Sirenomelia Phenotype in Bmp7;Shh Compound Mutants: A Novel Experimental Model for Studies of Caudal Body Malformations

Sirenomelia is a severe congenital malformation of the lower body characterized by the fusion of the legs into a single lower limb. This striking external phenotype consistently associates severe visceral abnormalities, most commonly of the kidneys, intestine, and genitalia that generally make the condition lethal. Although the causes of sirenomelia remain unknown, clinical studies have yielded two major hypotheses: i) a primary defect in the generation of caudal mesoderm, ii) a primary vascular defect that leaves the caudal part of the embryo hypoperfused. Interestingly, Sirenomelia has been shown to have a genetic basis in mice, and although it has been considered a sporadic condition in humans, recently some possible familial cases have been reported. Here, we report that the removal of one or both functional alleles of Shh from the Bmp7-null background leads to a sirenomelia phenotype that faithfully replicates the constellation of external and internal malformations, typical of the human condition. These mutants represent an invaluable model in which we have analyzed the pathogenesis of sirenomelia. We show that the signaling defect predominantly impacts the morphogenesis of the hindgut and the development of the caudal end of the dorsal aortas. The deficient formation of ventral midline structures, including the interlimb mesoderm caudal to the umbilicus, leads to the approximation and merging of the hindlimb fields. Our study provides new insights for the understanding of the mechanisms resulting in caudal body malformations, including sirenomelia.


Introduction
Sirenomelia, or sirenomelia sequence, is a rare (1.1-4.2 births of 100.000) and striking congenital malformation of the lower body characterized by the fusion of the lower limbs [1,2,3,4,5,6] . This feature, which is pathognomonic for the diagnosis, gives fetuses and newborns a mermaid appearance, thence the alternative name of Mermaid Syndrome. Sirenomelia is a multisystemic condition because of the constant association of significant anomalies, mostly but not exclusively of the caudal body. It is a severe condition, usually lethal in the perinatal period [3,4] although less severe cases have recently survived with appropriate reconstructive surgery [7].
Sirenomelia shows an ample spectrum of clinical presentations due to the variability in the degree of fusion of the lower limbs and in the combination of visceral malformations [2,4,8,9,10]. Without exception, genitourinary and gastrointestinal malformations are found in sirenomelia, since they are features secondary to oligohydramnios [5,8,11,12]. Agenesis of the terminal colon with imperforate anus occurs in the majority of cases [3,4,5,8,11]. The kidneys, ureters, urinary bladder and external genitalia are always hypoplastic in variable degree whereas the gonads are usually unaltered. Characteristically, the umbilical cord has a single artery and major anomalies of the abdominal arterial system are also common [3,4,10,13].
Although the causes of human sirenomelia remain unknown, clinical studies have yield two main pathogenetic hypotheses: the defective blastogenesis hypothesis and the vascular steal hypothesis. The defective blastogenesis hypothesis is supported by the overall malformation of the caudal body in sirenomelia. This suggests an early defect, probably a deficient generation of mesodermal precursors at late gastrula stages, which globally impairs the development of the caudal embryonic body [14,15,16]. Under this view, sirenomelia is considered a severe manifestation of the complex malformation called Caudal Dysgenesis (CD; OMIM 600145; formerly referred to as Caudal Regression Syndrome). CD describes a heterogeneous variable association of malformations of the lower body that always includes some degree of sacral agenesis [12,15,17,18,19,20].
The vascular steal hypothesis posits a vascular origin for sirenomelia. This is based on the observation that fetuses with sirenomelia almost invariably exhibit a single umbilical artery (SUA), instead of the normal two [13,21]. The SUA originates quite high from the abdominal aorta, usually immediately beneath the celiac branch, and below its origin the abdominal aorta is highly hypoplastic and lacks most of its branches. It has been proposed that the SUA diverts the normal blood flow to the placenta leaving the lower part of the body with a severely deficient circulation incompatible with normal development [2,8,21]. The SUA has also been referred to as persistent vitelline artery to indicate its possible origin in the vitelline plexus and its presence is considered by some authors as pathognomonic of sirenomelia although it has occasionally been reported in other conditions [2,8,14,18,21,22,23,24]. In this view, sirenomelia and CD are considered two separate conditions. Therefore, some debate remains in the clinic on whether sirenomelia and CD are separate conditions or just different degrees of the same multisystemic malformation [12,24].
Besides clinical data, experimental data in mouse indicate a genetic origin for sirenomelia as this malformation has been found in several genetically modified mouse strains with either loss-offunction of bone morphogenetic protein (Bmp) signaling or with gain-of-function of retinoic acid (RA) signaling [25,26]. Yet, human sirenomelia has mainly been considered sporadic since familial cases have only been reported very recently [27]. However, it should be noted that the lack of evidence for a major genetic influence in human mermaid syndrome may rely, at least in part, on the poor documentation of abortions [4,28].
The implication of Bmp signaling in sirenomelia came from the observation that the double Bmp7;Tsg (Twisted gastrulation) mutant mouse displayed this malformation [26]. Tsg is a modulator of Bmp signaling that in the caudal body positively influences Bmp signaling. Therefore, the removal of one or both copies of Tsg from the Bmp7 background further reduces Bmp signaling resulting in caudal abnormalities and sirenomelia [26].
The implication of RA signaling in sirenomelia came from the observation that excess of administration of RA to pregnant mice led to caudal malformation of the offspring including sirenomelia in some cases [29]. This has been confirmed by the occasional presentation of sirenomelia in the genetic disruption of Cyp26a1, an enzyme that degrades RA [25,30] and in the disruption of other genes that interact with the Cyp26 family of enzymes [31,32,33,34].
Unfortunately, the low penetrance and early embryonic lethality of Bmp7;Tsg and of Cyp26a1 mutants exclude them as suitable models of human sirenomelia. The availability of an animal model of sirenomelia would be an invaluable tool to unravel its pathogenesis and it could provide insights on the development of the caudal body. Such a model will also permit to analyze the relationship between sirenomelia and CD and the influence of epigenetic factors including RA and maternal diabetes.
Here we identify and characterize the sirenomelia phenotype in Bmp7;Shh compound mutants. We show that Bmp7 2/2 ;Shh 2/2 and Bmp7 2/2 ;Shh +/2 mutants conjointly replicate the constellation of external and internal malformations in the human condition and thus we propose these mutants as the most suitable animal model for human sirenomelia to date.
The analysis we have performed shows the early onset of the malformation that is detectable at E9.5. The atrophic hindgut and cloaca together with the defective development of the caudal end of the dorsal aortas, which form by vasculogenesis [35,36], result in a defective midline and ventral mesoderm caudal to the umbilicus that leads to the junction of the hindlimb fields. Our results suggest that the lower limb fusion in sirenomelia, although giving the name to the condition, is a secondary event and that sirenomelia phenotypes may occur without significant impact in dorsal structures.

Mutant mice
Studies with mice were performed strictly following the EU regulations and 3R principles. The Bmp7 [University of Oxford, Oxford, UK; 37] and Shh [University Hospital Basel, Department of Biomedicine, Switzerland; 38] mutant mice lines used in this study were kindly provided by Elisabeth Robertson (University of Oxford, Oxford, UK) and Rolf Zeller (University of Basel, Switzerland) respectively. The mice were maintained on a mixed genetic background and genotyped using tail or embryonic membranes biopsies as previously reported. Wild type and mutant embryos of the desired embryonic stage were obtained following standard protocols.

Skeletal preparations, X-Gal staining and in situ hybridization
After removing skin and viscera, mouse embryos were fixed in 95% ethanol. Alizarin Red and Alcian blue skeletal staining was performed according to standard protocols cleared by KOH treatment and stored in glycerol. X-Gal staining in Whole-mount embryos was performed as previously described [37]. After staining, the embryos were fixed and processed for inclusion in paraffin if desired.
In situ hybridization (ISH) using digoxigenin labeled antisense riboprobes was performed in whole-mount and in sections [39].

Cell death and cell proliferation assays
Detection of cell death in sections of paraffin-embedded tissue was performed using terminal deoxynucleotidyl transferase mediated dUTP nick-end labelling (TUNEL) with the Apoptag Fluorescein Direct In Situ Apoptosis Detection Kit (Intergen) and following the manufactured instructions. Detection of cell proliferation in sections was performed by immunohistochemical assay using the anti phosphorylated histone H3 antibody (rabbit polyclonal Phospho H3 from Upstate Biotechnology, USA) diluted at 1/100.

Reduction of Shh from the Bmp7 null background results in sirenomelia
In the context of other studies on limb development [40] we found that the reduction of one or both Shh functional alleles from the Bmp7-null background led to sirenomelia. To study this phenotype in depth, mice heterozygous for the Bmp7 null allele (Bmp7 +/2 ) were crossed to mice heterozygous for the Shh null allele (Shh +/2 ) and the double heterozygous (Bmp7 +/2 ; Shh +/2 ) mice obtained were subsequently crossed between them.
Up to E12.5 all genotypes were recovered at the expected Mendelian ratios, but subsequently the frequency of recovery of double homozygous mutants (Bmp7 2/2 ;Shh 2/2 ) progressively declined to be less than 2% at late gestational stages (6.25% expected). Thus, double homozygous shows an earlier embryonic lethality than single Bmp7 2/2 mutants, which are recovered at the expected Mendelian rates at birth and to single Shh 2/2 mutants, which die at the end of the gestational period. Of the 170 neonates obtained, only 3 were double homozygous mutants while Bmp7 2/2 ;Shh +/2 were recovered at the expected Mendelian ratio (Table S1). These two types of mutants died in the first few hours after birth due to the multisystemic visceral malformations.
Double null Bmp7 2/2 ;Shh 2/2 mutants displayed sirenomelia with complete penetrance (100%) in addition to the typical traits of the Shh null phenotype such as cyclopia, proboscide and truncated forelimbs Fig. 1B; [38]. Bmp7 2/2 ;Shh +/2 mutants, in which one functional copy of Shh remains, displayed a milder form of sirenomelia with about 50% penetrance (11 out of 21 neonates; Fig. 1C). The hindlimb fusion in Bmp7 2/2 ;Shh +/2 mutants ranged from a partial merging at proximal/tight level (Fig. 1D) to a complete fusion all along the proximo-distal length of the hindlimbs (Fig. 1E). It should be mentioned that in those Bmp7 2/2 ;Shh +/2 mutants that did not display a clear mermaid phenotype, the position of the hindlimbs was closer to the ventral midline than normal (compare panel A with F in Fig. 1).
External inspection of double homozygous mutants neonates, in comparison with wild type littermates, revealed a complete absence of genital tubercle and imperforate cloacal membrane (compare Fig. 1G with Fig. 1H). Bmp7 2/2 ;Shh +/2 mutants displayed a variable range of underdeveloped genital tubercle and anal malformations including imperforated and prolapsed anus (Fig. 1I). The rudimentary genital tubercle and any evidence of anus, if present, were always located caudal to the fused hindlimbs ( Fig. 1H-I).

Skeletal phenotype of sirenomelia mutants
To examine the skeletal morphology of the fused hindlimbs, we performed skeletal preparations of neonates (Alizarin red and Alcian blue and Victoria blue staining, see M&M). The skeletal preparations, in comparison with wild type ( Fig. 2A), demonstrated that the single medial hindlimb typical of the severe sirenomelia of double homozygous (Bmp7 2/2 ;Shh 2/2 ; Fig. 2B) was the result of the fusion of two independent hindlimbs. It consisted of two proximally fused femora followed by two tibia arranged in parallel with a single fibula between them. At its proximal (knee) and distal (ankle) ends it was possible to discern that the single fibula derived from the fusion of the two bilateral fibulae. In the foot one or two digits were scored as corresponded to the Shh mutant phenotype that only displays one digit in each hindlimb [43,44]. Interestingly, the zeugopod (intermediate limb segment) showed an improved morphology compared to Shh null embryos [43,44] indicating that, at least part of the Shh-limb phenotype was due to increased Bmp signaling (Fig. S1). The hindlimb fusion exhibited by Bmp7 2/2 ; Shh 2/2 double homozygous corresponds to Types IV-V in Stocker and Heifetz classification [3,10].
In Bmp7 2/2 ;Shh +/2 mutants the hindlimb fusion was mainly of soft tissue with each of the fused hindlimbs showing a complete set of skeletal elements (Fig. 2C). In the majority of cases the two fused limbs diverged from the ankle and nine or ten digits were scored depending on whether the two pinky digits, which were in medial position, were fused. Bone fusion between tarsal-metatarsal elements of the two bilateral feet was also observed (Fig. 2C). The sirenomelia of Bmp7 2/2 ;Shh +/2 mutants can be classified as Type I of Stocker and Heifetz classification [3,10].
Importantly, all degrees of sirenomelia showed a characteristic absence of hindlimb rotation that resulted in the medial position of the fibulae (or single fibula) between the two tibiae ( Fig. 2B-C). The patella, when present, was consistently found in a luxated medial position (arrowheads in Fig. 2C) and the soles of the feet faced anteriorly. This very same skeletal arrangement is found in human cases of sirenomelia [4,10,28]. An early merging of the left and right hindlimb fields in the midline along their posterior borders would prevent the subsequent rotation of the hindlimbs.
Abnormalities of the pelvic skeleton were also scored in mermaid newborns ( Fig. 2D-E). The pelvic (hip) bone is formed by the fusion of three distinctly developed skeletal elements: the ilium, the ischium and the pubis (see drawing in Fig. 2D). Bmp7 2/2 ;Shh 2/2 double homozygous showed the complete absence of caudal vertebral column described for the Shh mutant phenotype ([38], data not shown) whereas the vertebral column, the sacrum and the tail showed a normal development in Bmp7 2/2 ;Shh +/2 mutants except for the distal kinked tail typical of Bmp7 homozygous mutants (data not shown). Interestingly, Bmp7 2/2 ;Shh +/2 mutants showed pelvic defects characterized by a variable degree of truncation of the pubis and midline fusion of the two bilateral ischia whereas the ilia were normal (Fig. 2E). This led to an anomalous lesser pelvis that must impact the perineum and hindlimb musculature. This particular morphology of the pelvis, with both ischia fused in the midline and the pubes laterally separated, likely derives from the impossibility of the fused hindlimbs to rotate.
To sum up, our analysis shows a variable degree of approximation and fusion of the leg and pelvic skeletal elements in correlation with the genotype; fusion of leg bones was only observed in double homozygous. In all cases there is a constant absence of the normal rotation of the hindlimbs, which influences the arrangement of the skeletal elements including the pelvic bone. Finally, it should be stressed that, the mild sirenomelia of Bmp7 2/2 ;Shh +/2 occurs with a normal development of the sacrum and tail.

Visceral phenotype of sirenomelia mutants
To study the visceral component in our mutants with sirenomelia, we first performed a gross anatomy dissection of the abdominal organs in newborn specimens (Fig. 3). The macroscopic dissection showed absence of obvious anomalies in the liver, stomach and midgut. We focused our analysis on the Bmp7 2/2 ;Shh +/2 genotype because the milder phenotype makes it more informative. In Bmp7 2/2 ;Shh +/2 newborns the gross anatomy malformations concentrated in the urogenital track and hindgut. The kidneys and ureters were always severely affected, highly hipoplastic and frequently hydronephrotic and the bladder was consistently absent (Fig. 3A-B). The hindgut was narrower than normal and frequently ended blind. In some cases the intestine and stomach were distended with gas (not shown) probably secondary to the anal atresia and/or to a possible tracheo-esophageal fistula as has been shown for Bmp7 mutants [45]. In contrast, the gonads and adrenal glands were easily identifiable and no obvious gross abnormalities of the gonadal ducts of either sex were observed ( Fig. 3A-B; scheme).
A careful histological analysis of serial sagittal sections at E12.5 and E15.5 (Fig. 3) confirmed the results of the anatomic dissection in newborns. Whereas the urogenital septum separated the urogenital and anorectal tracts in wild type embryos (asterisk in Fig. 3C), it was never observed in mutants. Only the atrophic and frequently ending blind hindgut was distinguishable in Bmp7 2/2 ;Shh +/2 mutants (Fig. 3D). Interestingly, a blind termination of the urachus (an epithelial duct connecting the allantois and the bladder), due to the absence of bladder was also observed (Fig. 3D).

Molecular defect in the sirenomelia mutants
The phenotype of our mutants must result from reduced Bmp and Shh signaling. To understand and localize the origin of the defect, we first investigated the pattern of expression of Shh and Bmp7 during the early normal embryonic development of the caudal body.
Shh expression was analyzed by whole mount in situ hybridization (ISH) followed by section analysis. As previously described, Shh transcription occurred in the notochord and ventral hindgut [38], both at E8.5 (10-12 pairs of somites; before the turning of the embryo) and at E9.5 ( Fig. 4A-E).
As the Bmp7 null allele carries LacZ [37,41], we used X-gal staining in Bmp7 heterozygous embryos to detect Bmp7-expressing cells. At E8.5, Bmp7 expression was restricted to the notochord as can be observed in the whole mount staining and in corresponding transverse section (Fig. 4F-G). At E9.5, Bmp7 expression was mainly detected in the ventral ectoderm, caudal to the allantois, and corresponding to the level of the cloaca as can be appreciated in the whole-mount staining and medial sagittal section (Fig. 4H-J) [46,47]. This area of expression overlaps the ventral ectodermal ridge (VER), a remnant of the primitive streak [46]. A faint X-gal staining also remained in the notochord at this stage. A scheme depicting the pattern of expression of Shh (redish) and Bmp7 (blue) at E9.5 is shown in Fig. 4K.
To identify the state of Bmp and Shh signaling in our mutants, we compared the expression of well-established downstream targets of each pathway in wild type and mutant embryos at E9.5 focusing on the caudal embryo body. First, as expected for Shh heterozygous, we found that the domain of Shh expression was similar in Bmp7 2/2 ;Shh +/2 mutants and wild type embryos (Fig. 4L). The expression of Ptc1 was used as reporter of Shh activity. In wild-type embryos, Ptc1 transcripts were detected in the ventral neural tube and notochord, medial somites, splanchopleure, hindgut and the ventral caudal mesoderm in control embryos (Fig. 4M). In Bmp7 2/2 ;Shh +/2 mutant embryos, Ptc1 expression in the ventral caudal mesoderm was clearly downregulated suggesting a decrease in hh signaling at this level (Fig. 4M). The expression of Msx2, a bona fide target of Bmp signalling, that normally occurs in the ventral caudal mesoderm and ectoderm of control embryos, was dramatically downregulated in Bmp7 2/ 2 ;Shh +/2 embryos (Fig. 4N). The downregulation in Msx2 in compound mutants reflects the loss of Bmp7 signalling as is similarly observed in Bmp7 single mutants but not observed in Shh single mutants (not shown). Also, the expression of Bmp4 and Bmp2 ligands was not significantly changed in our mutants ( Fig. S2 and not shown).
Thus, our analysis localizes the region with defective signaling in the ventral caudal mesoderm between the allantois and the cloaca. This corresponds to the pericloacal ventral region and is encircled in Fig. 4K.

Onset of the mutant phenotype
To get insights into the way the mermaid phenotype was established, we performed a histological analysis in our allelic series of embryos. The analysis focused on the caudal body, and was performed at E8.5 (10-12 Somites) and E9.5 (22-23 Somites) before posterior morphogenesis was heavily altered. For a maximum detail, it was performed in semithin (1 micron thick) transverse serial sections of araldite embedded embryos.
Transverse sections through the caudal region of E8.5 Bmp7 2/2 ; Shh +/2 and Bmp7 2/2 ; Shh 2/2 embryos failed to detect any abnormality compared with equivalent stage wild type embryos (Fig. 5A-C). At this stage, the hindgut, the dorsal aortas and the primitive omphalomesenteric artery, a single vessel that forms immediately ventral to the hindgut as the splanchnopleural folds coalescence and fuse, showed similar morphology in all genotypes.
However, morphological differences between wild type and mutant embryos were evident by E9.5 (Fig. 5D-F0). In wild type embryos, transverse sections at the most caudal level showed the centrally positioned cloaca, the bilateral coelomic cavities and the dorsal aortas and their ventral recurved portions (Fig. 5D). A section through a more cranial level revealed the omphalomesenteric artery derived from the fusion of the two ventral recurved aorta portions (Fig. 5D9). The omphalomesenteric artery that, at this stage, was of big caliber, contacted the hindgut dorsally and the ventral body wall ventrally (Fig. 5D9) dividing the coelomic cavity into two independent bilateral cavities, at this low abdominal (pelvic) level.
Our study revealed similar defects in Bmp7 2/2 ;Shh 2/2 and Bmp7 2/2 ;Shh +/2 mutants at this stage. First, the diameter of the mutant hindgut and cloaca was significantly reduced compared to wild type embryos ( Fig. 5E-F compare with Fig. 5D). In fact, the final dilatation typical of the cloaca was only observed in much fewer sections than in wild type littermates indicating its poor development (Fig. 5E-F). Second, the final recurved portions of the paired dorsal aortas were not identified in the mutants and neither was the omphalomesenteric artery derived from their fusion (Fig. 5E-E9, F-F9). In contrast, the omphalomesenteric artery continued ventral to the hindgut, as in previous stages. It showed a variably poor development and in the majority of cases did not contact the ventral body wall thus leaving the coelomic cavity undivided in mutants (Fig. 5E9-F9).

Abnormal caudal vascular development in sirenomelia embryos
The above-described vascular phenotype was confirmed by the expression of a specific endothelial marker, cadherin5 (Cdh5, also VE-cadherin, [48]), in whole mount ISH followed by section analysis of hybridized embryos (Fig. 6). The expression of Cdh5 marks the distribution of endothelial cells and revealed the poor development of the local plexus of the caudal ventral body wall in compound mutants (arrowheads in Fig. 6D-E). The recurved portions of the aorta were not identified and the omphalomesenteric artery was always narrower and dorsally positioned in the compound mutants than in the other genotypes ( Fig. 6A9-E9). It should be noted that the morphology of the allantois was normal in the complete genotypic series.
In summary, our results demonstrate that sirenomelia has an early onset as the defects are present at E9.5. There is a concomitant impact on the hindgut, which is narrower than normal and shows a reduced cloacal dilatation, and in the caudal development of the dorsal aortas that fail to elongate and form the recurved portions. As a result the caudal midline of the embryo results highly affected.

Analysis of cell death and cell proliferation
Changes in cell proliferation or abnormal cell death in mesodermal precursors could in principle explain the hindgut and vascular phenotype. Indeed, a shortage in mesenchymal cells is considered the primary gut defect in Shh mutants [49]. Therefore, we performed a sequential analysis of cell death and  proliferation during the time the phenotype is being generated (E8.5 and E9.5). The TUNEL assays in transverse sections failed to detect differences in the amount or distribution of apoptotic cells between wild type and mutant embryos at E8.5 (Fig. S3). By E9.5, the analysis of serial transverse sections showed some scattered apoptotic cells in the hindgut and in the ventral caudal mesoderm of single Bmp7 2/2 , single Shh 2/2 and double Bmp7 2/2 ; Shh +/2 mutants while no cell death was detected in these locations in wild type embryos (Fig. S3). At this stage, cell death was more prominent in double homozygous Bmp7 2/2 ;Shh 2/2 embryos all across the dorso-ventral midline of the embryo (Fig. S3).
To study cell proliferation we processed sections for antiphosphorylated histone H3 immunohistochemistry that detects cells in mitosis. Our analysis failed to detect obvious differences between wild type, single and double mutants at E8.5 (not shown) and E9.5 (Fig. S4).
The finding that Bmp7 2/2 ; Shh +/2 mutants, which display sirenomelia, only showed a minimal increase in cell death intensity suggests that abnormal cell death or proliferation are not a major cause for the phenotype. Nevertheless, differences in proliferation or cell death of a few critical precursors at an undefined early stage may still occur.

Normal development of the leg buds except for their merging
It has already been suggested that the fusion of the lower legs in sirenomelia, although giving the name to the condition, is secondary to the absence of the midline structures that usually separate the left and right hindlimb fields [16,50,51]. Our results support this hypothesis.
To explore the situation of the interlimb region, we analyzed the expression of Islet1 (Isl1), a gene that is expressed in this location and that is important for hindlimb development. Isl1 is a LIM homeodomain transcription factor with multiple functions in cardiac, neural and hindlimb development [52,53,54,55,56]. It has been shown that its expression is regulated by Shh in neural tissues [57] and by Bmp4 in the oral epithelium [58]. Interestingly, Isl1 expression was highly downregulated in the caudal body of Bmp7 2/2 ;Shh +/2 mutants compared with wild-type littermates (Fig. 7A-B). At E9.5 Isl1 transcripts were found in the ventral hindgut, splanchopleure and ventral interlimb mesoderm of wildtype embryos whereas only some transcripts were detected in the splanchopleure of mutants. At E10.5 Isl1 was prominently expressed in the interlimb tissue particularly around the cloaca and recurved portions of the aortas (Fig. 7C) whereas in mutant embryos the expression was much reduced around the hindgut and in the splanchopleure (Fig. 7D). These results support the hypothesis that the midline tissues including the interlimb ventral body wall are not present in sirenomelia (Fig. 7D, compare with Fig. 7C).
We next examined Tbx4 and Pitx1 expression, specific markers of the hindlimb mesenchyme [54]. At E10.5 both Tbx4 and Pitx1 were conspicuously expressed in the hindlimb mesoderm and interlimb anterior mesoderm of wild-type embryos (Fig. 7E, G). In sirenomelia mutants, the expression of Tbx4 and Pitx1 occurred in a continuous domain reflecting the joining of both hindlimbs (Fig. 7F, H).
Except for their fusion, the development of the hindlimbs followed a normal course. The expression of Fgf8, a marker of the apical ectodermal ridge (AER) that is a signaling center crucial for limb development, was detected in mutants in a continuous stripe (shown for homozygous mutants in Fig. 8G, compare with wildtype in Fig. 8B) as the fusion of the leg buds along their posterior border resulted in the continuation of the AERs of both hindlimbs. Bmp4 and Msx2 were normally expressed both in the AER (Fig. 8 C-D and Fig. 8H-I) and mesoderm (not shown) of double homozygous mutants. Finally, Lmx1b expression, a marker of the dorsal mesoderm, also revealed the abnormal dorso-ventral position of the fused hindlimbs ( Fig. 8E and J). These results  allow the conclusion that limb development is not impaired in sirenomelia in agreement with the complete hindlimb development exhibited by Bmp7 2/2 ;Shh +/2 newborns (Fig. 1D-F and Fig. 2. C). The impaired final limb phenotype of double homozygous (Bmp7 2/2 ;Shh 2/2 ) corresponds to that of the Shh deficit with the above described improvement at zeugopod level due to the decreased in Bmp signaling [43,44].

A new animal model for human sirenomelia
Our results show a genetic interaction between Bmp7 and Shh that results in a spectrum of phenotypes including leg fusion and visceral malformations that are morphologically identical to human cases of sirenomelia. Based on the similarity of the phenotypes we consider that Bmp7 2/2 ;Shh 2/2 and Bmp7 2/2 ; Shh +/2 mutants conjointly constitute an excellent animal model in which study the mechanisms involved in the development of this devastating human malformation. Because double homozygous associates the traits typical of the Shh deficiency (Cyclopia, proboscide, abscence of ribs… [38]), they may be a less suitable model although it is important to note that human cases of sirenomelia with cyclopia have been reported, including a recent cluster in Cali Colombia [9,28,59,60]. This novel animal model has allowed us to investigate the onset of the sirenomelia malformation. Our results show a defective formation of midline structures in the caudal embryonic region coincident with the area in which decreased Bmp and Shh signaling is detected. The signaling defects result in a hypoplastic hindgut and precludes the appropriate vasculogenesis of the caudal end of the dorsal aortas and the remodeling of the primitive vascular plexus. The atrophic hindgut and cloaca, together with the vascular defect, results in a defective midline that leads to the junction of the limb fields.

The merging of the lower limbs is a secondary event in sirenomelia
Our results strongly suggest that the approximation and merging of the hindlimb fields are the result of the early visceral defect that leads to a considerably midline reduction. Therefore, the limb fusion phenotype, although giving the name to the condition, is a secondary event in sirenomelia as it has already been proposed [16,50,51].
The diagnosis of sirenomelia is certain since it relies on the fusion of the lower limbs. The grade of lower limb fusion considerably varies generally in correlation with the severity of the visceral associated malformations [3,10]. The mildest cases, types I and II of Stocker and Heifetz classification, are characterized by the midline merging of two almost normal legs except for their malposition. The most severe cases, types VI and VII of Stocker and Heifetz classification, are characterized by a single central tapering appendage that misses clear leg features. These two extreme phenotypes are seen in our mutants: sirenomelia type I and II is typically found in Bmp7 2/2 ;Shh +/2 mutants, while type V is typically found in Bmp7 2/2 ;Shh 2/2 double homozygous. Interestingly, the mild phenotypes show that the approximation and merging of the lower limbs do not interfere with normal limb morphogenesis except for the absence of rotation. This conclusion is supported by the gene expression study of limb development we have performed in the double mutants. Our analysis also indicates that the most severe lower limb phenotypes may result from late deficits in limb development independent of the merging event. For example, in our mutants, the deficit in Bmp and Shh signaling has an early impact in generating sirenomelia and an additional later effect in limb development.
One particularly interesting feature of the phenotype, which is constantly observed in all cases of sirenomelia, is the abnormal position of the legs with the soles of the feet facing anteriorly and the great toes in a lateral position. This position corresponds to a 180u rotation with regard to the normal position and results from the absence of the rotation that hindlimbs normally experience during embryonic development. Very little is known about the mechanisms that direct this rotation, but it seems obvious that the merging of the limb buds would prevent them from operating. This lack of rotation may also contribute to the altered pelvic morphology of our mutants, with both ischia fused in the midline and the pubes laterally separated. In wild type embryos, the two bilateral hip bones progressively join in the anterior midline forming the pubic symphysis by E16.5 [61,62,63]. In our mutants, the approximation of both pubes to the midline does not occur and therefore the pubic symphysis does not form whereas both ischia fuse in the midline. Thus, it seems reasonable to propose that the early merging of the leg buds would preclude the normal morphogenetic movements that reposition the pelvic girdle and lower limb and therefore leading to the observed phenotype.

A reevaluation of the etiopathogenic hypotheses of sirenomelia
As a result of the clinical studies of series of fetuses and newborns with sirenomelia, two main hypotheses have been proposed to explain this devastating malformation. First, based on the general impact in the caudal body, sirenomelia is envisaged as resulting from a deficient blastogenesis that failed to produce sufficient mesodermal precursors [3,8,14]. Second, based on the almost constant abnormal development of the umbilical vessels, sirenomelia is considered the result of a vascular defect that leaves the lower part of the body with a severely deficient circulation incompatible with normal development [2,21,64,65].
According to the deficient blastogenesis hypothesis, sirenomelia may be considered as a severe form of caudal dysgenesis (CD). CD is a heterogeneous assemblage of caudal anomalies that always include some degree of sacral agenesis [17,20]. The milder sirenomelia phenotype of Bmp7 2/2 ;Shh +/2 mutants, although associating major visceral defects in the urogenital and low intestinal tracts, exhibits a normal development of the sacrum and tail. This allows the conclusion that sirenomelia may exist as a pathologic entity distinct from CD. Interestingly, human cases with minimum or even absent dorsal skeletal and neurological component have also been reported [66,67].
Sirens resulting from reduced Bmp signaling (Bmp7;Tsg and our mutant) exhibit minor, if any, malformation of the sacrum and tail indicating normal anterior-posterior axis elongation. However, sirens resulting from excess RA signaling present impaired elongation of the caudal body [25,68] and therefore they could be considered part of the CD. These observations indicate the existence of at least two different presentations of sirenomelia depending on the presence of dorsal malformations and suggest that different regions (dorsal and ventral) of the caudal body may be subject to different developmental regulation. Some of these signals may be restricted to particular caudal regions while others may have a broader effect. Bmp and Shh signaling appear to be particularly important for the development of the ventral midline and body wall while Wnt signaling and RA signaling may have a more general effect. Thus, the heterogeneity in sirenomelia clinical presentations likely reflects different etiologies that distinctly affect different signaling pathways.
Regarding vascular steal hypothesis, our study reveals an early vascular defect in sirenomelia. We have detected a failure in the formation of the recurved portions of the aortas that leads to a misorganization of the ventral caudal vasculature. This abnormal arrangement of the caudal vessels is likely the origin of the abnormal Single Umbilical Artery (SUA) described in human cases of Sirenomelia [21,65,69]. This results from the persistence of the primitive omphalomesenteric artery due to the failure or underdevelopment of the recurved portions of the aorta. However, our data do not support a vascular origin for sirenomelia since the concomitant onset of the vascular and visceral defects precludes a causal relationship. The absence of the recurved portions of the aorta certainly contributes to the reduction of midline structures and subsequent merging of the hindlimb fields but no through a hypoperfusion mechanism as originally proposed by the vascular steal hypothesis.
To sum up, our study and the available experimental and human data fit with a genetic origin of sirenomelia that affects some of the signaling pathways involved in the development and organization of the pelvic organs and ventral body wall. The fact that each of these pathways involves multiple components, it provides an explanation for the genetic heterogeneity and phenotypic variability.

Bmp signaling and sirenomelia
Sirenomelia was first associated with reduced Bmp signaling when it was reported that the compound Bmp7;Tsg mutants exhibited this phenotype [26]. It should be noted that single Bmp7 mutants do not exhibit a mermaid phenotype and that a further reduction in Bmp signaling caused by the removal of one or two functional copies of the Bmp positive modulator Tsg is required to get the phenotype [26,41,70]. Also, it has been shown that the removal of the BMP receptor 1a (Bmpr1a), specifically from the interlimb mesoderm with the Isl1Cre line, leads to the approximation of both hindlimbs, but it does not to sirenomelia [53]. Thus, an appropriate level of Bmp signaling seems to be absolutely required for normal ventral caudal mesoderm development. This requirement is conserved across species as it has been confirmed in Xenopus laevis and in zebrafish [26,71].
The similarity of the sirenomelia phenotype of our mutants and that of Bmp7;Tsg mutants suggest a common origin. Bmps are wellaccepted targets of Shh in several systems including the gut [72,73,74,75]. Therefore, it is possible that the decrease of Shh dose in the absence of Bmp7 results in a further reduction of Bmp signaling similarly to the double Bmp7;Tsg mutant. However, our gene expression analysis do not support this hypothesis because the expression of Msx2 is similarly reduced in Bmp7 single and Bmp7;Shh compound mutants. Furthermore, a reduction in Shh signaling could contribute by itself to hindgut and vascular malformations [49,76,77]. Therefore, we favor the interpretation that the sirenomelia phenotype in our mutants is the result of the summation effect of reduced Bmp and reduced Shh signaling. Interestingly, the double Tsg 2/2 ;Bmp4 +/2 mutant presents holoprosencephaly, a characteristic of Shh loss of function [38,78]. Accordingly, Shh expression was undetectable in the fore and midbrain of Tsg 2/2 ;Bmp4 +/2 mutants, explaining the holoprosencephaly phenotype. Thus, the possibility that Shh function is reduced in Bmp7;Tsg sirenomelia mutants should be explored. In any case, the crosstalk between Bmp and Shh signaling pathways in the development of the caudal body merits further investigation.
Besides, the co-localization of the anatomic vascular defect and the site of maximum signalling defect suggest a causal relationship. Particularly, the progression of the dorsal aorta is thought to occur by the direct assembly of endothelial precursor cells (vasculogenesis) under the influence of paracrine signals that regulate the positioning and remodeling of embryonic blood vessels. Interestingly, members of the hedgehog family [77,79,80] and bone morphogenetic proteins (BMP) [81,82] are among the best-known paracrine factors that promote vascular development.
Collectively, our findings provide compelling evidence that the vasculogenesis of the recurved portions of the dorsal aortas and the reorganization of the vessels in the pericloacal area and ventral mesoderm caudal to the umbilicus require appropriate Bmp and Shh signaling. This is a region in which appropriate signaling is absolutely required to permit the normal development and organization of pelvic organs and ventral body wall.

Implications for human sirenomelia
Our data support the notion that the primary morphological defect in sirenomelia is an insufficient formation of midline structures, mainly the hindgut and caudal vasculature. This fits very well with the phenotypes observed in human cases of sirenomelia and contributes to a better understanding of the mechanisms that lead to this malformation.
Although most human cases of sirenomelia are sporadic, the information lately gathered from animal models, including our model, suggests a complex multigenetic basis for sirenomelia, probably requiring several genetic defects even in heterozygosis. This can explain the variability in phenotypic traits as well as the low prevalence in human populations. The recent report of familial descriptions of sirenomelia further supports this notion (27).
Our study also helps managing the malformations of the caudal body by determining candidate genes to establish screenings or develop markers to assist identifying potential couples at a higher risk of having an infant with caudal defects. We anticipate that in the future it will be particularly important to assess the contribution of human BMP7 gene to the genetic risk for that spectrum of developmental defects. Figure S1 Comparison of zeugopod development in Shh 2/2 versus double Bmp7 2/2 ;Shh 2/2 mutants. Alizarin Red-Alcian Blue skeletal preparations are accompanied by an schematic representation of the skeletal elements. Note the improvement in the morphology of the zeugopod in the double mutant compared to that of the single Shh 2/2 mutant. Abbreviations  Figure S4 Cell proliferation analysis. The immunohistochemistry with anti pH3, which marks cells in mitosis, failed to detect obvious differences between genotypes. (TIF)