Loss of Bmp7 and Fgf8 signaling in Hoxa13-mutant mice causes hypospadia

In humans and mice, reproduction occurs by internal fertilization, a strategy facilitated by the formation of specialized external genitalia. Developmentally, these structures are derived from the genital tubercle (GT),a ventrocaudal structure composed of bilateral shelves of mesenchyme bisected by a central epithelial plate (Cunha,1972; Murakami,1987; Kurzrock et al., 1999). At approximately 50 days of human gestation (embryonic day 14-15 in mice), the sexually ambiguous genitalia begin to exhibit secondary sexual characteristics, a process initially believed to be mediated solely by the presence (male) or absence (female) of androgen signaling (Cunha and Lung,1978; Murakami,1987; Murakami and Mizuno,1986).

Recently, new evidence suggests that androgens are not the sole regulators of GT patterning. In particular, many genes whose proteins provide the instructive signals necessary for limb development also play a predominant role during GT formation, including members of the fibroblast growth factor(Fgf8 and Fgf10), hedgehog (Shh) and Hox(Hoxa13 and Hoxd13) gene families(Haraguchi et al., 2000; Haraguchi et al., 2001; Stadler et al., 2001; Zhao and Potter, 2001; Warot et al., 1997; Mortlock and Innis, 1997;Podalesk et al., 1997; Podalesk et al., 1999; van der Hoeven et al., 1996a; van der Hoeven et al., 1996b). Finally, human androgen receptor mutations are generally rare in individuals exhibiting some of the more common genitourinary (GU) malformations such as hypospadia, a condition characterized by arrested growth, formation and closure of the external genitalia(Sutherland et al., 1996; Allera et al., 1995). In males,hypospadia can affect several different regions of the developing glans penis. These malformations range from distal defects, impacting the placement and formation of the urethral opening or meatus (coronal hypospadia), to more proximal defects, affecting the growth and closure of the ventral urethra(penile hypospadia), scrotum (scrotal hypospadia) and peritoneum (peritoneal hypospadia) (Baskin, 2000; Silver and Russel, 1999; Sutherland et al., 1996; Allera et al., 1995).

To date, little is known about the cellular and molecular mechanisms underlying this malformation; however, the incidence of hypospadia appears to be increasing dramatically, affecting as many as 1 in 125 live male births each year (Baskin et al., 1998; Baskin et al., 2001; Gallentine et al., 2001; Paulozzi et al., 1997). Recently, mutations in the transcription factor Hoxa13 were shown to cause Hand-Foot-Genital Syndrome (HFGS), an autosomal dominant disorder that profoundly affects the development of many genitourinary structures, including the uterus, bladder, ureters and cervix, as well as causing hypospadia(Stern et al., 1970; Pozanski et al., 1975; Giedion and Prader,1976; Mortlock and Innis,1997; Goodman et al.,2000).

Using gene targeting in mice, similar mutations that phenocopy HFGS(Stadler et al., 2001;Fromental-Ramain, 1996; Warot et al.,1997) have been generated. From these initial studies, a role for Hoxa13 in patterning both limb and caudal GU structures has been established. In particular, mice lacking Hoxa13 exhibited gross malformations of the rectum, Müllerian ducts, ureters and bladder(Warot et al., 1997).

Recognizing both the increasing frequency of hypospadia in the human population, as well as its association with HFGS, we hypothesized that many of the cellular and molecular mechanisms underlying this malformation could be identified in Hoxa13-mutant mice. In this report, we present evidence that mice mutant for Hoxa13 phenocopy the hypospadia in HFGS, causing malformation of the distal glans and meatus, as well as closure defects of the ventral urethra. At the molecular level, we demonstrate that Hoxa13function is essential for the normal expression of Fgf8 and Bmp7 in the developing urethral epithelium, as well as for the repression of noggin (Nog) expression in the mesenchyme flanking the urethral plate epithelium (UPE). Next, we demonstrate that supplementation with Fgf8 in the urethral epithelium is sufficient to restore proliferation in the developing GT. Furthermore, we show that Bmp7 provides an apoptotic signal to the UPE and flanking mesenchyme, as antibody blocking of Bmp7 is sufficient to reduce most programmed cell death (PCD) in these regions. Interestingly, the blocking of either Bmp4 or Bmp7 signaling during GT outgrowth recapitulated many of the defects exhibited by Hoxa13^GFP homozygous mutant mice, confirming that perturbations in Bmp signaling contribute to the coronal and penile hypospadias exhibited by these mice. Finally, a novel role for Hoxa13in mediating the vascularization of the glans penis was also identified during this study.

The Hoxa13^GFP allele originally described by Stadler et al. (Stadler et al., 2001) was utilized in all experimental procedures. Heterozygous intercrosses were utilized to produce both homozygous Hoxa13^GFP-mutant(–/–) and wild-type (+/+) embryos. Examination of the developing genitourinary regions revealed no differences between heterozygous mutant and wild-type embryos with respect to growth, morphology and gene expression patterns. Gestational ages were defined as embryonic days (E), where the detection of a vaginal plug denotes E 0.5. Embryo genotypes were determined by PCR using yolk sac-derived DNA and primers as described previously(Stadler et al., 2001). Embryonic sex was determined by PCR of yolk sac DNA with forward(5′-GAGAGCATGGAGGGCCAT-3′) and reverse(3′-CCACTCCTCTGTGACACT-5′) primers specific for murine Sry. All care and analysis of Hoxa13^GFP mice was carried out in accordance with an approved mouse handling protocol.

GFP, CY5 or Texas-Red fluorescence was assessed using a Bio-Rad MRC 1024 confocal imaging system fitted with a Leica DMRB microscope, using filter sets provided by the manufacturer. Whole-mount images represent 150 to 200 μm z-series stepped at 2 μm intervals, whereas cryosections were imaged in single planes using the GFP-tagged nucleus as the primary focal plane. A Kalman digital averaging filter was used to reduce random noise.

Primary antibodies used were Cytokeratin 14 (#RDIMCYTOK14abr, Research Diagnostics), Cytokeratin 5 (#RDIMCYTOK5abr, Research Diagnostics),Cytokeratin 8/18 (#RDI-Progp11, Research Diagnostics), anti-phosphohistone H3(#06-570, Upstate Biotechnology), Pecam (CD31; #550274 BD, PharMingen).

E11.5-15.5 male embryos were fixed for 1 hour in 4%paraformaldehyde/1×phosphate buffered saline (PBS) at 4°C. The fixed tissues were rinsed four times (for 5 minutes each time) in 1×PBS, and then blocked in PBS containing 1% Triton X-100 (PBX), 2% skimmed milk and 5%donkey serum. Primary antibodies were incubated with the blocked tissues at 4°C overnight using previously determined concentrations. Embryos were then rinsed in 1×PBX, 2% skimmed milk for 3 hours at room temperature,followed by overnight incubation with either Texas-Red or CY5-conjugated secondary antibodies (Jackson Immuno Research) in PBX, 2% skimmed milk, 5%donkey serum at 4°C.

Embryos were fixed as described above, followed by a 30 minute rinse in 1×PBS. Cryoprotection of the tissues was achieved using a sequential series of 10-40% sucrose/1×PBS. The embryos were oriented in OCT(Tissue-Tek) and frozen rapidly on dry ice. The embryos were sectioned to 15μm using a Leitz Kryostat 1740, and mounted on Superfrost plus slides(Fisher) for analysis. Sections were rinsed for 10 minutes three times in 1×PBS to remove the OCT. Next the sections were permeabilized and blocked for 1 hour in PBX, 2% skimmed milk, 5% donkey serum. After blocking,primary antibodies were incubated on the sections for either 1 hour at room temperature or overnight at 4°C, using the same blocking solutions at previously determined concentrations. The sections were then rinsed in 1×PBX, 2% skimmed milk for 2 hours. Species-specific secondary antibodies conjugated with either Texas-Red or CY5 fluorochromes (Jackson Immuno Research) were incubated with the sections overnight at 4°C.

Antisense riboprobes specific for Bmp4, Bmp7, Msx1, gremlin(Cktsf1b1–Mouse Genome Informatics) and Fgf8 were generated using plasmids kindly provided by B. Hogan (Bmp4), R. Beddington (Bmp7), Y. Chen (Msx1), R. Harland (gremlin) and A. Moon (Fgf8). Hoxa13, Shh, Bmpr1a, Bmpr1b, Bmpr2 and Msx2 riboprobes were produced using PCR amplifications of gene-specific exons. The amplifying primers were:

• Hoxa13, 5′-GCTGCCCTACGGCTACTTC-3′ and 5′-CAGGGCTGATAGCTTTCCAT-3′;

• Shh, 5′-GCGCTCTTTGCCAGCCGCGT-3′ and 5′-TCAGCTGGACTTGACCGCCATT-3′;

• Bmpr1a, 5′-CTGCATCAAGACTCCAATCCT-3′ and 5′-ACATGGAGTTTGGCCTTCAC-3′;

• Bmpr1b, 5′-TGCAGATATCAAAGGGACTGG-3′ and 5′-AAGCCCAGGTCTGCTATGCA-3′;

• Bmpr2, 5′-GACTTCACACAGGCTGCAAA-3′ and 5′-AAACGATCCAGAACCCCTTC-3′; and

• Msx2, 5′-TCAGTCTGCCCTTCCCTA-3′ and 5′-ATATAAAAAGTCTTATATTTTATTAT-3′.

Amplified exons were cloned into a t-tailed vector containing RNA polymerase T3 and T7 promoters. Embryo preparation, hybridization and analysis was performed as described by Manley and Capecchi(Manley and Capecchi,1995).

TUNEL analysis was performed as described by Stadler et al.(Stadler et al., 2001). Genital tubercles from E12.5 heterozygous- and homozygous-mutant embryos were examined by confocal analysis of 200 μm z-series as described above. TUNEL analysis on frozen sections followed the same approach as whole-mount TUNEL assays using 20 μm sections through the genital tubercle.

Heparin acrylic beads (#H-5263, Sigma), containing either 0.1 mg/ml BSA(control) or 0.1mg/ml Fgf8b (R and D Systems), were placed in the UPE of E11.5 male embryos. Tissue explants containing the bead-implanted genital tubercles were placed on cellulose membranes supported by stainless steel mesh in 60 mm organ culture dishes containing BGJb media (Sigma-Aldrich), supplemented with 50 U/ml penicillin, 25 μg/ml streptomycin and 0.1 mg/ml ascorbic acid, and grown for 24 hours in a incubator at 37°C containing air supplemented with 5% CO₂. The GT explants were examined for changes in cell proliferation using the anti-phosphohistone-H3 antibody (Upstate Biotechnology).

Extended growth response of the GT to Fgf8b was performed using BGJb media,supplemented with 0.1 mg/ml Fgf8b or BSA, in a BTC Engineering rotary culture apparatus. E11.5 GTs were dissected from Hoxa13^GFPheterozygous and homozygous mutant embryos and grown for 72 hours. After culture, the developing genitalia were examined for changes in GT outgrowth and morphology, using Hoxa13^GFP expression as a marker of the relevant tissues.

Dihydroxytestosterone (DHT; Sigma) was dissolved in 100% ethanol and diluted to 10 nM in BgJb media. GT explants were grown for 6 hours in the presence of 10 nM DHT or the equivalent carrier volume of ethanol. After culture, the explants were fixed in 4% paraformaldehyde and processed for in situ hybridization using the Shh or Hoxa13 riboprobes as described above.

E12.5 genital tubercles were dissected from Hoxa13^GFPheterozygous mutant embryos and grown for 72 hours in the BTC rotary culture apparatus. Culture media consisted of BGJb media (Sigma-Aldrich) buffered with 10 mM HEPES and supplemented with 0.1 mg/ml ascorbic acid. The sealed culture bottles were continuously supplied with humidified gas containing 40%O₂, 5% CO₂ and 55% N₂. Polyclonal antibodies specific for Bmp4 (αBmp4) or Bmp7(αBmp7; R and D Systems) were added to the media at a concentration of 4 μg/ml, whereas control cultures were treated with 2μg/ml of whole-goat immunoglobulins (IgGs). After 36 hours, 50% of the culture media was replaced with new aliquots of antibody or control IgGs. After 72 hours, the genital tubercles were examined for changes in meatus formation, as well as defects in urethral groove closure using Hoxa13^GFP expression as a marker. To assess changes in PCD in response to antibody blocking, GT explants from E12.5 Hoxa13^GFP-heterozygous embryos were grown, using the same rotary culture conditions, for 24 hours in media containing 4 μg/mlα Bmp7 or control IgGs. After culture, the explants were fixed in 4% paraformaldehyde and assessed for changes in programmed cell death by TUNEL analysis of frozen sections as described.

As hypospadia predominantly affects the development and closure of the urethra and its meatus, we first examined whether mice lacking Hoxa13exhibited any changes in mitosis or programmed cell death that could contribute to the hypospadiac phenotype. Examination of the Hoxa13^GFP allele (this work)(Stadler et al., 2001)indicates that the GFP expression patterns recapitulate previously reported patterns of Hoxa13 expression in mouse and other vertebrate species(Warot et al., 1997; Haack and Gruss, 1993; Yokouchi et al., 1991). In the GT, Hoxa13^GFP expression is detected in the surrounding mesenchyme as well as the UPE, a structure previously identified as an essential signaling center for GT outgrowth(Fig. 1A,B)(Perriton et al., 2002; Haraguchi et al., 2000; Haraguchi et al., 2001; de Santa Barbara et al., 2002). Examination of the GT at E12.5 revealed no major differences in the early formation of the UPE or mesenchymal shelves, with the exception that the distal UPE in homozygous mutants was consistently thicker(Fig. 1A,B). Analysis of cell proliferation in the GT also revealed distinct differences between homozygous-mutant and heterozygous-mutant embryos. In particular, homozygous mutants consistently demonstrated less proliferation in the mesenchyme flanking the proximal UPE (Fig. 1C; arrowheads), as well as in the lateral shelf tissues, which exhibited 46% (±6%) fewer mitotic cells when compared with controls(Fig. 1C,D; boxed regions). Interestingly, little proliferation could be detected in the UPE(Fig. 1C,D) for any genotype examined. To determine the impact of reduced mesenchymal shelf proliferation on urethral growth and closure, we examined the formation and closure of the distal urethra and meatus. In male embryos, differentiation of external genitalia begins at approximately E14.5 with growth of the glans penis and formation of the urethral meatus proceeding from the distal portion of the GT(Kaufman and Bard, 1999; Murakami, 1987). In E15.5 heterozygotes, the urethral groove epithelium was less visible because of the progression of the surrounding mesenchyme(Fig. 1E). In the distal urethra, fusion of the urethral folds was clearly visible, establishing the site of the urethral meatus (Fig. 1E). In homozygous mutants, the urethral groove remained exposed,with no fusion of the distal mesenchyme leaving the placement and formation of the distal meatus undefined (Fig. 1F).

In the caudal GT, Hoxa13 also regulates the formation of the rectum. In the heterozygous controls, rectal development proceeds from a centralized core of epithelium (Fig. 1E; arrowhead) flanked by mesenchyme expressing Hoxa13^GFP. In homozygous mutants, both the placement and maturation of the rectal epithelium is affected, resulting in the formation of a more ventralized rectum one gestational day earlier(Fig. 1F).

During GT development, programmed cell death (PCD) precedes the crucial fusion events necessary for the growth and closure of the penile urethra(Haraguchi et al., 2001; van der Werff et al., 2000). To determine if changes in PCD also contribute to the hypospadiac phenotype,apoptosis in the developing GT was measured using a TUNEL assay. By E11.5, PCD was readily seen in the distal and proximal UPE of wild-type (not shown) and heterozygous Hoxa13^GFP-mutant embryos(Fig. 2A). By contrast,homozygous mutants consistently exhibited reduced PCD in the distal UPE,whereas the proximal UPE showed no difference in PCD when compared with controls (Fig. 2A,B). Quantitation of PCD in the distal UPE of homozygous mutants revealed, on average, a 90% (±8%) loss in TUNEL-positive cells when compared with controls (Fig. 2A,B).

By E12.5, both homozygous mutant and control embryos exhibited a shift in PCD from the UPE to the surrounding mesenchyme(Fig. 2C,D). A comparison of PCD levels in the mesenchymal tissue revealed an 85% (±5%) reduction in PCD in homozygous mutants (Fig. 2C,D) when compared with heterozygous controls. TUNEL-positive cells were no longer detectable in the UPE or surrounding mesenchyme by E14.5 in both homozygous-mutant and control embryos (data not shown).

To address mechanisms underlying reduced PCD in the mutant GT, we first determined the localization and timing of pro-apoptotic signals in this region. In particular, the expression of Bmp4 and Bmp7 was examined, as these factors regulate many of the apoptotic events necessary for normal limb (Macias et al.,1997), craniofacial (Graham et al., 1996) and inner ear(Fekete et. al., 1997)development. Although no differences in Bmp4 expression could be discerned between homozygous-mutant and control embryos (data not shown),notable changes in Bmp7 expression were seen in the UPE and lateral shelf mesenchyme (Fig. 3). At E12.5, wild-type embryos exhibited high levels of Bmp7 expression along the entire axis of the UPE, as well as in more proximal lateral shelf mesenchyme (Fig. 3A). By contrast, mutant embryos express Bmp7 only in distal UPE, with little expression seen in the lateral shelf mesenchyme(Fig. 3B). Mutant embryos also exhibit a thickened layer of epithelium(Fig. 3B) that physically separates the growing glans from the more ventral UPE.

By E14.5, Bmp7 expression was highest in the distal glans at the site of meatus formation, as well as in the urethral epithelium that is covered with a thin layer of mesenchyme and in the developing preputial glands(Fig. 3C). In homozygous mutants, Bmp7 expression was noticeably reduced in the distal glans,whereas no expression could be detected in the urethral epithelium that lacked a covering layer of mesenchyme in all embryos examined (4/4; Fig. 3D).

We next examined whether homozygous-mutant embryos exhibited any changes in the expression of Bmp antagonists or Bmp-target genes that would augment reduced Bmp7 expression in the GT and contribute to the hypospadiac phenotype. In E11.5 wild-type embryos, Nog expression was strongest in the lateral shelf mesenchyme, with no expression in the UPE or mesenchyme adjacent to the UPE (Fig. 3E). By contrast, homozygous mutants exhibited ectopic Nog expression in the mesenchyme immediately flanking the proximal UPE(Fig. 3F). By E12.5, Nog expression was most noticeable in the lateral shelf mesenchyme and mesenchyme flanking the distal UPE(Fig. 3G). In homozygous mutants, ectopic Nog expression was most visible in the mesenchyme adjacent to the medial and proximal UPE(Fig. 3H), a region that overlaps with the normal expression pattern of Bmp4(Fig. 3H; inset). A second Bmp antagonist, gremlin, exhibited no changes in GT expression between mutant and wild-type embryos (data not shown).

Next we examined whether reduced Bmp7 expression as well as ectopic Nog expression in the GT were sufficient to affect the expression of the Bmp-target genes Msx1 and Msx2. In E11.5 wild-type embryos, Msx1 expression was highest in the mesenchyme immediately adjacent to the proximal UPE(Fig. 3I). In homozygous mutants, Msx1 expression was greatly reduced in the mesenchymal tissues flanking the proximal UPE (Fig. 3J). By E12.5, Msx1 expression is restricted to the distal glans region (Fig. 3K),which is smaller in the mutant male embryo(Fig. 3L).

Interestingly, Msx2 expression is seen throughout the UPE, as well as in the mesenchyme flanking the distal UPE of E11.5 wild-type embryos(Fig. 3M). In homozygous mutants, Msx2 expression is absent in the proximal UPE in the same region as Bmp7 expression is reduced(Fig. 3B) and adjacent to the site of ectopic Nog expression(Fig. 3H). In the rostral GT, Msx2 expression appears normal, which is consistent with the levels of Bmp7 expression in this region(Fig. 3N). By E12.5, Msx2 expression is highest in the developing glans and throughout the UPE (Fig. 3O), whereas in homozygous mutants, no Msx2 expression could be seen in the more proximal UPE (Fig. 3P). The lack of Msx2 expression in the proximal UPE is consistent with the loss of Bmp7 expression in the proximal UPE of Hoxa13^GFP-homozygous mutants(Fig. 3B,D).

We next examined the expression of Bmp-receptors Bmpr1a, Bmpr1b,and Bmpr2 in GT of Hoxa13^GFP embryos. At E12.5,all three receptors are uniformly expressed at low levels throughout the genital shelf mesenchyme (GSM) and UPE of wild-type male embryos(Fig. 4A,C,E). However, in Hoxa13^GFP-homozygous mutants, elevated levels of Bmpr1b expression was consistently seen in the condensing rectal mesenchyme (compare Fig. 4C and D). At E 13.5, Bmp-receptor expression was confined to the developing glans, UPE and preputial gland condensations in wild type and homozygous mutants (Fig. 4G-L). Interestingly, in homozygous mutants, Bmpr1b expression was reduced in the mid-proximal UPE (Fig. 4; compare I and J), whereas Bmpr2 expression appeared elevated in the distal UPE when compared with controls(Fig. 4K,L).

To correlate changes in Bmp signaling with the defects associated with loss of Hoxa13 function, we examined whether blocking Bmp signaling during GT outgrowth could recapitulate the homozygous-mutant phenotype in unaffected Hoxa13^GFP-heterozygous embryos. Blocking antibodies specific for Bmp4 (αBmp4) or Bmp7 (αBmp7) were applied to cultured GTs from E12.5 Hoxa13^GFP heterozygous male embryos. In control experiments, GT explants treated with whole-serum immunoglobulins (IgGs) exhibited normal development of the urethra and glans,which after 72 hours were similar in size and morphology to the external genitalia of E15.5 male embryos (Fig. 1). In the developing glans and urethra, treatments with control IgGs had no affect on the formation of a meatus (M) or the progression of the mesenchyme necessary for urethral closure in 4 out of 5 embryos tested(Fig. 5A). Next, although Bmp4 expression was unaffected in the GT Hoxa13^GFP homozygous mutants, we hypothesized that its function in the GT may be antagonized by the ectopic expression of Nog in the mesenchyme flanking the UPE(Fig. 3F,H), a site that highly expresses Bmp4 (Fig. 3H; inset). To test this hypothesis, Hoxa13^GFP-heterozygous explants were treated withα Bmp4. After 72 hours, dramatic changes in the growth and closure of the urethra were observed, which caused either partial (3/8) or complete ablation (5/8) of the developing meatus, as well as affecting the progression of urethral mesenchyme, causing a fistula in the sub-glans region in all embryos (8/8) examined (Fig. 5B). Similarly, explants treated with αBmp7 also exhibited partial (2/5) or complete (3/5) ablation of the meatus, as well as defects in glans morphology, which appeared rounded in all (5/5) treated explants (Fig. 5C). A comparison of the GT phenotypes elicited by αBmp7 andα Bmp4 treatments with the malformations exhibited by Hoxa13^GFP-homozygous mutants(Fig. 1F) reveal a high degree of similarity, particularly in meatus where loss of Hoxa13 function causes either partial or complete ablation of the meatus in all homozygous-mutant embryos examined (Fig. 1F).

Explants treated with αBmp7 also exhibited a thickening of the epithelial plate (Fig. 5C),a phenotype detected by E12.5 and persisting to E15.5 in Hoxa13^GFP homozygous male mutants(Fig. 1B,F). Interestingly, the effects on meatus development elicited by αBmp4 andα Bmp7 treatments were identical in female GT explants (data not shown), which suggests that the signals required for meatus cavitation function prior to the sex-specific differentiation of these tissues. Finally,we examined whether the blocking of Bmp7 signals could also reproduce changes in PCD seen in the UPE and GT mesenchyme of Hoxa13^GFP-homozygous mutants(Fig. 2). Indeed, heterozygous GT explants treated with αBmp7 for 24 hours exhibited dramatic reductions in PCD in the distal UPE and adjacent mesenchyme when compared with heterozygous explants treated with control IgGs(Fig. 5D,E). These results confirm that PCD in the GT mesenchyme requires Bmp7 signaling, which is reduced in the UPE of Hoxa13^GFP-homozygous mutants.

As Fgf8 produced by the UPE directs many of the proliferative events during GT development (Haraguchi et al.,2000), we examined whether loss of Hoxa13 function affects Fgf8 expression and signaling in this region. In the GT, Fgf8 expression was highest at E11.5 in both wild type and homozygous mutants (Fig. 6A,B). Section analysis of Fgf8 expression in the UPE revealed marked differences in the localization of Fgf8 transcripts between wild-type and homozygous mutant embryos. Specifically, wild-type embryos express Fgf8 throughout the UPE,providing a proliferative signal along the entire GT axis(Fig. 6C). By contrast, Hoxa13^GFP-homozygous mutants exhibit a dramatic reduction of Fgf8 expression in the proximal UPE, whereas expression in the distal UPE appears unaffected (Fig. 6D). Interestingly, the genital shelf mesenchyme immediately adjacent to the proximal UPE is also the site of poor proliferation in E11.5 embryos lacking Hoxa13 (Fig. 6; compare I,J with K,L), which suggests that reduced Fgf8 expression in this region causes this defect in proliferation.

To test whether the loss of Fgf8 expression in the proximal UPE of homozygous-mutant embryos is sufficient to account for reduced proliferation in genital shelf mesenchyme, heparin acrylic beads soaked in BSA or Fgf8b were placed in the proximal UPE of homozygous mutant GT explants. In all cases(4/4), Fgf8b-treated beads rescued the proliferation defects seen in the surrounding mesenchyme (Fig. 6K,L), restoring mitotic levels throughout the GT(Fig. 6H). Identical treatments using BSA-treated beads caused no change in the proliferative defect(Fig. 6G), which closely resembled the levels of proliferation exhibited by E11.5 homozygous-mutant embryos (Fig. 6K). Extended culture of E11.5 mutant GTs for 3 days in the presence of Fgf8b had no discernable effect on urethra or meatus development when compared with BSA-treated controls (data not shown). This suggests that Fgf8expression in the UPE provides mainly the initial proliferative impetus to the surrounding mesenchyme.

To determine if the loss of Fgf8 expression in the proximal UPE reflects a change in the polarization of the UPE, we examined the expression of Shh in the GT of E12.5 Hoxa13^GFP embryos. Analysis of Shh expression in the UPE showed no differences in expression between wild-type and homozygous-mutant embryos(Fig. 7A,B), which confirms that polarizing signals are normally produced in the mutant UPE. The presence of Shh expression in the mutant proximal UPE strongly suggests that Hoxa13 is required for Fgf8 expression in the proximal UPE,as a more generalized effect would have also affected Shh expression in this region.

Section analysis of the UPE of E12.5 embryos revealed distinct differences in epithelial cell morphology, stratification and cytokeratin expression patterns. In particular, epithelial cells in the wild-type (not shown) and heterozygous UPE appeared more mature, as defined by their cuboidal shape,stratification and expression of cytokeratin (K) 14 (K-14)(Fig. 7A,C). By contrast, the cells within the mutant UPE appeared less mature, as defined by their rounded morphology, poor stratification and reduced K-14 expression(Fig. 7B,D). Finally, the expression of K-8 and K-18 in these same tissues(Fig. 7E,F) confirms that reduced K-14 expression in the epithelia of homozygous mutants is specific for this maturation marker and is not a result of a generalized loss in epithelial cell identity, as K-8/18 are typically expressed in most simple epithelia(Kuzrock et al., 1999).

Next, because changes in Tgfβ signaling can also affect tissue vascularization, we examined whether changes in the expression of the Tgfβ family member Bmp7 and its antagonist, Nog, might also affect GT vascularization. Section analysis of the distal urethra and glans revealed dramatic differences between heterozygous- and homozygous-mutant embryos in the vascularization of these tissues. In E13.5 heterozygous and wild-type (not shown) embryos, the mesenchyme immediately flanking the UPE is vascularized by fine capillary vessels 10-15 μm in diameter (Fig. 8A). By contrast, the same region in homozygous mutants was consistently (10/10)vascularized by vessels typically ranging from 70-150 μm in diameter(Fig. 8B). Cellular analysis of these enlarged vessels confirmed them to be functional vascular tissue,containing an endothelium expressing the angiogenic marker Pecam-1 (CD31;Pecam–Mouse Genome Informatics) (Fig. 8A,B) as well as red blood cells(Fig. 7B). Interestingly,vessel endothelial cells of heterozygous embryos strongly co-express Pecam and Hoxa13^GFP as determined the yellow co-localized signal(Fig. 8A), whereas in homozygous mutants the degree of Pecam and Hoxa13^GFPco-localization was greatly reduced in the endothelial cells(Fig. 8B).

Section analysis of the distal GT at E13.5 also provided a possible explanation for distal closure defects seen in Fig. 1F, where the protrusion of the UPE throughout the distal urethra places a physical barrier of epithelium at the site where progressing mesenchymal shelves meet and fuse(Fig. 8B). By contrast, the UPE in heterozygous littermates was consistently regressed below the level of the progressing mesenchyme, providing a clear path for the progressing mesenchyme to join and fuse (Fig. 8A).

To characterize the relationship between Hoxa13 and the androgen signals necessary for GT growth and development, we examined the expression of the androgen receptor (AR) in the developing glans and urethra of E15.5 littermates. In the distal urethra of heterozygous embryos, AR proteins are detected in the surface ectoderm as well as in the condensing mesenchyme proximal and lateral to the urethral epithelium(Fig. 9A). In homozygous mutants, AR expression is reduced in the surface ectoderm as well as in the mesenchyme proximal and lateral to the hypospadiac urethral epithelium(Fig. 9B). Interestingly the medial mesenchymal condensation (Fig. 9A,C,D) reflects the anlagen from which the penian bone (P)develops postnatally, under the control of androgen signaling(Murakami, 1987). In Hoxa13^GFP-homozygous mutants, this condensation is absent in the more distal sections of the penis(Fig. 9B), and appears disorganized in more proximal sections when compared with heterozygous controls (Fig. 9C,D). This lack of mesenchymal condensation is remarkably similar to the phenotype exhibited in the autopod of Hoxa13^GFP-mutant mice(Stadler et al., 2001), which suggests a similar role for Hoxa13 in regulating mesenchymal cell adhesion and provides an explanation for hypoplasia of the os-penis in Hypodactyly mice (Post and Innis,1999).

In more proximal sections, AR expression co-localizes with Hoxa13^GFP in the penial mesenchymal condensation and in the lateral mesenchyme flanking the urethral epithelium(Fig. 9C). By contrast,homozygous mutants exhibit little AR expression in the lateral mesenchyme adjacent the urethral epithelium, although strong AR-Hoxa13^GFPco-localization is detected in an ectopic region approximately 50-80 μm from the urethral epithelium (Fig. 9D). Finally, we evaluated whether Hoxa13 could also be regulated by the AR ligand, dihydroxytestosterone (DHT). DHT treatments of cultured genital tubercles (10 nM) from wild-type or homozygous-mutant male embryos (E13.5) showed no elevation in Hoxa13 expression relative to ethanol treated controls (data not shown).

In this study, we present evidence that hypospadia in mice lacking Hoxa13 is caused by defects in UPE maturation, signaling and proliferative induction of the surrounding genital tubercle mesenchyme. Recently, studies by Haraguchi et al.(Haraguchi et al., 2000; Haraguchi et al., 2001) and Perriton et al. (Perriton et al.,2002) indicate that the UPE functions as a signaling center,secreting Shh and Fgf8, which are necessary for GT outgrowth and closure. To date, the unique functions of these factors during GT formation have been difficult to discern as mice lacking Shh fail to form a defined UPE from which Fgf8 is produced (Haraguchi et al.,2001). Our analysis of GT development in Hoxa13^GFP-homozygous mutants separates the individual function(s) of Shh and Fgf8, as well as identifies Bmp7 as a third UPE signaling molecule whose function is essential for the development and closure of the meatus and urethra.

In a model of Hoxa13 function during GT development(Fig. 10A,B,D), we conclude that Hoxa13 is required for the normal expression of Fgf8and Bmp7 in the UPE, as well as the repression of Nog in the mesenchyme flanking the UPE. In the absence of Hoxa13 function(Fig. 10C,E), defects in the external genitalia can be attributed to perturbations in three essential processes. First, reduced Fgf8 signaling from the proximal UPE causes a dramatic decrease in the initial proliferation of the GT mesenchyme, a process previously shown to be essential for the development of the external genitalia(Haraguchi et al., 2000; Perriton et al., 2002). Second, reduced Bmp7 signaling lowers the amount of Msx1 and Msx2 expression in the mesenchyme flanking the proximal UPE, causing a decrease in PCD necessary for the removal of overlying ectoderm, urethral tube closure and meatus formation. Finally, Hoxa13 is required for the repression of Nog in the mesenchyme flanking the medial-proximal UPE, a site where Bmp4 signaling has an essential role in mediating urethral tube closure. Surprisingly, we did not see an elevation in PCD in the distal urethra and glans, which are noticeably smaller in Hoxa13^GFP homozygous mutants. This finding is consistent with the normal levels of Shh expression in the UPE of Hoxa13^GFP mutants, which was previously shown to confer cell survival in tissues derived from the hindgut or foregut endoderm(Perriton et al., 2002; Litingtung et al., 1998). By this same mechanism, elevated levels of Shh in the UPE of Hoxa13^GFP-homozygous mutants could contribute to the reduced PCD seen in these tissues, although we see no evidence for elevated Shh expression in the UPE of Hoxa13^GFP homozygous mutant embryos.

Clearly, the signals emanating from the UPE are complex, as both proliferative (Fgf8) and pro-apoptotic (Bmp7) signals impact the developing GT in overlapping temporal and spatial domains. What is intriguing about this signaling is how the GT mesenchyme interprets the appositional roles of Fgf8 and Bmp7 to regulate the growth and closure of the external genitalia. In Hoxa13^GFP-mutant mice, small changes in these appositional signals, manifest as either coronal or coronal/urethral hypospadia. Recognizing the sensitivity of the developing GT to changes in Bmp and Fgf signaling, it is possible that perturbations in these same signaling events contribute to the rising incidence in hypospadias affecting the human birth population(Paulozzi et al., 1997). Warranting close scrutiny would be the role of environmental factors in perturbing signaling between Fgf8, Bmp7 or Bmp4 and their receptors during GT development.

Similarly, because no changes in Fgf10 or Wnt7aexpression could be detected in GT of Hoxa13^GFP mutants(data not shown), it is likely that Bmp7, Bmp4 and Fgf8function as the predominant factors required for GT outgrowth. This conclusion is supported by the rescue of GT proliferation by ectopic Fgf8 application(this work) (Haraguchi et al.,2000), as well as studies in chick where the overexpression of truncated Hoxa13 proteins caused reduced Fgf8 expression and severe malformations of the gut and genitourinary regions(de Santa Barbara and Roberts,2002). Recently, studies of β-catenin function in the developing limb (Barrow et al.,2003) also revealed a role for this molecule in mediating GT outgrowth, as mice lacking β-catenin in the GT exhibit hypoplasia of the GT and its derivatives (J. Barrow, personal communication).

By E12.5 the UPE exhibits changes in cell morphology and cytokeratin expression that are indicative of maturation and stratification of the epithelial signaling center. In particular, cells lining the UPE shift from a layer of simple epithelium to several cuboidal cell layers (stratified) that now express K-14, a cytokeratin that is normally found in differentiating epithelium (Kuzrock et al.,1999; Kivela and Uusitalo,1998; Coulombe and Omary,2002). In homozygous-mutant littermates, stratification and K-14 expression are compromised, resulting in the disorganized arrangement of rounded cells in the UPE. This phenotype may be attributed to the focal loss of K-14 expression in the UPE, as mice lacking K-14 exhibit defects in the stratification of the skin and corneal epithelium(Lloyd et al., 1995). Interestingly, the loss of epithelial keratins is often associated with elevated PCD (McGowan et al.,2002; Oshima,2002). However, in proximal UPE of Hoxa13^GFP-homozygous mutants no significant changes in PCD were observed. This result is explained by the maintenance of K-8/18 expression in mutant UPE, which protects epithelium from PCD by directly binding pro-apoptotic proteins, including the Tradd domain of Tnf receptor 1,as well as providing structural support to minimize PCD initiated by mechanical stress (Inada et al.,2001; Marceau et al.,2001; Baribault et al.,1993; Baribault et al.,1994). Similarly, changes in the stratification and keratin expression patterns in the UPE could also affect cell-cell interactions between the progressing mesenchyme and the underlying epithelium, preventing the efficient movement of mesenchymal cells necessary for urethral closure.

An examination of PCD during GT development suggests a highly dynamic interplay between mesenchymal proliferation and apoptosis. Using a TUNEL assay, PCD was first detected in the cloacal membrane (CM) covering the GT(data not shown). Within 12 hours, a rapid shift in PCD had occurred, moving from CM ectoderm to the endodermal tissues of the UPE. What is intriguing about this shift in PCD is its coincidence with peak Fgf8 expression in the UPE. Developmentally, the localization of PCD in the UPE could serve two important functions. First, the removal of the UPE by PCD could modulate levels of Fgf8 expression, providing a signal for differentiation to occur; a similar use of PCD is seen in the developing limb, where removal of the AER reduces Fgf8 signaling to stimulate differentiation of the underlying mesenchyme (Zwilling,1955; Saunders et al.,1957; Macias et al.,1997; Niswander et al.,1994). Second, to achieve closure of the urethra, the progressing mesenchyme must cover the medial epithelial layer. In Hoxa13^GFP-homozygous mutants, the epithelial layer is elevated and thickened, which would inhibit closure of the medial urethral by the proliferating mesenchyme. After UPE removal, PCD shifts again to the mesenchymal tissues immediately flanking the UPE, occurring as early as E12.5 and persisting in the mesenchyme until closure of the penile urethra(Baskin et al., 2001; van der Werff et al., 2000). Mechanistically, the reduction of PCD in the urethral mesenchyme of Hoxa13^GFP-mutant mice helps to explain the pathology of the hypospadiac phenotype, where the persistence of thickened urethral epithelium places a physical barrier between the progressing mesenchymal shelves.

Our analysis of GT development in the presence of Bmp4 and Bmp7 blocking antibodies strongly suggests a role for these factors in mediating growth and closure of the meatus and urethra. Here, an analysis of GT development in embryos lacking either Bmp7 or Bmp4, or combinations of both, would provide the most corroborative data to define the complete function of these molecules during the formation of the external genitalia. To date, no characterizations of Bmp7 function in the external genitalia have been reported, although Bmp7 has been shown to play an essential role in mediating kidney tubule formation, as well as in the production of meiotic germ cells (Luo et al.,1995; Dudley et al.,1995; Dudley and Robertson,1997; Zhao et al.,2001). Similarly, the function of Bmp4 in the developing GT is unknown as embryos lacking Bmp4 die between E6.5 and E9.5(Winnier et al., 1995). Recognizing the similarity in phenotypes between blocked Bmp signaling and those exhibited by Hoxa13^GFP-homozygous mutants, it is probable that the defects in meatus formation and urethral tube closure seen in Hoxa13^GFP-mutant mice reflect the combined effects of reduced Bmp7 expression and those elicited by Nogantagonism.

Normally, vascularization of the developing glans is provided by two well-defined rings of capillary beds supplying both the prepuce and the mesenchyme immediately adjacent the UPE. In Hoxa13^GFPmutants, capillary vessel morphology, placement and diameter are dramatically altered in the distal glans. Interestingly, this phenotype is also present in the glans clitoris of homozygous mutant female mice (data not shown),reflecting the common origin of these two structures and conservation of the mechanisms underlying this gross change in vessel diameter. Aberrant vascularization of the glans is also associated with human hypospadias, and was interpreted as a default state stemming from the aborted differentiation of the glans, meatus and urethral spongiosum(Baskin et al., 1998; Baskin, 2000). Clearly the enlargement of the capillary vessels in the glans of Hoxa13^GFP homozygous mutant mice reflects more than a default vascular state as only the distal vessels supplying the glans are affected, whereas the entire glans and urethra are hypoplastic in these mice.

Mechanistically, changes in Tgfβ or Smad signaling also affect blood vessel development, morphology and diameter (reviewed by Dennler et al., 2002; ten Dijke et al., 2002; Weinstein et al., 2000; Oh et al., 2000; Vargesson and Laufer, 2001). In particular, mice lacking Smad5 (Madh5–Mouse Genome Informatics)exhibit many of the same defects shown by Bmp4- and Tgf-beta 1- (Tgfb1–Mouse Genome Informatics) null mice, including enlarged blood vessels (Yang et al.,1999; Chang et al.,1999). Our examination of Bmp4 and Bmp7 function in the developing GT indicates that although Bmp7 is uniquely expressed in the UPE, both molecules function similarly in the glans to regulate the formation of the meatus. Here the combinatorial effects of perturbed Bmp4 and Bmp7 signaling may contribute the enlargement of the glans capillary vessels. We cannot exclude the possibility that changes in Bmp-receptor expression in the developing glans, could also contribute to the enlarged capillary vessel phenotype, although the expression of receptors Bmpr1a, Bmpr1b and Bmpr2 in the glans was not significantly different between E13.5 homozygous-mutant and control embryos.

Our examination of AR localization in the developing murine GT revealed a high degree of co-localization between Hoxa13 and the AR in the proximal urethral mesenchyme. By contrast, only minor levels of AR protein could be detected in the same mesenchymal regions of Hoxa13^GFP-homozygous mutants. This reduction in AR expression appears specific to the mesenchyme immediately adjacent to the urethral epithelium, and suggests a role for Hoxa13 in mediating the responsiveness of these tissues to androgens by modulating the levels of AR in the urethral epithelium. This modulation in AR expression would provide a developmental link between hypospadias associated with environmental antagonists of the AR and hypospadias associated with loss of Hoxa13function, placing Hoxa13 as an important modulator of AR signaling during the development of the external genitalia(Baatrup and Junge, 2001; Curtis, 2001; Baskin et al., 2001). Alternatively, environmental compounds could affect Hoxa13 expression directly, in a manner similar to the ectopic expression of Hox genes in response to exogenous retinoids (Simeone et al., 1991; Whiting,1997).

In mice, the prenatal response of the GT to AR signaling is the growth and closure of urethra and glans penis, as mice lacking AR signaling also exhibit hypospadia (Murakami, 1987). Postnatally, AR signaling is required for the formation of the os-penis(Murakami, 1987). Here, the function of Hoxa13 in mediating the postnatal development of the external genitalia is unknown as Hoxa13 mutations derived from homologous recombination are not compatible with life beyond E15.5(Stadler et al., 2001; Fromental-Ramain et al.,1996). An examination of os-penis formation in hypodactyly mice(Post and Innis, 1999)suggests that Hoxa13 mediates similar mesenchymal condensation events as described in the limb (Stadler et al.,2001). However, it is important to note that the hypodactyly mutation in Hoxa13 also affects the expression of Hoxd13causing phenotypes more similar to those of Hoxa13/Hoxd13-compound mutants (Robertson et al., 1996; Robertson et al., 1997;Fromental-Ramian, 1996).

Using conditional mutagenesis, Hoxa13 function could be completely assessed throughout genitourinary development. This analysis would provide important clues regarding the developmental pathology underlying hypospadias beyond E15.5 while also providing an important resource to examine both neonatal development and adult maintenance of genitourinary tissues affected by loss of Hoxa13 function.

The authors wish to thank Gen Yamada, Nancy Manley and Alan Godwin for the critical reading of this manuscript. We would also like to thank Mario Capecchi for allowing our continued investigation of phenotypes in the Hoxa13^GFP-knockout mouse. This work was supported by grants to H.S.S. from the National Institutes of Health/NIDDK (5R01DK5915-01) and the March of Dimes Birth Defects Foundation (Basil O'Connor Starter Scholar Research Award Grant number 5-FY00-521).