Expression of the mouse anti-Müllerian hormone gene suggests a role in both male and female sexual differentiation

During mammalian sex determination and differentiation, the gonads and the reproductive tract develop in fundamentally different ways. The genital ridge, which arises as a thickening on the mesonephros, is bipotential, developing as a testis if the Y chromosome is present due to the action of the testis-determining gene on the Y chromosome, Tdy. In the absence of the Y, it follows the default pathway giving rise to an ovary. In contrast the anlagen of the male and female reproductive tracts, the Wolffian and Müllerian ducts respectively, are both present within the mesonephroi prior to sex determination. These anlagen are unipotential and the survival and development of one versus the other depends on the type of gonad that differentiates. In the female, the Woffian duct system degenerates and the Müllerian ducts give rise to the oviducts, uterus and upper vagina. This does not depend on any factors produced by the ovary and so can be considered part of the default pathway (Austin and Edwards, 1981). In the male, therefore, two processes have to occur. The Wolffian ducts have to be maintained and stimulated to differentiate into the male tract and accessory organs, the vas deferens, seminal vesicles and epididymides. This is due to the influence of testosterone produced by Leydig cells in the testis. Additionally, the Müllerian duct system has to regress, and this is due to the action of anti-Müllerian hormone (AMH), also known as Müllerian inhibiting substance (MIS).

AMH is the first product known to be secreted by Sertoli cells in the developing testes (Tran and Josso, 1982). It is a member of the TGF-βfamily of growth factors (Rentrop et al. 1986 for review). The carboxyterminal part of the protein functions as a dimer (Pepinsky et al. 1988) through interaction with a receptor(s) that is as yet undefined. Its effect on Müllerian ducts can be reproduced in vitro with organ cultures of mesonephroi, and these cultures can also serve as a biological assay for the protein (Picard and Josso, 1984).

In addition to its function during Müllerian duct regression, AMH may play a role in testis differentiation itself as first suggested by observations of Freemartinism. Freemartins are masculinised female cattle embryos which are thought to arise due to the passage of AMH from a male twin via placental anastomoses (Jost et al. 1975). Freemartins not only show the absence of Müllerian duct derivatives but also exhibit masculinisation of their ovaries, namely the formation of seminiferous tube-like structures, containing morphologically normal Sertoli cells. This effect can also be reproduced in vitro by exposing rat ovaries to either testis tissue or purified AMH (Vigier et al. 1987; Charpentier and Magre, 1990), or in transgenic mice overproducing human AMH (Behringer et al. 1990). It has been attributed to the action of AMH either directly on the somatic cells of the developing gonad, or indirectly by the elimination of germ cells entering meiosis (McLaren, 1990), since in females germ cells are thought to be necessary to ensure the normal differentiation and maintenance of follicle cells. In the complete absence of germ cells, ovaries fail to develop (Merchant, 1975; Merchant-Larios and Centeno, 1981).

In contrast to the masculinising effects described above, AMH may also have a normal role in females. Its activity can be detected in follicular fluid of the adult ovary, and immunohistochemistry suggests that it is produced by follicle cells (Vigier et al. 1984; Takahashi et al. 1986; Bezard et al. 1987). Its function here is not understood, but it is clear that the timing and dosage of AMH expression in the ovary is crucial, otherwise it could exert a masculinising effect on the female reproductive tract or ovaries themselves.

In order to understand more about AMH gene function in both male and female development, and as a preliminary to study its regulation, it is important to establish the precise timing and location of AMH transcription. We have decided to undertake this analysis using the mouse as a model system to facilitate future genetic manipulation. In this report, we describe the cloning and sequencing of the mouse AMH gene and present a detailed analysis of its embryonic and postnatal expression using in situ hybridisation. We also discuss the timing of AMH expression with respect to cellular events occurring in both male and female gonad development.

Embryos and tissues for RNA isolation and in situ hybridization studies were prepared from Parkes outbred mice and from mice homozygous for the extreme allele of the dominant white spotting mutation (W^e, Mintz and Russell, 1957). The W^e/W^e genotype of 14.5 days post coitum (dpc) embryos was confirmed by liver morphology (Borghese, 1959). The day of vaginal plug was taken to be 0.5 dpc, and the morphology of the hindlimb bud served as a reference for staging the embryo. Embryos at 11.5 dpc were sexed by staining for sex chromatin in amnion cells (Burgoyne et al. 1983).

The partial bovine and human cDNA clones (Picard et al. 1986) were obtained from Nathalie Josso (INSERM, Paris) and the partial rat cDNA clone was obtained from Richard Cate (Biogen, Boston). The mouse genomic libraries were obtained from A. M. Frischauf (ICRF, London) and Keith Willison (Chester Beatty, London).

Genital ridges and male and female gonads were dissected from 11.5 dpc and 12.5 to 17.5 dpc embryos, respectively. Total RNA was isolated as described (Auffray and Rougeon, 1980). Tissues were homogenized in 6 M urea, 3 M LiCl. After overnight incubation at 4°C, the homogenate was centrifuged and the pellet obtained was washed twice in the same solution. The pellet was resuspended in 10nw Tris pH 7.5, 0.5% sodium dodecyl sulphate (SDS) containing 50 μgml^-1 proteinase K and incubated for 30 – 120 min at 37°C. After a number of phenol extractions, total RNA was ethanol precipitated and resuspended in water.

Total RNA (10 μg) was loaded on 1 % agarose-formal-dehyde gels, electrophoresed at 40 mA in 1 × MOPS buffer and transferred to Hybond-N filters as described by Maniatis et al. (1982). The filters were baked at 80°C for 30 min and UV-crosslinked. Probes were labelled with ³²P using a multiprime labelling kit (Amersham) according to the manufacturer’s instructions, denatured by heating, added to the filters in 7% SDS, 0.5 M sodium hydrogen phosphate pH6.5, 1mM EDTA and hybridised for 12 –16h at 65°C. Filters were washed at 65°C in 40 mM NaP₁ pH 6.5, 1 % SDS and exposed to Kodak XAR film for 7 days.

Total RNA was prepared from approximately 500 pairs of testes, collected from 14.5 dpc embryos, giving a yield of 500 μg. Poly(A)⁺RNA was prepared using Hybond mAP paper (Amersham) according to the manufacturer’s instructions. 1 μg poly(A)⁺RNA from 14.5 dpc fetal testes served as starting material for the synthesis of cDNA which was ligated into a bacteriophage lambda gt10 vector, using cDNA synthesis and cloning kits (Amersham). The resulting library contained 1 x 10⁶ recombinant lambda clones of which approximately 200000 were plated in 0.7% top agarose containing 10 mM MgSO₄ after adsorption to plating bacteria (E. coli strain NM538, Frischauf et al. 1983) for 15 min at 37 °C. Plating bacteria were prepared by resuspending an overnight culture of NM538 in ice-cold 10 mM MgSO₄. Phage DNA was transferred to Hybond-N filters, denatured in 0.5 M NaOH, 1.5 M NaCl, neutralised in 40 mM NaH₂PO₄/Na₂HPO₄ (NaP₁) pH6.5, baked, UV crosslinked, and hybridised as follows: probes from partial cDNAs of human, bovine (Cate et al. 1986; Picard et al. 1986) or rat AMH were labelled with ³²P using a multiprime labelling kit (Amersham) according to the manufacturer’s instructions, denatured by heating, added to the filters in 7% SDS, 0.5 M sodium hydrogen phosphate pH6.5, 1mM EDTA and hybridised for 12 –16h at 65°C. Filters were washed at 65°C in 40 mM NaP₁ pH 6.5, 1 % SDS and exposed to Kodak XAR film. Positive clones were picked, replated and rescreened until a pure population was obtained.

The entire coding region of the mouse AMH gene was sequenced using a random sequencing approach. A 2.5 kb SmaI fragment from the mouse AMH genomic clone XY-1 and the mouse AMH cDNA pBAM5 (see results), were self ligated and then sonicated on ice for 4 ×1 min. The ends were repaired using T4-polymerase and Klenow fragment (Boehringer). The DNA was separated by electrophoresis through a 1.5% agarose gel, and three different size fractions were isolated (200 – 400 bp, 400 – 600 bp, >600 bp) and cloned into M13mpl9 (New England Biolabs) linearized with SmaI. After transformation into E. coli strain TG-1 (Gibson, 1984) recombinant M13 plaques were isolated. Single-stranded DNA was prepared and sequenced using the dideoxy chain termination method (Sanger et al. 1977) using T7 DNA polymerase (Pharmacia). Both strands were sequenced at least once. The sequences were aligned using Staden-Plus software (Staden, 1980), and analysed using the University of Wisconsin Genetics computer group software package (Devereux et al. 1984).

Transcription reactions were performed in vitro to generate sense and antisense RNA probes, labelled with ³⁵S-UTP. The mouse AMH cDNA clone pBAM5 was linearized with PvuII to serve as template for T7 RNA polymerase (Promega) giving rise to an AMH antisense RNA probe. To obtain the sense RNA probe, the plasmid was digested with XbaI and transcribed with T3 RNA polymerase (Promega). The transcription was performed in 20 μl volume at 37°C for 1 h. The reaction mixture contained 40 mM Tris – HCl pH 8.25, 6mM MgCl₂, 2mM spermidine, 10 mM DTT, 0.25 mM each of GTP/ATP/CTP ribonucleotides; 50 units RNAsin (Promega), 1 μg DNA template and 5 units of either T3 or T7 RNA polymerase. The DNA template was removed by digestion with 40 units of RNAase-free DNAase I (Boehringer), for 15 min at 37°C. The RNA transcripts were hydrolysed to an average length of 0.1 kb in 40 mM NaHCO₃/ 60 mM Na₂CO3, at 60°C for t min (where t=(L – 0.1)/ 0.011 ×L; L=original probe length in kb; Cox et al. 1984). After neutralizing with an equal volume of 10% acetic acid, 1 μl of yeast tRNA (10 mg ml^-1) was added and the RNA probe was separated from unincorporated nucleotides by fractionation over a Sephadex G50 column (column buffer: 10mM Tris – HCl, 1 mM EDTA pH7.5, 10mM DTT, 0.1% SDS). Fractions (200 μl) were collected and counted by liquid scintillation. The first peak was pooled and precipitated on dry ice by adding 0.75 volumes of 4 M ammonium acetate and 2.5 volumes ethanol. The pellet was dissolved in 10mM DTT at a concentration of 1 ×10⁶cts min^-1 μl^-1. After addition of 9 volumes hybridisation buffer (50% formamide, 0.3M NaCl, 20 mM Tris – HCl, 5 mM EDTA pH 8.0, 10% dextran sulphate, 1 ×Denhardt’s solution, 0.5 mg ml^-1 yeast RNA) the probes were stored at – 70°C.

Embryos were fixed in 4% paraformaldehyde in PBS overnight at 4 °C, and then successively incubated for at least 30 min in PBS (twice at 4°C), 50% ethanol/50% PBS, 70%, 70% (overnight, 4°C), 85 %, 96% and 100% ethanol (twice). Next, the embryos were transferred through Histoclear (National Diagnostics, twice for 1 h) and paraffin wax (three changes of 25 min each at 60°C using Fibrowax, BDH). The embryos were orientated in a glass embryo dish under a dissection microscope before allowing the wax to set. Sections 5- or 7 μm were mounted on slides treated with 3-aminopropyl ethoxysilane (TESPA; Rentrop et al. 1986).

Wax was removed from mounted sections by two 10 min treatments in Histoclear and the sections rehydrated through an ethanol series, and postfixed in 4 % paraformaldehyde in PBS for 20min. The sections were treated for 5min with 10 μgmi^-1 proteinase K, the postfixation was repeated and the slides were treated for 10 min with 0.25% acetic anhydride in 0.1M triethanolamine, pH 8.0. Then the slides were dehydrated through an ethanol series and air-dried. Hybridisation under coverslips using ³⁵S-labelled anti-sense or sense control RNA probes (2 ×10⁵ctsmin^-1 μl^-1) was carried out at 55 °C overnight in a moist chamber. The coverslips were removed in 5 ×SSC, 10mM DTT at 55°C for 1 h and the slides subjected to a high-stringency wash in 50% formamide, 2 ×SSC, 10mM DTT at 65°C for 30min. The slides were then washed in 0.5 M NaCl, 10mM Tris/HCl, pH8.0, 10mM EDTA prior to incubation at 37°C for 30min with 20 μg μl^-1 ribonuclease A in the latter buffer. The high-stringency washing was repeated once more before the slides were dehydrated through an ethanol series and air-dried.

The slides were dipped in Ilford K5 nuclear track emulsion diluted 1:1 with 2% glycerol at 42°C, and dried slowly at room temperature for 4 h before exposure at 4 °C. Typically, slides were exposed for 6 days and then developed in Kodak D19 developer (2 min), rinsed in 1 % acetic acid/1 % glycerol (1min) and fixed in 30% sodium thiosulphate (2 min). Following rinsing in H₂O (twice for 10 min), the sections were stained with toluidine blue (0.02% solution, 1.5 min) and mounted using Permount (Fisher Scientific).

Screens of genomic libraries using partial bovine and human cDNAs as probes yielded no mouse AMH clones. Genomic Southern analysis suggested (data not shown) that the mouse gene was fairly divergent from the probes used, and that other sequences in the genome crosshybridized. This was probably due to the GC-rich nature of the probes. To overcome this problem, we decided to construct a cDNA library from an appropriate stage of mouse testis development. To define the stage of peak levels of AMH transcripts, we performed northern analysis using the human probe on mouse RNA samples (Fig. 1A). AMH transcripts were first detected in embryonic testes at 13.5 dpc and the level of expression appeared to be similar at 14.5 dpc. No transcripts were found in developing ovaries or adult testes or ovaries. A cDNA library was constructed, using poly(A)⁺ RNA from 14.5 dpc fetal testes. This cDNA library was screened using the bovine, human and rat AMH probes independently on triplicate filters. By this strategy, eight cDNA clones that hybridized to all three probes were isolated. The longest of the phage clones contained a 1 kb EcoRI insert. This was subcloned into Bluescript KS-(Stratagene) and named pBAM5. When probes derived from pBAM5 were used on a northern blot, they showed the expected pattern of AMH expression (Fig. 1B). Transcripts were found in fetal testes from 13.5 dpc up to 17.5 dpc, the last fetal stage examined. The hybridizing band was approximately 2.2 kb as expected for AMH by analogy with other species and was not detected in fetal ovaries.

The cDNA clone pBAM5 was used to derive probes for screening two genomic mouse libraries. Five independent phage clones were isolated. These fell into two different classes, which were largely overlapping. One of the phage clones, called XY-1, contained a 12 kb SalI insert. This was subcloned into Bluescript KS- and a detailed restriction map was derived, part of which is shown in Fig. 2.

From Southern analysis, the pBAM5 cDNA clone was shown to hybridize to a 2.5 kb Smal fragment of XY-1 (data not shown). This fragment together with the pBAM5 cDNA were sequenced. It was established that the partial cDNA is missing the 5 ’ end and starts within the third exon. The genomic clone contains all five exons; exon –intron boundaries have been defined by comparison to the cDNA and the human gene (Cate et al. 1986). The start site of transcription has not been defined but a putative TATA-box is indicated (Fig. 3). This TATA-box is in a similar position to the one that was defined by S1 mapping in the bovine gene (Cate et al. 1986) and the human gene (Guerrier et al. 1990). The predicted amino acid sequence of the mouse protein was derived and compared to the bovine and human proteins (Fig. 4). The analysis showed an overall homology of 68 % between mouse and human and 70 % between mouse and bovine proteins. The carboxy terminal part of the proteins are extremely highly conserved with only 5 (mouse/human) and 6 (mouse/ bovine) amino acid changes within the last 106 residues. This equals a 95 % and 94 % homology, respectively. Two possible N-glycosylation sites have been conserved in all three species. The cysteine residues and adjacent amino acids that are conserved within the TGF-β family of growth factors (Cate et al. 1986) are also present in the mouse AMH protein.

The sensitive method of in situ hybridisation was used to look precisely at the distribution of AMH transcripts in the developing embryo and in postnatal testes and ovaries. This technique allows the examination of many tissue types at once. In addition, it is possible to identify the cell type expressing the gene in complex tissues. The mouse AMH probe used in this study is described in Fig. 2.

AMH transcripts were not detected in 11.5 dpc genital ridges (data not shown), but were clearly present at high levels in 12.5 dpc, 13.5 dpc (data not shown) and 14.5 dpc testes (Fig. 5). The distribution of the signal over the testes suggested that it is the Sertoli cells of the testis that express AMH. No signal was found in the germ cells, which are surrounded by the cytoplasm of Sertoli cells. These are aligning into testis cords at this stage of development. No specific signal was seen outside the testis cords where interstitial cells and Leydig cells are found, or in any other tissue in the embryo (Fig. 6). Furthermore AMH transcripts were not detected in embryonic ovaries or in any other fetal female tissues. Hybridisation of sense RNA probes showed no signal above background (not shown).

To support further the finding that AMH is expressed in the Sertoli cells of the testis but not in germ cells, we looked at the expression in mice mutant in the W locus. The characteristic phenotype of mutations at this locus can be attributed to the failure of stem cell populations to migrate and/or proliferate effectively during development. Melanoblasts, hematopoetic stem cells and primordial germ cells are affected to various degrees, depending on the molecular nature of the mutation (Mintz and Russell, 1957). In severe forms, such as W extreme (W^e), the germ cells completely fail to reach and colonize the gonad in homozygous embryos. However, the testes show normal somatic development. In situ hybridisation to 14.5 dpc W^e/W^e testes indicated the pattern of AMH expression in the absence of germ cells (Fig. 7). In this case, the signal occurred over all cells within the testis cords, confirming the localisation of AMH expression to the Sertoli cells of the developing gonad. In addition, it shows that its production in Sertoli cells is not dependent on interaction with germ cells.

Just after birth AMH was found to be expressed at high levels in the Sertoli cells (Fig. 8). Transcript levels decrease during the following three weeks and are absent in adult testes (not shown). On the day of birth, the testis cords of normal males are filled with maturing germ cells easily recognizable by their completely round nucleus in high magnification bright-field micrographs (Fig. 9A,C,E). They were clearly negative for AMH expression. At around 5 to 7 days post partum (Fig. 9C,D) the male germ cells migrate from the centre of the cords to the periphery. This is reflected in the distribution of the silver grains over the testis cords which at this stage only contain Sertoli cell cytoplasm in the centre. Even later, at 11 to 14 days after birth (Fig. 9E,F), some germ cells are entering meiosis and populating the lumen of the testis cords. AMH was no longer detectable in these cords, but remains present in those where the germ cells are still mitotic and located in the periphery.

In female mice, AMH was expressed after birth in the granulosa cells of the follicle and was first detectable in ovaries at 6 days (Fig. 10A,B). AMH expression was dependent on the degree of follicular maturation and not on the age of the ovary. The signal was not detectable in primordial or small follicles (stages 1, 2 and 3a; Pedersen and Peters, 1968) which are characterized by a small oocyte (less than 15 μ m in diameter) with either no or only a few follicle cells attached to its surface (stages 1 and 2, respectively). The beginning of AMH expression in the follicles was correlated with the growth of the oocyte in medium-size follicles (stages 3b, 4 and 5a), which are found in the centre of the early postnatal ovary, whereas the small follicles are found mainly in the periphery. It was clearly the granulosa cells in the growing follicles that were expressing AMH. No silver grains were detectable over theca cells or the oocyte itself. Whilst maturing, follicles migrate from the centre of the ovary to its periphery. The oocyte in these follicles is still arrested in meiosis but is increasing in size and transcribing many genes. It is surrounded by many layers of granulosa cells and AMH transcripts were evenly distributed over all the granulosa cells. In large antral follicles (stage 6 and 7) AMH was still expressed at high levels. These follicles contain a fully grown oocyte, which is arrested in the first meiotic division, and a cavity filled with follicular fluid. This cavity divides the granulosa cells into two distinct populations: some layers at the periphery of the follicle and some around the oocyte. AMH transcipts were clearly present in granulosa cells surrounding the oocyte; however, the signal was significantly lower in the granulosa cells in the periphery of the antral follicles (Fig. 10E,F). This result is in agreement with immunohistochemical studies in adult sheep ovaries where the authors describe heterogeneity in the distribution of AMH protein in the granulosa cell population (Bezard et al. 1987). Just before ovulation, when the oocyte completes the first meiotic division and becomes a secondary oocyte AMH transcripts are no longer detectable (Fig. 10E,F).

To help understand the role of AMH in sex differentiation, we wished to look precisely at the timing and the sites of AMH expression during male and female development. Previous biochemical and immunohistochemical experiments had shown that AMH protein is found in Sertoli cells and in testicular and follicular fluid (Tran and Josso, 1982; Vigier et al. 1984). However, we were interested in determining when AMH is first transcribed during testis formation and whether the Sertoli cell is the only site of expression in male embryos. We wanted to clarify further the timing of AMH expression in the female. It was hoped that correlating AMH expression with cellular events in the developing gonads would shed some light on how AMH functions.

Using in situ hybridisation, we have shown that AMH transcripts are present in the Sertoli cells of the testis when they first differentiate at 12.5 dpc. At this time testis and ovaries show the first morphological differences due to the Sertoli cells beginning to align into testis cords. Within the next few days of development, the effect of AMH on the epithelial cells of the Müllerian ducts becomes visible and they degenerate. If AMH is not expressed in the male at the right time during gestation the Müllerian duct persists (Guerrier et al. 1989).

These studies also confirmed that the Sertoli cells are the only site of expression in male embryos. Extragonadal sites were negative for AMH mRNA. In situ hybridisation to 14.5 dpc W^e/W^e testis (Fig. 6) confirmed the localisation of AMH transcripts to Sertoli cells, and demonstrated that expression of AMH is independent of gene action in the male germ cells.

The level of AMH expression in the Sertoli cells remains roughly constant until several days after birth. As regression of the Müllerian ducts is completed during fetal life, the persistance of AMH expression implies additional functions for AMH. During later fetal stages of testis development, the Sertoli cells continue to differentiate and form seminiferous tubules harbouring the maturing germ cells in their lumen. In the first stages of spermatogenesis, the germ cells undergo multiple rounds of mitosis, before the first generation enters meiosis. It is interesting to note that AMH expression is switched off between 2 and 3 weeks after birth, coincident with the first wave of spermatids undergoing meiosis. AMH may interfere with male germ cells entering and completing meiosis. A possible function of AMH in fetal testes might be to eliminate any germ cells that enter meiosis precociously despite the inhibiting effect of the testis cords (McLaren, 1990). Evidence from female transgenic animals ectopically expressing human AMH at very high levels supports this idea. Behringer et al. (1990) found that the ovaries of these transgenic mice had fewer germ cells at birth than normal. The oocytes were then lost during the next few weeks and the somatic cells developed as structures resembling testis cords. This suggests that timing, site and levels of AMH expression are very important. Early in development, and at high levels, AMH appears to be toxic for germ cells in meiosis (Vigier et al. 1989). However, at moderate concentrations and at a later stage in development, AMH may be required to keep the oocyte in meiotic arrest as it starts growing. The onset of AMH expression in the female is correlated with the onset of oocyte growth postnatally. AMH transcripts are only detected in granulosa cells and expression continues through maturation of the follicle until the oocyte completes the first meiotic devision just before ovulation. A better understanding of AMH function will be achieved by introducing mutations into the gene by homologous recombination.

It has been suggested that Sertoli cells and granulosa cells differentiate from the same precursor cell lineage (Vigier et al. 1984). Both types of cells carry a common lineage-specific surface antigen (Ciccarese and Ohno, 1978), form a diffusion barrier shielding germ cells from the somatic environment, and release their secretion products into a fluid that bathes germ cells during normal maturation. The fact that both Sertoli and granulosa cells express AMH supports this assumption. However, the transcriptional regulation in both cases must be different since expression in the female occurs with a significant time delay, and strictly controlled expression is very important in order to achieve appropriate male and female gonadal and genital tract development.

The pattern of AMH expression allows us to speculate on its possible regulation. It is clear that the presence of AMH is absolutely required to give rise to the complete male phenotype, and it has been thought for a long time that AMH is a very likely target for direct regulation by the Y-linked testis-determining gene Tdy. Tdy is thought to be autonomously expressed in Sertoli cells (Burgoyne et al. 1988; Palmer and Burgoyne, 1991), and a candidate gene for Tdy known as Sry has recently been cloned (Gubbay et al. 1990; Sinclair et al. 1990). This has proven to be sufficient to cause sex-reversal in transgenic mice (Koopman et al. 1991). The Sry gene product contains a putative DNA-binding domain. Sry is expressed in the somatic portion of the genital ridge just prior to testis differentiation from 10.5 dpc to 12.5 dpc (Koopman et al. 1990). AMH transcripts were first seen using in situ analysis at 12.5 dpc. The timing and location of expression of the two genes so far does not exclude the possibility of AMH being a target for regulation by Sry. However, the delay of 48 h between the first appearance of Sry and AMH transcripts suggests an indirect rather than a direct interaction. In vitro studies can be used to distiguish these possibilities. If Sry does play a role in inducing AMH expression in male embryos, it is certainly not required for its maintenance, since Sry expression is switched off at 12.5 dpc whereas AMH continues to be expressed.

During embryonic development of the female, it is very important that AMH expression is not induced. One way to ensure the silencing of AMH in females would be the lack of an essential transcription factor, for example Tdy. However, AMH seems to be required at a later stage in female development and the inductive function of Tdy may be substituted by a factor released by the oocyte. A palindromic element that is nearly identical to the consensus sequence of estrogen response elements (ERE) has been identified upstream of the transcriptional start site of the human AMH gene (Guerrier et al. 1990). The authors conclude from in vitro studies that this element may play a regulatory role in granulosa cells. The cloning of mouse AMH will now allow in vivo analysis of regulatory elements in transgenic mice.

We thank A. McLaren and M. Buehr for allowing us the use of W^e mouse stocks, B. Skene for help in initiating the project, D. Wilkinson for assistance with in situ hybridisation, R. Buxton for help with the compilation of the sequence data, P. Koopman, J. Gubbay, B. Capel and D. Guerrier for helpful comments on the manuscript and our colleagues for discussion and encouragement. We thank N. Josso and R. Cate for probes and A. M. Frischauf and K. Willison for genomic libraries. A. M. is supported by the Gottlieb Daimler and Carl Benz Stiftung.