While trying to uncover the molecular mechanisms controlling steroidogenic genes expression, two independent research teams identified an AGGTCA (Ad4) motif in the promoter region of these genes that was able to bind a putative member of the nuclear receptor superfamily 12. This 53 kDa protein called Ad4BP (Adrenal 4 Binding Protein) or SF-1 (Steroidogenic Factor 1) is encoded by a cDNA that was succesively cloned by the two teams 13 and is specifically expressed in steroidogenic tissues 1. SF-1 shows high homology with the drosophila Ftz-F1 transcription factor, which controls fushi tarazu homeotic gene expression 4. The gene encoding SF-1, which is conserved amongst metazoans, 56, was called Ftz-f1. It encodes four different proteins ELP1, ELP2, ELP3 and SF-1 by using alternative promoters and splicing. Although ELPs are likely to play a repressive role on certain nuclear receptors, 78, study of mice that were selectively ablated for SF-1 has shown that it was the key activator of steroidogenic endocrine function 9.

Nuclear receptors usually display common features such as a DNA-binding domain (DBD), a ligand binding domain (LBD) and two activation domains, amino-terminal AF-1 (activation function 1) and carboxyterminal AF-2, whose activity is normaly dependent on the presence of a ligand 10 (figure 1A). In all the species studied so far, SF-1 harbors a classical DBD characterized by two Cys₂-Cys₂ zinc fingers in the N-terminal region 11. However, as opposed to a majority of nuclear receptors, SF-1 binds DNA as a monomer, in a manner reminiscent of NGFI-B (Nur 77), ROR or LRH-1/CPF binding 101213. This binding is stabilized by an A box or Ftz-F1 box which recognizes nucleotides flanking the AGGTCA nuclear receptor core binding sequence on its 5' side (figure 1B). This protein domain defines the binding specificity of a nuclear receptor, as a function of its responsive element 1214. The role of A box in SF-1 activity is illustrated by a human mutation that results in sex reversal and adrenal failure 15. Nuclear receptors usually shuttle from the cytoplasm where they bind their cognate ligands to the nucleus where they activate transcription. This shuttling is dependent on nuclear localisation signals (NLS). One NLS is present on SF-1, downstream of the DBD (amino acids 89 à 101). It is required for transcriptional activity 14. Although SF-1 harbors a putative LBD highly conserved across species, it is classified as an orphan receptor because no bona fide SF-1 ligand was identified so far 1617. Recent experiments show that SF-1 LBD helices 1 and 12 can adopt an active conformation independently of a ligand, in response to phosphorylation of a serine residue at position 203 18. A conserved AF-2 domain that recruits coactivators, is present in SF-1. It is necessary but not sufficient for SF-1 transcativating activity 14192021. In a majority of nuclear receptors, the AF-2 domain cooperates with the constitutive amino-terminal AF-1 domain in order to activate transcription. Such N to C interactions are essential for ligand-activated nuclear receptors function such as PPARγ 22 or AR 2324. SF-1 amino-terminal region upstream of the DBD is very short and does not possess a classical AF-1. In fact SF-1 AF-2 cooperates with two activating domains downstream of the DBD. The proximal activation domain (pAF: proximal activation function) overlapping the hinge region and helix H1 of the putative LBD (amino acids 187–245) is required for maximal SF-1 activity with the coactivator SRC-1. It harbors a serine residue at 203, the phosphorylation of which is essential for SF-1 activity 182025. The FP region (amino acids 78 to 172), composed of the Ftz-F1 box and of a proline-rich region, interacts with c-jun and TFIIB for maximal activity 14.

As expected from its function as an essential regulator of steroidogenesis, SF-1 is expressed in the testes and ovaries as early as their anlages appear (figure 2). In the ovary SF-1 expression rapidly decreases during development and increases after birth 262728. Though the placenta is a major steroid producing tissue during pregnancy, it only shows slight SF-1 expression, detected by RT-PCR 29, but not by in situ hybridisation 30 nor by northern-blot 31. Moreover, SF-1 knock-out mice do not show placental development nor placental steroid synthesis defects 29. Ben-Zimra et al. have confirmed that SF-1 is dispensable in the placenta by showing that AP2 can substitute for SF-1 to induce P450scc expression in this tissue 32. SF-1 expression has also been detected in the pituitary anlage, in the precursors of gonadotrope cells that control reproductive function and in the ventro-medial hypothalamus (VMH) which regulates the gonadotrope axis as well as some aspects of metabolism 333435. At last, SF-1 has also been detected in skin 36 where it is associated to steroidogenic enzymes 373839 and in the spleen 3140.

The SF-1 and ELP isoforms were knocked-out in mice by three teams simultaneously 293450. Homozygous knock-out offsprings are observed in normal proportions at birth, but die by eight days because of acute glucocorticoid and mineralocorticoid deficiency. Corticosterone concentrations are very low and correlate with a marked increase in plasma ACTH, confirming primary adrenal failure and normal function of the pituitary corticotropes, in the absence of negative feedback by glucocorticoids 950. Corticosteroid injections allow survival of knock-out mice 9. Homozygous knock-outs also show female external genitalia, regardless of their genetic sex. Internal observation of SF-1 ablated mice shows a complete lack of adrenal glands and gonads and a persistence of Müllerian ducts, regardless of the genetic sex 2950. Absence of testes in homozygous mutant male mice precludes MIS (müllerian inhibiting substance) and androgen production, which probably accounts for the observed phenotypes. It is noteworthy that the abnormalities observed at birth are not observed during precocious development. Indeed, at E10.5, when sex determination has not yet occured, the bipotential gonad is normal and colonized by primordial germ cells in knock-out animals. However at E12-E12.5, when sex determination normally occurs, mutant mice gonads regress by apoptosis. The adrenal primordium also forms and progressively regresses at E11.5 50. Selective ablation of SF-1 but not of the ELP1-3 transcripts resulted in the same phenotypes, demonstrating the essential role of SF-1 in their etiology 9.

Altogether, these results show that SF-1 is indeed required for differentiation and maintenance of the primordia for adrenals and gonads, but that its presence is not required for their early formation. The mechanisms underlying the complete degeneration of the primordia are largely unknown.

SF-1 is normally expressed in the VMH and pituitary. The histological appearance and physiology of both tissues have been studied in knock-out animals. SF-1 -/- adult mice neither express LHβ nor FSHβ, two markers of pituitary gonadotrope cells, and do not show the characteristic structures of the VMH 3334355152. These are present at E17 but progressively regress thereafter, an observation reminiscent of the situation in the gonads and adrenals 35. The cause of the absence of gonadotropins in gonadotrope cells is not clear. Absence of GnRH in SF-1 -/- mice is not probable, as hypothalamic GnRH-producing neurons are present and normally deliver it to the medial eminence of the pituitary 3553. Pituitary-specific ablation of SF-1 confirms the pituitary origin of the absence of gonadotropins. Indeed these animals show the same gonadotrope defects though their VMH is perfectly normal 535455. A direct effect of SF-1 in the pituitary is thus probable, especially when considering that it is able to control the in vitro expression of LH, FSH and GnRH receptor 5657585960616263. However, although GnRH receptor is not detected in knock-out mice 3353, supra-physiologic doses of GnRH are able to induce LH and FSH expression through an unknown SF-1-independent mechanism 3553.

Primary consequences of SF-1 ablation at the level of pituitary and hypothalamus have secondary consequences in their target organs. Indeed, pituitary-specific knock-out mice show marked hypogonadism which is characterized by a 95% decrease in male and female gonad mass and absence of sexual maturation, resulting in sterility 355354. However, hypoplastic gonads have normal morphology of immature gonads, indicating that their precocious determination and differentiation normally occur. This phenotype, which is equivalent to the phenotype of gonadotropes-ablated mice 64, can be reversed by PMSG injection, indicating that this is indeed gonadotropins absence which is responsible for gonadal hypoplasia in pituitary-specific knock-outs 53. This is confirmed by observation of an attenuated gonadal phenotype in mice with a pituitary-specific SF-1 hypomorphic allele 55.

Lesions of the ventro-medial hypothalamus obtained by stereotaxic surgery induce hyperphagia and obesity, suggesting that the VMH could participate to satiety and feeding control 65. It is noteworthy that SF-1 -/- mice that are maintained by adrenal transplantation, develop marked obesity after 8 weeks. Around 6 months, their body mass is nearly twice as wild type and this is correlated to a marked increase in adipose tissue. Although the molecular basis for such a phenotype is not known, obesity seems to be correlated to a decrease in daily exercise observed as early as 7 weeks 52. These results confirm that the VMH may be implicated in the etiology of certain obesity phenomenons, and suggest that SF-1 may have broader roles in metabolism control.

Collectively, these data indicate that SF-1 is required for the differentiation of gonadotrope cells of the pituitary and of the ventro-medial hypothalamus. Its presence does not seem to be required for formation of the VMH of which the anlage is present in knock-out mice, but may participate to its maintenance during developement.

The discovery that SF-1 is expressed in the spleen raises the question of its potential role in non-steroidogenic peripheral tissues. SF-1 ablation induces defects in the establishment of splancnic vascularization as well as defects in erythropoïesis, although most hematopoïetic cell lines are correctly matured and differentiated 40. Whether SF-1 could be implicated in the vascularization of other tissues such as the adrenals or gonads is an interesting question. A defect in the formation of the endothelium could at least in part, be responsible for the involution of gonadal and adrenal primordia in knock-out mice. It is noteworthy that a steroidogenic tissue-specific endothelial growth factor - EG-VEGF - has recently been identified 66. A potential implication of SF-1 in EG-VEGF or EG-VEGF receptor expression has not been studied so far.

SF-1 was initially identified for its capacity to interact with and activate the promoters of the steroidogenic enzymes P450SCC, CYP11B1 and CYP21 12, through the consensus AGGTCA sequence. In the last ten years, transient transfection studies have shown that SF-1 participates to the expression of all steroidogenic enzymes in the adrenal cortex and gonads (table 2). In any case except for CYP11B2 which synthesizes aldosterone in the zona glomerulosa 78, SF-1 activates basal expression of these genes. Because of its tissue-restricted pattern of expression, SF-1 is obviously a key factor in the tissue-restricted expression of steroidogenic enzymes. Its implication in their cAMP responsiveness is still a subject of debate. At least, cAMP-sensitive promoter regions often overlap with SF-1 responsive elements the mutation of which alters cAMP responsiveness 47787980818283848586. These experiments indicate that SF-1 may be required to mediate cAMP responsivenes, but they do not show that it is sufficient for this process. Indeed, inside P450scc promoter, a proximal SF-1 responsive element (-40) is solely required for basal activity, whereas a distal site (-1600) is required for hormonal sensitivity in vivo 79. If SF-1, was intrinsically a mediator of cAMP signaling, it would be difficult to understand how it would behave differently from one site to another. This underlines the essential role of the DNA context of the SF-1 responsive element in its ability to transduce activation of the cAMP pathway to promoters.

Besides steroidogenic genes, SF-1 is also able to modulate cholesterol delivery to steroidogenic reactions, by controlling expression of the HDL-receptor SR-BI, of intracellular cholesterol transporter SCP2 87, and of StAR which transfers cholesterol from the outer to the inner mitochondrial membrane 88. SF-1 can also stimulate de novo cholesterol synthesis in steroidogenic tissues through activation of HMG-CoA synthase, irrespective of intracellular cholesterol concentrations 89. SF-1 also controls expression of the receptors for ACTH and FSH, the hormones that stimulate adrenal and gonadal steroidogenesis, respectively. Our group has also demonstrated that SF-1 could also control akr1-b7 expression in vitro and in vivo. This gene whose expression is controlled by ACTH in the adrenal cortex, encodes a protein which detoxifies isocaproaldehyde, produced by the cleavage of cholesterol through action of P450scc 4290919293. It is noteworthy that a new class of genes coding proteins able to reduce toxic side-effects of steroidogenic reactions, is emerging. Indeed, expression of superoxyde-dismutase 2 (SOD2), which protects adrenal cells against free radicals generated during steroidogenesis, is also controlled by ACTH 94. However Chinn and colleagues have not studied the role of SF-1 in the control of SOD2 expression in adrenocortical cells. Altogether, these results confirm the major role of SF-1 as a key activator of steroidogenesis. However, most of these studies rely upon transient transfections in heterologous cells that do not necessarily represent a good model enough for studying the role of SF-1 on gene transcription. As SF-1 null mice completely lack steroidogenic tissues, the easiest alternative to study these mechanisms in a physiological context, is to generate transgenic mice harboring wild-type or mutant constructs of the promoter being studied.

Analyses of SF-1 null mice have shown the key role of this factor for the establishment of the hypothalamus-pituitary-adrenal and hypothalamus-pituitary-gonadal axes. Neuronal NO synthase participates to the secretion of GnRH, the neuropeptide responsible for the synthesis and secretion of LH and FSH by pituitary 95. Although SF-1 knock-out mice do not show obvious GnRH secretion defect 3553, SF-1 is able to control neuronal NO synthase gene transcription in vitro 96. This raises the question of a more general role of NO in the etiology of the phenotypes in null mice. Pituitary glycoproteins LH and FSH are composed of a common subunit, αGSU, which is associated to a β chain, specific to LH or FSH. Gonadotropins production and secretion is controlled by GnRH which activates Ca²⁺ and PKC pathways through its receptor 97. In vitro, SF-1 controls expression of the GnRH receptor, αGSU and LHβ, but not of FSHβ. All these proteins are absent from SF-1 -/- mice pituitaries suggesting that SF-1 participates to their expression in vivo 5354. However, null mice treatment with supra-physiological GnRH doses induces LH and FSH expression in the absence of detectable amounts of GnRH receptor 3553. Although the question of GnRH action in SF-1 null mice is as yet unresolved, these surprising results shed light on transcription factors other than SF-1, that may participate to LH and FSH expression in vivo.

In mammals, male sex determination is triggered by the Y chromosome-borne SRY gene. SRY activation in turn triggers expression of SOX9 which stimulates MIS transcription in Sertoli cells. This hormone from the TGF-β family, is responsible for the regression of Müllerian ducts that normally form oviducts, uterus and the upper third of the vagina in females 98.

SF-1 expression in the urogenital ridge at E9.0 26 precedes expression of SRY in pre-sertoli cells at E10.5 in mouse 99100. SF-1 has recently been shown to regulate human and porcine SRY expression. One SF-1 responsive site at -327 is required for a modest activation of human SRY promoter 101 whereas two sites are essential for porcine SRY promoter activity in porcine genital ridge cells 102. Although the physiological relevance of such an observation is not established yet, treatment of NT2/D1 embryonic carcinoma cells with cAMP induces SF-1 phosphorylation that prevents its binding to the SRY promoter, thus downregulating human SRY expression 101. Nonetheless, it is noteworthy that residual SRY expression is observed in SF-1 knock-out mice 103, suggesting that SF-1 may not be essential for triggering SRY expression.

SF-1 and MIS are coexpressed in developing Sertoli cells 104, and SF-1, in association with SOX9 105, GATA-4 106 and WT-1 107 is able to stimulate MIS promoter activity in transient transfections (figure 3). Male SF-1 null mice show Müllerian duct persistence 293450. However, the absence of gonads in these animals does not allow a conclusion on a direct effect of SF-1 to be drawn. This difficulty was overcome by mutating the proximal SF-1 responsive element of the MIS promoter by homologous recombination with the endogenous MIS locus. This mutation reduces MIS accumulation albeit to an extent which is insufficient to prevent Müllerian duct regression, whereas when the SOX9 binding site is mutated by the same procedure, there is Müllerian duct persistence in male mice (figure 3). These results suggest that SOX9 triggers MIS expression whereas SF-1 acts as a quantitative regulator 108.

Gonadal descent is a process associated to male sexual differentiation which is dependent on regression of the cranial suspensory ligament and proliferation of the gubernaculum specifically in males. The knock-out of the Insl3/RLF gene, which codes a protein produced by Leydig cells, results in a defect in testes descent. It is noteworthy that SF-1, at least in vitro, is able to control RLF expression through three responsive-elements 109110. Confirmation of the role of SF-1 in testis descent via RLF should come from observation of testes position in SF-1 +/- mice, or of gubernaculum proliferation in null mice.

The number of SF-1 potential target genes is rapidly growing. However few studies are based on the in vivo demonstration that SF-1 responsive elements are indeed required for putative target genes expression. Analysis of putative target genes in SF-1 haplo-insufficient mice may confirm the role of SF-1 in their expression. However, compensatory mechanisms as those observed for StAR expression in SF-1 +/- mice, may mask the effects of SF-1 dosage reduction 69.

The presence of gonadal and adrenal anlages in SF-1 knock-out mice and their rapid disappearance during development suggests that SF-1 regulates genes that are implicated in cell survival and/or proliferation. The majority of known SF-1 target genes is responsible for the maintenance of differentiated function rather than survival of steroidogenic tissues. Degeneration of steroidogenic tissues in null-mice is due to apoptosis 50. It was recently shown that glucocorticoids can protect glandular tissues such as ovarian follicular cells against apoptosis although they have a pro-apoptotic effect on hematopoïetic cells 111. As SF-1 controls the expression of StAR 112 and P450scc 79, two enzymes that are indispensable for glucocorticoids production, apoptosis in SF-1 null-mice may be due to glucocorticoids defects. In fact, this is rather unlikely because StAR 113 or P450scc 114 genetic ablation results in histological defects of the adrenals that are linked to progressive cholesterol accumulation, but does not result in adrenal regression during development.

POMC is the pituitary peptide that is cleaved to produce ACTH, pro-γ-MSH and β-LPH. Interestingly, POMC knock-out mice show defective adrenal development 115. The role of ACTH in trophic and mitogenic stimulation of the adrenal cortex is still a subject of intense debate 76116117. The N-POMC(1–52) peptide, derived by cleavage of pro-γ-MSH shows highly mitogenic activity on adrenocortical cells. AsP, the adrenal-specific protease which is responsible for its cleavage, and its cognate receptor have been cloned recently. These are specifically expressed in the outermost regions of the cortex and the protease is required for Y1 cells growth 76118. An important question is now to determine whether SF-1 regulates AsP and its receptor expression in the adrenal cortex.

During the last decade, numerous orphan nuclear receptors have been cloned without any known ligand. Some, like LXR, FXR, PXR or CAR, have since been attributed a ligand, whereas others like Nur77/NGFI-B, Nurr-1, LRH-1 or DAX-1, have no known ligand so far and seem to be activated by other mechanisms 510. 25-hydroxycholesterol, an hydroxylated cholesterol derivative is able to activate CYP21 promoter transcription in a SF-1 dependent manner in heterologous CV-1 cells, indicating that it could be a SF-1 endogenous ligand 119. Nonetheless, in steroidogenic MA-10 cells that naturally express SF-1, both exogenous and endogenous 25-hydroxycholesterol are unable to stimulate endogenous P450scc expression as well as six SF-1-responsive reporter genes 16. Collectively, these results tend to prove that 25-hydroxycholesterol is not a bona fide ligand for SF-1 in steroidogenic cells. This is confirmed by the ability of SF-1 LBD to adopt an active conformation independently of any ligand 18.

A small 90 bp SF-1 proximal promoter is specifically expressed in adrenocortical cells in transient transfections. This region encompasses an E-box (-87/-82), a Sp1 binding site (-30/-24) and a CAT box that binds CBF (-68/-59) 120121 (figure 4). If the role of the latter is not clearly demonstrated, transcriptional control of SF-1 expression at least requires the E-box which is conserved in human and which is functional in both steroidogenic and non-steroidogenic cells 120122123124. This element is able to bind the ubiquitous USF factor contained within pituitary αT3-1 cells, steroidogenic Y1 and JEG-3 cells, as well as CV1 and HeLa cells 123. None of these interactions however, can account for tissue-restricted SF-1 expression. SF-1 itself is able to bind a site which is present in its own first intron in rat and human genes (+156/+163) 124125 (figure 4). Whereas Nomura et al., 125 have shown the role of this sequence for the expression of a reporter gene in Y1 cells, or in response to SF-1 overexpression in heterologous CV-1 cells, Woodson et al., using the rat gene and Oba et al., using the human gene, were unable to obtain the same results in either steroidogenic or non-steroidogenic cells 120124. However, it is worth considering that some sequences contained within the first intron might be required for SF-1 expression, though their specificity is not yet established 124. This is confirmed by the use of different lengths of SF-1 regulatory regions and intragenic sequences in transgenic mice 49126. SF-1 is expressed in the urogenital ridge as early as E9.5 at similar levels in males and females. When sex determination occurs between E10.5 and E12.5, SF-1 expression strongly decreases in the ovary until E18.5, whereas it remains elevated in the testis. After birth, expression rises in the ovary although it is reduced in adult testis 2627. Pod-1/Capsulin is a transcription factor of the b-HLH family which is able to heterodimerize with other factors of the family by binding to E-boxes (figure 4). It participates to kidney and lung differentiation 127 and displays a sexually dimorphic pattern of expression in the gonad, which is reminescent of SF-1 gonadal expression. However, Pod-1 is expressed in gonadal regions where SF-1 is not expressed (i.e. coelomic epithelial cells, peritubular myoid cells and epithelial-like cells). In fact, it seems that Pod-1 may act as a repressor of SF-1 promoter activity through interaction with the previously described E-box 128. Sox9 and SF-1 are colocalized in somatic cells of the testis and follow parallel expression patterns during development 129. They both participate to transcriptional activation of the MIS promoter in males 105. Recent results show that Sox9 is able to induce SF-1 expression in heterologous cells and that a Sox9 binding site (-110/-104) is required for SF-1 expression in Sertoli and Y1 cells 130 (figure 4). GATA-4 is also dimorphically expressed in the gonad. Whereas its expression is high in the undetermined gonad it decreases as ovary differentiation starts, although it is maintained in the testis. This dimorphism may participate to the control of MIS expression 131, but GATA-4 is also able to moderately activate SF-1 expression via a conserved site at -177/-172 (figure 4). This activation is dependent on the cell type and seems to be essentially restricted to Sertoli cells where SF-1 participates to MIS expression 132. Although most elements required for SF-1 expression in steroidogenic tissues are likely to be localized in the 5' flanking regions and first intron of the gene, a GFP/SF-1 fusion, the expression of which is directed by a 50 kb BAC comprised of SF-1 promoter, first exon and first intron, is not expressed in pituitary gonadotropes 49. Although a single transgenic line was studied so far, this indicates that further downstream sequences might be required for pituitary expression. However, there is no mention of pituitary expression in the work of Zubair et al., who used SF-1 regulatory regions extending from the first intron to the seventh exon 126. Altogether, these data allow a better understanding of SF-1 tissue-specific expression especially in the gonads but do not establish a link between SF-1 transcription and the cAMP signalling pathway.

In vivo, SF-1 protein accumulation in the adrenals and gonads is unchanged by a four weeks hypophysectomy in rats 133. However an eighty hours treatment of mice with dexamethasone, induces a marked reduction of SF-1 mRNA accumulation in the adrenals 134, suggesting that compensatory mechanisms may have masked the effects of long term hypophysectomy on SF-1 accumulation. Nonetheless, lipopolysaccharide treatment which induces increases in circulating ACTH concentrations, has no effect on SF-1 accumulation 134. The problem is far more complex in cell culture systems. Whereas numerous papers show that SF-1 mRNA accumulation is unchanged in Y1, MA-10, theca or bovine granulosa cells in response to forskolin or PKA catalytic subunit overexpression 135136137138, one paper describes a slight increase in bovine adrenocortical cells treated with ACTH 139. At last, in human granulosa cells 140 or in forskolin-treated Y1 cells, SF-1 protein accumulation increases independently of an increase in its mRNA accumulation 137. Based on this observation, Aesoy et al. have proposed a post-translational model in which PKA could stabilize SF-1 protein 137. However, this was only demonstrated in a heterologous cell system overexpressing both SF-1 and the PKA catalytic subunit. Interestingly, we have not been able to obtain similar results with both Y1 and ATC-1 141 adrenocortical cell cultures treated with forskolin or ACTH, respectively. Our unpublished data rather suggest that SF-1 accumulates into cell nucleus in response to cAMP, without a concomitant increase in overall protein accumulation (Bruno Ragazzon, personal communication). This increase in SF-1 nuclear accumulation correlates with increased SF-1 binding in gel shift experiments (Christelle Aigueperse, personal communication). Whatever the changes in SF-1 expression may be, simple modulations of SF-1 accumulation are unlikely to account for some SFREs being implicated in cAMP-responsiveness while others only support basal promoter activity in vivo 4279. This differential activity may be achieved by more subtle mechanisms that control SF-1 activity, such as interaction with other transcription factors and / or cofactors.

Ligand-dependent nuclear receptors activate their target genes transcription through interactions with coactivators and/ or corepressors that link receptors to the transcription machinery. Accordingly, SF-1 harbors an AF2 activation domain in its LBD (figure 1B). This motif (LLIEML, consensus LLXXL) is necessary but not sufficient for transactivation 14192021 which also depends on two amino-terminal regions of the protein, the FP region 14 and a proximal activation domain 2025. SF-1 proteins bearing mutations in their AF2 domain have dominant negative properties on CYP17 promoter activation by PKA in Y1 cells. This suggests that SF-1 AF2 is implicated in the transduction of the cAMP signal 83. As ligand-activated nuclear receptors, SF-1 interacts with numerous coactivators such as SRC1 1920, RIP140 142, PNRC and PNRC2 143144, hMBF1 145, TIF2 25, p/CIP 146 and GCN5 147. These interactions, independent of an exogenous ligand, are dependent on AF-2 and for some of them, on proximal interaction domain integrity. Most of these interactions are mediated by LXXLL motifs found on coactivators. PNRC and PNRC2 are quite unique in that they interact with SF-1 and other nuclear receptors through SH3 proline-rich motifs 143144. None of these coactivators is specific for SF-1 and none of them shows a steroidogenic tissue-restricted pattern of expression. Nonetheless, two of them may be required for integration of the cAMP signalling (figure 5). TIF2 and p/CIP interact with SF-1 through the AF-2 and proximal activation domain 25146. Overexpression of TIF2 or p/CIP in heterologous or Y1 cells, stimulates transcription of a reporter gene driven by four copies of a SF-1 responsive-sequence of the bovine CYP17 promoter. However, whereas p/CIP increases sensitivity to PKA overexpression in the presence of SF-1, overexpression of the catalytic subunit of the PKA inhibits potentiation of SF-1 activity by TIF2, through a decrease in TIF2 protein accumulation 146. In a more physiological system, one might imagine that on the sites where it participates to cAMP-responsiveness, SF-1 would preferentially associate to p/CIP, whereas where it participates to basal promoter activity, SF-1 would rather associate to TIF2. What could allow SF-1 to choose between those two coactivators? Although the sequence of the SF-1 responsive element may participate to this choice, it is possible that adjacent transcription factors, endowed with cAMP sensing capacity, may modulate cofactors recruitment by SF-1. This would allow modulation of SF-1 activity in response to extra-cellular signals.

DP103 is a DEAD-box protein which is highly expressed in steroidogenic tissues (table 3). Although it has intrinsic RNA helicase properties as other proteins of its family 148, DP103 physically interacts with SF-1 through a newly described repressive domain, located in the vicinity of the proximal activation domain 149. DP103 harbors a C-terminal repression domain that by itself, represses SF-1 activity. DP103 repressive function is independent of its helicase activity and can decrease P450scc and P450c21 promoter activity, inducing a significant reduction in progesterone production by Y1 cells 148. One interesting question is now to analyze whether DP103 is implicated in the cAMP responsiveness of steroidogenic enzymes genes.

Apart from interaction with bona fide cofactors, SF-1 also interacts with numerous transcription factors (table 3) that modulate its activity either by binding to adjacent DNA sequences or by interacting with SF-1 without binding to DNA 106107150151152. The regions of interaction between SF-1 and other transcription factors are rather broadly delineated. These can overlap with previously described regions such as AF-2 or the proximal activation domain 14160 but also extend to the DBD 107153 and LBD 107154 or distal 152 and proximal 148 repression domains. These interactions result in variable responses depending on the promoter or cell type under experiment.

A first class of factors, illustrated by the studies on MIS and LH-β promoters, encompasses proteins that will restrict SF-1 target genes expression in a very narrow region of the organism in a particular context. Indeed, the combinatory interactions between SF-1, WT-1 107, GATA-4 106150, SOX-9 105108 and DAX-1 107 allow MIS expression specifically by Sertoli cells of the male gonad, just before regression of the Müllerian ducts (figure 3). Also, this is an interaction between Egr-1, Ptx1 and SF-1 that allows expression of the LH-β subunit in pituitary gonadotrope cells in response to GnRH stimulation 155156157158159.

Another group of factors is more implicated in SF-1 target genes response to external stimuli. Indeed, SF-1 interaction with TReP132 160161, Sp1 153162 or CREB 86163164 allows recruitment of the CBP/p300 coactivator which interconnects multiple cell signalling pathways 165. This may participate to the cAMP responsiveness of some SF-1 target genes.

At last, a third interaction group allows repression of SF-1 activated genes. DAX-1 is the paradigm of such repressors. It encodes a peculiar nuclear receptor devoid of a classical DNA binding domain, which is replaced by three-and-a half repeats of an alanine and glycine-rich motif 166167. DAX-1 overexpression subsequent to Xp21 duplication, induces male to female sex reversal in humans 168169, a phenotype that can be mimicked by Dax-1 overexpression in transgenic mice with a poor Sry allele 170. DAX-1 mutations in humans are responsible for adrenal hypoplasia congenita as well as hypogonadotrophic hypogonadism suggesting that DAX-1 may perform similar functions as SF-1 171. Indeed, SF-1 and Dax-1 are coexpressed in steroidogenic tissues as well as in the VMH and pituitary, early in development 47172173. However, when overexpressed in Y1 adrenocortical cells, DAX-1 markedly impairs steroidogenic output and transcriptionaly represses the promoters of SF-1 target genes such as StAR, P450scc, 3β-HSD 174175 and akr1-b7 92. Essentially two mechanisms have been proposed for DAX-1 mediated transcriptional repression, although they are probably not mutually exclusive. DAX-1 can repress SF-1 target genes expression either by binding DNA to hairpin-like structures 176, or by physically interacting with SF-1, independently of the DNA context 107. This interaction implies a carboxy-terminal repression domain (437 to 447) and a proximal interaction domain of SF-1 (226 to 230) 152 (figure 1) that contact amino-terminal LXXLL domains of DAX-1 151177. This physical interaction allows the recruitment of the corepressors NcoR 152 and Alien 178 to SF-1-responsive genes promoters. The physiological relevance of such a negative functional interaction between Dax-1 and SF-1 is as yet unclear as Dax-1 knock-out in mice does not lead to marked adrenal defects 179. However, it is noteworthy that Dax-1 ablation in SF-1 haploinsufficient mice allows a reversion of the histological and functional adrenal defects, suggesting that the two receptors interact in vivo 180. Furthermore, we have been able to observe Dax-1 downregulation in either ACTH-treated mice or adrenocortical ATC-1 cells that naturally express Dax-1, indicating that this receptor may be implicated in hormonal responsiveness in vivo (B. Ragazzon, personal communication).

An unexpected repressor of SF-1 activity is the androgen receptor. The increase in plasmatic LH concentrations which is induced by GnRH leads to increased sex steroid production by the gonads. In turn, sex steroids exert a negative feedback on GnRH and LH synthesis (figure 6). Recently, Jorgensen and Nilson, have elegantly demonstrated that androgen receptor was able to repress LH-β promoter transactivation by interacting with SF-1 LBD 154. Under low GnRH conditions, AR blocks the functional interaction between SF-1 and Ptx1/Egr-1. When GnRH concentrations in pituitary gonadotropes increase, Egr-1 expression is induced 156 and allows recruitment of Ptx-1 as well as AR displacement, resulting in the formation of an activating complex composed of SF-1, Egr-1 and Ptx1. When circulating androgens increase, activated AR displaces Egr-1 and Ptx1, thus repressing LH-β transcription in response to SF-1 154. This model (figure 6) perfectly illustrates the complex interactions that are required fo SF-1 target genes activation. It is likely that such mechanisms may participate to the control of SF-1 target genes expression in response to cAMP increases.

Although it is clear that SF-1 transcriptional activity requires complex interactions with numerous cofactors, the mechanisms underlying the recruitment of these partners are still unclear. Because PKA is implicated in the stimulation of most of SF-1 target genes expression, its role in SF-1 activity has been extensively investigated. In vivo, SF-1 is phosphorylated in response to granulosa cells stimulation by FSH 86. In vitro, PKA can phosphorylate SF-1 LBD and N-terminal region 101135. Although SF-1 seems to be implicated in the cAMP responsiveness of the rat CYP17 gene, its in vitro phosphorylation by PKA decreases its DNA binding capacity, an observation which is not compatible with its activating role 101135. In heterologous cell systems SF-1 is phosphorylated on serine 203 by the MAPK ERK2, regardless of the presence of cAMP. This residue is situated in the proximal activation domain, the integrity of which is essential for SF-1 activity. Serine 203 phosphorylation is required for SF-1 transcriptional activity and allows the recruitment of two cofactors, the coactivator GRIP1/TIF2 and the corepressor SMRT 25. Recently, Desclozeaux et al., have shown that SF-1 hinge and helix 1 of the LBD were able to set helices 2 to 12 of the LBD in an active conformation reminiscent of ligand-activated nuclear receptors, albeit in the absence of a ligand. Helices 2 to 12 recruitment is enhanced by MAPK stimulation and decreased by mutating serine 203, or in the presence of MKP-1, a MAPK specific phosphatase. At last, serine 203 phosphorylation stabilizes SF-1 LBD, an observation reminiscent of ligand-activated nuclear receptors 18. This shows that apart from being implicated in cofactor recruitment 25 serine 203 phosphorylation also structures and stabilizes SF-1 LBD. However, these experiments do not establish a link between cAMP stimulation and SF-1 activity. This may depend on activation of the MAPK pathway by PKA 181. Indeed, cAMP-induced StAR transcription is dependent on activation of the MPAK pathway in Y1 and MA-10 cells. This activation induces SF-1 phosphorylation in vivo in Y1 cells, resulting in an increase in SF-1 binding in EMSA 182. However, cAMP-induced P450scc expression does not seem to depend on MAPK activation in the same experiments 182. Furthermore, cAMP pathway stimulation in H295 cells results in a decrease in SF-1 overall phosphorylation 183, indicating that PKA/MAPK crossovers may be gene and cell-specific. Thus in H295 cells, MAPK inhibition by the specific inhibitor PD98059 does not reduce, but on the contrary, stimulates hCYP17 transcription 183. Indeed, StAR and hCYP17 cAMP-stimulated transcription in H295 cells is dependent on protein phosphatase activities 183184. In H295 cells, MKP-1 that can be phosphorylated by PKA in vitro, is overexpressed in response to cAMP stimulation. MKP-1 overexpression leads to an increase in hCYP17 expression, whereas MKP-1 inhibition by an antisense RNA prevents hCYP17 induction by cAMP. Whether MKP-1 is directly implicated in SF-1 activity is still unclear. However, it is noteworthy that phosphatase inhibitors decrease SF-1 transcriptionnal activity on the hCYP17 promoter, indicating that at least one phosphatase is required for promoter activation by SF-1 185.

Altogether, these results seem rather contradictory. Although it is probable that the MAPK pathway influences SF-1 activity, it is still unclear whether this implicates direct phosphorylation of SF-1 by a MAPK. Also, it seems that such a mechanism would not necessarily apply to all cell types or promoters. One pitfall of these experiments is that the MAPK pathway is highly sensitive to extracellular changes and temporal variations in the experimental setting. Completely different experimental conditions may thus account for the contradictory observations. At last, future experiments may distinguish between overall SF-1 phosphorylation and phosphorylation on specific residues such as serine 203. When these mechanisms are decyphered, it will be interesting to study the effect of one particular phosphorylation or dephosphorylation on the recruitment of cofactors such as p/CIP versus TIF2 in response to cAMP stimulation.

Another emerging activation control mechanism for nuclear receptors is their acetylation 186. SF-1 is acetylated in vivo in a heterologous cell system and interacts with the acetyl-transferase GCN5, which can acetylate SF-1 in vitro. Although SF-1-dependent activation of a reporter gene may depend on SF-1 acetylation in the presence of GCN5, experimental results obtained by mutating the potential acetylation sites or by the use of the deacetylase inhibitor trichostatin A, are contradictory 147. If more substantial data is obtained regarding the role of acetylation on SF-1 activity, it will be interesting to study the implication of certain histone acetyl transferase (HAT) coactivators such as CBP/p300, in SF-1 acetylation in vivo.