Antagonistic regulation of Cyp26b1 by transcription factors SOX9/SF1 and FOXL2 during gonadal development in mice

Abstract

Sex determination in fetal germ cells depends on a balance between exposure to retinoic acid (RA) and the degradation of RA achieved by the testis-specific expression of the catabolic cytochrome P450 enzyme, CYP26B1. Therefore, identification of factors regulating the expression of the Cyp26b1 gene is an important goal in reproductive biology. We used in situ hybridization to demonstrate that Cyp26b1 and transcription factor genes steroidogenic factor-1 (Sf1) and Sry-related HMG box 9 (Sox9) are coexpressed in Sertoli cells, whereas Cyp26b1 and Sf1 are coexpressed in Leydig cells in mouse fetal testes. In the mouse gonadal somatic cell line TM3, transfection of constructs expressing SOX9 and SF1 activated Cyp26b1 expression, independently of the positive regulator RA. In embryonic gonads deficient in SOX9 or SF1, Cyp26b1 expression was decreased relative to wild-type (WT) controls, as measured by quantitative RT-PCR (qRT-PCR). Furthermore, qRT-PCR showed that Cyp26b1 up-regulation by SOX9/SF1 was attenuated by the ovarian transcription factor Forkhead box L2 (FOXL2) in TM3 cells, whereas in Foxl2-null mice, Cyp26b1 expression in XX gonads was increased ∼20-fold relative to WT controls. These data support the hypothesis that SOX9 and SF1 ensure the male fate of germ cells by up-regulating Cyp26b1 and that FOXL2 acts to antagonize Cyp26b1 expression in ovaries.—Kashimada, K., Svingen, T., Feng, C.-W., Pelosi, E., Bagheri-Fam, S., Harley, V. R., Schlessinger, D., Bowles, J., Koopman, P. Antagonistic regulation of Cyp26b1 by transcription factors SOX9/SF1 and FOXL2 during gonadal development in mice.

Sexual differentiation in mammals is a unique process in that two completely different organs, testes and ovaries, arise from a common precursor, the bipotential genital ridge. The Y-linked gene Sry is the master switch of mammalian male determination (1), and its major role is to up-regulate an autosomal but related gene, Sry-related HMG box 9 (Sox9; ref. 2). SOX9 directly up-regulates a number of male-specific genes, such as Amh, Pgds, Vnn1, and Sox9 itself, in conjunction with the transcription factor steroidogenic factor-1 (SF1; also known as NR5a1 and Ad4BP), during Sertoli cell differentiation (2–5). Sertoli cells orchestrate testis development, including the differentiation of androgen-producing Leydig cells, testis vascular cells, and other interstitial cells. In the absence of Sry, female genes, such as Wnt4 and Forkhead box L2 (Foxl2), are up-regulated by default, leading to ovarian development (6, 7).

As the progenitors of the gametes, germ cells also undergo sexual differentiation in mammalian gonads. The earliest known sex-specific event in germ cell development is entry into meiosis in ovaries and not in testes. Approximately 13.5 days post coitum (dpc), germ cells stop proliferating and enter the prophase of meiotic division in mouse fetal ovaries, whereas, in testes, germ cells arrest in G₀ or G₁ of the mitotic cycle, resuming mitosis after birth (8, 9). Studies using XX↔XY chimeric mice showed that both XX and XY germ cells enter meiosis in developing ovaries (10), suggesting that sexual differentiation of bipotential germ cells is not cell-autonomous but instead that the tissue environment determines the initial sex differentiation of germ cells (11, 12).

Recent studies identified retinoic acid (RA) as an environmental factor that controls germ cell meiosis (13, 14). RA is a metabolite of vitamin A that mediates many physiological functions, including embryogenesis and organogenesis. RA binds to nuclear receptors, particularly retinoic acid receptors (RARs), which form heterodimers with retinoid X receptors (RXRs). The RAR/RXR complex binds to retinoic acid response elements in the regulatory regions of target genes (15, 16). It is believed that fetal gonads are exposed to RA that is synthesized in the adjacent mesonephros in a non-sex-specific manner. In the gonad, the concentration of RA is regulated by a P450 cytochrome enzyme, CYP26B1, that catabolizes all-trans-RA into inactive oxidized metabolites (13, 17, 18). Cyp26b1 is initially expressed in fetal testes and ovaries and then becomes male-specifically expressed at 11.5–12.5 dpc (13). In developing testes, germ cells are prevented from entering meiosis by the presence of the RA-degrading enzyme, CYP26B1. Hence, CYP26B1 plays a key role in determining whether or not germ cells enter meiosis by controlling local distribution of RA (13, 14).

Despite its important role in the control of germ cell sexual differentiation, the mechanisms of Cyp26b1 gene regulation are not known. Because Cyp26b1 is up-regulated in developing testes as early as 11.5 dpc and is expressed by both Sertoli and interstitial cells (13), we hypothesized that Cyp26b1 is regulated by the transcription factors SF1 and/or SOX9, both being expressed during the early stages of gonadal sex differentiation.

We report here that the spatial and temporal expression patterns of Cyp26b1 and Sf1 or Sox9 overlap during testicular development in vivo. Further, we find that in TM3 cells, a model of fetal testicular somatic cells, exogenous Sox9 or Sf1 significantly activated Cyp26b1 expression. Cyp26b1 regulation by SOX9 and SF1 did not require RA, another positive regulator of Cyp26b1. In AMH-Cre:Sox9^flox/flox mice (19), which lack Sox9 expression, or in Cited2-null mice, a mouse model with impaired Sf1 expression (20–22), Cyp26b1 expression was significantly decreased at 13.5 and 12.25 dpc, respectively. Furthermore, we find that one of the female sex-determining factors, FOXL2, attenuates Cyp26b1 up-regulation by SOX9 and SF1 in TM3 cells and that Cyp26b1 expression was significantly increased in XX gonads of Foxl2-null mice. Our results indicate that the testicular transcription factors SOX9 and SF1 and the ovarian transcription factor FOXL2 have mutually antagonistic actions that ensure correct sexual differentiation of fetal germ cells through regulation of Cyp26b1 expression.

MATERIALS AND METHODS

Animals

Protocols and use of animals were approved by the Animal Welfare Unit of the University of Queensland. Mouse embryos were collected from timed matings of the Swiss Quackenbush and CD1 outbred strain, with noon of the day on which the mating plug was observed designated as 0.5 dpc. Embryos were sexed at 10.5–11.5 dpc using an X-linked GFP marker (23) and at 12.5–13.5 dpc by gonadal morphology. AMH-Cre:Sox9^flox/flox mice were obtained as described previously (19). In brief, Sox9^flox/flox mice produced by homologous recombination in embryonic stem cells were bred to AMH-Cre transgenic mice on a C57BL/6 background, and the resulting AMH-Cre:Sox9^flox/+ offspring were backcrossed to Sox9^flox/flox mice to obtain AMH-Cre:Sox9^flox/flox embryos. The Cited2-knockout mice on a C57BL/6 background have been described previously (24), and embryo sex was determined by UBEX1 PCR analysis (25). The generation of Foxl2-null mice was reported previously (26), and these mice were maintained on a mixed C57B6/J/129/SVJ genetic background.

Transfection into TM3 cells

The murine testicular somatic cell line TM3 was obtained from American Type Culture Collection (Manassas, VA, USA) and cultured in DMEM (Invitrogen, Carlsbad, CA, USA) with 10% FBS (AusGenex, Loganholme, QLD, Australia) at 37°C in 5% CO₂. Cells were plated at 2.5 × 10⁵/well in 6-well plates 12 h before transfection. Cells were transfected with 0.5, 1, or 2 μg of expression vector (pSG.Sox9, ref. 27; pcDNA.Sf1, ref. 28; pcDNA.HA-Sry, ref. 29; or pcDNA.Foxl2, ref. 30) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. The total amount of transfected plasmid was standardized using the empty expression vector as control. All-trans-RA was purchased from Sigma-Aldrich (R2625; Sigma-Aldrich Corp., St. Louis, MO, USA). RAR antagonist AGN193109 was provided by Vitae Pharmaceuticals (Fort Washington, PA, USA).

Real-time quantitative RT-PCR (qRT-PCR)

qRT-PCR analysis of TM3 cells and mouse gonads (except Foxl2-null mice) was conducted as follows. Total RNA was collected at 48 h post-transfection, except for experiments depicted in Fig. 2A. For tissue analysis, embryonic gonads without mesonephroi were dissected in ice-cold PBS at the appropriate stages. Total RNA from cells and tissue was isolated using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA) or Micro Kit (Qiagen) including DNase treatment. Total RNA (500 ng for TM3 or 300 ng for gonads) was used as a template for synthesis of cDNA using SuperScript III (Invitrogen) and random primers (Invitrogen), according to the manufacturer's instructions. cDNA samples were diluted 1:4, and 1 μl was used in each 25 μl of qRT-PCR reaction, containing SYBR Green PCR Master Mix (Applied Biosystems, Foster City, Ca, USA). Transcript levels were analyzed on an ABI Prism 7500 Sequence Detector System (Applied Biosystems) over 40 cycles of 95°C for 15 s and 60°C for 1 min in a 2-step thermal cycle, preceded by an initial 10-min step at 95°C. Ribosomal protein S29 (Rps29) served as the normalizing gene to standardize qRT-PCR data (31). For analysis of Foxl2-null mice, we collected gonadal samples at 13.5 dpc, and total RNA was obtained from dissected gonads by enzymatic extraction (MELT system; Ambion, Austin, TX, USA) followed by linear RNA amplification using Ovation Pico (NuGEN, San Carlos, CA, USA). qRT-PCR (TaqMan; Applied Biosystems) was performed using an ABI 7900HT system (Applied Biosystems). Sdha was used as the housekeeping gene to standardize the data as reported previously (32), and the relative expression levels to XX wild-type (WT) mice were calculated. All qRT-PCR primers are listed in Table 1. The means ± se of 3 biological replicates measured in triplicate were calculated. We used an unpaired Student's t test to demonstrate statistical significance of differences between the control and the given sample.

Figure 2.

SOX9 and SF1 up-regulate Cyp26b1 expression in TM3 cells. A) TM3 cells transfected with constructs expressing Sox9 (1 μg) or Sf1 (1 μg) show induction of endogenous Cyp26b1 expression in a time-dependent manner. Cyp26b1 transcript levels were measured at 0, 6, 12, 24, and 48 h post-transfection. B–D) Cyp26b1 mRNA levels in TM3 cells transfected with constructs expressing Sox9 (B; 0, 0.5, 1.0, and 2.0 μg), HA-Sry (B; 1 and 2 μg), Sf1 (C; 0, 0.5, 1.0, and 2.0 μg), or Sox9 and/or Sf1 (D; 0, 0.5, and 1.0 μg). Cyp26b1 responded additively to SOX9 and SFI, but not to SRY. All data sets represent qRT-PCR analysis of Cyp26b1 mRNA expression relative to Rn18s (means±se of 3 biologically independent experiments performed in triplicate). *P < 0.05, **P < 0.01 vs. control (CTRL); unpaired Student's t test.

Table 1.

Primers used for qRT-PCR analysis

Gene	Description	Forward primer	Reverse primer
SYBR qRT-PCR
Rn18s	18S rRNA	GATCCATTGGAGGGCAAGTCT	CCAAGATCCAACTACGAGCTTTTT
Rps29	Ribosomal protein S29	TGAAGGCAAGATGGGTCAC	GCACATGTTCAGCCCGTATT
Cyp26b1	Cytochrome P450 26B1	TGGACTGTGTCATCAAGGAGGT	GTCGTGAGTGTCTCGGATGCTA
Sox9	Sry-related HMG box 9	AGTACCCGCATCTGCACAAC	TACTTGTAATCGGGGTGGTCT
Sf1	Steroidogenic factor 1	TCCAGTACGGCAAGGAAGA	CCACTGTGCTCAAGCTCCAC
Foxl2	Forkhead box L2	GCTACCCCGAGCCCGAAGAC	GTGTTGTCCCGCCTCCCTTG
TaqMan qRT-PCR
Sdha	Succinate dehydrogenase complex, subunit A, flavoprotein	Mm01352366_m1
Cyp26b1	Cytochrome P450 26b1	Mm00558507_m1
Sox9	Sry-related HMG box 9	Mm00448840_m1
Sf1 (Nr5a1)	Steroidogenic factor 1	Mm00496060_m1

In situ hybridization and immunohistochemistry

In situ hybridization was performed as described previously (33). In brief, sequential sections of paraformaldehyde-fixed, paraffin-embedded embryos were dewaxed, rehydrated, incubated in proteinase K, refixed with 4% paraformaldehyde, acetylated, and prehybridized. Hybridization was performed overnight at 60°C. After 2 h of blocking, anti-digoxigenin antibody (Roche Diagnostics, Indianapolis, IN, USA) at 1:2000 in blocking solution was added, and sections were incubated overnight at 4°C. After washing, sections were equilibrated in NTM buffer and incubated in color solution.

Immunofluorescence of a 7-μm paraffin section was performed as described previously (33) using 12.5 dpc mouse fetus samples fixed in 4% paraformaldehyde. The primary antibody used for this study was rabbit anti-SF1 (1:1000, kindly provided from Dr. Ken-Ichirou Morohashi, Kyushu University, Fukuoka, Japan). The secondary antibody (goat anti-rabbit Alexa Fluor 488 or goat anti-rabbit Alexa Fluor 594) was obtained from Molecular Probes (Invitrogen) and used at 1:200 dilution.

RESULTS

Cyp26b1, Sf1, and Sox9 have overlapping expression profiles in fetal testes

First, to clarify the relationship between Cyp26b1 and Sox9/Sf1, we performed a time-course expression analysis from 10.5 to 13.5 dpc using qRT-PCR analysis. Sox9, Sf1, and Cyp26b1 were expressed in a testis-specific manner (Fig. 1A). In particular, Cyp26b1 and Sox9 had very similar temporal expression profiles: Cyp26b1 started to be expressed in male genital ridges at 11.5 dpc and reached a peak level at 12.5–13.5 dpc, as did Sox9 (Fig. 1A). Sf1 was expressed more strongly in testes than ovaries by 11.5 dpc, and expression continued to increase until 13.5 dpc in testes (Fig. 1A).

Figure 1.

Cyp26b1, Sf1, and Sox9 display overlapping expression profiles in fetal testes. A) Temporal expression profile of Sf1, Sox9, and Cyp26b1 during mouse gonadal development by qRT-PCR analysis. From 10.5 dpc, all three genes are up-regulated in fetal testis compared with fetal ovaries. Data sets represent mRNA expression relative to Rps29 (means±se of 3 biologically independent experiments performed in triplicate). Red and blue lines indicate XX and XY gonads, respectively. B) Spatial expression profiles of Sox9, Sf1, and Cyp26b1 transcripts in fetal testes by in situ hybridization. Serial sagittal sections of 13.5 dpc testes show that Cyp26b1 expression overlaps with that of the Sertoli cell marker Sox9 and with the Sertoli/Leydig cell marker Sf1. Arrows indicate interstitial cells expressing both Sf1 and Cyp26b1. Scale bar = 50 μm.

SOX9 and SF1 are transcription factors and, if both are involved in Cyp26b1 regulation, spatial expression patterns of Sox9/Sf1 and Cyp26b1 should overlap. Therefore, we analyzed the expression of each gene in sequential sections using in situ hybridization. Sox9, Sf1, and Cyp26b1 are known to be expressed in Sertoli cells (13, 34–36). Thus, expression of all three genes was detected in irregular-shaped cells within the testis cords, consistent with the features of Sertoli cells and distinct from the large, round germ cells also found within testis cords (Fig. 1B). In addition, Sf1 is known to be expressed in Leydig cells residing in the interstitial space; in our analyses, both Sf1 and Cyp26b1 were expressed in cells with a similar number and distribution in the interstitium, consistent with coexpression in Leydig cells (Fig. 1B, arrows). These findings support the hypothesis that SOX9 and SF1 are together involved in Cyp26b1 regulation in Sertoli cells and that SF1 might regulate Cyp26b1 in Leydig cells, during testicular development.

SOX9 and SF1 up-regulate endogenous Cyp26b1 expression in TM3 cells

To investigate potential functional involvement of SOX9 and SF1 in Cyp26b1 regulation, we performed in vitro experiments using the mouse gonadal cell line TM3 (29, 37). These cells express several genes characteristic of Sertoli cells, as assayed nonquantitatively by RT-PCR (29, 37). However, SOX9 protein expression was reported as undetectable by immunofluorescence in these cells (29), and using qRT-PCR we found that the expression levels of endogenous Sox9, Sf1, and Cyp26b1 were extremely low compared with those in fetal testes (Supplemental Fig. S1). Therefore, any effects on Cyp26b1 regulation in this cell culture system must reflect the effects of transfected rather than endogenous SOX9/SF1.

First, we investigated the time course of Cyp26b1 expression after transfection with Sox9 or Sf1 expression constructs. Cyp26b1 expression started to increase 12 h after Sox9 or Sf1 transfection and continued to increase until at least 48 h post-transfection (Fig. 2A). On the basis of these data, we used 48 h after transfection as our standard time point for further analyses.

We next introduced varying amounts of Sox9 or Sf1 expression vector into TM3 cells. Exogenous Sox9 and Sf1 expression induced Cyp26b1 expression in a dose-dependent manner (Fig. 2B, C). In contrast, Sry did not up-regulate Cyp26b1 expression, suggesting that Sry is not directly involved in Cyp26b1 regulation (Fig. 2B). It has been reported that SOX9 and SF1 target genes, such as Sox9 itself, Vnn1, and Amh, are synergistically up-regulated by SOX9 and SF1 (2, 4). To investigate a potential cooperative action of SOX9 and SF1 in Cyp26b1 regulation, we introduced both Sox9 and Sf1 expression vector into TM3 cells. Individual introduction of 1 μg of Sox9 or Sf1 expression vector increased Cyp26b1 expression by ∼6 and ∼10-fold, respectively (Fig. 2D). Simultaneous introduction of both Sox9 and Sf1 increased Cyp26b1 expression by ∼18-fold, indicating that SOX9 and SF1 regulate Cyp26b1 expression additively in TM3 cells (Fig. 2D), consistent with the similar temporal profile of Cyp26b1 transcriptional response to either SOX9 or SF1 (Fig. 2A).

Up-regulation of Cyp26b1 by SOX9 or SF1 does not require RA action

It has been reported that cytosolic class-1 aldehyde dehydrogenase (Aldh1a1) is expressed in a male-specific manner in fetal gonads and that Aldh1a1 is genetically downstream of Sox9 (38). Previous studies involving mRNA injection into Xenopus embryos have established that ALDH1A1 can catalyze RA synthesis (39) and RA has been shown to positively regulate Cyp26 genes, including Cyp26b1 (40, 41). In agreement with published results, we found that exogenous RA strongly up-regulated Cyp26b1 expression in TM3 cells (Fig. 3A). This up-regulation was abolished by the synthetic antagonist AGN193109, which suppresses RA action by binding to RARs but not RXRs (42), confirming that Cyp26b1 up-regulation by RA is mediated by RARs (Fig. 3A).

Figure 3.

RA is not required for Cyp26b1 up-regulation by SOX9 or SF1. A) qRT-PCR analysis of Cyp26b1 mRNA levels in TM3 cells treated with RA (0.01, 0.1, and 1.0 μM), RA (1.0 μM) together with the RAR antagonist AGN193109 (RAR ant, 5μM), or AGN193109 alone (5 μM). RA stimulated Cyp26b1 expression, whereas AGN193109 counteracted this effect. B) qRT-PCR analysis of Sf1 and Sox9 mRNA levels in TM3 cells treated with RA (0.01, 0.1, and 1.0 μM), RA (1.0 μM) and AGN193109 (5 μM), or AGN193109 alone (5 μM). Neither treatment affected Sf1 or Sox9 expression levels. C) qRT-PCR analysis of Cyp26b1 mRNA levels in TM3 cells transfected with Sox9 or Sf1 expression vector (1 μg each) and/or treated with 5 μM AGN193109. RAR antagonist did not interfere with Cyp26b1 up-regulation stimulated by SOX9 or SF1. Data sets represent mRNA expression relative to Rn18s (A, C) or control (no RA or AGN193109; B); means ± se of 3 biologically independent experiments performed in triplicate. Level of statistical significance: *P < 0.05, **P < 0.01, ***P < 0.001 vs. control (CTRL); unpaired Student's t test.

These observations suggest the possibility that RA synthesized by ALDH1A1 mediates the Cyp26b1 regulation by SOX9 and SF1 observed in our experiments. However, neither exogenous RA nor RAR antagonist affected Sox9 or Sf1 expression levels (Fig. 3B), and RAR antagonist had no effect on the ability of transfected Sox9 and Sf1 expression vectors to up-regulate Cyp26b1 expression (Fig. 3C). Therefore, the effects of SOX9 and SF1 are not mediated by RA action involving RARs and represent an independent mechanism of regulating Cyp26b1 during testis development.

Cyp26b1 expression requires SOX9 and SF1 in vivo

We next sought to determine whether Cyp26b1 is regulated by SOX9 or SF1 in vivo. We have shown previously that Cyp26b1 expression is impaired in Ck19-Cre;Sox9^flox/flox mice (38), in which the Sox9 gene is deleted before the time of sex determination (43). However, in those mice, the XY gonads develop as ovaries, in which expression of all male-specific genes is presumably suppressed. To investigate SOX9 action on Cyp26b1 expression in vivo more specifically, we used AMH-Cre:Sox9^flox/flox mice. In these mice, Sox9 expression is markedly reduced at 13.5 dpc, after the time of sex determination, and absent by 14.5 dpc. In contrast to Ck19-Cre:Sox9^flox/flox mice, the XY gonads of AMH-Cre:Sox9^flox/flox mice are phenotypically testes, maintaining normal expression levels of early testis-specific genes, such as Fgf9 and Sox8 (19). We found that Cyp26b1 expression was significantly reduced in AMH-Cre:Sox9^flox/flox XY gonads at 13.5 dpc (Fig. 4A), supporting our hypothesis that SOX9 is required for positive regulation of Cyp26b1 expression.

Figure 4.

Cyp26b1 expression is significantly reduced in mice with reduced SOX9 or SF1 expression. A) qRT-PCR analysis of Cyp26b1 expression at 13.5 dpc of WT (+/+) XY (n=4), AMH-Cre:Sox9^flox/flox (flox/flox) XY (n=3), and WT XX gonads (n=5). B) qRT-PCR analysis of Cyp26b1 expression at 12.25 dpc of WT XY (n=3), Cited2-knockout (−/−) XY (n=3), and WT XX (n=3) gonads. Data sets represent mRNA expression relative to Rps29 (means±se). **P < 0.01, ***P < 0.001 vs. WT XY; unpaired Student's t test.

In Sf1-knockout mice, regression of the gonads after sex determination precludes analysis of the roles of this gene in gonadal differentiation (44, 45). For this reason, we also studied Cyp26b1 expression in mice deleted for the gene encoding CBP/p300-interacting transactivator with Glu/Asp-rich carboxyl-terminal domain 2 (Cited2; refs. 20–22). CITED2 is a transcriptional coactivator or repressor that cooperates with WT1 to stimulate Sf1 transcription (20); in Cited2-knockout mice, Sf1 expression is temporarily reduced (20–22). The delay in testis development in Cited2-null mice is caused by the temporal loss of Cited2-mediated enhancement of Sf1 levels, rather than a direct effect by loss of Cited2 (20). Therefore, Cited2-knockout mice are a useful proxy with which to study loss of SF1 function during gonad differentiation. We chose 12.25 dpc as the time point of examination because the reduced Sf1 expression recovers by 12.5 dpc in fetal gonads (22). Analysis by qRT-PCR showed that in Cited2-knockout mice, Cyp26b1 expression was significantly reduced at 12.25 dpc in fetal testes (Fig. 4B). The observations from AMH-Cre:Sox9^flox/flox and Cited2-knockout mice strongly suggest that SOX9 and SF1 up-regulate Cyp26b1 expression during testicular development in vivo.

FOXL2 attenuates Cyp26b1 up-regulation by SOX9 or SF1

Unlike the situation in developing testes, germ cells in the developing ovaries require exposure to RA to ensure entry into meiosis and initiation of the oogenic pathway (13, 14). Thus, Cyp26b1 is up-regulated during testis development and down-regulated during ovarian development (ref. 13 and the present study; Fig. 1A). In view of our findings that SOX9 and SF1 can up-regulate Cyp26b1 expression, the lack of Cyp26b1 expression in the ovary might in theory be explained by a lack of SOX9 and SF1 in that tissue. However, we found that SF1 mRNA is expressed at moderate levels in the fetal ovary (Fig. 1A). Moreover, SF1 protein was readily detectable by immunofluorescence in the ovary at 12.5 dpc (Fig. 5A). In view of these observations and because of the critical requirement to suppress CYP26B1 activity in the developing ovary, we reasoned that additional mechanisms might operate in developing ovaries to actively suppress Cyp26b1 transcription, consistent with the emerging concept that balanced opposing signals act to ensure correct testis or ovarian development (46, 47).

Figure 5.

FOXL2 negatively regulates Cyp26b1 expression. A) Immunofluorescence of sagittal sections of paraffin-embedded fetuses of 12.5 dpc showing ovarian expression of SF1 (green). Dotted line encompasses the ovary. Scale bar = 200 μm. B) Time course of Cyp26b1 and Foxl2 gene expression during mouse ovarian development, with maximum expression levels adjusted to 100%. The two genes showed complementary expression. C) qRT-PCR analysis of Cyp26b1 mRNA levels in TM3 cells transfected with Sox9 (1 μg) or Sf1 (1 μg) with or without cotransfection of Foxl2 expression vector (1 μg). FOXL2 suppressed Cyp26b1 up-regulation by both SF1 and SOX9. D) qRT-PCR analysis of Cyp26b1, Sox9, and Sf1 mRNA levels in XX gonads of 13.5 dpc Foxl2-null mice. Relative expression levels to Sdha of each gene were compared with those of XX wild-type controls. Absence of FOXL2 caused strong up-regulation of Cyp26b1. Data sets represent mRNA expression relative to Rps29 (B), Rn18s (C), or Sdha (D); means ± se of 3 biologically independent experiments performed in triplicate. NS, nonsignificant. *P < 0.05, **P < 0.01, ***P < 0.001; unpaired Student's t test.

FOXL2 is a forkhead transcription factor expressed mainly in somatic cells of female gonads and in developing eyelids (48). FOXL2 is considered essential for ovarian folliculogenesis and granulosa cell development (26, 48, 49). Furthermore, recent studies have suggested that FOXL2 antagonizes SOX9-dependent pathways (50) and suppresses SF1 target gene activation (51). Therefore, we tested whether FOXL2 actively suppresses Cyp26b1 induction. First, we compared the temporal expression pattern of Foxl2 and Cyp26b1. In XX gonads, Cyp26b1 expression decreased from 12.5 dpc and, consistent with our hypothesis, coincided with a dramatic up-regulation of Foxl2 expression (Fig. 5B).

To further investigate potential involvement of FOXL2 in Cyp26b1 regulation, we transfected TM3 cells with a Foxl2 expression plasmid together with Sox9, Sf1, or a combination of both expression plasmids. Sox9 and Sf1 expression constructs do not contain binding sites for FOXL2 in their regulatory regions, and, accordingly, expression levels of these constructs were unaffected by the coexpression of FOXL2 (data not shown). However, FOXL2 strongly suppressed endogenous Cyp26b1 expression induced by SOX9, SF1, or both together (Fig. 5C).

Finally, we analyzed Cyp26b1 expression in 13.5 dpc XX gonads of Foxl2-null mice. We found that Sox9 and Sf1 expression levels increased ∼2-fold (Fig. 5D). In contrast, Cyp26b1 expression was increased ∼20-fold relative to that in controls (Fig. 5D). The Cyp26b1 expression levels in Foxl2-knockout ovary were 5-fold lower than those in WT testes (data not shown), as would be expected, because the Foxl2-knockout ovaries lack positive regulators such as SOX9, and consistent with the report that germ cells enter meiosis in these mutant ovaries (49). In these experiments, absence of FOXL2 caused up-regulation of Cyp26b1 to a greater extent than that likely to be caused by up-regulation of Sox9 and/or Sf1. The magnitude of the increase of Cyp26b1 expression in Foxl2-null mice is consistent with relief of active suppression by FOXL2 during ovarian development in vivo, consistent with in vitro data observed in TM3 cells (Fig. 5C).

DISCUSSION

Despite its critical role for germ cell sexual differentiation, the mechanisms of Cyp26b1 regulation during gonadal development are not known. In the present study, we present cell-based gain-of-function data, together with in vivo loss-of-function analyses in knockout mice, implicating the gonadal transcription factors SOX9, SF1, and FOXL2 in regulating Cyp26b1 during gonadal development.

SOX9 was previously suggested to be acting upstream of Cyp26b1 (38). We found that Cyp26b1 was up-regulated in the testicular somatic cell line TM3 after transfection with the Sox9 expression construct and that Cyp26b1 expression was significantly reduced in XY gonads of AMH-Cre:Sox9^flox/flox mice. In contrast to SOX9, SRY did not affect Cyp26b1 expression levels in TM3 cells, indicating that activation of Cyp26b1 transcription is not a general property of SOX transcription factors. Taken together, our data support the conclusion that SOX9 positively regulates Cyp26b1 expression during testicular development.

Previous studies have established that Cyp26b1 is expressed in both Sertoli and interstitial cells (13). However, Sox9 expression is limited to Sertoli cells in fetal testes, indicating that some other molecules expressed in interstitial cells must be involved in Cyp26b1 up-regulation. SF1 was considered a good candidate because Sf1 is expressed not only in Sertoli cells but also in interstitial Leydig cells and is one of the earliest genes expressed in a male-specific manner in fetal gonads (52). We showed that the spatial expression patterns of Sf1 and Cyp26b1 overlapped each other in cells inside and outside testis cords; in TM3 cells, Sf1 transfection robustly up-regulated Cyp26b1 expression; and in Cited2-knockout mice, a mouse model with impaired expression of Sf1, Cyp26b1 expression was significantly reduced. These data implicate SF1 as a second factor involved in the positive regulation of Cyp26b1 expression during testicular development. Furthermore, our data suggest that the previously unidentified interstitial cells expressing Cyp26b1 (13) are Leydig cells. Taken together, our data show that in fetal Sertoli cells, both SOX9 and SF1 contribute to Cyp26b1 regulation, whereas in fetal Leydig cells, Cyp26b1 is regulated by SF1 (Fig. 6).

Figure 6.

Model for the regulation of Cyp26b1 expression and germ cell sexual fate. Distribution of RA in gonads is determined by the RA-degrading enzyme, CYP26B1. Cyp26b1 is expressed in a male-specific manner, and resulting RA levels are high in ovary and low in testes. In testes, Cyp26b1 is up-regulated by SF1 in Leydig cells and by SF1 and SOX9 in Sertoli cells. Although SF1 could be expected to up-regulate Cyp26b1 expression in ovaries, this expression is suppressed by the female-specific transcription factor FOXL2. Thus, male and female pathways compete in the regulation of Cyp26b1.

Considering the rapid up-regulation of Cyp26b1 by SOX9 and SF1 in TM3 cells; the similar temporal expression profile of Cyp26b1, Sox9, and Sf1 in fetal testes; and the fact that this regulation does not require RA, it seems likely that SOX9 and SF1 directly up-regulate Cyp26b1 during testicular development. CYP26B1 is expressed in various organs during fetal development, such as limb, hindbrain, and the first and the second branchial arches (17, 53), indicating that the regulation of the Cyp26b1 gene is complex, and many trans-acting factors and cis-regulatory sequences are likely to be involved. Substantial homology is maintained between the sequence of mouse Cyp26b1 intron-1 and the human CYP26B1 5′ flanking region (data not shown), suggesting that these sequences may be important for Cyp26b1 gene regulation. However, neither 3 kb of 5′ flanking sequence nor intron-1 or intron-2 of mouse Cyp26b1 responded to mSOX9 and mSF1 in our TM3 culture assay (data not shown); the gonad-specific enhancer of Cyp26b1 is therefore presumed to be located some distance from the Cyp26b1 transcription unit itself. Further studies will be necessary to identify the gonad-specific enhancers of Cyp26b1 and to specifically address whether SF1 and/or SOX9 directly bind to these elements.

We found that Sf1 mRNA and protein are expressed in fetal ovaries at 12.5 dpc at higher levels than previously recognized (52); this expression and its functional significance are supported by observations that the ovary regresses after sex determination in Sf1-knockout mice (44). Consistent with previous studies (52), we also found lower Sf1 expression levels at 13.5 dpc. The presence of SF1 in the ovaries suggested that mechanisms might exist in the developing ovary to counter the possible up-regulation of Cyp26b1 in that tissue.

Our data implicate a third regulatory factor, FOXL2, as part of such a mechanism. FOXL2 is a key transcription factor in ovarian development, and the phenotype of mammalian models of Foxl2 deficiency varies from XX sex reversal in goats (54) to ovarian failure in mice and humans (26, 48, 49). Foxl2 is expressed in somatic cells during ovarian development and is involved in the regulation of other female-specific genes, such as Aromatase, Follistatin, and Bmp2 (30, 55, 56). It has been reported that FOXL2 represses SF1-induced Cyp17 transcription by interacting directly with SF1 (51); that observation not only substantiates the concept that FOXL2 can act as a negative transcriptional regulator but also supports our finding that it antagonizes SF1 regulation of Cyp26b1. The absence of FOXL2 in vivo caused up-regulation of Cyp26b1 to a greater extent than that likely to be caused by the observed up-regulation of Sox9 and/or Sf1. Therefore, our findings suggest that down-regulation of Cyp26b1 during ovarian development may occur through a combination of decreasing SF1 levels and up-regulation of Foxl2 expression.

Suppression of Cyp26b1 expression by of FOXL2 may provide a failsafe mechanism to ensure that ovarian germ cells are not exposed to RA. Such a mechanism would support and extend our recent findings that fibroblast growth factor 9 (FGF9) acts directly on germ cells to inhibit meiosis (57). Those findings indicate that two independent and mutually antagonistic pathways involving RA in the ovary and FGF9 in the testis determine mammalian germ cell sexual fate commitment (57). In addition to that model, our present study suggests a further level of antagonism between male and female germ cell pathways, namely suppression of Cyp26b1 expression by FOXL2. A model for this regulatory network is depicted in Fig. 6, providing a framework for further studies of the interaction between somatic and germ cells in fetal gonads and the control mechanisms that underpin development of germ cells.