Two-Step Regulation of Ad4BP/SF-1 Gene Transcription during Fetal Adrenal Development: Initiation by a Hox-Pbx1-Prep1 Complex and Maintenance via Autoregulation by Ad4BP/SF-1

Abstract

The orphan nuclear receptor Ad4BP/SF-1 (adrenal 4 binding protein/steroidogenic factor 1) is essential for the proper development and function of reproductive and steroidogenic tissues. Although the expression of Ad4BP/SF-1 is specific for those tissues, the mechanisms underlying this tissue-specific expression remain unknown. In this study, we used transgenic mouse assays to examine the regulation of the tissue-specific expression of Ad4BP/SF-1. An investigation of the entire Ad4BP/SF-1 gene locus revealed a fetal adrenal enhancer (FAdE) in intron 4 containing highly conserved binding sites for Pbx-Prep, Pbx-Hox, and Ad4BP/SF-1. Transgenic assays revealed that the Ad4 sites, together with Ad4BP/SF-1, develop an autoregulatory loop and thereby maintain transcription, while the Pbx/Prep and Pbx/Hox sites initiate transcription prior to the establishment of the autoregulatory loop. Indeed, a limited number of Hox family members were found to be expressed in the adrenal primordia. Whether a true fetal-type adrenal cortex is present in mice remained controversial, and this argument was complicated by the postnatal development of the so-called X zone. Using transgenic mice with lacZ driven by the FAdE, we clearly identified a fetal adrenal cortex in mice, and the X zone is the fetal adrenal cells accumulated at the juxtamedullary region after birth.

The adrenal 4 binding protein/steroidogenic factor 1 (Ad4BP/SF-1) (NR5A1) (42) is an essential transcription factor required for reproduction and endocrine organogenesis (39, 44, 57). Adrenal glands and gonads do not develop in mice with a targeted disruption of the Ad4BP/SF-1 gene (32, 49). Additionally, the ventromedial hypothalamus (VMH), pituitary gonadotrope, and spleen developed abnormally in the mice (22, 38, 54). Moreover, the Ad4BP/SF-1 gene dosage appears critical for adrenal development; heterozygous mouse embryos developed hypoplastic adrenal glands (7) and displayed aberrant early gonadal development (43). Humans appear more sensitive to reduced Ad4BP/SF-1 expression because individuals with a single mutated copy of this gene exhibit defective adrenal and/or gonadal development and function (2, 5, 11, 15). Because the gene dosage dramatically affects the target tissues, it is essential for the body to tightly control levels of Ad4BP/SF-1 gene expression.

Several factors regulating the expression of the Ad4BP/SF-1 gene have been identified. An E-box binding site within the basal promoter is critical for Ad4BP/SF-1 expression (41). Using reporter gene assays, Pod1/capsulin was indicated to suppress Ad4BP/SF-1 gene transcription through binding to the E box (56). The disruption of the Pod1/capsulin gene confirmed the suppressive function of the E box (12). A study using transgenic (Tg) mice harboring the 5′ upstream region of Ad4BP/SF-1 indicated a role for Wt1 in controlling the gene expression (59), and Sox9 is also required for Ad4BP/SF-1 gene expression in cultured cells (52). Finally, gene disruption of Lhx9 (6) clearly indicated that this transcription factor lies upstream of Ad4BP/SF-1. Although these factors are assumed to regulate Ad4BP/SF-1 gene transcription, studies of the factors all focused on the basal promoter region but not the upstream or downstream regulatory region. Recently, Shima et al. (53) demonstrated for the first time that a distal enhancer in intron 6 that is conserved among animal species is important for VMH-specific Ad4BP/SF-1 gene transcription. However, the regulatory regions for the other tissues remained elusive.

PBX1 and its Drosophila equivalent, EXD, are three-amino-acid loop extension (TALE) class homeodomain proteins that heterodimerize with other members of the TALE homeodomain subfamily, PREP, MEIS, and HTH (8). Heterodimerization modifies both transcriptional activity and subcellular localization (1, 34, 45, 47). Additionally, PBX interacts with HOX to activate gene transcription. Interestingly, the transcriptional activity of the PBX-HOX complex is enhanced through interactions with PREP or MEIS (13, 23, 46, 50). Targeting studies demonstrated a critical role for Pbx1 as a developmental regulator for multiple tissues. Notably, the gonads of Pbx1-deficient mice exhibited only rudimentary sexual differentiation and decreased cell proliferations, while the adrenal glands did not develop at all. Consistent with other data, Ad4BP/SF-1 expression in the adrenal primordium was virtually undetectable in the mice, suggesting that Pbx1 is an essential upstream regulator of Ad4BP/SF-1 during adrenal development (51).

The adult adrenal gland consists of two functionally distinct cell layers, the outer cortex and inner medulla. Cortex development and differentiation gradually occur after birth, and the fetal cortex is comprised of cells that are fundamentally different from those of the adult cortex. During the transition from fetal cortex to adult cortex, the fetal cortex undergoes extensive regression, while cells of the adult cortex proliferate to occupy the entire cortical region. Although this transition has been confirmed in various mammalian species, in mice it has been controversial (27).

Given the essential role of Ad4BP/SF-1 in adrenal development, we attempted to determine the distal regulatory sequences governing adrenal-specific gene expression by using Tg mouse assays. Successfully, an enhancer was identified in intron 4 that drives Ad4BP/SF-1 gene transcription in a fetal adrenal-specific manner by two successive steps, initiation and maintenance, which are achieved through the Hox-Pbx1-Prep1 complex and Ad4BP/SF-1, respectively. Moreover, we used Tg mice expressing lacZ driven by this enhancer to demonstrate a link between the fetal adrenal cortex and the X zone in the mouse adrenal gland.

MATERIALS AND METHODS

DNA construction for Tg assay and FISH analysis.

The mouse Ad4BP/SF-1 gene has multiple noncoding first exons, Ia, Ig, and Ix (28), and the first methionine is located 11 bp upstream of the SplI site in the second exon. A 5.8-kb KpnI-SplI DNA fragment which is comprised of the 1.9-kb upstream region, exon 1, intron 1 (containing exons Ig and Ix), and the region of exon 2 upstream of the SplI site, was used as a basal promoter. β-Galactosidase (lacZ), together with the simian virus 40 poly(A) signal, was inserted into the SplI site to generate the Ad4BP-lacZ cassette. The Ad4BP-lacZ cassette was introduced into the NotI site of the SuperCos-1 cosmid vector (Stratagene) to produce cAd4BP-lacZ (Fig. 1A). The cassette was also introduced into pBluescript (Stratagene) to produce pAd4BP-lacZ. Yeast artificial chromosome (YAC) (approximately 400 kb) and bacterial artificial chromosome (BAC) (approximately 200 kb) clones encompassing the entire Ad4BP/SF-1 gene were purchased from GenomeSystem, Inc. and Incyte Genomics, respectively. DNAs prepared from the YAC and BAC clones were partially digested with Sau3AI and inserted into cAd4BP-lacZ cosmid vector. Randomly selected cosmid clones were subjected to PCR analyses using primers specific for Gcnf exon 5 (GGAGCCACATTACCACGTTTC and GCTGTCCTGGAATTCACTATG), Gcnf exon 9 (GACCTGGGAACCGGAACTTAC and GCTCTTGCCACCACCTACTCA), Ad4BP/SF-1 exon 4 (TGGCTGGCTACCTCTATCCTG and TAAAGACCATGCACCTTCGTG), Ad4BP/SF-1 exon 7 (ATGCCACTGCCTCCAAAAGAC and GGGTTAGGGCAGGAATGTTGG), and Psmb7 exon 1 (TGGGGTTGCCTTGTCCTCGTG and CTGGGAAGGGGAAACTGACGA). Both ends of the inserted genomic DNA were sequenced to confirm the regions included in the clones. Ad4, Pbx/Prep (PP), and Pbx/Hox (PH) sites in the fetal adrenal enhancer (FAdE) were mutated with oligonucleotide used in the electrophoretic mobility shift assay (EMSA) described below. All mutated fragments were subcloned into the SpeI/SalI site of pAd4BP-lacZ (Fig. 1). An EGFP DNA fragment (BamHI/SspI) was prepared from pEGFP-N1 (Clontech), blunt ended, and inserted into the SplI site to generate pAd4BP-EGFP. The Hsp68 basal promoter (29) with the EGFP gene (hsp68-EGFP) was prepared from the plasmid nestinhsp68EGFP (26). The promoter fragment (SacI/NcoI) was ligated to lacZ to construct hsp68-lacZ. These reporter fragments were fused to FAdE to construct FAdE-hsp68-lacZ or FAdE-hsp68-EGFP. Fluorescence in situ hybridization (FISH) analysis was performed using fluorescein isothiocyanate-labeled YAC DNA containing the Ad4BP/SF-1 gene as a probe as described previously (40).

FIG. 1.

Localization of FAdE in intron 4 of the Ad4BP/SF1 gene. (A) Ad4BP-lacZ and Ad4BP-EGFP cassettes carrying lacZ and EGFP, respectively, the simian virus 40 poly(A) signal, and the 5.8-kb fragment were constructed as described in Materials and Methods. The Ad4BP-lacZ cassette was placed into cosmid and plasmid vectors to construct cAd4BP-lacZ and pAd4BP-lacZ, respectively (indicated in panel E). Genomic DNAs prepared from YAC and BAC clones were ligated into the SalI site downstream of the lacZ reporter genes. (B and C) The Ad4BP-lacZ (B) and Ad4BP-EGFP (C) cassettes were used for transient Tg mouse assays at E11.5. Arrows indicate gonads (Go). (D) To examine whether chimeric recombination occurred in the YAC clone, FISH analysis was performed using DNA prepared form the YAC clone as a probe. Arrowheads indicate hybridization signals. (E) Cosmid and plasmid clones used for the Tg assays are presented schematically. The structure of Ad4BP/SF1, together with the Gcnf, Gpr144, and Psmb7 genes, is shown at the top. Cosmid clones (cGcnf4, cGcnf5, cIA3, cIVC6, and cIIE6) and plasmid clones (SB7.5, HB, SH4.5, XhH, KH, SA, SXh, and SacK0.6) are indicated. Red circles indicate the location of FAdE, while closed boxes indicate exons. S, SalI; Xh, XhoI; Sac, SacI; K, KpnI; A, ApaI; H, HindIII; B, BamHI. (F) The Tg constructs listed were used for Tg assays, and the expression of lacZ was examined at E12.5 or E13.5. The total number of Tg fetuses examined (Tg) and the number of fetuses showing lacZ staining at the fetal adrenal gland (Ad) or VMH are summarized. nd, not determined. (G) The lacZ expression patterns of the mouse fetuses carrying cGcnf5, cIA3, SH4.5, or SacK0.6 are shown. The activity of the SacK0.6 fragment was examined with the hsp-lacZ cassette (hsp-lacZ-SacK0.6). Arrows indicate adrenals (Ad) and gonads (Go). (H) The homology plot of intron 4 between mouse and human is shown. The sequences of intron 4 of the mouse (AL844842, 115558-107693) and human (AL844842, 14055-4256) Ad4BP/SF-1 genes were compared. Arrows indicate the locations of exon 4 and FAdE.

The generation of Tg mice and lacZ staining.

Tg mice were generated as described previously (19). The cosmid and plasmid DNAs were digested with NotI and NotI/SalI, respectively, and the linear DNA fragments were used for microinjection. The founder animals were subjected to PCR with primers for lacZ and enhanced green fluorescent protein (EGFP). The primers for lacZ were GCCGAAATCCCGAATCTCTATC and GATTCATTCCCCAGCGACCAG, while those for EGFP were GAGCTGGACGGCGACGTAAAC and CACCTTGATGCCGTTCTTCTGC. LacZ activities of the fetuses were examined as described previously (19). Animals were maintained and cared for in accordance with the guide by the Institutional Animal Care and Use Committee of the National Institute for Basic Biology. All experimental protocols were approved by the committee.

Immunohistochemistry.

Samples were fixed overnight in phosphate-buffered saline (PBS) containing 4% paraformaldehyde (PFA) at 4°C and embedded in paraffin. Five-micrometer sections were incubated with rabbit anti-Ad4BP antiserum (1/2,000) (37). After being washed three times with PBS containing 0.1% Triton X-100 for 10 min, the sections were incubated with Cy3-anti-rabbit immunoglobulin G secondary antibody (1/250; Jackson ImmunoResearch) for 1 h, washed three times with PBS containing 0.1% Triton X-100 for 10 min, and then incubated with the anti-β-galactosidase (1/1,500) antibody. Alexa-anti-rabbit immunoglobulin G (1/250; Molecular Probes) was used as the second antibody, and samples were washed three times with PBS containing 0.1% Triton X-100 for 10 min.

Electrophoretic mobility shift assays.

Double-stranded oligonucleotides containing 5′ protruding ends were labeled with ³²P-dCTP and Klenow polymerase. The nucleotide sequences for the sense strands of the probes were GCCTGTTGGCCTTGGCTGCC for Ad4-site 1, GCCTGTTGGaaTTGGCTGCC for mAd4-site 1, GCTGTGTCAAGGCTCCAAGAT for Ad4-site 2, GCTGTGTCAAttCTCCAAGAT for mAd4-site, GACTGCAAATGACCTTGTTCC for Ad4-site 3, GACTGCAAATGAaaTTGTTCC for mAd4-site 3, GTTCCCAAGGTGGCAAGAGG; for Ad4-site 4, GTTCCCAAttTGGCAAGAGG; for mAd4-site 4, AGTGCGTCTGTCAGTACTGGGCTGATAAATAATTGCTCAGCC for PP/PH, AGTGCGTCTcgagGTACTGGGCTccTAccccATTGCTCAGCC for mPP/mPH, AGTGCGTCTGTCAGTACTGGGCTccTAccccATTGCTCAGCC for PP/mPH, AGTGCGTCTcgagGTACTGGGCTGATAAATAATTGCTCAGCC for mPP/PH, AGTGCGTCTGTCAGTACTGGG for PP, AGTGCGTCTcgagGTACTGGG for mPP, GTACTGGGCTGATAAATAATTGCTCAGCC for PH, and GTACTGGGCTccTAccccATTGCTCAGCC for mPH. Lowercase letters in the nucleotide sequences indicate mutation sites. All proteins, Ad4BP/SF-1, Pbx1a, Pbx1b, and Prep1 were prepared using a coupled TNT transcription and translation kit (Promega). Pbx1 and Prep1 were translated simultaneously to facilitate complex formation. Two-microliter reticulocyte lysates containing Pbx1/Prep1 with or without 2-μl reticulocyte lysates containing Hox proteins were incubated with ³²P-labeled oligonucleotide probe in 15 μl binding buffer as described previously (3). After 30 min of incubation on ice and 60 min thereafter at room temperature, DNA-protein complexes were separated (36).

cDNA cloning for Pbx1, Prep1, and Hox expressed in fetal adrenal glands.

cDNAs for Pbx1a, Pbx1b, Prep1, and Hox were cloned by reverse transcription- PCR (RT-PCR). The primers used were as follows: P1-PBX1a/S (354)-BamHI (gttcggatccGAGCCTTCAGAGATGGACGAG) and P2-PBX1a/AS(1737)-HindIII (gccccaagcttGCTTCGACCTCCAGTCTGAC) for Pbx1a, P1-PBX1a/S (354)-BamHI (gttcggatccGAGCCTTCAGAGATGGACGAG) and P3-PBX1b/AS(1633)-HindIII (gccccaagcttCCTGCGGACTGTACATCTGAC) for Pbx1b, P1-Prep1/(57)S-SplI (ccgtacgTCTCACTAACCATGATGGCG) and P2-Prep1/(1412)AS (GCAAACAGTGAAGAGTCCAGC) for Prep1, mHoxA2/S(178)EcoRI (cctagaattcGGAGAAGGCCATGAATTACGA) and mHoxA2/AS(1344)Xba (gccgtctagaCAAAGCCACCTGGTCAAAGGC) for HoxA2, mHoxA7/S(1066)Hind (ctgcaagcttGGGGTTTGGTGTAAATCTGGG) and mHoxA7/AS(1878)XbaI (gccgtctagaTTTCCAACTGTCCTGTGCAGC) for HoxA7, mHoxA9/S(995)EcoRI (cctacgaattCTGCGCGATCCCTTTGCATA) and mHoxA9/AS(2065)XbaI (gccgtctagaGGCTCACTCGTCTTTTGCTCG) for HoxA9, mHoxB5/S(88)Kpn (ggtaggtaccTCCAAAATCACCCAAATGAGC) and mHoxB5/AS(899)XbaI (gccgtctagaGATGGGCTCAAGGTTGGAAGG) for HoxB5, mHoxB9/S(12)Hind (ccgcaagcttAGCGCGCGGATAATGTCTGAG) and mHoxB9/AS(906)XbaI (gccgtctagaGACAGGCTCCAGGTGCAGATG) for HoxB9, mHoxC5/S(68)Hind (ccgcaagctTCCCCGCCATGAGCTCCTACG) and mHoxC5/AS(857)ApaI (tatagggcccATGCGCTTTTGTCTGGGAGGC) for HoxC5, mHoxC6/S(107)Hind (ccgcaagcttCGACCAGGTAAAGGCAAAGGG) and mHoxC6/AS(880)XbaI (gccgtctagaAGTGTGTGATTCGGGGAGCTG) for HoxC6, and mHoxC8/S(220)EcoRI (cctacgaattCGCGGGTTTTCATGTACCCAG) and mHoxC8/AS(1001)XbaI (gccgtctagaAACTTCAAGGGAGTTGCTGGG) for HoxC8. The nucleotides indicated by lowercase letters were artificially added for cloning. cDNAs encoding Pbx1a, Pbx1b, and Prep1 were cloned into pCMX (provided by Kazuhiko Umesono), while cDNAs coding for Hox were cloned into pcDNA3.1 (Invitrogen) to construct expression plasmids.

RNA preparation from fetal adrenal glands and RT-PCR.

Tg mice carrying FAdE-hsp68-EGFP were used to isolate adrenal primordial cells. The EGFP-labeled cells in the Tg fetuses at embryonic day 10.0 (E10.0) were dissected and incubated in 100 μl PBS containing 0.1% trypsin, 0.1% bovine serum albumin, and 0.75 mM EDTA for 15 min at 37°C. After the addition of 100 μl PBS containing 20% calf serum, the EGFP-labeled cells were collected. Total RNA prepared from the cells was used for cDNA synthesis. A region corresponding to the first to third alpha-helices in the homeodomain was amplified using the degenerate PCR primers, Hox/S (cgcatggatcCARNYSNTRGARYTRGARAARGART) and Hox/AS (acggatctcgagTTCAYBCKNCGRTTYTGRAACCA) (R, A/G; Y, C/T; S, C/G; N, A/C/G/T; B, T/C/G; K, T/G). The amplified fragments were subcloned between the BamHI and XhoI sites of pBluescript and subjected to sequencing analysis. The Hox gene expression was also examined with RNA prepared from the adrenal glands of E13.5 fetuses. RT-PCRs for Pbx1 and Prep1 were performed with total RNA prepared from adrenal primordial cells at E10.5. The PCR primers for Pbx1 were GAGCCTTCAGAGATGGACGAG and GAACTTGCGGTGGATGATGCTC, while those for Prep1 were CAGCACATAGGGCATCCCTA and CAAACAGTGAAGAGTCCAGC.

RESULTS

Localization of tissue-specific enhancers in the Ad4BP/SF-1 gene locus.

The Ad4BP/SF-1 gene has multiple noncoding first exons (Ia, Ig, and Ix), and exon usage is differentially regulated among tissues (28). Therefore, we examined the transcriptional activity of a 5.8-kb DNA fragment upstream of exon 2. This 5.8-kb fragment was ligated to lacZ to produce the Ad4BP-lacZ cassette, and we generated Tg mice with this construct (Fig. 1A). Mouse fetuses 11.5 to 13.5 (E11.5 to E13.5) were examined. LacZ was not induced in 12 fetuses examined at E11.5 (Fig. 1B). Interestingly, however, when EGFP replaced the lacZ reporter gene (Ad4BP-EGFP), EGFP expression was observed in 4 of 11 Tg fetal gonads at E11.5 (Fig. 1C). Although the two constructs, Ad4BP-lacZ and Ad4BP-EGFP, gave inconsistent expression patterns, the 5.8-kb region was unable to drive the reporter gene expression in the adrenal cortex, pituitary gonadotrope, and VMH areas (data not shown), where Ad4BP/SF-1 is expressed endogenously. It is possible that the observed differences between Ad4BP-lacZ and Ad4BP-EGFP could arise from the increased sensitivity of detection of EGFP. Thus, lacZ could have been produced but at levels below the threshold of detection. While less sensitive, the Ad4BP-lacZ construct was expected to be appropriate for the identification of tissue-specific enhancers in the Ad4BP/SF-1 gene.

We expected the Ad4BP/SF-1 tissue-specific enhancers to lie within the gene locus, and consequently we used YAC (approximately 400 kb long) and BAC (approximately 200 kb long) clones containing this locus as starting materials. YAC clones frequently recombine unexpectedly. Therefore, we examined whether the YAC clone carried a single chromosomal locus by FISH analysis. As shown in Fig. 1D, a FISH signal was detected on only chromosome 2, indicating that chimeric recombination did not occur in this clone. Using these YAC and BAC clones, cosmid libraries were generated with the cosmid vector Scos1 carrying the Ad4BP-lacZ cassette (Fig. 1A). Several overlapping cosmid clones were isolated covering approximately 120 kb of the genomic region (Fig. 1E). These cosmid clones were used to produce Tg mice, and lacZ reporter gene expression was examined in the mouse fetuses. As summarized in Fig. 1F, cGcnf5 reproducibly induced lacZ expression in the fetal adrenal gland at E12.5 or E13.5, while cIA3 induced expression in both the fetal adrenal gland (Fig. 1G) and VMH (53). In addition, cIA3 induced lacZ expression in the pituitary gonadotrope at E14.5 (data not shown). This pattern of transgene expression suggests that the overlapping region between cGcnf5 and cIA3 contains an adrenal-specific enhancer. Thus, this region (SB7.5) was subcloned into the Ad4BP-lacZ plasmid (Fig. 1E) and subjected to Tg assays. As expected, SB7.5 induced lacZ expression in the fetal adrenal gland in a manner similar to those of cGcnf5 and cIA3 (Fig. 1F).

In general, functional genomic regions are conserved among animal species (58). Therefore, the mouse SB7.5 DNA fragment was compared with the corresponding region of the human gene. As shown in Fig. 1H, this region contains several conserved intronic regions in addition to exon 4. Using this homology plot as a guide, we prepared several deletion constructs. As summarized in Fig. 1F, the 7.5-kb (SB7.5) fragment and the 5′ half of this fragment (SH4.5) induced lacZ expression in the fetal adrenal gland (Fig. 1G). Further investigation using a series of subcloned constructs revealed that a 0.6-kb SacI-KpnI fragment (SacK0.6) in intron 4 controlled lacZ expression in the fetal adrenal gland (Fig. 1F and G). Interestingly, this active fragment had a high degree of homology between mouse and human. Further studies were undertaken to examine whether SacK0.6 drives adrenal-specific expression when ligated to the hsp68 basal promoter (29). Indeed, this fragment was able to induce lacZ expression in the fetal adrenal gland with the heterologous promoter (Fig. 1G).

Fetal adrenal enhancer.

To characterize the activity of the adrenal-specific enhancer, we generated three Tg mouse lines each with SacK0.6 and SH4.5. LacZ expression patterns in the adrenal glands were similar in all Tg lines at both E11.5 and postnatal day 12 (P9) (data not shown). Thus, we chose one of them (SacK0.6 line 9) to study the transgene expression pattern during adrenal development. As shown in Fig. 2A, lacZ expression was observed at the medial side of the urogenital ridges at E10.5. Based on our previous studies, the gonad and adrenal gland form a single cell group, the adreno-gonad primordium, at this developmental stage. Cross sections of rat fetuses (16) and chick embryos (60) revealed that the adreno-gonad primordium localizes from the coelomic epithelia (ventral domain) to the area proximal to the dorsal aorta (dorsal domain). When the Tg mouse fetuses stained with lacZ were cross-sectioned, the lacZ signal clearly localized to the area corresponding to the dorsal domain of the primordium (Fig. 2B).

FIG. 2.

LacZ expression in developing adrenal glands of Tg mice harboring pAd4BP-lacZ-SacK0.6. Tg mouse lines were generated with pAd4BP-lacZ-SacK0.6, and the expression of lacZ was examined throughout the developmental stages. (A) LacZ expression was examined at E10.5. The signal was localized at the adrenal primodium (AdP arrowhead) but not the gonad primordium (GP arrowhead). (B) The fetus was cross-sectioned at the level indicated for panel A. Arrowheads indicate the locations of the adrenal and gonad primordia. The lacZ signal was localized clearly at the AdP. (C) At E17.5, the lacZ signal was observed in the adrenal gland (Ad arrowhead) but not in the testis (Tes arrowhead) or kidney (Kid). (D) The adrenal gland was cross-sectioned at the level indicated for panel C. The lacZ signals were observed in the inner part of the cortical but not the medullary region. (E and F) At P12, lacZ-positive cells were scattered throughout the medulla in both sexes. (G) At P53, the lacZ signal was still observed in only the female at the innermost cortical layer. (H) By P35, the lacZ signal disappeared from the male adrenal gland. (I and J) To compare the expression patterns of lacZ and endogenous Ad4BP/SF-1, the adrenal gland at P9 was examined with antibodies for Ad4BP/SF-1 (αAd4BP/SF-1) and lacZ (αlacZ). (K and L) Merged views show that lacZ staining colocalized with Ad4BP/SF-1 staining. Panel L shows the inset of panel K. Bars indicate 100 μm in panels E, F, I, J, K, and L, and 200 μm in panels G and H. c, cortical; m, medullary.

The adreno-gonad primordium subsequently divides into two primordia, the adrenal primordium and gonadal primordium. The lacZ activity unexpectedly became weak at E17.5 (Fig. 2C). Cross sections revealed that only a certain population of cortical cells proximal to the medulla were lacZ positive (Fig. 2D), and this staining pattern was maintained until the neonatal period. Fetal transgene expression did not show any sexual dimorphism. Surprisingly, at P12, lacZ signals were distributed sporadically in the medulla of both sexes (Fig. 2E and F). Thereafter, by P35, lacZ-expressing cells were mostly absent from the male tissue (Fig. 2H) but, although decreased in number, lacZ-positive cells persisted at the innermost cortical layer until P53 in females (Fig. 2G). These cells completely regressed during the first pregnancy (data not shown). When examined by immunohistochemistry using antibodies to Ad4BP/SF-1 and lacZ, all lacZ-expressing cells at P9 were immunoreactive for Ad4BP/SF-1 (Fig. 2I through L).

Function of Ad4 sites in FAdE.

In order to identify the sequences that are responsible for FAdE function, we compared the nucleotide sequences of the enhancer regions between mouse and human. Among the sequences conserved between the two animals (Fig. 3A), we noted the presence of two potential Ad4BP/SF-1 binding sites (36), Ad4 site 1 and Ad4 site 2. Additionally, Ad4 site 3 and Ad4 site 4 were present in only the mouse sequence. When analyzed by EMSA, Ad4BP/SF-1 associated with all four potential binding sites, although binding to site 4 was much weaker (Fig. 3B).

FIG. 3.

Potential binding sites for Ad4BP/SF-1 and homeobox proteins in FAdE. (A) The nucleotide sequences of the mouse and human FAdEs are aligned. Potential binding sites for Ad4BP/SF-1 (Ad4 sites 1 through 4 in which site 1 and site 3 are in reverse directions) and for PP and PH are shown. Asterisks indicate conserved nucleotides between the two animals. (B) EMSAs were performed with wild-type (Ad4 sites 1 to 4) or mutated (mAd4 sites 1 to 4) oligonucleotides corresponding to the mouse gene. The in vitro translation product of Ad4BP/SF-1 was used as the protein source. An arrow shows the DNA-protein complex.

Ad4BP/SF-1 activates gene expression by binding to recognition sites. The presence of multiple Ad4 sites in FAdE suggested that Ad4BP/SF-1 has a role in the transcriptional regulation of gene expression in the fetal adrenal gland. To examine this hypothesis, we generated Tg mice harboring a construct carrying mutations in all Ad4 binding sites (Ad4-1234-mut). When we examined Tg mouse fetuses at E11.5, lacZ expression in the Ad4-1234-mut was similar to that in the wild-type construct (Fig. 4A and B). However, at E17.5, the lacZ signals disappeared from the Tg fetuses with the mutated construct. To determine which Ad4 sites were responsible for lacZ expression at E17.5, mutations were introduced at site 1 and site 2 or site 3 and site 4 (Fig. 4B). Tg assays with these constructs revealed that Ad4 sites 1 and 2 are essential for this expression pattern, and we subsequently examined the enhancer activities of constructs with a single site mutation (Fig. 4B). Since neither mutation led to the disappearance of the lacZ signal, at least a single site, Ad4 site 1 or site 2, is sufficient to drive lacZ expression in the fetal adrenal gland.

FIG. 4.

Function of the Ad4 sites in FAdE. (A) LacZ expression in fetuses harboring FAdE (wild type) or FAdE mutated at the Ad4 sites (Ad4-1234-mut) was examined at E11.5 and E17.5. LacZ expression was observed in the adrenal primordia (AdP arrowhead) of fetuses harboring both the wild-type and Ad4-1234-mut constructs at E11.5, whereas the signal was absent from the adrenal glands of the fetuses harboring the mutated construct by E17.5 (arrowheads). The signal was still evident in the adrenal glands with the wild-type construct (arrowheads). Kid; kidney. (B) LacZ expression of a series of Ad4 site mutants is summarized. The Ad4 sites in FAdE were mutated in combination (Ad-1234-mut, Ad-34-mut, or Ad-12-mut) or singly (Ad-1-mut or Ad-2-mut). LacZ expression driven by these constructs was examined at E11.5 and E17.5. The numbers of the fetuses harboring transgene (Tg) and those with lacZ expression at the adrenal primordia (Ad) are listed. n.d., not determined. (C) Tg mouse lines established with pAd4BP-lacZ-SacK0.6 were mated with Ad4BP/SF-1 heterozygous KO mice (Ad4BP/SF-1^+/−). Representative lacZ staining for wild-type mice carrying the transgene (Tg:Ad4BP^+/+), Ad4BP/SF-1 KO mice carrying the transgene (Tg:Ad4BP/SF-1^−/−), and Ad4BP/SF-1 KO mice carrying no transgene (non-Tg:Ad4BP/SF-1^−/−) is shown. Arrowheads indicate the location of the adrenal primordium.

Taken together, these results strongly suggest that Ad4BP/SF-1 binding to the Ad4 sites in the FAdE participates in the autoregulation of Ad4BP/SF-1 gene expression in the adrenal cortex at later stages of fetal development. However, given that the Ad4-1234-mut construct induced lacZ expression at E11.5, the Ad4 sites appear unnecessary for FAdE activity at earlier developmental stages. To confirm this hypothesis, we mated the SacK0.6 (line 9) Tg mouse (Fig. 2) with a heterozygous Ad4BP/SF-1 gene-disrupted mouse to obtain an Ad4BP/SF-1-null mouse with the lacZ transgene. As shown in Fig. 4C, lacZ was expressed in the adrenal primordium of the wild-type fetuses (Tg:Ad4BP^+/+) at E11.5. Consistent with our hypothesis, the Ad4BP/SF-1-null mouse with the transgene (Tg:Ad4BP^−/−) remained lacZ positive, albeit with a weaker intensity. The decreased signal likely arises from the tissue degeneration that begins as early as E11.5 in Ad4BP/SF-1-deficient animals (7, 32). No lacZ signal was seen in the non-Tg Ad4BP/SF-1-null mouse (non-Tg:Ad4BP^−/−). Thus, the FAdE drives a two-step regulation, initiation, and maintenance of Ad4BP/SF-1 gene transcription.

Hox/Pbx1 and Pbx1/Prep1 binding sites in FAdE.

The data presented above argue that Ad4BP/SF-1 maintains its own expression at later stages of development, but it remains unclear how Ad4BP/SF-1 expression is controlled prior to the maintenance phase of transcription. To address this question, we attempted to identify functional cis elements other than the Ad4 sites present in the FAdE. Because both Ad4 site 3 and site 4 are not required for lacZ expression and the 5′ region of the Ad4BP/SF-1 enhancer is not highly conserved, we generated a DNA fragment containing nucleotides 88 to 467 (Fig. 3A). As described below, this truncated fragment promoted a level of lacZ activity similar to that of the original enhancer.

When the sequence of this DNA fragment was examined in detail, we found potential binding sites for the Pbx/Prep and Pbx/Hox heterodimers present in the upstream region proximal to the Ad4 site 1 (Fig. 3A). Since both sites are conserved in human and chick (data not shown), we examined whether those factors are expressed in the adrenal primordium. However, adrenal primordia are difficult to isolate at early developmental stages and thus we generated two Tg mice expressing EGFP under the control of the FAdE in conjunction with the basal Ad4BP/SF-1 or hsp68 promoter. As shown in Fig. 5A, the Ad4BP/SF-1 basal promoter drove high levels of EGFP expression in the adrenal primordium and lower levels in the gonadal primordium at E10.5. Due to the large difference in fluorescence between these two primordia, we could easily isolate the adrenal primordia at E10.5. RNAs prepared from the isolated tissue were used for RT-PCR analyses to examine candidate gene expression. Primers specific for Pbx1 and Prep1 successfully amplified their corresponding transcripts (Fig. 5B).

FIG. 5.

Expression of Pbx1, Prep1, and Hox in the adrenal primordia. (A) Representative expression pattern of EGFP in Tg fetuses harboring FAdE-Ad4BP-EGFP is shown. Strong EGFP expression was observed in the fetal adrenal gland (Ad arrowhead), while weak expression was observed in the fetal gonad (Go arrowhead) at E10.5. (B) Total RNA prepared from the adrenal primordia was used for RT-PCR with primers for Pbx1 and Prep1 (RT-prod). Genomic DNA was used as a control (genomic). M, marker. (C) EGFP-positive cells were isolated from E10.0 fetuses harboring FAdE-hsp68-EGFP, while the adrenal glands were prepared from wild-type fetuses at E13.5. RNAs were subjected to RT-PCR using degenerate primers. The amplified DNA fragments were cloned and sequenced. The numbers of the cloned Hox are shown on a schematic map of Hox gene clusters.

In order to identify Hox genes expressed in the adrenal primordium at stage E10.0, the primordial cells were collected from Tg fetuses carrying the hsp68 basal promoter with the FAdE. RNA prepared from these cells was subjected to RT-PCR with degenerate primers corresponding to a highly conserved homeodomain to amplify all Hox family members expressed. RNA prepared from E13.5 fetuses was also used for the analysis, and the amplified fragments were cloned and sequenced. Seventy and 95 clones for E10.0 and E13.5 fetuses, respectively, were analyzed, and the results were overlaid on a schematic map of the mouse Hox gene clusters (Fig. 5C). Hox classes 5 to 9 were abundant in both E10.0 and E13.5 adrenals. In particular, clones carrying Hoxa2, Hoxa7, Hoxa9, Hoxb5, Hoxb9, Hoxc5, Hoxc6, and Hoxc8 were isolated with greater frequency than other Hox gene sequences in both developmental stages.

Binding of the Hox-Pbx-Prep complex to Pbx/Prep and Pbx/Hox sites in FAdE.

EMSAs were performed to examine whether the Pbx/Prep and Pbx/Hox sites were recognized by in vitro-synthesized Pbx1, Prep1, and Hox proteins. When an oligonucleotide containing the Pbx/Prep site (Fig. 6A) was used as a probe, a signal was detected in the presence of Pbx1a and Prep1, while the signal disappeared upon the addition of unlabeled PP. A mutated oligonucleotide (mPP) was unable to compete with this binding (Fig. 6B). Pbx1b bound in a similar manner. Using an oligonucleotide probe containing the PH site, we examined the binding activity of each Hox protein in the presence or absence of Pbx1b. Hoxa2, Hoxb5, Hoxb9, and Hoxc5 bound the probe in the absence of Pbx1b, but they failed to bind the mPH (Fig. 6C). In the presence of Pbx1b, three Hox proteins, Hoxb5, Hoxb9, and Hoxc5, gave shifted signals, strongly suggesting heterodimer formation between Pbx1b and these Hox proteins. Although Hoxc6 gave no signal by itself, it gave a signal in the presence of Pbx1b.

FIG. 6.

Binding of Pbx, Prep, and Hox proteins to the candidate sequences in FAdE. (A) Oligonucleotide sequences used for EMSAs are listed. Potential binding sites for Pbx/Prep and Pbx/Hox are underlined. Mutated nucleotides are shown with lowercase letters. (B) Pbx1a or Pbx1b, together with Prep1, was incubated with the PP probe containing the Pbx/Prep site. Fiftyfold and 250-fold molar excess amounts of unlabeled PP or mPP were used as competitors. (C) Hoxa2, Hoxa7, Hoxa9, Hoxb5, Hoxb9, Hoxc5, Hoxc6, and Hoxc8 were used for EMSAs. These Hox proteins, together with Pbx1b, were incubated with the PH probe containing the Pbx/Hox site. Mutated oligonucleotide mPH was used as a control. The locations of Hox and Pbx-Hox heterodimer are indicated by arrowheads. (D) EMSAs with Hoxa2, Hoxb5, Hoxb9, and Hoxc5 were performed in combination with Pbx1b and Prep1. The proteins and probes used are indicated at the top. A ternary complex (Hox-Pbx1-Prep1) was formed when the three proteins were incubated with the PP/PH, PP/mPH, and mPP/PH probes. Arrowheads indicate the ternary complex with PP/mPH.

Prep1 can influence the Pbx-Hox complex through direct interactions with Pbx (3, 4). Indeed, Pbx/Prep sites are frequently localized near Pbx/Hox sites in some genes (13, 23, 48). Similar to these genes, FAdE contains both Pbx/Prep and Pbx/Hox sites with an 8-bp interval (Fig. 3A and 6A). Therefore, we examined the binding of Pbx1b, Prep1, and Hox proteins to a probe (PP/PH) containing both binding sites (Fig. 6D). Hoxa2, a protein unable to interact with Pbx1b, was used as a reference. When Hoxb5, Hoxb9, or Hoxc5 was mixed with Pbx1b and Prep1, two signals were detected. The signal with the highest mobility was very similar to that given by the Pbx1b-Prep1 heterodimer, and the Pbx1b-Hox heterodimer gave a signal, albeit substantially weaker, at a similar location. Thus, the majority of the high-mobility signal arose from the interaction of the Pbx1b-Prep1 heterodimer with the probe. Consequently, the slower migrating signal likely corresponds to a Hox-Pbx1b-Prep1 complex. As expected, Hoxa2 failed to form larger complexes. To characterize Hox-Pbx1b-Prep1 complex formation, we next performed EMSAs with probes carrying mutations at the Pbx/Hox site (PP/mPH), the Pbx/Prep site (mPP/PH), or both the Pbx/Prep and the Pbx/Hox sites (mPP/mPH). As expected, the mPP/PH probe failed to give a signal corresponding to the Pbx1b-Prep1 heterodimer but clearly gave a signal corresponding to the Hox-Pbx1b-Prep1 complex when incubated with Hoxb5, Hoxb9, or Hoxc5 together with Pbx1b and Prep1, indicating that Prep1 interacts with Pbx1-Hox heterodimer in the absence of Prep1 DNA binding. When PP/mPH was used for binding, the high-mobility complex consisting of the Pbx1b-Prep1 heterodimer was seen with all Hox proteins examined. Additionally, a slower-migrating weak signal was seen with Hoxb5 and Hoxc5, likely corresponding to Hox-Pbx1b-Prep1 complexes (Fig. 6D). All signals disappeared when mPP/mPH was used. Taken together, these data suggest that the close proximity of the Pbx/Hox and Pbx/Prep sites synergistically enhances complex binding. Moreover, in the absence of either binding site, heterotrimeric Hox-Pbx1b-Prep1 complexes are able to bind the DNA with reduced efficiencies.

Function of Pbx/Prep and Pbx/Hox binding sites in vivo.

We examined whether the binding sites contribute to the Ad4BP/SF-1-independent transcriptional activity of the FAdE. We used the short form of FAdE (88 to 476 bp) containing Ad4 sites 1 and 2 as well as the Pbx/Prep and Pbx/Hox binding sites for Tg assays (Fig. 7A). This short form of the enhancer (PP/PH/Ad) reproducibly induced lacZ activity as well as the original enhancer. Moreover, similar to the wild-type construct (PP/PH/Ad), a PP/PH/mAd construct carrying mutations at both Ad4 site 1 and Ad4 site 2 induced lacZ activity in the E11.5 fetal adrenal gland (Fig. 7B). Thus, we examined whether the retained activity of PP/PH/mAd is driven by the Pbx/Prep and/or Pbx/Hox binding sites. When either the Pbx/Hox (PP/mPH/mAd) or the Pbx/Prep (mPP/PH/mAd) binding site was mutated, weak lacZ signals were still observed in the fetal adrenal tissue. Expectedly, when both the Pbx/Hox and Pbx/Prep sites were mutated (mPP/mPH/mAd), lacZ expression was completely absent from the adrenal primordia.

FIG. 7.

Functional analyses of Pbx/Prep and Pbx/Hox sites with Tg assays. (A) Tg constructs and their activities to drive lacZ expression in the adrenal primordia are summarized. The short form of the FAdE (88 to 467 bp) was used in this study. Pbx/Prep and Pbx/Hox sites mutated singly (PP/mPH/mAd or mPP/PH/mAd) or together (mPP/mPH/mAd) were combined with mutated Ad4 sites. The wild-type construct (PP/PH/Ad) was used as the control. The numbers of the Tg fetuses (Tg) obtained and those showing a lacZ signal in the adrenal primordia (Ad) at E11.5 are shown. (B) Representative lacZ expression in the Tg fetuses at E11.5 harboring the constructs listed in panel A is shown. Similar to PP/PH/Ad and PP/PH/mAd, the lacZ signal was detected in the fetuses harboring PP/mPH/mAd or mPP/PH/mAd. However, mPP/mPH/mAd failed to show a signal. Arrowheads, adrenal primordia.

DISCUSSION

Fetal adrenal-specific enhancer of the Ad4BP/SF-1 gene.

The Ad4BP/SF-1 gene is essential for the proper development and function of the adrenal cortex, gonads, pituitary gonadotrope, VMH, and spleen (38, 44, 57). Therefore, Ad4BP/SF-1 expression should be differentially regulated by tissue-specific enhancers and thus the identification of these tissue-specific enhancers seemed to provide insight into the regulation of tissue development. Toward this end, Tg mouse studies performed using BAC and YAC clones containing the Ad4BP/SF-1 gene locus successfully reproduced endogenous gene expression (24, 55). However, these studies did not identify any tissue-specific enhancers. An additional Tg mouse study used the proximal upstream region from −589 to +85 that was capable of driving transgene expression in the sexually indifferent fetal gonad (E11.5) and studies with cultured cells demonstrated that Wt1 and Lhx9 binding sites in this region are essential for Ad4BP/SF-1 gene transcription (59). However, whether this upstream region is sufficient to control gene expression at later developmental stages in the gonads or other target tissues remains unclear. Interestingly, Shima et al. recently identified a VMH-specific enhancer in intron 6 of the Ad4BP/SF-1 gene, strongly suggesting that other tissue-specific enhancers are localized in the locus (53).

In the present study, we newly identified FAdE in intron 4 of the Ad4BP/SF-1 gene by using Tg mice assays. Similar to the VMH enhancer, the nucleotide sequence of the FAdE is conserved among animal species. However, this enhancer was unable to drive transgene expression in the adult adrenal cortex. Thus, the fetal and adult adrenal cortices appear to use different systems for the regulation of gene expression, and this may explain the distinct structures and functions observed between fetal and adult cortices. Interestingly, FAdE activity appears to be controlled by an autoregulatory loop through the interaction of Ad4BP/SF-1 with internal Ad4 sites in the FAdE. However, before autoregulation can begin, Ad4BP/SF-1 gene transcription must occur in an Ad4BP/SF-1-independent fashion. This hypothesis was clearly supported by the ability of Ad4 binding-deficient FAdE to drive transgene expression at E11.5, and moreover, the FAdE remained active in the adrenal primordium of Ad4BP/SF-1 knockout (KO) mice. After extensive investigation, we identified binding sites for Pbx-Hox and Pbx-Prep in FAdE and demonstrated that these sites are essential for the initial expression of Ad4BP/SF-1 prior to the establishment of autoregulation.

The autoregulatory system likely maintains the intracellular concentration of Ad4BP/SF-1 at an appropriate level, and indeed the system is known to control the expression of transcription factors that are critically required for the development of certain tissues (14, 18, 61). Several lines of evidence suggest that Ad4BP/SF-1 gene dosage and subsequent protein expression levels are critical for target tissue development. Fetuses that are heterozygous for Ad4BP/SF-1 deficiency develop hypoplastic adrenal glands (7), and the same phenomenon is observed in the M33 KO mouse in which Ad4BP/SF-1 expression is downregulated to approximately half the level of the expression in wild-type mice (25). Interestingly, this defect is caused by decreased cell numbers of the adrenal primordia (7), suggesting that Ad4BP/SF-1 activates cell proliferation dose dependently. Thus, the expression levels of Ad4BP/SF-1 in the fetal adrenal gland appear to be fine tuned in a developmental stage- and tissue-specific manner by the function of the FAdE, although there are possible tissue differences in Ad4BP/SF-1 function.

Correlation of fetal and X zones in the developing mouse adrenal cortex.

The fetal adrenal cortices of many mammalian species consist of two distinct cell groups, small and tightly packed outer cells (adult zone) and larger and irregularly aligned inner cells (fetal zone) (27). During pregnancy, the adult zone remains undifferentiated, but the fetal zone enlarges and synthesizes steroids. After birth, the fetal zone regresses, while the adult zone increases in size. Although the rat and mouse fetal adrenal cortices develop two distinct layers like other mammalian species, their adrenal glands have been described as lacking a true fetal zone (30). In fact, the degeneration of the fetal zone and growth of the adult cortex starts in the late fetal stage in these rodents. In addition, the mouse adrenal cortex develops a unique islet of eosinophilic cells, the so-called X zone, after birth (21, 35). This particular layer disappears during puberty in male mice (21), whereas it persists for a longer period in females and undergoes regression during the first pregnancy (20).

In this study, we examined lacZ reporter gene expression driven by the FAdE. LacZ expression was detected at E10.5 at the presumptive adrenal cortex. Thereafter, the expression was maintained until puberty in males but persisted until after sexual maturation in females. Interestingly, lacZ expression in females completely disappeared after the first pregnancy. This expression pattern is remarkably similar to the kinetics of X-zone regression, strongly suggesting that the lacZ-expressing cells comprise the X zone. However, lacZ-expressing cells were present during the fetal stage and the X zone has been thought to emerge after birth. This discrepancy may arise from the unique distribution of X-zone (lacZ-positive) cells at the postnatal stage. Shortly after birth, lacZ-positive cells are distributed sporadically in the medulla and thereafter they accumulate at the juxtamedullary region, where the X zone is thought to develop (17). Previous studies examining the X zone have relied on merely structural and histological observations, not marker gene expression, and this renders firm conclusions of the origins of the cells that comprise the X zone difficult. Therefore, it is reasonable to conclude that the fetal adrenal zone is present in the mouse fetus and is maintained after birth as X-zone cells. All of these data demonstrate that the developmental processes of the adrenal gland are extremely similar in mouse and other mammals.

Role of Hox-Pbx1-Prep1 as upstream regulators of Ad4BP/SF-1.

Hox transcription factors direct the patterning of various structures during the embryonic development of vertebrates and invertebrates through regulating numerous target genes (9, 33, 45). Based on the expression profile and binding specificity, Hoxb5, Hoxb9, Hoxc5, and possibly Hoxc6 were thought to regulate the Ad4BP/SF-1 expression in the fetal adrenal gland. The adrenal cortex is known to be derived from a certain part of the intermediate mesoderm lying along the anterior to posterior axis (16). To specify the adrenal region, these Hox gene products are thought to induce Ad4BP/SF-1 expression at this particular region of the mesoderm. As it has been established that Hox genes control anterior to posterior axial identity through regulating target gene expression, the location of the adrenal cortex could be determined by the combined expression of the particular set of the Hox genes.

The binding and heterodimerization of Hox proteins with Pbx have been extensively investigated. The Pbx-Hox heterodimer binds a bipartite sequence comprised of two adjacent half sites in which Pbx contacts the 5′ half, while Hox contacts the more variable 3′ half (31). The Pbx/Hox site (CTGATAAATAA) found in the FAdE is highly similar to a consensus binding sequence for Pbx and certain Hox paralogue members (NTGATTNATNN) (10). In addition, the FAdE contains a Pbx/Prep (or Pbx/Meis) site near the Pbx/Hox site. The proximity of these two sites allows efficient formation of a ternary Hox-Pbx-Prep complex in vitro. Moreover, considering that a mutation at either of these two sites reduced the enhancer activity in vivo, both sites are crucial for complete functioning of the FAdE. Consistent with our current observations, the combined presence of these two binding sites in the regulatory regions of the Hoxb1, Hoxb2, and Phox2b was reported to direct cooperatively rhombomere-restricted expression (13, 23, 46, 50).

Pbx1 appears to play critical roles during the early stages of adrenal development. Interestingly, mice deficient in the Pbx1 gene completely lack adrenal glands (51) as do Ad4BP/SF-1 gene-deficient mice. In the absence of Pbx1, Ad4BP/SF-1 expression is substantially reduced. Thus, considering the dose dependence of adrenal development on Ad4BP/SF-1 expression, it is likely that the low levels of Ad4BP/SF-1 in Pbx1-deficient mice are responsible for the defective adrenal development. Although the knockout study strongly suggested that Pbx1 regulates Ad4BP/SF-1 gene expression in the adrenal primordium, the molecular mechanism governing this regulation remained unclear. The present study greatly clarifies this mechanism and provides valuable insight into the overall mechanisms regulating cell fate determination in the fetal adrenal cortex.

In summary, we have identified a fetal adrenal enhancer in the Ad4BP/SF-1 gene. An investigation of cis elements revealed that homeobox protein-containing complexes initially activate Ad4BP/SF-1 gene expression, which subsequently establishes an autoregulatory loop. These findings substantially provide a deeper understanding of adrenal development.