DAX-1 and SOX6 molecular interplay results in an antagonistic effect in pre-mRNA splicing

Authors

  • Kenji Ohe,

    1. Institut de Génétique et de Biologie Moléculaire et Cellulaire, CNRS-INSERM-ULP, Strasbourg, France
    Current affiliation:
    1. Institute for Comprehensive Medical Science, Fujita Health University, Toyoake, Aichi 470-1192, Japan
    Search for more papers by this author
    • Kenji Ohe and Katherine T. Tamai contributed equally to this work and should be considered jointly as first author.

  • Katherine T. Tamai,

    1. Institut de Génétique et de Biologie Moléculaire et Cellulaire, CNRS-INSERM-ULP, Strasbourg, France
    Current affiliation:
    1. Centre for Cell and Molecular Dynamics, Department of Cell and Developmental Biology, University College London, London WC1E 6DE, UK
    Search for more papers by this author
    • Kenji Ohe and Katherine T. Tamai contributed equally to this work and should be considered jointly as first author.

  • Martti Parvinen,

    1. Department of Anatomy, University of Turku, Turku, Finland
    Current affiliation:
    1. Department of Physiology, University of Turku, Kiinamyllynkatu 10 B 601, 20520 Turku, Finland
    Search for more papers by this author
  • Paolo Sassone-Corsi

    Corresponding author
    1. Institut de Génétique et de Biologie Moléculaire et Cellulaire, CNRS-INSERM-ULP, Strasbourg, France
    • University of California, Department of Pharmacology, 2115 Gillespie Neuroscience Research Facility (GNRF), Mail Code: 4625, Irvine, CA 92697
    Search for more papers by this author

Abstract

DAX-1 (DSS-AHC Critical Region on the X Chromosome-1) is a member of the nuclear hormone receptor superfamily that has an important role in steroidogenesis and gonadogenesis. To investigate the role of DAX-1 in the testis, a yeast two-hybrid screen was performed and SOX6, member of the Sry box (SOX) protein family, was cloned as candidate. The interaction was confirmed biochemically and expression of SOX6 overlapped with that of DAX-1 in the developing gonad, as well as in Sertoli cells of the adult testis. We show here that DAX-1 is able to inhibit splicing in in vivo and in vitro splicing assays, and this inhibition is relieved by the addition of SOX6. SOX6 appears to inhibit the interaction between DAX-1 and the splicing machinery, thus providing a likely mechanism for functional interference. We conclude that DAX-1 and SOX6 proteins interact, have overlapped expression in the testis, and act antagonistically during pre-mRNA splicing. Developmental Dynamics 238:1595–1604, 2009. © 2009 Wiley-Liss, Inc.

INTRODUCTION

DAX-1 is an unusual member of the nuclear receptor superfamily, containing a unique triple repeat motif in the DNA-binding domain (DBD) at the N terminus and a C terminus ligand-binding domain (LBD) resembling that of several nuclear receptors (Muscatelli et al.,1994). Nuclear receptors are known to exert their function in the nucleus where they bind the DNA elements of promoters and regulate their transcriptional activity. In addition to being involved in cell- and promoter-specific transcription silencing activity (Lalli et al.,1997) as a transcription factor, DAX-1 was found to be associated with actively translating polyribosomes in the cytoplasm. Biochemical fractionation of the steroidogenic adrenal and gonadal cell lines shows that DAX-1 is present in the polyribosome fraction of the cytoplasm. The cytoplasmic fraction was purified by an oligo(dT) column and showed DAX-1 to be associated with polyadenylated RNA complexes. RNA binding of DAX-1 was tested by RNA homopolymer binding assays and Northwestern blot analyses: they revealed two differential RNA binding domains existing in the N and C termini of the protein. DAX-1 was suggested to have a role in post-transcriptional regulation (Lalli et al.,2000).

DAX-1 is expressed in tissues related to steroidogenesis and gonadogenesis. It functions as a global negative transcriptional regulator of steroid hormone production, through the C terminus of the protein, by repressing the expression of multiple genes involved in the steroidogenic pathway (Lalli et al.,1997; Zazopoulos et al.,1997). DAX-1 mutations in humans result in a defect that causes a congenital disease, adrenal hypoplasia congenita, consistently associated with hypogonadotropic hypogonadism (Muscatelli et al.,1994). These mutations have an altered C terminus resulting in cytoplasmic retention of the mutant protein, which results in impaired transcriptional silencing by DAX-1. This manifests an important example of a human disease caused by a defect in nuclear localization of a transcription factor (Lehmann et al.,2002,2003). Importantly, we have found that these mutations also have a defect in RNA binding (Lalli et al.,2000), which implies that the DAX-1 protein is involved in RNA metabolism. Patients harboring DAX-1 mutations show a defect in spermatogenesis, likely from a defect in Sertoli cell function. DAX-1 was shown to be expressed in rat Sertoli cells being regulated during spermatogenesis with a peak during the androgen-sensitive stages of the seminiferous epithelial cycle (Tamai et al.,1996). Impairment in spermatogenesis is also found in Dax1-deficient male mice (Yu et al.,1998), where Sertoli cells are aberrantly located, and can be rescued by a Müllerian inhibiting substance-DAX1 transgene (Jeffs et al.,2001).

Partners or co-regulators for DAX-1 have been discovered and characterized. The LXXLL motif found in the N terminus of DAX-1 was found to mediate specific interaction with other nuclear receptors such as activated estrogen receptors and was found to inhibit its activation (Zhang et al.,2000). Ad4BP/SF-1 is important for the formation of steroidogenic tissues and control of steroidogenesis and also interacts with LXXLL motifs found in DAX-1 (Suzuki et al.,2003). The C terminus of DAX-1 is known to interact with co-regulators, N-CoR (Crawford et al.,1998), and Alien (Dressel et al.,1999; Altincicek et al.,2000), by which it mediates its transcriptional silencing activity.

DAX-1 was shown to antagonize SRY, the testis-determining gene, and a sex-reversed phenotype developed when Dax-1 was expressed as a transgene in mice with weak alleles of the sex-determining Y-chromosome gene Sry (Swain et al.,1998). SRY is a member of the SOX (Sry box) protein family, which comprises transcription factors important for embryonic development (Wilson and Koopman,2002). SOX proteins were also reported to be involved in pre-mRNA splicing. Immunodepletion of SOX6 from HeLa nuclear extracts resulted in impaired splicing in an in vitro splicing assay. SOX6 as well as SRY and SOX9 were able to reconstitute splicing by the addition of each recombinant protein to the SOX6-depleted nuclear extracts (Ohe et al.,2002). This result, together with results of WT-1 isoforms, which associate and interact with components of the spliceosome (Larsson et al.,1995; Davies et al.,1998), provide us with a reminiscent pathway between sex determination in mammals and Drosophila melanogaster, where a cascade of splicing factors and transcription factors regulate a pathway leading to sex determination (Lalli et al.,2003). The molecular mechanism by which DAX-1 and SOX proteins function as testis-determining factors is still yet to be deciphered and controversial results of DAX-1 and SOX proteins in finding a common antagonistic target in transcriptional regulation have led us to pursue the antagonistic function at post-transcriptional levels such as pre-mRNA splicing. Bioinformatics data have shown that the testis has high levels of alternative splicing (Yeo et al.,2004). Since SOX proteins play a role in pre-mRNA splicing (Ohe et al.,2002) and DAX-1 plays a role in post-transcriptional events (Lalli et al.,2000), we performed a functional investigation of DAX-1 to the level of splicing.

In this study, we cloned SOX6 as a partner for DAX-1. SOX6 was expressed in close vicinity with DAX-1 in the testis. DAX-1 showed inhibition of pre-mRNA splicing for specific substrates, which was restored by SOX6. The mechanism of this antagonistic effect seems to be through DAX-1 interaction with U2AF65, which was abrogated by SOX proteins. We therefore propose an antagonistic pre-mRNA splicing function of DAX-1 and SOX6.

RESULTS

Association of DAX-1 and SOX6

In order to investigate the biological function of DAX-1 in the testis, the full-length human DAX-1 (Fig. 1A) fused to LexA (LexA/DAX-1) was used to screen a human testis library in order to find partners. Clones (0.5 × 106) were screened and 8 different classes of clones were isolated that interact with LexA + DAX-1, but not with LexA alone. One class was isolated 16 times and turned out to be partial and full-length clones of cDNAs encoding human SOX6 (Fig. 1A) by sequence analysis. A clone containing full-length SOX6 (pGAD10/SOX6) was retransformed with LexA/DAX-1 or pGBT9/DAX-1 into L40 (with LexA binding sites) (Fig. 1B) or HF7c (with GAL4 binding sites) (Fig. 1C) yeast strains, respectively, to confirm the interaction. Two hybrid analyses using different portions of SOX6 and DAX-1 showed that both the DNA-binding domain and ligand-binding domain of DAX-1 interact with the first coiled coil domain of SOX6 (between 208–428 amino acids) (Fig. 1D).

Figure 1.

Isolation of SOX6 as a DAX-1 partner in testis by yeast two-hybrid assay. A: Diagram of DAX-1 and SOX6 proteins. DAX-1 contains a triple repeat in the DNA-binding domain (DBD) in the N terminus and ligand binding domain (LBD) in the C terminus. SOX6 contains one HMG box in the C terminus and two coiled coil (cc) domains in the N terminus. The full-length DAX-1 fused to lexA was used to screen partners from a human testis library. B: Growth of transformants coexpressing DAX-1 and SOX6 on selective medium using the LexA two-hybrid system. DAX-1 and SOX6 showed interaction in media lacking leucine. LexA alone and Lamin (LAM) are negative controls. C: DAX-1 and SOX6 interaction in the GAL4 two-hybrid system. DAX-1 and SOX6 showed interaction in media lacking histidine. GAL4-DBD and lamin are negative controls. D: Table of the domains by which DAX-1 and SOX6 interact. SOX6 interacted with both the N and C terminal domain of DAX-1, while DAX-1 interacted with the first coiled coil domain of SOX6.

The interaction of DAX-1 and SOX6 was tested by GST-pulldown analyses. When recombinant DAX-1 (pET-DAX-1) was used as a source, GST-SOX6 (1–428aa) was able to pull down a significant amount of DAX-1 by direct protein-protein interaction (Fig. 2A). When whole cell extract of COS7 overexpressing DAX-1 was used as a source, GST-SOX6 (full-length) was able to pull down DAX-1 at normal to moderate salt washes (Fig. 2B). DAX-1 and SOX6 interaction was also tested by co-immunoprecipitation of the two proteins over-expressed in COS7 cells using specific antibodies of DAX-1 (Tamai et al.,1996) and SOX6 (Ohe et al.,2002). Extracts from cells transfected with DAX-1 and SOX6 expression vectors, individually or in combination, were immunoprecipitated using the anti-DAX-1 antibody. Western blot analysis revealed that SOX6 associated with DAX-1 in the immunoprecipitated complexes. No SOX6 protein was co-immunoprecipitated by anti-DAX-1 antibodies when SOX6 and DAX-1 were expressed separately. Conversely, DAX-1 and SOX6 association was also detected when SOX6 protein was immunoprecipitated from co-transfected cells, using the anti-Sox6 antibody and revealed by Western blot using the anti-DAX-1 antibody (Fig. 2C).

Figure 2.

SOX-6 and DAX-1 proteins interact in vitro and in transfected cells. A: GST pulldown of DAX-1 by GST-SOX6 (1–428), using recombinant DAX-1 protein (10% input is shown) as a source. B: GST pulldown of COS cells overexpressed with DAX-1 (10% input is shown) with GST-SOX6 (full-length) is resistant to moderate salt wash (compare NaCl 300 mM to NaCl 100 mM). C: DAX-1 coimmunoprecipitates with Sox6 after transfection in COS cells of DAX-1 and Sox6 expression vectors. Immunoprecipitations (IP) with DAX-1 and Sox6 antibodies revealed by Western blot with the reciprocal antibody. The DAX-1 antibody was able to precipitate the overexpressed DAX-1-SOX6 complex, as well as precipitating DAX-1 alone and not precipitating SOX6 when they were individually overexpressed (left four panels). The SOX6 antibody was able to precipitate the overexpressed DAX-1-SOX6 complex, as well as precipitating SOX6 alone and not precipitating DAX-1 when they were individually overexpressed (right four panels).

Overlapped Expression of Sox6 and Dax-1 in Sertoli Cells

The expression of Sox6 mRNA in the testis was detected by ribonuclease (RNase) protection assay (Tamai et al.,1996). Total RNA was isolated from rat tissues where Dax-1 is known to be expressed : hypothalamus, pituitary gland, adrenal gland, testis, and ovary. Sox6 transcript was detected in the testis but also slightly detected in the hypothalamus (Fig. 3A). Though Sox6 was thought to be exclusively expressed in the testis, it has also been shown, though at a lower level, to be expressed in various tissues (Hagiwara et al.,2000). Sox6 is up-regulated in the genital ridge at 11.5 and 12.5 dpc (Fig. 3B), when Dax-1 is known to be detectable (Tamai et al.,1996) and when Sry shows its peak expression (Swain and Lovell-Badge,1999). When we used whole testis from rats at different ages, Sox6 was detectable throughout the time-points examined with an increase at 2–4 weeks. This increase coincides with Dax-1 expression, where it is known to be detectable at day 15 (Tamai et al.,1996). While Dax-1 expression has been shown to peak during the first synchronized spermatogenic wave and decrease thereafter (Tamai et al.,1996), Sox6 maintains its expression up to 15 weeks (Fig. 3C). Though Sox6 protein is expressed throughout development, its expression is stage specific in rat seminiferous tubules. Sox6 is up-regulated during stage II–VI, with a peak at stages VII, VIII, and no expression is found after stage XII (Fig. 3D). The distribution of Sox6 in the seminiferous epithelium suggests its role in haploid cell transcription (Penttila et al.,1995). Sox6 expression is remniscent of Dax-1, which is known to be expressed at stages V–VII (Tamai et al.,1996). In situ hybridization shows expression of Sox6 in stage VII–VIII seminiferous tubules with scarce expression in latter stage tubules (Fig. 3E). Immunohistochemistry of Sox6 shows the main expression is in round spermatids and spermatogonia (Fig. 3F, left panel), and a significant amount of Sox6 is also detected in Sertoli cells (Fig. 3F, right panel, red arrowheads). Thus, Sox6 is expressed stage-specifically, developmentally and during spermatogenesis in germ cells and Sertoli cells. Sertoli cells are the site of interplay between Sox6 and Dax-1 (Dax-1 is known to be mainly expressed in Leydig and Sertoli cells but not in germ cells; Tamai et al.,1996), the interaction of which we have found biochemically in Figures 1 and 2. Finally, to confirm overlapped expression of Dax-1 and Sox6, immunofluorescence was conducted using primary culture of rat Sertoli cells. Dax-1 and Sox6 showed overlapped expression and localization (Fig. 3G).

Figure 3.

Analysis of Sox6 expression in the rat testis analysed by RNase protection analyses. A: Total RNA from rat tissues showed Sox6 is abundantly expressed in the testis and to a lesser degree in the hypothalamus. Scarce expression was found in pituitary gland, adrenal gland, or ovary where Dax-1 is known to be expressed. B: Sox6 shows increased expression at the genital ridge at 11.5 dpc. C: Sox6 is expressed in rat Sertoli cells with an increase at 3 weeks. D: Sox6 is expressed (bottom panel) stage specifically (Stage VII–IX) during the cycle of the rat seminiferous epithelium (red indicates nuclei where Sox6 is expressed in germ cells). E: In situ hybridization of Sox6 mRNA expression in rat seminiferous tubules shows expression in stages VII–VIII with scarce expression in stages thereafter. F: Left: Immunohistochemistry of Sox6 in rat seminiferous tubules shows the main expression of Sox6 in round spermatids and spermatogonia. Right: High magnification of F. Sox6 staining of nuclei of rat Sertoli cells (red arrowhead). G: Immunofluorescence of rat Sertoli cell primary cultured cells with anti-DAX-1 and anti-SOX6 antibodies.

Antagonistic Function of DAX-1 and SOX6 in Pre-mRNA Splicing

In our previous report, DAX-1 turned out to be associated with actively-translating polyribosomes (Lalli et al.,2000). It is known that splicing factors have a characteristic speckle localization, as is for SOX6 (Ohe et al.,2002). These speckles have been found to be maintained by protein–protein interactions rather than attachment to the interchromatin granule cluster-specific framework (Sacco-Bubulya and Spector,2002; Lamond and Spector,2003). Since DAX-1 interacts with SOX6, which is localized to speckles, it was tempting to test whether DAX-1 influences the function of SOX6 in pre-mRNA splicing. GST-DAX-1 was added to an in vitro splicing reaction using β-globin pre-MRNA as splicing substrate. GST-DAX-1 showed inhibition of β-globin pre-mRNA splicing in a dose-dependent manner (+:5 pmol, ++:10 pmol) (Fig. 4A), while equal mol of GST had no considerable effect. Next we used in vitro translated DAX-1 protein to see whether it has the same effect. In vitro translated DAX-1 protein (IVT-DAX-1) was added to the splicing reactions and clear inhibition of splicing by IVT-DAX-1 was observed, compared to reticulocyte cell lysate alone or by a control protein (IVT-CREB). This inhibition of splicing could be restored by pre-incubation of IVT-DAX-1 with anti-DAX-1 antibody (Fig. 4B). Inhibition of splicing by GST-DAX-1 was observed in all of the substrates we tried (δ-crystallin pre-mRNA, E1A pre-mRNA, fushitarazu [ftz] pre-mRNA) (data not shown). But inhibition by IVT-DAX-1 showed substrate specificity, i.e., the ftz substrate showed no inhibition (Fig. 4B) while the other substrates did (data not shown). To test whether this inhibition by DAX-1 was possible to be reversed by SOX6, we added in vitro translated SOX6 (HMG) protein (IVT-SOX6) (HMG) to the reaction in addition to IVT-DAX-1. To our surprise, reconstitution of splicing was observed (Fig. 4C). This antagonistic effect could not be observed using GST-SOX6 (HMG), which implicates an additional factor(s) found in reticulocyte cell lysate or post-translational modifications of the proteins to be involved. In order to check whether DAX-1 and SOX6 influence splicing activity in transfected cells, we transfected expression vectors of DAX-1 and SOX6 with an in vivo splicing mini-gene of tumour necrosis factor β (TNFβ) pre-mRNA (Neel et al.,1995). As shown (Fig. 4D), DAX-1 inhibited the stepwise constitutive splicing of TNFβ. Addition of SOX6 antagonized the effect of DAX-1, which correlates very well with our in vitro data. Another transcription factor involved in pre-mRNA splicing is WT1, essential for urogenital development and known to bind the splicing machinery through U2AF65 (Davies et al.,1998). According to this, we tested DAX-1 interaction with U2AF65, and though GST pull down of U2AF65 by GST-DAX-1 was marginal (Fig. 4E), anti-DAX-1 antibody was able to immunoprecipitate a considerable amount of endogenous U2AF65, in HeLa cells over-expressed with DAX-1 (Fig. 4F, lane 5) while mouse IgG (mIgG) could not (Fig. 4F, lane 2). Intriguingly, when Sox6 was co-transfected with DAX-1, U2AF65 immunoprecipitation by anti-DAX-1 antibody was abrogated (Fig. 4F, lane 6). In a different experiment transfecting CREB as a control, to rule out the possibility of artifacts due to over-expression, anti-DAX-1 antibody was still able to immunoprecipitate a considerable amount of endogenous U2AF65 (Fig. 4G, lane 2), and SOX6 reduced the interaction (Fig. 4G, lane 3). Other SOX proteins, SOX9 and SRY, also reduced the DAX-1-U2AF65 interaction (Fig. 4G, lane 4 and 5, respectively). Finally, we show the model for splicing regulation by the DAX-1-SOX6 complex (Fig. 4H), where SOX6 (and likely other SOX proteins) interact with DAX-1 in order to suppress its inhibitory splicing activity (through abrogation of DAX-1-U2AF65 interaction).

Figure 4.

Antagonistic effect of DAX-1 and Sox6 in pre-mRNA splicing. A: GST-DAX-1 inhibited in vitro splicing of a standard β-globin pre-mRNA substrate in a dose-dependent manner (compare lanes 4 and 5 to lanes 2 and 3). B: Inhibition of splicing was also observed by IVT-DAX-1 (lanes 5 and 6), and splicing was restored by pre-incubation with the anti-DAX-1 antibody (lanes 7 and 8). No inhibition by IVT-DAX-1 was observed when using a fushitarazu (ftz) pre-mRNA substrate (lanes 10 and 11). The exon at the bottom is a longer exposure. C: Inhibition of β-globin splicing by IVT-DAX-1 in a dose-dependent manner (lanes 2 to 4) was restored by adding IVT-Sox6 (HMG) in a dose-dependent manner (lanes 6 to 8). D: Effect of full-length DAX-1 and full-length Sox6 expressed in HeLa cells using an in vivo constitutive splicing substrate (TNFβ). DAX-1 overexpression showed decrease of F1, F2 splicing (lane 2), which was restored by DAX-1 and Sox6 overexpression (lane 3). E: GST pull-down of endogenous U2AF65 of HeLa cell nuclear extract by GST-DAX-1 revealed with anti-U2AF65 antibody. F: Immunoprecipitation of nuclear extract of HeLa cell overexpressed with indicated plasmids with mouse IgG (mIgG) and anti-DAX-1 antibody revealed with anti-U2AF65 antibody (top panel). DAX-1 immunoprecipitation was confirmed by revealing with anti-DAX-1 antibody (bottom panel). Sox6 overexpression interfered DAX-1-U2AF65 interaction. G: Interference of DAX-1-U2AF65 interaction by other members of the SOX protein family. Sox6 (lane 3), SOX9 (lane 4), and SRY (lane 5) inhibited DAX-1-U2AF65 interaction compared to control (CREB) (lane 2). H: Model for antagonistic function of DAX-1 and SOX proteins in pre-mRNA splicing. DAX-1 inhibits U2AF65 function in pre-mRNA splicing (indicated as “off” (of U2AF65) at bottom), which is restored by SOX protein (indicated as “on” (of U2AF65) at top). SOX proteins sequester DAX-1 to inhibit formation of DAX-1-U2AF65 inhibitory complex.

DISCUSSION

Transcriptional regulation is believed to govern the mammalian sex determination cascade (Swain and Lovell-Badge,1999; Clarkson and Harley,2002; Harley et al.,2003), and, in fact, TES (testis-specific enhancer of Sox9) has been recently characterized as a specific transcriptional target for SRY (Sekido and Lovell-Badge,2008). The findings in D. melanogaster, where regulation of sex determination at the post-transcriptional level is observed (Schutt and Nothiger,2000; Lalli et al.,2003), raises the question of whether any trace or function of this event can be found in mammals. However, regarding alternative splicing in vertebrates, cell-specific splicing occurs from the antagonistic effect on splice site usage by regulatory factors, which differs from the mechanism in Drosophila and yeast. Regulatory complexes are favored over gene-specific single factors (Smith and Valcarcel,2000; Charlet et al.,2002). This is consistent with the findings reported in this study that a DAX-1 splicing regulatory complex exists and functions through interaction with SOX proteins and U2AF65.

In this study, we have determined SOX6 as a novel partner for DAX-1 by yeast two-hybrid assays and co-immunoprecipitation. SOX6 had overlapped expression with DAX-1 in Sertoli cells developmentally and during spermatogenesis. Since DAX-1 has a role in mRNA metabolism (Lalli et al.,2000) and SOX proteins have a role in splicing (Ohe et al.,2002), we tested DAX-1 in pre-mRNA splicing and found an intriguing antagonistic effect between DAX-1 and SOX6. Splicing inhibition by DAX-1 was observed by both recombinant protein and in vitro translated protein, but the restoration of splicing by SOX6 was only observed when we used the in vitro translated DAX-1 protein. An additional factor in reticulocyte cell lysate or post-translational modification may have a crucial role in this phenomenon. The restoration by SOX6 was attributable to the HMG box, which is not the interaction domain (SOX6 coiled coil domain) with DAX-1. This eliminates the possibility of DAX-1 and SOX6 interfering with each other's function in splicing via protein–protein interaction. Rather, it seems that both proteins compete for a specific unknown RNA target or protein, by dosage for instance. Another explanation is that these two proteins are non-specific RNA-binding proteins, but once they are incorporated into a complex, this complex has a sequence-specific function. This well explains why only recombinant protein produced in reticulocyte cell lysate exhibited substrate specificity.

SOX6 is a member of the SOX protein family, whose function has been implicated in cell fate decisions in various developmental processes. These SOX proteins are known as transcription factors exerting their function by interacting with partners. We have located the DAX-1 interaction domain of SOX6 to its coiled coil domain, a leucine zipper followed by a glutamine-rich region. This does not create a new DNA-binding interface as in the case of the classic bZip proteins. Rather, this coiled coil domain of SOX6 is known to mediate protein–protein interaction (Wegner,1999). SOX6 is expressed in the CNS during embryogenesis, and associates with the type II collagen gene during chondrogenesis (Lefebvre,2002) and is involved in oligodendrocyte development (Stolt et al.,2006). Sox6 has also been implicated in a cascade of alternative splicing during stem cell differentiation (Yang et al.,2007). Interestingly, DAX-1 is also known to be involved in embryonic development (Niakan et al.,2006). It is conceivable that SOX2, which is known to be an essential factor to induce pluripotent stem cells from skin cells (Takahashi and Yamanaka,2006), exerts its effect through antagonizing DAX-1 splicing activity. There are reports suggesting that DAX-1 is expressed in the skin (Patel et al.,2001). Moreover, growing evidence that other HMG proteins are involved in splicing has been found. HMGA1a, previously called HMG I/Y, has also recently been implicated in sequence-specific binding and aberrant splicing of presenilin-2 as a cause of sporadic Alzheimer's disease (Manabe et al.,2003,2007).

U2AF65 and PTB (poly-pyrimidine tract binding-protein) are known to functionally compete for binding to the polypyrimidine tract (Sauliere et al.,2006). Since we show here that DAX-1 and U2AF65 can interact, we believe that DAX-1 inhibits splicing through interaction of U2AF65, which is blocked when SOX6 is expressed and binds to DAX-1. DAX-1-U2AF65 interaction is counteracted by the DAX-1-SOX6 interaction. This interaction of DAX-1-U2AF65 is reduced most by SRY, among the SOX factors we tested. Thus, DAX-1, known as a global transcriptional repressor (Lalli and Sassone-Corsi,2003), may ensure shutting down its target genes by inhibiting splicing, which is restarted by SOX factors during the vast developmental changes of global expression during Sertoli cell maturation.

The difficulty of accepting the transcription factors DAX-1 and SOX6 to be involved in splicing is a critical issue. Emerging evidence of functional coupling of transcription and splicing (Hirose and Manley,2000) has been extended to transcriptional co-regulators of nuclear receptors. Exon skipping was induced in a nuclear receptor–dependent fashion, which was not a result of promoter strength or saturation of splicing capacity. Nevertheless, the existence of splicing factors in purified transcriptional co-regulator complexes suggests a bi-functional role of nuclear receptors in transcriptional regulation and splicing (Auboeuf et al.,2002). Although there is no direct evidence of transcription-regulating complexes in reassembly or recruitment of factors to splicing complexes, co-regulators such as PGC-1 (Monsalve et al.,2000) or CoAA (Auboeuf et al.,2004) have been shown to have independent functions in transcription and splicing. PGC-1 was shown to exert its function through interaction with the C-terminal domain of RNA pol II, and CoAA has been proposed to be recruited to the splicing apparatus through its affinity and its conformational accessibility depending on its interaction with other factors on specific promoters. The antagonistic effect of DAX-1 and SOX6 in transfected cells using TNFα as a splicing substrate reiterates our findings in splicing rather than mRNA stability since this substrate is known for its stable mRNA in both the nucleus and the cytoplasm (Neel et al.,1995).

The antagonizing effect of DAX-1 and SOX proteins is also found in DAX-1 transgenic mice where a sex-reversed phenotype is observed in mice with weak Sry alleles. Delayed testis development can be observed in mice with normal Sry alleles. Additionally, from these transgenic models, it is unlikely that DAX-1 and Sry function upstream of each other because the Dax1 transgene led to female development when Sry was driven by either Dax1 or Sry regulatory elements (Swain et al.,1998). The transgenes of the HMG box of Sox3 and Sox9 and the Sox9 gene can induce mouse testis development (Bergstrom et al.,2000; Vidal et al.,2001), which is in agreement with our previous findings that the HMG domain of SOX6, SRY, and SOX9 can reconstitute splicing of SOX6-depleted HeLa nuclear extracts (Ohe et al.,2002). The data presented here extend our previous findings that the possibility of sex determination or sex reversal may be based on the antagonistic effects of DAX-1 and SOX proteins in pre-mRNA splicing. Further accumulation of data (particularly of DAX-1 and SRY or SOX9), determining quantitative, temporal, and spatial expression during Sertoli-cell development would provide further clues of deciphering the roles of these factors. Interestingly, a report has shown that the specific isoform of Fgfr2 in mice has an important role in Sertoli cell fate determination (Kim et al.,2007). It is well known that PTB regulates FGFR2 (Carstens et al.,2000) with other recently identified factors (Warzecha et al.,2009). Our study here implicates that DAX-1 and SOX proteins are involved in alternative splicing through polypyrimidine tract-binding proteins such as U2AF65 and PTB. Whether DAX-1 and SOX proteins are involved in Sertoli cell–specific alternative splicing awaits future studies of FGFR2 splice variants in the genital ridge of DAX-1 knockout mice or the discovery of a sequence-specific RNA target, which is shared by DAX-1 and SOX proteins.

EXPERIMENTAL PROCEDURES

Plasmids

Full-length human DAX-1 was subcloned into the LexA fusion vector pBTM116 to generate pLexA/DAX-1. The DAX-1 DNA-binding domain (DBD) and DAX-1 ligand-binding domain (LDB) were subcloned into pBTM116 to create plasmids pLexA/DAX-1 (DBD) and pLexA/DAX-1 (LBD), respectively. Full-length DAX-1 was also fused to the GAL4 DNA-binding domain to generate plasmid pGBT9/DAX-1. Full-length SOX6 and truncations encoding amino acids (aa) 1–428, 1–207, and 429–794 of SOX6 were subcloned into pGAD424 to generate pGAD424/SOX6 (full-length), pGAD424/SOX6 (1–428), pGAD424/SOX6 (1–207), and pGAD424/SOX6 (429–794). Expression vectors for DAX-1 (pSG.DAX-1) (Tamai et al.,1996) and SOX6 (pSG.SOX6) (Ohe et al.,2002) have been previously described. pET-DAX-1 was constructed by subcloning full-length human DAX-1 in pET15b (Novagen). GST-DAX-1 (full-length) and GST-SOX6 fusion proteins (full-length and aa 1–428) were generated in vector pGEX-1λT. All constructs were confirmed by sequencing on both strands.

Yeast Two-Hybrid Assays

A human adult testis cDNA library constructed in vector pGAD10 (Clontech) was cotransformed along with bait plasmid pLexA/DAX-1 in yeast strain L40. Yeast two-hybrid screening was performed on synthetic media minus histidine plus 10 mM 3-aminotriazole (3-AT) as described (Clontech). The interaction between DAX-1 and SOX6 was confirmed in both the LexA and GAL4 yeast two-hybrid systems (Clontech). Protein interactions were assayed by cotransforming yeast strain L40 (LexA) or HF7c (GAL4) with the appropriate plasmids indicated in Figure 1 and monitoring growth on synthetic media minus histidine plus 10 mM 3-AT.

GST Protein, GST Pull-Down Assays, and In Vitro Translated (IVT) Protein

GST and GST-DAX-1 protein were produced and purified according to the manufacturer's protocol. For GST-pull down assays, GST fusion protein (0.5 μg) bound to glutathione beads was incubated with 70 μg of protein extract from COS cells at 4°C for 30 min. Beads were washed and bound protein was analysed by Western blotting by Anti-U2AF65 (clone MC3) antibody, a kind gift of M. Carmo-Fonseca (Univ. Lisboa, Lisboa, Portugal). In vitro translated (IVT) protein, IVT-DAX-1, IVT-CREB, and IVT-SOX6 (HMG) proteins were synthesized in vitro using a TNT T7 transcription/translation kit (Promega) according to the manufacturer's instructions except in the presence of unlabelled methionine.

Immunoprecipitation and Western Blot Analysis

Anti-DAX-1 (Tamai et al.,1996) and anti-SOX6 (Ohe et al.,2002) antibodies were produced and verified as previously described. Immunoprecipitation assays were performed with COS cells transfected with expression vectors for DAX-1, SOX6, SOX9, or SRY by the calcium phosphate method. Transfected cells (Fig. 2C) were lysed in 1 ml EBC (50 mM Tris-HCl, pH 8.0, 170 mM NaCl, 0.5% NP-40, 50 mM NaF) containing 1 mM PMSF plus 10 μg ml-1 of aprotinin and leupeptin. Following pre-clearing with protein A-sepharose or protein G-sepharose, cell extracts were incubated at 4°C overnight with the appropriate antibody. The protein-antibody complexes were immunoprecipitated using protein A- or protein G-sepharose and then subjected to Western blot analysis.

RNase Protection Assays, In Situ Hybridization and Immunohistochemistry, and Immunofluorescence Microscopy

Total RNA was isolated from embryonic and adult rat tissues or from purified rat Sertoli cell cultures as previously described (Tamai et al.,1996). A SOX6 antisense probe was synthesized from a plasmid containing a 658-bp PstI-BamHI fragment of SOX6 and was used for both in situ hybridization and RNase protection assays, which were performed as described (Tamai et al.,1996). The rat seminiferous tubule segments at defined stages of the cycle were microdissected as described (Parvinen and Ruokonen,1982). Immunofluorescence of rat Sertoli cell primary culture was performed as described (Ohe et al.,2002).

In Vitro and In Vivo Splicing Analysis

Splicing substrates human β-globin (Krainer et al.,1984) and D. melanogaster fushi-tarazu (Rio,1988) were used for in vitro splicing assays, which were performed as described (Ohe et al.,2002), except that in Figure 4C, Mg2+ was reduced to 1.5 mM to obtain a clear reconstitution by IVT SOX6 (HMG). The in vivo splicing assays were performed in HeLa cells, which were transfected by the calcium phosphate precipitation method with equal amounts of DNA, normalized by the addition of pSG5. The splicing substrate TNFβ was a kind gift of Dr. François Dautry (IFSBM, Villejuif, France). The in vivo splicing assays were performed as described using specific primers (forward: 5′-GTTCTCCACATGACACTGCTCG-3′ and reverse: 5′-CTTCTGAGGGAGTGGATGGG-3′) and 5 μg of total RNA for the detection of TNFβ pre-mRNA and mRNA by RT-PCR (Neel et al.,1995).

Acknowledgements

We thank E. Lalli, B. Bardoni, L. Monaco, S.G. Lehmann, and Y. Sunami for helpful advice and assistance. We also acknowledge all the previous members of the Paolo Sassone-Corsi laboratory (at IGBMC), especially E. Heitz and M. Rastegar, for help, discussions, and gifts of material, as well as the IGBMC oligonucleotide synthesis, sequencing, and cell culture facilities for technical support. The splicing substrate TNFβ was the kind gift of Dr. François Dautry (IFSBM, Villejuif, France), and we thank P. Berta for the SRY clone and M. Carmo-Fonseca for the U2AF65 antibody.

Ancillary