomes, in formal terms, the best candidate for Z. Interspecific phenotypic differences due to FoxL2 mutation/misregulation could result from different timings during sex determination and interaction with different partners/pathways (i.e., ER in goat and Rspo1/Wnt4 in the mouse?). As proposed by Uhlenhaut et al., since the gonadal PIS phenotype appears comparatively earlier than that of XX Foxl2−/− mice, this would suggest that the ER pathway may have “replaced” the Rspo1/Wnt/beta-catenin pathway as the predominant anti-testis/pro-ovary mechanism during fetal development in goats (and perhaps in other mammals where estrogen synthesis takes place during gonadal development).

Before closing this section, it is tempting to propose the existence of a positive feedback loop involving FOXL2 that increases aromatase expression, whose product will activate ER, which might enhance FOXL2 expression and activity. Although published data are compatible with this idea, it remains to be experimentally shown.37

Stability and plasticity of a differentiated state

Targeted deletion of either Foxl2, Wnt4, or Rspo1 tends to “masculinize” the XX embryonic gonads without overt sex reversal. In contrast, robust testis development is elicited in XX mouse embryos without both Wnt4 and Foxl2.21 This has been interpreted as suggesting that Wnt4 (Rspo1) and Foxl2 operate rather independently to allow normal ovarian development. Thus, unlike the situation in males, in which a single gene (either Sry or Sox9) imposes the male fate, activation of at least two different pathways would be required for robust murine ovarian development.39 Although not impossible, it is difficult to conceive the existence of two (or more) completely independent pathways governing ovarian development (i.e., forming a simple “AND” gate, in computer terms). To ensure a coordinated ovarian development, these pathways should either respond to a common signal (yet to be discovered) or cross-talk at some point.

Assuming the existence of a master regulator, if all the signals/pathways mentioned above (Wnt4, Rspo1, and Foxl2) acted as activators, they would form a coherent feed-forward loop. This mechanism is capable of efficiently sorting out true signals from noise, and all the inputs are required to robustly produce a response. At present it is not clear whether Wnt4 and Foxl2 signals are convergent or complementary.1, 2, 39 Thus, if, on the contrary, some of these signals act as activators, and some as repressors, the system would form an incoherent feed-forward loop, a device for driving an outcome that depends on the fold change in the input signal, and not on its absolute level.40, 41 Further studies are required on this whole subject.

After gonadal determination, this fate has to be maintained and might involve positive feedback loops of critical fate-determining genes. This is obviously the case for Sox9 directly using the TESCO element (in synergy with Sf1, see Uhlenhaut et al.1) or via Ffg9 + PDG2 signaling, and is potentially the case for FOXL2 since the goat promoter is at least in vitro stimulated by FOXL2 itself.42 Downstream of this feedback loop, we can consider several ways of maintaining a differentiated state: (i) by expressing a series of activators responding to the master regulator that positively define the set of genes to be expressed, (ii) by expressing a series of repressors that define by exclusion the repertoire of expressed genes, and, (iii) more likely, a combination of both. Whatever the case, the work by Uhlenhaut et al.1 shows that “terminally” differentiated follicle cells can transdifferentiate into Sertoli/Leydig cells when FOXL2 is lacking.

The already mentioned absence of both Foxl2 and Sox9 for one day during the transdifferentiation process raises an interesting question: does transdifferentiation require a dedifferentiation step leading to a bipotential state followed by redifferentiation? It would be interesting to explore the transcriptome (and, more interestingly, the proteome) of follicular cells in the period when neither Foxl2 nor Sox9 are expressed, in order to assess whether these cells are temporarily “bifunctional,” as proposed in cases of transdifferentiation in the context of pituitary cell lineages.43–45 This would still be compatible with the idea of toggle switch, because a proteomic switch should take some time (i.e., synthesis of new proteins and decay of old ones). Transdifferentiation has been proposed to occur in the pituitary.44 For instance, induced hypothyroidism in adult rats leads to a reduced number of somatotroph cells that are gradually “replaced” by thyrotrophs, which share the same precursor. Co-expression of GH and TSH in the same cell and a reduction of GH signal in somatotrophs suggest that the increased number of TSH cells results, at least in part, from the transdifferentiation of somatotrophs.45 Conversion of lactotrophs into somatotrophs (which also share the same precursors) has also been reported.43

The rapidity of the transdifferentiation process of follicle cells is also worth noting. The current paradigm to explain why one genome gives rise to a wide variety of cells and tissues is based on facultative heterochromatinization. This state of the chromatin is often associated with DNA methylation and specific histone modifications. Removing methyl groups from DNA requires in most cases replication. In the present case the process seems to be much more dynamic than passive dilution of repressive DNA marks. Such a dynamic behavior might result from a toggle-like switch governing the recruitment of histone-modifying/demodifying enzymes able to rapidly change chromatin marks and, hence, conformation (Fig. 3).

Conclusions and open questions

The reports by Uhlenhaut et al.1 and Garcia-Ortiz et al.2 clearly show that early ovarian development and maintenance do not rely on a passive pathway. Indeed, the ovary is permanently struggling against transdifferentiation. Is the reverse also true? Is this type of battle of the sexes also taking place in the testis?

It is becoming increasingly clear that despite its early sex-specific expression pattern, Foxl2 seems to play a more prominent role in postnatal repression of the male pathway than in the initial commitment period during development. The fact that Foxl2 single mutants do not sex reverse embryonically makes room for some speculations. Does the block in granulosa cell differentiation prevent the dedifferentiation/masculinization process? If so, does this predict that in a conditional postnatal Foxl2 mutant, the granulosa cells of primordial follicles (which have not undergone the squamous to cuboidal transition) would not undergo transition into Sertoli-like cells? Or would this be prevented by the close proximity of the oocyte? Would the results of a conditional mutant differ from those previously obtained?27

The two major forms of the estrogen receptor, Esr1 (ERalpha) and Esr2 (ERbeta), are expressed in the mouse ovary. However, Esr2/ERbeta is predominantly expressed in granulosa cells (and ERalpha in theca cells46). Therefore, it would be interesting to study the interactions between Esr2 and Foxl2 on TESCO in the same way as reported for Esr1. One interesting molecular aspect yet to be defined concerns the nature of the Esr-Foxl2 binding sites. Do both Foxl2 and Esr bind DNA? If so, through which sites?

Both Uhlenhaut et al.1 and Garcia-Ortiz et al.2 claim, on justified grounds, that there is a masculinization of the transcriptomes of the Foxl2-depleted cells. However, a quick look at transcriptomic data of Foxl2-depleted ovaries1 shows that they still “remember” that they used to be ovaries. Indeed, the correlation between the transcriptomic profiles of Foxl2-depleted and wild-type ovaries is fairly strong and significant (R = 0.5, p ≪ 0.001). On the contrary, the correlation with the testis transcriptome is much lower and non-significant (R = 0.1, based on the data from the table PIIS0092867409014330.mmc2.xls). This behavior probably comes from the confounding effects of the transcriptomes of other cell type in the tissue analyzed. Thus, it would be interesting to explore the transcriptome of purified supporting cells in order to gather statistical evidence for their interesting findings.

Concerning the seemingly autonomous transdifferentiation process of granulosa cells following oocyte loss, which can take weeks,47 it would be interesting to continue the observation for a longer time after treatment with diphtheria toxin to further substantiate the role of the oocyte in this process.

Finally, and given the seeming independence of Wnt and Foxl2 pathways at early time points, it would be interesting to conditionally alter the beta-catenin pathway in adult ovaries to explore whether follicular cells transdifferentiate into testis-like ones or not. This would test the idea that Wnt signaling, known to be of great importance for repressing aspects of testis differentiation in the XX embryo, is, like Foxl2, also required to repress testis-specific gene expression and morphology in the adult.


The author thanks two anonymous referees for their very helpful comments. The author is grateful to B. Benayoun, M. Pannetier, E. Pailhoux, M. Penrad, and C. Guigon for their insights on the manuscript. R.A.V. and his lab is supported by University Paris Diderot-Paris 7, the Centre National de la Recherche Scientifique (CNRS), the Institut Universitaire de France (IUF), and the Association pour la Recherche contre le Cancer.