Mice with targeted disruption of Hoxb-1 fail to form the motor nucleus of the VIIth nerve

Mice and humans contain at least 39 Hox genes distributed on four linkage groups, designated Hox A, B, C and D, on four separate chromosomes. Based on DNA sequence similarities and on the positions of the genes on their respective chromosomes, individual members of the four linkage groups have been classified into 13 paralogous families. These genes encode transcription factors belonging to the Antennapedia homeodomain class. The homologous genes in Drosophila are used to pattern the developing embryo (Lewis, 1978). Mutational analysis in the mouse has demonstrated that these genes, alone or in concert with other Hox genes, are used to regionalize the embryo along its major axes (Chisaka and Capecchi, 1991; Lufkin et al., 1991; Chisaka et al., 1992; LeMouellic et al., 1992; Dollé et al., 1993; Gendron-Maguire et al., 1993; Jeannotte et al., 1993; Ramirez-Solis et al., 1993; Rijli et al., 1993; Small and Potter, 1993; Davis and Capecchi, 1994; Kostic and Capecchi, 1994; Davis et al., 1995; Horan et al., 1995; Manley and Capecchi, 1995; Rancourt et al., 1995; Satokata et al., 1995; Suemori et al., 1995; Boulet and Capecchi, 1996; Davis and Capecchi, 1996; Favier et al., 1996; Fromental-Ramain et al., 1996). Thus, mutations in 3′ Hox genes affect the formation of anterior embryonic structures, whereas disruption of 5′ genes gives rise to posterior abnormalities. In addition to being used to regionalize the embryo along the major axes, Hox genes also appear to be used to regionalize subcomponents of the embryo such as the gut, the reproductive organs and the limbs (Yokouchi et al., 1991, 1995; Roberts et al., 1995; Dollé et al., 1989, 1991; Izpisúa-Belmonte et al., 1991). Regionalization of the embryo by Hox genes appears to be accomplished by the controlled temporal and spatial activation of these genes such that a 3′ gene is activated prior to and in a more anterior region of the embryo than its 5′ neighbor (Duboule and Dollé, 1989; Graham et al., 1989). This cascade of gene activation generates a nested set of Hox gene transcripts along the embryonic axes and in the embryonic subcomponents. As a result, each region is characterized by expression of a specific complement of Hox genes. The complement of Hox genes in turn activates a region-specific developmental program and thereby orchestrates the morphological regionalization of the embryo. Indeed, through their ordered position on the chromosomes, a fundamental property of Hox genes may be their ability to convert a series of temporally ordered transcriptional events into a directed, ordered series of morphological events.

In this report, we examine the effects on mouse development of disrupting the hoxb-1 gene, the most 3′ member of the Hox B linkage group. After 8.5 days of gestation (E8.5), rostral expression of hoxb-1 is limited in the hindbrain to rhombomere 4 and to the neural crest cells migrating from this rhombomere. Hoxb-1 mutant mice do not have any apparent defects in tissues derived from this neural crest cell population. However, formation of the somatic motor component of the facial nerve is severely compromised in these mutants. Defects are also seen in the facial musculature. Both of these defects may be tied to a failure to specify the population of motor neurons that is normally generated in rhombomere 4 and destined to contribute to the facial motor nucleus in the hindbrain.

Genomic DNA surrounding and including the hoxb-1 locus was isolated from a 129Sv mouse library in lambda FIX II (Stratagene). Exon I was inactivated by insertion of a 12 bp BamHI linker into the unique EagI site. The homeodomain in exon II was disrupted by replacement of a 371 bp BglII fragment in this exon (Frohman et al., 1990) with the neomycin resistance (neo^r) gene from MC1neopA (Thomas and Capecchi, 1987). The mutated sequences included in 12 kb of genomic DNA were cloned into a pUC-based vector containing the herpes simplex virus thymidine kinase genes, TK1 and TK2 (Fig. 1A).

The targeting vector described above was linearized and electroporated into R1 ES cells (Nagy et al., 1993). Following positive/negative selection in G418 and FIAU (Mansour et al., 1988), Southern blot analysis was used to identify the homologous recombinants (Fig. 1 and below). The accuracy of the recombination events was verified by a variety of restriction enzyme digests and three different probes. The latter included the 5′-flanking probe, a probe specific for the neo^r gene, and a 750 bp internal probe extending from the EagI site in exon I to an XbaI site in the intron. Positive cell lines were injected into blastocysts. The resulting male chimeric mice were bred with C57BL/6J (BL/6) females, and offspring carrying the mutation in the germ line were identified by Southern blot analysis (Fig. 1B and below).

Adult mice were genotyped by Southern blot analysis or by gel analysis of PCR-amplified sequence from DNA purified from tail biopsies (Thomas et al., 1992). For Southern blot analysis, the DNA was digested with BamHI and probed with the 5′-flanking probe. While the wild-type sequences were contained on a 6.7 kb fragment, those of hoxb-1^neo and hoxb-1^neoB were detected as 7.7 kb and 5.2 kb fragments, respectively (Fig. 1B).

The primers used for PCR amplification were as follows: Hoxb-1 sense primer, 5′AAGGTGTCCGAGCTGGGACTG (nucleotides 663-684, Frohman et al., 1990); hoxb-1 antisense primer, 5′CCAGC-CATCAATCATCCCTCCA (nucleotides 1182-1161, Frohman et al., 1990); neo sense primer, 5′GCCTGCTTGCCGAATATCATGG (225 bp from the 3′-end of the neo gene, Beck et al., 1982). PCR cycling times were 94 °C, 30 seconds; 61 °C, 30 seconds; 72 °C, 30 seconds for 30 cycles with a 1 second/cycle extension at 72 °C. Amplified products were analyzed on a 1.5% TreviGel 500 (Trevigen). The wild-type band was 520 bp; that of the neo allele was 373 bp.

Yolk sacs dissected from embryos were analyzed by PCR. After incubation at 55 °C overnight in PCR lysis buffer (Carpenter et al., 1993) containing proteinase K at 1 mg/ml, the samples were boiled for 5 minutes and chilled on ice; an aliquot was added to the PCR reaction mixture and analyzed as described above.

For paraffin sectioning, newborn mice were killed by CO₂ asphixiation, fixed in Bouin’s and embedded in Paraplast X-Tra. 10 μm sections from 3 hoxb-1^neoB heterozygous and 3 hoxb-1^neoB homozygous animals were stained with hematoxylin and eosin, mounted in DPX and photographed as previously described (Mansour et al., 1993).

For vibratome sectioning, SWR/J embryos (E9.0 to E10.5) were fixed at 4 °C overnight in 4% paraformaldehyde in phosphate-buffered saline (PBS), washed in PBS and embedded in 15% gelatin. The embedded embryos were refixed at 4 °C overnight and washed with PBS before sectioning.

For detection of neurofilament protein, paraffin sections were dewaxed with HemoD, hydrated through a graded ethanol series and then treated for 30 minutes with 0.6% hydrogen peroxide in methanol to inactivate endogenous peroxidases. The sections were blocked in 3% goat serum in PBS (PBSG) for 60 minutes, rinsed with PBS and incubated at 4 °C overnight with the 2H3 antibody directed against the 155×10³M_r subunit of the neurofilament protein (Dodd et al., 1988) diluted 1:1 in PBSG. Sections were washed in PBS and then incubated 2 hours with a horseradish peroxidase (HRP)-conjugated goat antimouse secondary antibody (Jackson Immunoresearch) diluted 1:200 in PBSG. For the HRP reaction, sections were incubated in 0.5 mg/ml diaminobenzidine (DAB) in PBS plus 0.03% hydrogen peroxide. After rinses in PBS and water, sections were counterstained in nuclear fast red, mounted in DPX and photographed as previously described (Mansour et al., 1993).

Immunostaining of hoxb-1 protein in embryos and vibratome sections of embryos was carried out with a hoxb-1 affinity-purified polyclonal antibody directed against a GST-fusion protein encoded by hoxb-1 exon I (Manley and Capecchi, 1995). An HRP-conjugated donkey anti-rabbit secondary antibody (Jackson Immunoresearch) was used for detection of the antigen-antibody complex. The primary and secondary antibodies were used at dilutions of 1:150 and 1:400, respectively. For whole-mount staining, E9.5 embryos were fixed in 4% paraformaldehyde in PBS overnight at 4°C. Conditions for inactivation of endogenous peroxidases, incubation with the primary and secondary antibodies, and the HRP-reaction using nickel chloride were as described in Wall et al. (1992). For hoxb-1 immunostaining of vibratome sections, inactivation of endogenous peroxidases was carried out in PBS, 0.5% Triton X-100:hydrogen peroxide (4:1).

A krox-20 polyclonal antibody was generated as described by Manley and Capecchi (1995). A protein containing 87 amino acids from the beginning of the krox-20 protein (BclI-MscI 257 base pair fragment, amino acids 8-95) fused to the glutathione-S-transferase gene (GST, Smith and Johnson, 1988), was made in E. coli and purified by adsorption to GST-sepharose beads. Purified fusion protein was injected into rabbits using a 750 mg initial lymph node injection with 750 mg intramuscular boosts. Antibodies against the krox-20 protein were affinity-purified using the krox-20-GST fusion protein after subtraction of antibodies against GST and bacterial proteins. Absence of antibodies to GST was confirmed by western blot.

Immunostaining of krox-20 protein in E8.5 (9 –12 somites) embryos was performed as described above for hoxb-1 protein, except that the embryos were fixed in methanol:dimethylsulfoxide (4:1). The krox-20 primary and secondary antibodies were diluted 1:50 and 1:200, respectively.

Immunolocalization of the CRABP 1 protein was performed on E9.0 embryos as previously described (Manley and Capecchi, 1995) using a polyclonal antibody generously provided by U. Eriksson (Eriksson et al., 1987). Whole-mount embryos and vibratome sections were photographed with Ektachrome 160T film with a Wild dissecting microscope.

In situ hybridization of E9.5 embryos was performed as previously described (Manley and Capecchi, 1995). The hoxb-2 probe was a 900 bp PstI/HindIII fragment isolated from a cDNA clone, containing 130 bp of 3′-coding sequences and extending into the 3′-untranslated region. The probe was labelled with digoxigenin and detected with alkaline phosphatase-conjugated anti-digoxigenin Fab fragments (Jackson Immunoresearch) diluted at 1:5000.

Procedures used for the injection of the carbocyanine dyes, DiO and DiI (Molecular Probes), into cranial ganglia (V, VII/VIII, IX/X) have been described previously (Carpenter et al, 1993). Embryos were injected either bilaterally or unilaterally. For photography, the embryos were mounted in depression slides and viewed using fluorescein and rhodamine filter systems on a Leitz Ortholux microscope equipped with epifluorescence illumination. Animals were photographed with Kodak Ektachrome 400 film. In addition, embryos were viewed through a rhodamine filter set using a laser scanning BioRad MRC600 confocal microscope and COMOS software. Serial optical sections were combined in a superimposed ‘Z series’ image.

Dye injections of embryos and embryo culture were performed as previously described (Serbedzija et al., 1992; Manley and Capecchi, 1995) and analyzed by confocal microscopy.

Two mutant alleles of the hoxb-1 gene, hoxb-1^neo and hoxb-1^neoB, were generated by homologous recombination between a replacement targeting vector and one of the hoxb-1 loci in embryo-derived stem (ES) cells (Fig. 1A). The vector contained 12 kb of genomic DNA sequences flanked at the 5′ and 3′ ends by the herpes simplex virus thymidine kinase genes. Exon I in the vector was disrupted by insertion of a 12 bp BamHI linker. This caused a frameshift mutation closely followed by termination codons in all three reading frames. Exon II of hoxb-1 was inactivated by introduction of MC1neopA (Thomas and Capecchi, 1987) into the BglII site in helix III of the homeodomain.

The targeting vector was linearized and electroporated into R1 ES cells. Homologous recombinants were enriched by positive/negative selection (Mansour et al., 1988). Southern blot analysis of 150 selected cell lines revealed that, in five, the normal hoxb-1 sequences had been accurately replaced with those containing MC1neopA from the vector. In four of these five, the BamHI linker in exon I had also been cotransferred. The two alleles could be distinguished from each other and from the wild-type sequences by Southern transfer analysis (Fig. 1B). All positive cell lines were further analyzed with several restriction enzyme digests and internal probes to ensure that the cell lines contained no detectable rearrangements (data not shown).

Two of these cell lines, one containing only the MC1neo sequences (1g) and the other containing both the MC1neo sequences and the BamHI linker in exon I (2h), were injected into blastocysts. Male chimeric mice were derived from both of these cell lines and subsequently bred to BL/6 females; analysis of the resulting progeny revealed that the mutant alleles had been transmitted through the germ line. Animals heterozygous for either mutant hoxb-1 allele are viable and fertile.

The effects of the two types of mutation on the expression of hoxb-1 protein were examined in E9.5 embryos of each class subjected to whole-mount immunohistochemistry with antibody directed against amino acids encoded within exon I (Manley and Capecchi, 1995). As shown in Fig. 2, animals heterozygous for either mutant allele exhibited the expected pattern of expression in rhombomere 4 (r4), in the caudal region of the tail and in the neural crest emanating from r4. A very low level of hoxb-1 immunoreactivity was detectable in hoxb-1^neo mutant homozygous embryos, but none in the hoxb-1^neoB mutant homozygotes (Fig. 2). As shown below, the phenotypes of mice containing the two different mutant alleles are very similar; however, they are distinguishable with respect to survival.

Offspring generated from matings between hoxb-1^neo or hoxb-1^neoB animals were found to be present in the expected Mendelian ratios. This was true regardless of whether the crosses were performed between heterozygotes or between homozygous and heterozygous animals. Of the 20 mutant animals in the hoxb-1^neo colony allowed to mature to weaning, all but one survived. This animal died at 15 days after birth. However, 3/35 newborn hoxb-1^neoB mutant mice were found dead at birth and another 7/29 that were allowed to mature died between 15 and 24 days.

A striking characteristic of all of the adult mutant mice carrying either mutant allele is paralysis of the muscles of the face. They have little or no movement of the whiskers and nose, they do not close their eyelids in response to touch, nor do the ears move in response to touch or sound. The mutant phenotype is further evidenced by the runting of some animals, by a narrow face and by a receded lower lip, resulting in the exposure of the lower two front teeth. These lower teeth are longer and more rounded than those of control mice, suggesting reduced chewing on solid food by these animals. Although the expressivity of the whisker movement, receded lower lip and runting phenotypes was variable among mutant animals within each colony, no significant difference could be detected between the two mutant colonies. Thus, in surviving adult mice, the phenotypes of the two alleles could not be distinguished. Although the cause of death is unknown, the period with the highest death rate in the hoxb-1^neoB colony (15 –24 days) corresponds to the period when the animals are changing to a solid food diet and the altered appearance of the lower teeth in mutant animals suggests that feeding behavior may be impaired. Extensive facial paralysis could also interfere with early feeding proficiency. Consistent with this hypothesis, mutant mice that died shortly after birth did not have milk in their stomachs.

Many of the symptoms described above are similar to features seen in humans suffering from Bell’s Palsy or Moebius Syndrome (Kumar, 1990), which are associated with damage to or congenital absence of the VIIth (facial) cranial nerve, respectively. In newborn animals, the facial motor nucleus is located in the pons. Sections were made of the head and caudal hindbrain region from five mutant and six control animals. The sections were analyzed either by immunohistochemistry with the 2H3 antibody (Dodd et al., 1988) directed against a subunit of the neurofilament protein (Fig. 3A,B) or by histological staining with hematoxylin and eosin (Fig. 3C,D). In control and mutant animals, the abducens (VIth) nucleus is visible in the dorsomedial aspect of the pons. The large facial motor nucleus is seen at the ventrolateral margin in the sections of wild-type control animals (Fig. 3A,C), but is not visible in the corresponding sections of the hoxb-1 mutants (B,D). Of the five mutant animals examined, the facial motor nuclei were undetectable in three and severely reduced in the other two. Other histological sections of these animals showed the VIIth as well as the VIIIth (vestibulocochlear) ganglia to be normal in size and shape (data not shown).

The VIIth nerve is a complex nerve having a sensory, vis-ceromotor and somatic motor component, of which the latter is the most prominent. The somatic motor component controls the muscles of facial expression. Peripherally, the VIIth nerve emerges from the skull through the stylomastoid foramen where the somatic motor component gives rise to six major branches diagrammed in Fig. 4A. In adult animals these branches are readily visible through a dissecting microscope following exposure by surgical dissection of the head and face. Five control animals and six hoxb-1 homozygous mutant animals were dissected to reveal these nerves and associated muscles (see below). In controls, the pattern of the VIIth motor nerve branches was consistent between animals. However, among the mutant mice the pattern of nerve branches differed from mouse to mouse and even on opposite sides of the same mouse (Fig. 4B-D). All of the nerve branches that could be identified in the hoxb-1 mutant homozygotes were reduced in diameter approximately 2-3 fold and appeared to terminate prematurely. Several branches were missing, including the branch innervating the orbicularis oris muscle (B,C,D), the mandibular branch (C,D) and the buccal branch (B,D). The mutants shown in Fig. 4 had been scored as having a severe phenotype. One mutant mouse having a less severe phenotype was also dissected. In this animal, all of the major motor nerve branches were detected but reduced, 2-3 fold, in diameter. Although variability in the expressivity of the defects in the VIIth motor nerve branches is evident in these mutant mice, the severity of the defects among survivors did not correlate with the presence of the hoxb-1^neo or hoxb-1^neoB mutant alleles.

The sensory component of the VIIth nerve does not appear to be affected by either hoxb-1 mutation. Sensory neurons, whose cell bodies are located in the VIIth (geniculate) ganglion, innervate taste buds on the soft palate and the anterior two-thirds of the tongue as well as the skin of the external ear. The VIIth ganglion was found to be normal in size and shape in hoxb-1 mutants when detected either with the 2H3 neurofilament antibody in E10.5 mutant embryos or in histological sections of newborn mutant animals (data not shown). In addition, the density of taste buds, which require innervation for their development, on the anterior two-thirds of the tongue in hoxb-1 mutants was found to be comparable with those of heterozygotes both in newborn animals (analyzed in histological sections) and in adults (analyzed by microscopy using a camera lucida) (data not shown). The visceromotor or parasympathetic component of the VIIth nerve, however, may be affected by the hoxb-1 mutation, and these defects may contribute to the loss of viability observed among hoxb-1^neoB mutant homozygotes following weaning. Parasympathetic neurons innervate the glands of the nasal mucosa and the lacrimal glands through the greater petrosal nerve and the pterygopalatine ganglion, and the salivary glands through the chorda tympani and the submandibular ganglion. The latter neuronal circuit is responsible for activating secretion of saliva following the ingestion of solid food. Interestingly, one of the hoxb-1 mutant homozygotes that died 15 days after birth showed gross deficiencies of the lacrimal, sublingual and submandibular glands (data not shown).

The faces of hoxb-1 mutant homozygotes are characteristically narrower than those of their wild-type or heterozygous littermates. The narrowing of the face is apparent two weeks following birth and then becomes progressively more severe with age. Since the skull bones, including the mandible and maxilla, appear normal in hoxb-1⁻/hoxb-1⁻ mice, the narrowing of the face is most likely a result of facial muscle degeneration. Consistent with this hypothesis is the fact that all of the facial muscles in hoxb-1 mutant homozygotes appear normal at birth (the facial muscles of three mutant and three control animals were examined in histological, transverse sections). To test for facial muscle integrity in adult animals, the dissections used to expose the branches of the facial nerves, described above, were extended to permit examination of the facial muscles (Table 1). In these hoxb-1 mutant homozygotes, the zygomatic, buccinator, depressor anguli oris and caudal digastric muscles were reduced approximately two-fold in size compared with those of nonmutant littermates. In both mutants, traces of the nasolabialis muscle could not be found and, in one mutant, the labii maxillaris muscle was not detected. Interestingly, in both mutants, the mass and shape of the temporalis and masseter muscles were abnormal. These muscles are innervated by the Vth cranial nerve. The defects in these muscles most likely arose as an indirect consequence of degeneration of the facial muscles of expression (i.e., abnormal modeling either as a consequence of improper tensions among the facial muscles, decreased usage or both). All of the facial muscles could be detected in a mutant mouse with a less severe phenotype. In this atypical mutant, only the levator nasolabialis and levator labii maxillaris were reduced.

An additional muscle innervated by the somatic motor component of the VIIth nerve is the stapedius muscle associated with the stapes of the middle ear. The normal function of this muscle is to dampen the movement of the stapes in response to loud noises. In control animals, the stapedius muscle, the stapes and the VIIth nerve are in tight association (Fig. 5A,C). In mutants, the stapes and stapedius muscle appear normal; however, the nerve innervating this muscle is reduced in size and dissociated from the muscle (Fig. 5B,D).

The results described above are consistent with the interpretation that the hoxb-1 phenotype results from defects in the VIIth nerve. However, in mice mutant for the paralogous gene, hoxa-1, the neuronal defects result from disorganization of the hindbrain, including the absence of rhombomere 5 (r5) and major defects apparent from r3 through r8 (Carpenter et al., 1993; Mark et al., 1993). Therefore, it was important to examine the integrity of hindbrain organization in the hoxb-1 mutant mice. In addition, since hoxb-1 expression is detected in neural crest cells migrating from r4 into the 2nd pharyngeal arch, it was also important to examine mutant mice for defects in the production or migration of r4 neural crest cells or defects in tissues derived from the 2nd pharyngeal arch.

The integrity of rhombomeres in hoxb-1 mutant mice was examined by the use of a series of molecular markers expressed in the developing hindbrain including krox-20, hoxb-2, hoxb-3 and hoxa-2. Krox-20 is a zinc-finger transcription factor that is transiently expressed in the early hindbrain in rhombomeres 3 and 5 (Wilkinson et al., 1989a). Hoxb-1 heterozygous and homozygous mutant E9.0 embryos were subjected to immuno-histochemistry with an antibody directed against the krox-20 protein. Production of krox-20 protein in rhombomeres 3 and 5 was indistinguishable in the two embryos (Fig. 6A,E). Thus the r3/r4 and r4/r5 boundaries appear to be intact in the hoxb-1 mutants. In addition, normal krox-20-labelled neural crest cells, migrating from the hindbrain, were apparent in both embryos.

The expression of hoxb-2 was examined using RNA in situ hybridization of whole embryos. Fig. 6B,F show results of this analysis in hoxb-1 heterozygous and homozygous mutant embryos, respectively. No significant differences from controls could be detected in the hoxb-1 mutant embryo, either in the r2/r3 boundary of hoxb-2 expression, or in the level of expression in r3, r4 and r5, or in the neural crest migrating from r4 (Wilkinson et al., 1989b; Hunt et al., 1991). This analysis also demonstrated that the hoxb-1 mutation does not alter the expression pattern of hoxb-2, the gene adjacent to hoxb-1 in the Hox B linkage group. In these preparations, as well as many others, the overall rhombomeric morphology of hoxb-1 mutant embryos was indistinguishable from that observed in non-mutant embryos. Expression patterns of hoxa-2 and hoxb-3 in the hindbrain and neural crest of E9.5 embryos were also found to be similar in hoxb-1 control and mutant embryos (data not shown).

The pattern of neural crest migration from rhombomere 4 was further examined by two additional means. First, hoxb-1 heterozygous and homozygous mutant E9.0 embryos were examined by immunohistochemistry with an antibody against the CRABP1 protein, which labels neural crest cells migrating from rhombomeres 2, 4 and 6 (Maden et al., 1992). In the heterozygous control embryo (Fig. 6C), a stream of neural crest cells migrating from rhombomere 4 into the second pharyngeal arch is seen. A similar pattern of migration is present in the homozygous mutant embryo (Fig. 6G).

The second method used to examine neural crest from rhombomere 4 involved introducing DiI into the neural tube of hoxb-1 heterozygous and homozygous mutant E8.5 embryos and analyzing the migration of the resulting DiI-labelled neural crest after 24 hours of in vitro culture (Serbedzija et al., 1992; Manley and Capecchi, 1995). Examples of this type of analysis are shown in Fig. 6D,H. Again, one can see a stream of neural crest cells migrating from rhombomere 4 in both the heterozygous and homozygous mutant embryos. Finally, hoxb-1 mutant mice were carefully examined for defects in tissues and bones derived from r4-mesenchymal neural crest cells but none was observed (data not shown).

The position and early migration patterns of the cell bodies in the hindbrain associated with the VIIth nerve can be determined by retrograde labelling of these cells with fluorescent carbocyanine dyes. We previously described (Carpenter et al., 1993) the distribution of efferent neurons in the hindbrain of E11.5 wild-type embryos following injection of DiI into the VII/VIIIth (facial/vestibulocochlear) cranial ganglion complex to label neurons in r4 and r5, and injection of a second carbocyanine dye, DiO, into the Vth (trigeminal) cranial ganglion to label neurons in r1-3. Results from control embryos double-labelled in this way are shown in Fig. 7. Since rhombomere boundaries are no longer clearly visible at E11.5, the DiO-labelling patterns in rhombomeres 1-3 define the level of the r3/r4 boundary. Fig. 7A shows the ventral surface of the right side of the control hindbrain that was photographed with fluorescein optics, which allows viewing of both DiI and DiO. In Fig. 7B, the same hindbrain was photographed with a rhodamine filter which allows visualization only of DiI. In this figure, one can see the typical ‘fan-shaped’ pattern formed by the cell bodies and their fibers in rhombomeres 4 and 5. Axons from the medial cell bodies in rhombomere 4 extend laterally, exiting the hindbrain at the level of rhombomere 4 in tightly bound fascicles. In rhombomere 5 there are two populations of neurons. First, there is a medial group whose cell bodies are adjacent to the ventral midline and whose axons project anteriorly and then laterally. These cells are likely to contribute to the facial motor nucleus. The second group has more laterally positioned cell bodies and projects axons laterally and then anterolaterally. Fibers from both populations exit the hindbrain at the level of rhombomere 4. The more lateral population of neurons in rhombomere 5 is likely to contribute to the parasympathetic preganglionic cells that will form the superior salivatory and lacrimal nuclei just caudal to the facial motor nucleus in the pons.

Many of the medial cell bodies in r4 and r5 appear to be migrating caudally and then ventrolaterally. Details of this migration pattern are more apparent at higher magnification. Migrating cells extend leading processes in the direction of their migration, and then elongate their cell bodies along the same direction. Such leading processes and their migrating cell bodies can be seen in Fig. 7E (elongated migrating cells and their fibers are indicted by the arrows in Fig. 7C-E).

DiO and DiI were also injected into the cranial ganglia of hoxb-1 homozygous mutant embryos. Fig. 8A,B show the ventral surface of the right side of a hoxb-1 mutant hindbrain injected with DiO in the Vth ganglion (GV) and DiI into the VII/VIIIth complex. Cell bodies associated with GV efferent neurons are again restricted to r1-3, while those projecting axons to the GVII complex are present in r4,5. Thus, rhombomere integrity appears to be intact in the mutant embryos. This is in contrast to hoxa-1 mutant embryos in which rhom-bomere 5 is missing and the efferent neurons labelled by injection of DiI into the VIIth ganglion are detected anteriorly as far as r3 and posteriorly as far as r8 (Carpenter et al., 1993).

Fig. 8D-F shows the DiI-labelling patterns in rhombomeres 4 and 5 in three additional hoxb-1 mutant embryos. The most striking feature common to the mutant embryos is that the neurons normally present along the medial margin of r4 and r5 are either absent, greatly reduced in number, or different in character. Anteriorly directed axon fibers could not be detected from the medial cell bodies in either rhombomere 4 or 5. Furthermore, neuronal cell bodies are observed to be dispersed more laterally in rhombomere 4 and appear to be migrating laterally. The identity of these neurons is not known. Finally, the fibers in rhombomere 4 of the mutant embryos exit the hindbrain without forming the tightly bound fascicles seen in control embryos. Fig. 8C shows a higher magnification of the embryo in B emphasizing the lack of caudal and ventrolateral migration of the medial neurons around the parasympathetic neurons in rhombomere 5.

Whether the pattern of parasympathetic preganglionic neurons in rhombomere 5 is affected by the hoxb-1 mutation is more difficult to discern. Examination of this component shows it to be less robust and more diffuse in mutants than in controls. Also the cell bodies, especially those of the embryo shown in Fig. 8D, appear to extend more medially, possibly because they are not displaced by migrating medial motor neurons. DiI was also injected in the IXth (glossopharyngeal) and Xth (nodose) cranial ganglia in 4 hoxb-1 mutant homozygous embryos. The resulting labelling patterns in r6-8 were indistinguishable from those of non-mutant animals.

In the absence of hoxb-1, formation of the facial motor neurons is severely affected, suggesting that the hoxb-1 gene product is required for specification of these neurons. Although the penetrance of the phenotype is 100%, some variation in expressivity is apparent, suggesting that other Hox genes may compensate for this function. Between E9.0 and E9.5, hoxb-1 expression in the rostral portion of the embryo is restricted to rhombomere 4 (Fig. 9A). At E10.0 a medial finger of hoxb-1 expression begins to be extended caudally into r5 and becomes more pronounced by E10.5 (Fig. 9B-C). A similar pattern of hoxb-1 expression in r5 has been observed in E10.5 transgenic mice which contained a lacZ reporter gene (Marshall et al., 1992). The dynamics of hoxb-1 expression in normal embryos suggest that the facial motor neurons are first specified in r4 and then migrate caudally along the medial axis into r5. On reaching the r5/r6 boundary these neurons turn abruptly and migrate laterally (see Fig. 7). This migration pattern is diagrammed in Fig. 9D-F. Consistent with the hypothesis that the facial motor neurons are generated in r4, Lumsden and Keynes (1989) have previously noted that, in the hindbrain, neuron maturation occurs earlier in even-numbered rhombomeres than in odd-numbered rhombomeres. Also, mice mutant for krox-20, which are deficient in the formation and/or maintenance of rhombomeres 3 and 5, have a morphologically normal facial motor nucleus (Schneider-Maunoury et al., 1993).

The somatic motor component of the VIIth nerve is characterized by its unusual ‘looped’ course through the pons (Altman and Bayer, 1982). Thus, although the motor nucleus is located ventrolaterally in the pons, its axons project dorso-medially around the VIth (abducens) nuclei and then turn ventrolaterally to exit the pons (diagrammed in Fig. 9G). The course of this loop can be understood in terms of the early migration pattern of these neurons during development proposed here (Fig. 9D-F).

We generated mice with two separate targeted disruptions of hoxb-1. The first, hoxb-1^neo, disrupts the homeodomain, and the second, hoxb-1^neoB, disrupts the first exon of hoxb-1, as well as the homeodomain. Among survivors, phenotypes of mice homozygous for the two mutant alleles were not distinguishable, arguing that the homeodomain, which confers DNA-binding capability, is an essential component of hoxb-1 protein function. However, mutants for the hoxb-1^neoB allele did show greater morbidity at birth and at the age of weaning, suggesting that this mutation is a more severe loss-of-function allele.

The inability to distinguish these two mutant alleles among survivors may result from both mutant alleles showing similar variability in the expressivity of the mutant phenotype, thereby masking their differences. This variability is likely to result from overlap of function from other Hox genes. At least two Hox genes can provide compensation for the hoxb-1 mutation, hoxa-1 and hoxb-2. Thus, mice mutant for both hoxa-1 and hoxb-1 show a very exacerbated hoxb-1 mutant phenotype (Rossel and Capecchi, unpublished results). Similarly, mice expressing neither hoxb-1 nor hoxb-2 show a stronger and less variable hoxb-1 mutant phenotype than do hoxb-1^neoB homozygotes (Barrow and Capecchi, unpublished data).

The predominant feature of the hoxb-1 mutant homozygous phenotype among survivors is paralysis of the muscles controlling facial expression. This feature is also observed in humans suffering from Bell’s Palsy or Moebius Syndrome. These human diseases result from damage to, or congenital absence of, the VIIth (facial) cranial nerve, respectively. Hoxb-1 mutant homozygotes also have defects in the VIIth nerve. Thus, sections of the pons show a marked reduction or absence of the VIIth nerve motor nucleus. Also, surgical examination of the facial periphery shows that the major VIIth nerve somatic motor branches are either absent or reduced 2-3 fold in diameter. Finally, at E11.5 retrograde labelling of neurons in hoxb-1 mutant hindbrains by injection of DiI into the VII/VIIIth ganglionic complex fails to label cells thought to give rise to the VIIth nerve motor nucleus.

The entire hoxb-1 phenotype could be explained in terms of a defect in the formation of the VIIth nerve. This interpretation would postulate that the primary role of hoxb-1 is to specify the neurons that will form the somatic motor component of the VIIth nerve. Consistent with this hypothesis, we have not observed any defects in tissues or structures derived from neural crest cells in hoxb-1 mutant mice or from the 2nd pharyngeal arch, both sites of hoxb-1 expression. The rhombomeric pattern of the embryonic hindbrain also appears normal. This is in sharp contrast to hoxa-1 mutant mice, which exhibit gross defects in the metameric organization of the hindbrain (Carpenter et al., 1993; Mark et al., 1993; Capecchi, 1997).

Whether the hoxb-1 mutation affects other components of the VIIth nerve is less clear. The sensory component does not appear to be affected since both the VIIth ganglion and taste buds on the anterior two-thirds of the tongue appear normal in hoxb-1 mutant homozygotes. On the contrary, the parasympathetic component of the VIIth nerve may be affected in these mutant mice, but changes in this component show incomplete penetrance and broad variability in expressivity. A number of survivors have been examined histologically for defects in the lacrimal, sublingual and submandibular glands and none were observed. However, a hoxb-1 mutant homozygote that died 15 days after birth had gross defects in these glands. The morbidity of hoxb-1^neoB homozygotes following their transfer to solid foods could be explained in terms of defects in the parasympathetic component of the VIIth nerve.

Hoxb-1 mutant mice have narrow faces relative to control littermates. This phenotype becomes apparent two weeks after birth and then becomes progressively more severe with age. Since the skull bones in these mutant mice appear normal, this defect has been attributed to facial muscle degeneration. Consistent with this hypothesis, we found that a number of facial muscles either had reduced mass or could not be detected in hoxb-1 adult mutant mice but were histologically normal in mutant newborn animals. However, we were surprised that there was not complete degeneration of the facial muscles. Survival of the remaining facial muscle fibers may result from more extensive arborization of the remaining VIIth nerve fibers, from abnormal innervation by other cranial nerves (i.e., the Vth or IXth nerve) or both.

Hoxb-1 mutant mice fail to form the VIIth nerve motor nucleus. The fate of the cells that, in the presence of hoxb-1 protein, would normally give rise to these motoneurons has not been determined in mutant embryos. They may fail to proliferate or they may die. Alternatively, they may have acquired a new fate. Injection of DiI into the VII/VIIIth ganglionic complex of hoxb-1 mutant mice labels nuclei in the more lateral aspect of rhombomere 4 that are not observed in control wild-type or heterozygous embryos. The identity of these cells has not been determined. However, they send axons to the VII/VIIIth ganglionic complex. Determination of whether these cells represent cells with an altered fate or an altered motoneuron behavior must await further experimentation.

The differences in phenotype between mice with targeted disruptions in the two paralogous genes hoxa-1 and hoxb-1 are extensive. At E8.0 both genes have the same anterior limit of expression at the presumptive r3/r4 rhombomere boundary. However, whereas hoxa-1 expression rapidly recedes, such that by E8.5 hoxa-1 expression is no longer detectable in the hindbrain, hoxb-1 expression remains behind in a stripe of expression in presumptive rhombomere 4 (Murphy and Hill, 1991). This expression is upregulated in rhombomere 4 and maintained via autoregulation well into mid-gestation (Pöpperl et al., 1995). As already mentioned, in the absence of hoxa-1 function, hindbrain development is severely affected, rhombomere 5 is missing and defects are apparent from rhom-bomere 3 through rhombomere 8. The phenotype suggests either that rhombomere 5 is not formed or not maintained. Such a role in segmentation is in contrast to the situation in Drosophila, in which mutations in HomC genes do not alter the pattern of segmentation but rather determine the identity of cells within parasegments. On the contrary, the hoxb-1 mutation does not alter the rhombomere pattern but rather appears to alter the fate of a restricted set of cells within rhom-bomere 4. In this sense hoxb-1 function parallels more closely the function of HomC genes in Drosophila.

In summary, mice mutant for hoxb-1 have defects in the formation of the somatic motor component of the VIIth (facial) nerve. The abnormality appears to result from a defect in cell specification. The analysis of these mutant mice has provided a model for when and where these motor neurons are specified and for how the early migration patterns in the developing hindbrain lead to the circuitous axonal projection that typifies the facial motor nucleus in the pons. The phenotype of hoxb-1 mutant mice resembles symptoms found in humans with Bell’s Palsy or Moebius Syndrome. These mutant mice may allow study of the etiology of these diseases in greater detail than is possible in humans as well as serve as a model for developing more effective therapies.

We especially acknowledge and thank J. Barrow for his many contributions to this work, including in situ hybridization and immunostaining of vibratome sections. B. Condie prepared the krox-20 GST-fusion construct used to generate the krox-20 antisera and K. Thomas provided DNA vectors. We thank E. Carpenter for help with neuroanatomy, M. Allen, C. Lenz, G. Peterson, S. Barnett, E. Nakashima and M. Wagstaff for excellent technical assistance, and L. Oswald for help with preparation of the manuscript. The CRABP 1 antibody was generously provided by U. Eriksson. The 2H3 neurofilament hybridoma was obtained from the Developmental Studies Hybridoma Bank under contract N01-HD-6-2915 from the NICHD.