Severe nasal clefting and abnormal embryonic apoptosis in Alx3/Alx4 double mutant mice

The vertebrate skull is classically subdivided in chondro-, neuro- and viscerocranium, which have different embryonic and evolutionary origins. Based on lineage studies in avian embryos, Couly et al. (Couly et al., 1993) recognise a ‘chordal’ and a ‘prechordal’ (or ‘achordal’) skull. While the chordal skull is derived from paraxial and cephalic mesoderm, the prechordal skeleton is entirely derived from the cephalic neural crest (LeDouarin et al., 1993). This implies that the bones and connective tissues of the vertebrate face originate entirely from neural crest-derived tissues.

Facial development starts with the emergence of the five facial primordia: the frontonasal process and the paired mandibular and maxillary processes that are derived from the mandibular arch. These processes consist mainly of neural crest-derived mesenchyme covered by epithelium. Outgrowth of the primordia occurs through controlled proliferation of mesenchymal cells, which depends in part on signalling from the overlying epithelium (Wedden, 1987; Richman and Tickle, 1992). As the facial prominences grow out, the nasal processes fuse in the midline with each other and laterally with the maxillary processes. Frequently occurring facial cleft disorders, as known from human pathology and several mouse mutant models, are in many cases thought to result from these fusion events going awry (Thorogood, 1997).

Morphogenesis of neural crest-derived structures depends on the integrated action of several types of patterning genes (Schilling, 1997). Genes responsible for early anteroposterior patterning of the central nervous system, e.g. Hox and Otx genes, are to some extent responsible for the pre-programming of neural crest cells as they emigrate from the neural tube. Many other genes, including members of different homeobox gene families, exert their function in the mesenchymal cells that populate the facial processes. Other genes encode signalling factors, including members of the Fgf and Wnt families and sonic hedgehog (Shh) are locally expressed in the epithelia that cover the processes or in adjacent endoderm. Crosstalk between these two types of gene is likely to underlie in many cases the reciprocal epitheliomesenchymal interactions known to be crucial in the development of the craniofacial skeleton (Ferguson et al., 2000). Aristaless-like homeobox genes form a distinct gene family whose members are characterised by a paired type homeobox and the presence of a small conserved C-terminal domain in the proteins encoded, known as aristaless or OAR domain; see review by Meijlink (Meijlink et al., 1999). A subset of these genes, including Prx1 and Prx2, Alx3, Alx4, and Cart1, are expressed during embryogenesis in similar patterns in neural crest-derived mesenchyme of developing craniofacial regions and in the mesenchyme of developing limbs (Leussink et al., 1995; Zhao et al., 1994; Qu et al., 1997; Ten Berge et al., 1998a). Mutant studies have established their partially redundant functions in these embryonic regions. For example, studies of Prx1/Prx2 double mutant mice demonstrated the roles of these genes in craniofacial and appendage development. These mutants have severe reductions of mandibular arch derived structures, and in the zeugopod of the limbs (Ten Berge et al., 1998b; Lu et al., 1999a; Lu et al., 1999b). On the basis of sequence similarity and embryonic expression patterns, Alx3, Alx4 and Cart1 appear to represent a distinct subgroup (Ten Berge, 1998a). Mice with mutations in Alx4, including the Strong’s Luxoid (lst) alleles, have strong preaxial polydactyly, mild craniofacial abnormalities in anterior components of the skull base and parietal and frontal bones, as well as gastroschesis (Forsthoefel, 1963; Qu et al., 1997). Cart1 mutant mice have major cranial defects including acrania and meroanencephaly (Zhao et al., 1996).

This study describes the generation and analysis of Alx3 null mutant mice. Alx3 single mutants appear to be normal, but Alx3/Alx4 double mutant mice have severe craniofacial malformations that are never seen in Alx4 single mutants.

Double mutants are born with midfacial clefting and many neural crest-derived skeletal elements are severely malformed. The anatomical abnormalities leading to the cleft nose phenotype are first detectable at E10.5 in medial craniofacial regions where Alx3 and Alx4 are strongly co-expressed. Although in wild-type embryos the medial nasal processes fuse at E11.5, in Alx3/Alx4 double mutant embryos they develop in an abnormal lateral position, resulting in their failure to fuse. These defects correlate with abnormal and localised apoptosis at E10.0 in the nasal processes.

A fusion construct was made in pUC19 containing 3 kb lacZ coding sequence and a 3 kb loxP-pGK-TKNEOpA-loxP cassette that was transcribed in the reverse direction and flanked by loxP sites. A pUC19 vector containing pSDK-lacZ (gift of J. Rossant) was digested with KpnI, blunted and ligated to NotI adaptors. Subsequently, the linearised fragment was digested with SmaI. The loxP-pGK-TKNEOpA-loxP cassette was isolated by digestion with NotI and SmaI of pHR56 (a gift from P. Kastner and F. Rijli) and was cloned into the linearised pUC19 vector containing pSDK-lacZ. In pBR-GEM11 (a gift from H. te Riele). The SalI site in the polylinker was destroyed by partial SalI digestion, filling in of the overhangs and re-ligation, generating pBR-GEM11ΔSalI. Genomic fragments containing part of the Alx3 gene (see Fig. 1) were isolated by screening of a 129 SVJ mouse genomic library cloned in phage lambda FIX® II vectors (Stratagene) with part of the Alx3 cDNA. A 1.9 kb genomic Alx3 PstI-NcoI fragment containing exon 2 was cloned in the pSDK-lacZ-loxP-pGK-TKNEOpA-loxP fusion construct, resulting in an in frame fusion of Alx3 (exon 2) and lacZ sequences. The 7.4 kb NotI-SalI fragment containing the Alx3 (exon2)-lacZ-loxP-pGK-TKNEOpA-loxP cassette was ligated into the NotI and XhoI site of pBR-GEM11ΔSalI. Finally, a 5.5 kb genomic SalI fragment containing sequences 3′ of Alx3 exon 4 was inserted into the XhoI site of Alx3 (exon 2)-lacZ-loxP-pGK-TKNEO-pA-loxP-pBR-GEM11ΔSalI, generating the complete targeting vector, pAlx3^lacZ, containing 1.9 kb 5′ and 5.5 kb 3′ homologous sequence to the Alx3 locus (see Fig. 1).

Strong’s luxoid-J mice (official strain symbol Alx4^lst-J) were obtained from the Jackson Laboratory (Bar Harbor). These mice carry an Alx4 allele containing a 16 bp deletion in the homeobox (Qu et al., 1998; Takahashi et al., 1998).

A subclone from the E14 embryonic stem (ES) cell line was obtained from C. Brouwer, P. Krimpenfort and A. J. M. Berns (The Netherlands Cancer Institute, Amsterdam). 100 μg SalI linearised pAlx3^lacZ was electroporated into 4×10⁷ ES cells. The ES cells were grown on irradiated mouse embryonic fibroblasts (MEFs) in 10 ml GMEM containing 1000 units/ml leukaemia inhibiting factor (LIF) in a 5% CO₂ humidified incubator. From 2 days after electroporation, ES cells were selected in medium supplemented with 60% BRL-conditioned medium containing G418 (200 μg/ml) and LIF (1000 units/ml) without feeder cells to isolate 400 individual neomycin resistant clones. Targeted ES cell clones were identified for correct 3′ integration by digestion of their genomic DNA with StuI. Southern analysis was performed using a unique NcoI-XhoI fragment of the 3′ Alx3 locus. The NcoI-XhoI probe binds to sequences of a 9.5 kb fragment of the wild-type locus and to a 7.5 kb fragment of the mutant Alx3 locus. Positive ES cell clones were subsequently checked for correct 5′ integration by digestion of their genomic DNA with EcoRI and Southern analysis using a unique SstI-StuI fragment of the 5′ Alx3 locus. The SstI-StuI probe binds to sequences of a 9.5 kb fragment of the wild-type locus and to an 11.5 kb fragment of the mutant Alx3 locus (see Fig. 1).

Chimaeric mice were obtained by injection of ES cells into C57BL/6 blastocysts. Chimaeric males were crossed with FVB/N females and heterozygous offspring was identified by coat colour and Southern analysis of tail DNA. Mice carrying the mutant Alx3 allele were crossed to homozygosity. The lst^J mutant was obtained from the Jackson laboratory. This mutant has been shown to have a 16 bp deletion in the homeobox of the Alx4 gene, resulting in loss of function (Qu et al., 1998). Alx3/Alx4 double mutant mice were bred on a mixed genetic background with contributions of OLA/129, FVB/N, C57BL/6J and B6C3H.

All animal experiments were conducted under the approval of the animal care committee of the KNAW (Royal Dutch Academy of Arts and Sciences).

Genotyping was carried out using PCR on tail tip or yolk sac DNA. Amplification of 326 bp of the Alx3 wild-type allele was performed by using: 5′-gaggctcaagaacaaggaagga-3′ (forward primer) and 5′-ctaggagcaggtcagagcaggaag-3′ (reverse primer). Amplification of 821 bp of the lacZ gene by using 5′-tcgagctgggtaataagcgttggcaat-3′ (forward primer) and 5′-agaccaactggtaatggtagcgac-3′ (reverse primer) was carried out to identify the mutant Alx3 locus. Either 340 bp of the wild-type Alx4 locus or 324 bp of the mutant Alx4 locus was amplified using: 5′-caagccccctggaaaagacc-3′ (forward primer) and 5′-ggtacattgagttgtgctgtcc-3′ (reverse primer) (Qu et al., 1998). To distinguish these products, they were restricted with PvuII and the products were separated on a 3% agarose gel. The wild-type locus produces a 195 bp product and the mutant locus a product of 179 bp.

Bone and cartilage staining was carried out according to a modified procedure of C. Fromental-Ramain (IGBMC, Strasbourg). Foetuses, and skinned and eviscerated new-born animals were fixed overnight in 96% ethanol, stained overnight for cartilage in 80% ethanol, 20% acetic acid + 0.5 mg/ml Alcian Blue (Sigma), rinsed twice in 96% ethanol for 3 hours and digested in 1.5% KOH for 2 hours, followed by bone staining in 0.5% KOH + 0.15 mg/ml Alizarin Red S (Sigma) overnight, destained in several changes of 20% glycerol + 1% KOH, and stored in 20% glycerol, 20% ethanol. Whole-mount X-Gal staining of E10.5 and E11.5 embryos was as described (Hogan et al., 1994). Whole-mount in situ hybridisation was performed as described (Leussink et al., 1995). Shh, Ptc and Fgf8 probes were kind gifts of C. Tabin (Boston, MA), R. Zeller (Utrecht) and G. R. Martin (San Francisco, CA), respectively. Sections hybridised with radioactive probes were photographed by double exposure of dark-field and bright-field images; a red filter was used during the dark-field exposure.

Cell proliferation was detected via analysis of BrdU incorporation into the DNA of dividing cells. Pregnant mice were treated with 1 ml 10 mM BrdU in PBS per 100 g body weight, and embryos were isolated 75 minutes later. Embryos were fixed overnight in 4% paraformaldehyde, dehydrated, embedded in paraffin and 4 μm thick sections were prepared according to standard techniques. Sections of each embryo were distributed over three parallel series of object glasses.

BrdU was detected using a fluorescein-labelled monoclonal anti-BrdU antibody (Roche) according to the instructions of the manufacturer. The signal was enhanced using the Alexa Fluor 488 signal amplification kit for fluorescein conjugated probes (Molecular Probes). Sections were counterstained with the fluorophore TO-PRO-3 (Molecular Probes), which stains nucleic acids. Each labelled section was scanned simultaneously for BrdU and TO-PRO-3 signals in two different channels using a Leica TCS NT confocal laser scanning microscope.

Apoptosis was detected by TUNEL assay using the In Situ Cell Death Detection KIT (AP) (Roche) according to the instructions of the manufacturer with some adaptations. The 4 μm sections were rehydrated and in contrast to the manufacturer’s instructions sections were not treated with proteinase K. DNA was extended by terminal-deoxynucleotidyl transferase in the presence of fluorescein-labelled dUTP and incorporated fluorescein-labelled dUTP was subsequently detected with an anti-fluorescein antibody conjugated with alkaline phosphatase (AP). After substrate reaction with Fast Red apoptotic cells were detected using a light microscope.

Alx3 mutant mice were generated through homologous recombination in ES cells. The targeting strategy is shown in Fig. 1. A lacZ sequence linked to a pGK-TKNEO cassette was inserted in frame in a site 28 bp upstream of the homeobox. Upon recombination the mutant allele was expected to express a fusion protein consisting of the first 132 amino acids of the Alx3 protein and the β-galactosidase protein in a pattern identical to the Alx3 expression pattern. LoxP sites flanked the selection cassette allowing its subsequent deletion from the allele. The mutant Alx3 allele is expected to be functionally inactive as the protein encoded lacks both its conserved domains. Southern blotting using 3′ and 5′ probes confirmed successful targeting in at least 12 out of the 400 isolated ES cell clones (not shown). Three of these ES cell clones contributed to the germline of chimaeric mice generated by blastocyst injection. Heterozygous mice for the Alx3 mutation appeared normal and were fertile. Two mouse lines were intercrossed and yielded wild-type, heterozygous and homozygous mutant offspring at the expected Mendelian frequencies. As the promoter region of the Alx3 locus including the large first intron remains intact in the knock-in allele, no major differences are expected in the lacZ expression, when compared with Alx3 mRNA expression. Analysis of the β-galactosidase expression pattern in heterozygous embryos of E9.5 to E12.5 showed essentially no differences with the endogenous Alx3 expression pattern, confirming correct targeting of the gene (compare Fig. 2A with Fig. 2B for E10.5 embryos).

Alx3 heterozygous and null mutant mice were born at expected frequencies and did not display noticeable abnormalities. They were fertile and appeared to have a normal lifespan. As Alx3 is expressed predominantly in mesenchyme of developing limbs and craniofacial regions (Ten Berge et al., 1998a; Beverdam and Meijlink, 2001) (Fig. 2) skeletons of null mutant newborn mice were thoroughly analysed, but no abnormalities were observed.

Alx3 and Alx4 encode highly related proteins (Qu et al., 1997; Ten Berge et al., 1998a); in addition, their expression patterns are highly similar, as we demonstrate in Fig. 2C-J (Beverdam and Meijlink, 2001). In facial primordia at early embryonic stages (Fig. 2C-H) there is considerable overlap of the expression domains, while the Alx4 expression domain extends more laterally. In limb buds expression of both genes is almost identical (Fig. 2F,J). It is therefore very likely that Alx3 and Alx4 have overlapping functions, and for that reason we decided to generate Alx3/Alx4 double mutant embryos. We crossed Alx3 mutants with Strong’s luxoid Jackson (Alx4^lst-J) mice, that are known to carry a defective Alx4 allele (Qu et al., 1998; Takahashi et al., 1998). Alx4 mutants have preaxial polydactyly, abnormalities in the skull and frequently gastroschesis (Forsthoefel, 1963; Qu et al., 1997).

Alx4 single and Alx3/Alx4^lst-J double mutant mice had preaxial polydactyly of the fore- and hindlimbs as described for Alx4 mutant mice (see Fig. 3) (Forsthoefel, 1963; Qu et al., 1997). Penetrance and severity of the Alx4 phenotype are known to depend greatly on genetic background (Forsthoefel, 1968). In contrast to Alx4/Cart1 double mutant mice (Qu et al., 1999), Alx3/Alx4^lst-J double mutant had no significant aggravation of the polydactyly. We noted that both Alx4 mutants and double mutants lacked the deltoid crests (arrows in Fig. 3A-C). These processes are located anteriorly on the humeri of normal forelimbs, and their absence may therefore be interpreted as a proximal expression of a mirror-image-duplication phenotype in the limb. This phenotype has not previously been described for any Alx4 mutant, but has been found as an aspect of the extensive mirror image duplication that can be caused by Hoxb8 overexpression (Charité et al., 1995). Alx3/Alx4^lst-J double mutant mice had a reduction of the medial (sternal) end of the clavicle, a phenotype that is absent from the Alx4^lst-J mutant (see arrowheads in Fig. 3A-F). Therefore in the shoulder girdle, but not in the limb itself, nor in the pelvic girdle, a function of Alx3 was demonstrable.

Although in Alx3^+/–/Alx4^lst-J/+ mice no craniofacial abnormalities were detected and Alx3^+/+/Alx4^{lst–J/ lst-J} mutant mice had only relatively mild craniofacial abnormalities, Alx3^–/–/”Alx4^{lst–J/lst-J}, Alx3^–/–/Alx4^lst-J/+ and Alx3^+/–/Alx4^{lst–J/lst-J} mice had a broader and shorter cranium and displayed variable facial clefting comprising the entire nose region (see Fig. 4). Generally, they died within a few hours after birth with empty stomachs and air in their intestines, apparently as a consequence of craniofacial defects leading to inability to breath or owing to the ventral body wall defects as described for the Alx4 mutant mouse (Qu et al., 1997). Eye lid abnormalities, as described by Forsthoefel (Forsthoefel , 1963) for Strong’s luxoid (Alx4) mutants, seemed to be enhanced in double mutants. In Alx3^–/–/Alx4^{lst–J/lst-J} newborn mice, the eyes were more often wide open (compare Fig. 4D with Fig. 4A-C). This aspect of the phenotype is apparently an indirect consequence of other cranial defects that make the eyes bulge out (Forsthoefel, 1963).

We performed a detailed analysis of skulls of newborns with various double mutant genotypes using skeletal staining (Fig. 5, Fig. 6) and histological sections (Fig. 7 for E12.5 and E13.5; see also Fig. 9 for earlier stages). A gradual increase in the severity of the phenotype was found, being weakest in the Alx3^+/+/Alx4^lst-J/lst-J mutant mice and becoming progressively stronger in Alx3^–/–/Alx4^lst-J/+, Alx3^+/–/Alx4^lst-J/lst-J and Alx3^–/–/”Alx4^lst-J/lst-J mutant mice, as demonstrated in Fig. 5. Double mutants displaying a mild phenotype had only a partially split nasal tip, but in mutants with extreme phenotypes, both lateral halves of the nose were anteriorly truncated and spaced wide apart. Most facial bones and many of the membranous bones of the vaults were affected in double mutants (compare in Fig. 6A,E with Fig. 6C,G). In homozygous Alx4 mutant mice the parietal and frontal bones are reduced (see Fig. 5) (Forsthoefel, 1963) and recent studies reveal that in humans, ALX4 haploinsufficiency is associated with ossification defects in the parietal bones (Wuyts et al, 2000; Wu et al., 2000; Mavrogiannis et al., 2001). We observed in Alx3/Alx4^lst-J double mutant newborn skulls an aggravation of the vault phenotype: besides a stronger reduction in the size of the frontal and parietal bones, the nasal bones were also affected, causing wide fontanels (compare Fig. 6A,B with Fig. 6C,D). The nasal capsule was divided in two separate halves both closed by remnants of the nasal septum while a medial nasal septum was absent (see white arrows in Fig. 5; Fig. 6C,G). The nasal labyrinths and the anterior part of the nasal capsule were severely malformed and strongly curved (Fig. 5, lower panels; Fig. 7, compare A,E with B,F for the situation at E12.5 and E13.5). Also the premaxilla and maxilla were strongly affected and positioned abnormally laterally, and the palatine was cleft (Fig. 6, ‘P’ in F,H). In the mandible the distal part of the dentaries was truncated, but the incisors were normally present (see Fig. 5). The proximal part of the mandible seemed unaffected. The more lateral skull bones consisting of the alisphenoid bones and squamosal bones were malshaped and reduced (see structures marked ‘As’ and ‘S’ in Fig. 6F,H). The neural crest-derived skull base, consisting of the basipresphenoid and the pterygoid processes was also severely malformed and appeared broader (see green arrowheads in Fig. 5 and asterisks in Fig. 6E,G). More posterior skull structures derived from cephalic and somitic mesoderm like the posterior elements of the sphenoid bone, the occipital bone and the otic capsule, tympanic ring, inner and middle ear structures were normal in double mutants (see Fig. 5, Fig. 6, Fig. 7). This region of the skull appears somewhat broader, which may be explained as a consequence of the anterior truncation of the skull that forces the more posterior skeletal elements outwards, in order to allow the brain to expand (Forsthoefel, 1963). Histological sections of E12.5 and E13.5 embryos demonstrate for these stages the normal morphology in posterior cranial structures, for example, the palatal processes and tongue in Fig. 7A-D), while more anteriorly the nasal cavities are malformed (arrows in Fig. 7A,B,E,F). Fig. 7G,H shows the cleft ‘nasal septum’.

The skeletal elements derived from second and third arch mesenchyme, which form the hyoid skeleton were unaffected (not shown).

To define when the craniofacial defects in Alx3/Alx4^lst-J double mutants first become apparent, control embryos and Alx3^–/–/Alx4^lst-J/+, Alx3^+/–/Alx4^lst-J/lst-J and Alx3^–/–/Alx4^lst-J/lst-J embryos were analysed from gestational day 9.5 to 15.5 (see Fig. 8).

At E9.5, craniofacial regions of wild-type embryos are composed of the primitive facial primordia that are arranged around the stomodaeum. The facial primordia consist of the frontonasal process and the protrusions of the first branchial arch. At E10.0 the nasal placodes become slightly indented, causing the formation of the medial and lateral nasal processes. Analysis of E9.5 and E10.0 Alx3^–/–/Alx4^–/– whole-mount and sectioned embryos revealed no abnormalities in the first branchial arch or frontonasal processes (not shown). At E10.5 in normal embryos, the lateral and medial nasal processes enclose the nasal pits; the medial nasal processes will fuse in the midline around E11.5 forming the nasal septum. The primary palate is formed by fusion of the inferior borders of the medial nasal processes. From the medial walls of the maxillary processes palatal shelves grow out that fuse in the midline with the inferior border of the nasal septum, which grows down from the frontal process. Anteriorly, the shelves fuse with the primary palate resulting in the formation of the secondary palate.

In E10.5 Alx3^–/–/Alx4^lst-J/lst-J embryos, the nasal processes are spaced more laterally than in wild-type embryos (compare arrowheads in Fig. 8A and Fig. 8E). The abnormally wide spacing became more pronounced later during development and resulted in defective merging of the nasal processes in the midline (compare arrowheads in Fig. 8B-D with Fig. 8F-H). Moreover, fusion of the maxillary processes with the lateral nasal processes occurred more laterally than in wild-type littermates, causing an abnormal shape and position of skeletal elements of the upper jaw. Fig. 7C,D,G,H (arrowheads) shows the presence of the nasal labyrinth, nasal septum remnants and vomeronasal organs within the unfused nasal processes in sections of E12.5 and E13.5 embryos. Instead of one medial mesenchymal condensation forming the cartilaginous nasal septum, two lateral condensations were detected, leaving an open space in the midline, which was continuous with the oral cavity owing to the cleft palate (Fig. 7E-H). In Fig. 8D,H, lacZ expression in E11.5 control embryos is shown to illustrate Alx3 expression in the medial nasal processes; this domain becomes much wider in the double mutant. It thus appears that the defects observed in the Alx3/Alx4^lst-J double mutant mice were directly or indirectly caused by the abnormal lateral position of the nasal processes in E10.5 embryos.

As a next step to understand the basis of the Alx phenotype, we compared the expression in normal versus double mutant embryos of a number of selected genes in whole-mount or sectioned embryos of E9.5 to E11.5.

Shh is expressed in a distinct region of the ectoderm covering the medial nasal processes from E9.5 onwards (see arrow in Fig. 9A). It is known to be upregulated in anterior limb bud mesenchyme of Alx4 mutants (Chan et al., 1995), and downregulated in mandibular epithelium of Prx1/Prx2 double mutants (Ten Berge et al., 2001). In double mutants at E10.5, but not at earlier stages, the expression domain in nasal epithelium is expanded to the same degree as the skull is wider (compare Fig. 9A with Fig. 9B,C). The expression pattern of the direct Shh target gene Ptc was changed accordingly (not shown). As the change of the expression pattern seems to follow the anatomical abnormalities, rather than vice versa, we do not believe that Shh and Ptc are downstream target genes of Alx3 and Alx4, but rather that the changes in expression domains are secondary to the morphological changes. Fgf8 is expressed in restricted regions in the oral ectoderm of the first branchial arch, in the ectoderm of the nasal pits and more distally in the medial nasal process ectoderm (see Fig. 9E). In Alx3/Alx4^lst-J double mutant embryos of E9.5 to E11.5, we could not detect obvious changes in the Fgf8 expression pattern (see Fig. 9E,F).

Pax9, Prx3 and Bmp2 have distinct expression patterns in mesenchyme of developing craniofacial primordia. We compared their expression domains in control embryos of E10.5 and E11.5 with those in Alx3/Alx4^lst-J double mutant embryos. Fig. 9G-L shows the results for Pax9, known to have an important role in patterning of mesoderm and endoderm of the pharynx, and which is expressed in nasal, mandibular and maxillary processes. At E10.5 the expression patterns in control versus wild-type embryos are indistinguishable, while at E11.5 the changes observed appear to be a consequence of the anatomical abnormalities that are clearly visible around this stage. For example, Pax9 is expressed laterally from the precartilage condensation in the midline of the nasal septum around the vomeronasal organs (Fig. 9H) and around the nasal cavities (Fig. 9I). In the highly abnormal nasal septum remnants the domain is split apart but essential characteristics, like its location with respect to the epithelium and surrounding the vomeronasal organs and nasal cavities appear to be retained (Fig. 9K,L). Similarly, Bmp2 (not shown) and Prx3 (Fig. 8M-R), a family member of the Alx genes also known as Shot or OG-12, which we selected because of the overlap of its expression domain with those of Alx3 and Alx4 in the medial nasal process, were normally expressed at E10.5, and at E11.5 in a way more or less predictable from the phenotype. This marker gene expression study therefore is informative because it contributes to the histological analysis of the mutant embryos, but does not provide clues as to the molecular mechanisms that underlie the phenotype.

To investigate the cellular basis of the facial phenotype of the Alx3/”Alx4^lst-J double mutant mice we performed BrdU incorporation assays to analyse proliferation patterns and TUNEL assays to detect apoptosis in embryos of around the developmental stages when the abnormalities first become obvious.

We studied cell proliferation in control and double mutant embryos of E9.5, E10.0 and E10.5. In both control and double mutant embryos, BrdU incorporation appeared highest in the mesenchymal cells flanking the deepening nasal placode and became lower towards more medial regions of the frontonasal process. We were not able to detect significant differences in proliferation rates between double mutant and control embryos (not shown).

We next studied apoptosis in transversal tissue sections of E9.5 and E10.0 control and double mutant embryos by performing TUNEL assays. In E9.5 embryos we did not observe any difference in cell death in control and double mutant embryos (not shown). In two double mutant embryos of E10.0, however, we observed two contralateral areas containing abnormally high numbers of apoptotic cells dorsolaterally in frontonasal mesenchyme adjacent to the placodes (Fig. 10A-G). We counted 30 adjacent 4 μm sections in which the number of apoptotic cells was significantly higher compared with corresponding regions in five control embryos (compare Fig. 10B,E with Fig. 10A,D). The part of the frontonasal process where the apoptotic cells were observed falls within the region where Alx3 and Alx4 are both highly expressed (see Fig. 2C-H). This region contributes to the development of the nasal processes that will ultimately give rise to the nasal capsule, premaxilla and primary palate. In E10.5 embryos we detected essentially the same patterns of apoptotic cells in mutants and wild types. In the region presumably corresponding to the area where we saw the enhanced apoptosis in E10.0 embryos, we saw the same very low levels of TUNEL-positive cells in wild types and mutants (Fig. 10H,I). In E11.5 embryos, we did not see any TUNEL-positive cells in nasal mesenchyme (Fig. 10K,L).

We conclude that the lack of Alx3 and Alx4 early during craniofacial patterning leads to loss of cell survival in a confined area during a small time window. We deem it possible that this seemingly minor effect suffices to subsequently lead to abnormal outgrowth of the medial nasal processes. We propose that this early developmental anomaly contributes significantly to the rather dramatic phenotype seen in newborns.

We report that mice that lack the aristaless-related homeobox gene Alx3 do not display anatomical abnormalities or diminished health. This suggested to us that loss of Alx3 is compensated by genes with overlapping functions, which was confirmed when we analysed the phenotype of mice double mutant for Alx3 and Alx4. These genes are structurally highly related and are expressed in similar expression patterns in developing embryos. Alx4 loss-of-function mutants, including Strong’s luxoid mutants and Alx4-targeted knockouts, have a complex phenotype including preaxial polydactyly, and mild abnormalities in the skull (Forsthoefel, 1963; Qu et al., 1998; Qu et al., 1999). Mice double mutant for Alx3 and Alx4^lst-J null alleles had severe abnormalities in the skull that are not seen in Alx4 mutants, although other aspects of the Alx4 phenotype were at the most marginally aggravated. Alx3/Alx4^lst-J double mutants have severe clefting of the nose region. We noted the earliest signs of anatomical abnormalities in the head region around gestational day 10.5, when the nasal processes grew out abnormally laterally. The wide lateral spacing of the processes resulted in failure to fuse, and ultimately in the cleft nose phenotype in newborn mice. Analyses of cellular parameters in early double mutant embryos demonstrated increased cell death in restricted regions of the presumptive nasal processes, which is most likely causally linked to the subsequent abnormal morphogenesis of the nasal processes.

Alx3 and Alx4 are expressed in overlapping domains in neural crest-derived mesenchyme of medial regions of the developing facial primordia (Qu et al., 1997; Ten Berge et al., 1998). These regions grow out to form the cartilages, bones and connective tissues of the face and the most anterior part of the skull base.

Cart1 is a third protein that is highly related to Alx3 and Alx4. These three proteins share very similar homeodomains, a common C-terminal domain called aristaless- or OAR domain, and additional sequence similarity between these two domains that sets them apart at the structural level from other aristaless-related proteins. Cart1 expression in craniofacial regions is spatially and temporally similar to that of Alx3 and Alx4, but Cart1 expression is detectable earlier in the head mesenchyme. Cart1 mutant mice have a strong cranial phenotype caused by cranial neural tube closure defects that lead to acrania and meroanencephaly at the time of birth (Zhao et al., 1996). In addition, these mice have other, relatively mild craniofacial defects. It is striking that, like Alx3/Alx4^lst-J mutants, Alx4/Cart1 double mutant mice have severe midfacial clefting and distally truncated mandibles (Qu et al., 1999). In contrast to our conclusions on the early anatomical basis of the cleft-nose phenotype of the Alx3/Alx4^lst-J mutants, the phenotype of the Alx4/Cart1 double mutant mice has been explained as a failure of the ‘nasal cartilages to fuse’ (Qu et al., 1999). In view of the similar molecular basis of the defect in both cases, we consider it quite likely, however, that in both double mutants the anatomical basis of the phenotype is the same. We have shown that the critical stage in the aetiology of the phenotype occurs before cartilage differentiation and even mesenchyme condensation. Unfortunately, no embryonic stages of Alx4/Cart1 double mutants have been described (Qu et al., 1999).

Comparison of the Alx4^lst-J single mutants and Alx3/4^lst-J double mutants showed a strong exacerbation of the phenotype in craniofacial, but not in other regions. The striking Alx4 expression pattern anteriorly in limb bud mesenchyme is also found for Alx3, so it is unexpected that in Alx3/Alx4^lst-J double mutants no enhancement of the phenotype is seen. Apparently, in Alx4 mutants, Alx3 can not compensate to a significant extent the function of Alx4. Strikingly, mutation of Cart1 has been shown to clearly exacerbate the polydactyly of Alx4 mutants (Qu et al., 1999). We have recently shown that Cart1 is expressed at low level in anterior limb bud mesenchyme at E9.5-E10.5 (Beverdam and Meijlink, 2001). It seems contradictory that low expression of Cart1 has more impact than that of Alx3; this paradox may be solved by taking into account possible specific functional properties of heterodimers probably formed by Alx3, Cart1 and Alx4 with each other and with other paired-related homeodomains (Qu et al., 1999; Tucker and Wisdom, 1999). It remains possible that in yet to be made Alx3/Alx4/Cart1 triple mutants, preaxial polydactyly is further aggravated. The expression patterns of Alx3, Alx4 and Cart1 do not straightforwardly suggest an explanation for Alx3 being apparently more important in the cranium than in the limbs. This explanation should perhaps be sought for in differences in interactions with other genes in these areas.

The only defect outside the skull exclusively found in the double mutant was the truncation of the clavicle. We previously noted Alx3 expression in the developing clavicle (shown in Ten Berge et al. (Ten Barge et al., 1998a); A. B., unpublished). The embryological origin of the clavicle is obscure, but recent evidence suggests that there is some neural crest contribution to the clavicle, in particular the site of the attachment point for the cleidohyoid muscle (McGonnell et al., 2001). This suggests that the lost piece of the clavicle is at least partly neural crest derived. It is striking that the function of Alx3 is almost exclusively manifest in the craniofacial areas where it appears to control morphogenesis of neural crest-derived skeletal elements, and that this exception concerns a structure also linked to neural crest-derived mesenchyme. It should be noted that the clavicle, the only remaining dermal bone of the mouse pectoral girdle, evolved originally as part of the head (McGonnell et al., 2001; Torrey, 1978)

As Alx3 is highly expressed in the body wall, gastroschesis was expected to increase in double mutants, but no significant differences in penetrance or severity were observed. It should be noted that a quantitatively accurate comparison of phenotypes is hampered by the heterogeneous genetic background of the mice used for the analysis and the sensitivity of Alx4 mutations for genetic background effects (Forsthoefel, 1968).

The data discussed here indicate that Alx3, Alx4 and Cart1 constitute a distinct subgroup of genes sharing identical or very similar roles in the patterning of the facial primordia, in particular the nasal process. A second subgroup of aristaless-like genes consists of Prx1 and Prx2, which share to some extent the characteristics of structure and expression of the Alx3/4 and Cart1 genes, but have functions that are primarily vital to correct patterning of the mandibular arch and consequently the formation of the jaws.

We determined that the abnormalities in Alx3/Alx4^lst-J double mutant embryos first become obvious around E10.5 in development, when the nasal processes become prominent structures. In double mutant embryos of E10.5, the nasal processes develop too wide apart and therefore fail to fuse at around E11.5, eventually causing a wide midline cleft by the time of birth. The crucial process of fusion of the facial primordial processes depends heavily on their proper outgrowth and morphogenesis. A lack of critical mass, or their abnormal position is likely to jeopardise the fusion process (Thorogood, 1997).

Within the nasal processes, most structures, including the nasal cavities, continue to develop in a relatively normal way, suggesting that most morphogenetic processes that regulate proliferation and differentiation of cells in the nose region do not depend on Alx3 or Alx4. To obtain clues at a molecular level on the nature of the defects, and in the hope of identifying target genes, we studied the expression pattern of several genes known to have distinct expression patterns in the facial regions, generally with negative results. Shh, which has been implicated in the outgrowth of facial processes (Hu and Helms, 1999), is expressed in normal embryos in a restricted pattern in the facial ectoderm covering the frontonasal and the medial nasal processes, and Ptc, a direct downstream Shh target, is expressed in ectoderm as well as mesenchyme that underlies the Shh-expressing ectoderm. We found that in double mutant embryos of E10.5, but not at earlier stages, the region of ectodermal cells expressing Shh had expanded and Ptc expression in ectoderm and mesenchyme had changed accordingly. As the aberrant expression seems to occur only after the emergence of anatomical abnormalities, it appears that this expansion is a result rather than a cause of the malformations, and there is no reason to suggest that Shh is a specific target of Alx genes. Conceivably, the ectopic Shh expression from E10.5 onwards contributes to the aetiology of the phenotype.

To understand at a cellular level the basis of the aberrant anatomy of E10.5 double mutants, we analysed cellular processes underlying growth of the nasal prominences by measuring proliferation and cell death patterns in embryos of these and earlier stages. We could not detect significant differences in proliferation, indicating that the defects are not caused by proliferation defects. We cannot, however, exclude the possibility that changes in proliferation too subtle to be detected in our assays do account for significant morphological defects.

In double mutant embryos at E10.0, but not at earlier or later stages, we did detect apoptosis in a very restricted region in the mesenchyme of the outgrowing frontonasal process where in control embryos of the same developmental stage no or very little apoptosis was detected. Apoptosis in double mutants was restricted to a region located dorsally within the process where Alx3 and Alx4 are both expressed. Lineage studies performed by McGonnell et al. (McGonnell et al., 1998) suggest that this region will contribute to the medial nasal processes. It is conceivable that local cell death in an outgrowing process leads to its subsequent malpositioning and therefore that the aberrant apoptosis seen in the Alx3/4 mutant is likely to be in part responsible for the later defects. It remains currently unclear which pathway links the expression of these transcription factors to the process of apoptosis. A causal link between aristaless-related genes and regulated cell death is not without precedent. In Cart1 mutants, 70% of E8.75 embryos have an anterior neural tube closure defect that appears to result from increased cell death in the mesenchyme surrounding the forebrain vesicles (Zhao et al., 1996). These results suggest that the Alx/Cart subgroup regulates craniofacial patterning by mechanisms that involve control of cell survival.

Our understanding of the molecular basis of craniofacial patterning is rapidly expanding. From gastrulation onwards Hox genes pattern the embryo along the anteroposterior axis from posterior up to the hindbrain regions. Formation and regionalisation of more rostral head structures, on the other hand, is dependent on genes like Otx1 and Otx2 and Emx1 and Emx2, which are expressed in nested patterns in mid- and forebrain before and during cephalic neural crest cell migration (Kuratani et al., 1997; Cecchi and Boncinelli, 2000). These genes may to some extent be responsible for determining genetic identity and preprogramming of these neural crest cells, which ultimately will populate the branchial arches and the frontonasal process. In a genetic hierarchy that appears to regulate craniofacial patterning, downstream from these neural genes, other gene families, including those related to Distal-less and aristaless, have local functions in patterning neural crest-derived mesenchyme. It appears that the two subgroups of aristaless-like genes that we distinguish exercise their function on different but slightly overlapping complementary anteroposterior levels. The most important defects in Prx mutants are in the derivatives of the distal mandibular arch that is populated by neural crest derived from the anterior hindbrain (Ten Berge et al., 1998; Lu et al., 1999a). By contrast, Alx3, Alx4 and Cart1 control patterning of the medial regions of the anterior skull base and of the nose and upper jaw regions (Qu et al., 1999) (this paper); these structures derive from midbrain neural crest. In agreement with this, Otx2 heterozygous mutants have defects in similar regions of the anterior skull base and distal jaws (Kuratani et al., 1997). In future research, we will test our hypothesis that Alx and Prx genes are directly or indirectly regulated by Otx2 and a rhombomere 1- and rhombomere 2-specific gene like Gbx2, respectively (Wassarman et al., 1997).

We thank Jacqueline Deschamps, Christine Mummery, Ronald Plasterk, Rolf Zeller and the anonymous referees for constructive criticism. We acknowledge Fons Verbeek and Marjo den Broeder for kindly helping preparing the diagrams of Fig. 6 and Willem Hage for help with analysis of proliferation assays. We thank Anthony Graham for communicating information before publication. We also thank Carla Kroon for technical assistance, Ferdinand Vervoordeldonk and Jaap Heinen for help in preparing the illustrations, and the animal care unit for their contributions.