ABSTRACT
Mice with a disruption in the hoxb-2 locus were generated by gene targeting. 75% of the hoxb-2 mutant homozygotes died within 24 hours of birth. While a majority of these mice had severe sternal defects that compromised their ability to breathe, some had relatively normal sternum morphology, suggesting that one or more additional factor(s) contributed to neonatal lethality. At 3-3.5 weeks of age, half of the remaining hoxb-2 homozygotes became weak and subsequently died. All of the mutants that survived to 3 weeks of age showed marked facial paralysis similar to, but more severe than, that reported for hoxb-1 mutant homozygotes (Goddard, J. M., Rossel, M., Manley, N. R. and Capecchi, M. R. (1996) Development 122, 3217-3228). As for the hoxb-1 mutations, the facial paralysis observed in mice homozygous for the hoxb-2 mutation results from a failure to form the somatic motor component of the VIIth (facial) nerve which controls the muscles of facial expression. Features of this phenotype closely resemble the clinical signs associated with Bell’s Palsy and Moebius Syndrome in humans. The sternal defects seen in hoxb-2 mutant mice are similar to those previously reported for hoxb-4 mutant mice (Ramirez-Solis, R., Zheng, H., Whiting, J., Krumlauf, R. and Bradley. A. (1993) Cell 73, 279-294). The above results suggest that the hoxb-2 mutant phenotype may result in part from effects of the hoxb-2 mutation on the expression of both hoxb-1 and hoxb-4. Consistent with this proposal, we found that the hoxb-2 mutation disrupts the expression of hoxb-1 in cis. In addition, the hoxb-2 mutation changes the expression of hoxb-4 and the hoxb-4 mutation, in turn, alters the pattern of hoxb-2 expression. Hoxb-2 and hoxb-4 appear to function together to mediate proper closure of the ventral thoracic body wall. Failure in this closure results in severe defects of the sternum.
INTRODUCTION
Hox genes encode transcription factors belonging to the Anten-napedia homeodomain class. Man and mouse contain at least 39 Hox genes distributed on four linkage groups designated Hox A, B, C and D. This organization is believed to have arisen early in vertebrate phylogeny by quadruplication of an ancestral complex common to vertebrates and invertebrates (Pendleton et al., 1993; Holland and Garcia-Fernandez, 1996). Based on DNA sequence similarities and on the position of the genes on their respective chromosomes, individual members of the four linkage groups have been classified into 13 paralogous families. 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; Satokata et al., 1995; Suemori et al., 1995; Boulet and Capecchi, 1996). Thus, mutations in 3 ′ Hox genes affect the formation of anterior structures whereas disruption of 5 ′ genes gives rise to posterior abnormalities. 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; Duboule, 1994; Capecchi, 1997). However, Hox genes function as highly integrated circuits such that paralogous genes, adjacent genes on the same linkage group and even non-paralogous genes in separate linkage groups interact positively, negatively and in parallel with each other to orchestrate the mor-phological regionalization of the embryo (Condie and Capecchi, 1994; Rancourt et al., 1995; Davis et al., 1995; Horan et al., 1995; Davis and Capecchi, 1996; Favier et al., 1996; Fromental-Ramain et al., 1996).
In this report, we examine the effects of disrupting the hoxb-2 gene on mouse development. The phenotype is complex, but can be understood in terms of the effect of inactivating hoxb-2 function, combined with the consequences of disrupting the expression of two neighboring Hox genes, hoxb-1 and hoxb-4. Hoxb-1 and hoxb-2 appear to function together in the specification of the motor component of the VIIth nerve. Hoxb-2 and hoxb-4 are apparently both required to mediate closure of the ventral thoracic body wall. It is suggested that a failure in this closure prevents fusion of the two sternal bands and subsequent abnormal formation of the sternum.
MATERIALS AND METHODS
Targeting vector and generation of hoxb-2- mutant mice
A 12 kb DNA fragment containing the hoxb-2 gene was isolated from an embryo-derived stem (ES) cell genomic library and was used to construct the targeting vector (Thomas and Capecchi, 1987). The fragment extended from an EagI site 5 ′ of the hoxb-3 gene to a NotI site 3 ′ of hoxb-2 (Fig. 1). The KT3NP3 neomycin resistance cassette was inserted into the EagI site in the hoxb-2 homeobox, thus disrupting the third helix in the homeodomain of the encoded protein. KT3NP3 is a 2.7 kb neo cassette that is identical to the previously reported KT3NP4 (Deng et al., 1993) but lacks 400 bp from the 3 ′ end of the polyadenylation signal. The hoxb-2/neo construct was flanked by the herpes simplex virus thymidine kinase genes, TK1 and TK2. The vector was linearized and electroporated into R1 ES cells (Nagy et al., 1993). The cells were then subjected to positive/negative selection using G418 and FIAU as previously described (Mansour et al., 1988).
77 G418/FIAU-resistant cell lines were isolated and screened for targeting events by Southern transfer analysis (Fig. 1). BamHI-digested DNA from each of the cell lines was transferred to nitrocellulose and hybridized with a 500 bp NotI/BamHI 3 ′ flanking probe (Fig. 1). DNA from four cell lines was found to have the expected electromobility shifts from 5.5 to 2.9 kb and each DNA sample was further analyzed with multiple restriction enzyme digests and hybridization probes in order to verify the fidelity of the homologous recombination event. Two of these cell lines, when injected into C57BL/6J blastocysts and implanted into foster mothers, gave rise to germ-line chimeras.
Genotyping of animals
Genotypic analysis was carried out on DNA extracted from tails of mice and performed either by Southern transfer analysis or by PCR amplification (Thomas et al., 1992). Southern blots were performed with the same digests and probes as described above and in Fig. 1. PCR analysis was performed using the following primers: hoxb-2 sense primer 5 ′GATTGCCAGAATGCGGCGGCA3 ′ derived from the 5 ′ end of the homeobox (Rubock et al., 1990); hoxb-2 antisense primer 5 ′CCCTGCGCGGCCTCGGCG3 ′ commencing 150 bases 3 ′ of the homeobox (N. Manley, unpublished sequence); RNA poly-merase II (pol II) antisense primer 5 ′TGTTCAAGCC-CAAGCTTTCGCG3 ′. Briefly, the first 16 bases of the primer were derived from the 5 ′ end (antisense) of the RNA polymerase II promoter (bases 2-17, accession number M14101) whereas the last 6 bases were taken from the polylinker of pIC19H (HindIII, NruI sites). The hoxb-2 sense and antisense primers amplify a 359 bp fragment from wild-type template whereas a 271 bp product is amplified from the neo allele with the pol II and hoxb-2 antisense primers. The PCR reactions were performed under the following cycling conditions: 95°C for 30 seconds, 60°C for 20 seconds and 72°C for 30 seconds for 35 cycles. The PCR amplified bands were resolved on 2% TreviGel500 (Trevigen) gels.
Embryos were genotyped from yolk sac DNA with the PCR primers and conditions described above. Yolk sacs were dissected from the embryos and incubated in PCR lysis buffer and proteinase K (1 mg/ml) at 50°C overnight. The yolk sac DNA samples were then boiled for 5 minutes, and an aliquot was used for PCR amplification.
RNA in situ hybridization and immunohistochemistry
RNA in situ analysis was performed with digoxigenin-labeled probes as described previously (Manley and Capecchi, 1995). The hoxb-1 probe was a 300 bp PstI/BamHI fragment beginning 20 bp 3 ′ of the homeobox and extending into the 3 ′ UTR. The hoxb-2 probe was a 900 bp fragment isolated from a cDNA clone, containing 130 bp of 3 ′ coding sequences and extending into the 3 ′ untranslated region (Goddard et al., 1996). The hoxb-3 in situ hybridization probe consisted of a 450 BamHI/EcoRI fragment generated from a cDNA clone. The 5 ′ end of the probe is located ∼330 bp downstream of thehoxb-3 homeobox. The hoxb-4 probe was a 450 bp NaeI/BamHI fragment containing the last nine hoxb-4 codons and continuing into the 3 ′ untranslated sequences (Boulet and Capecchi, 1996). The hoxa-2 probe was a 600 bp XhoI/EcoRV fragment extending from the 5 ′ end of the homeodomain to the end of the protein coding sequence.
Immunohistochemistry using antibodies directed against the hoxb-1 protein was carried out as described by Goddard et al. (1996). Krox- 20 immunohistochemistry was also performed as described by Goddard et al. (1996) except that the embryos were fixed in 4% paraformaldehyde. Immunodetection of neurofilament protein was performed on E10.5-E11.5 embryos with the use of the 2H3 anti-155×103Mr neurofilament monoclonal antibody (Dodd et al., 1988) as described by Chisaka et al. (1992).
Histology and skeletal preparations
Newborn mice or E18.5 fetuses were asphyxiated in CO2 and fixed in 4% formaldehyde/PBS for 24 hours or longer. The mice were then dehydrated, embedded in paraffin, sectioned and stained with hema-toxylin and eosin as previously described (Mansour et al., 1993).
Skeletons were prepared as described by Chisaka and Capecchi (1991).
RESULTS
Targeted disruption of the hoxb-2 locus
A 12 kb genomic DNA sequence encompassing the hoxb-2 gene was used to construct the hoxb-2 targeting vector (see Materials and Methods). A neomycin resistance cassette was inserted into the homeobox of hoxb-2, thereby destroying the ability of the encoded protein product to bind to specific DNA sequences (Fig. 1). The targeting vector was electroporated into R1 ES cells and positive-negative selection was used to enrich for cells containing the desired homologous recombination events (Mansour et al., 1988; Fig. 1). Four independent cell lines that carried a targeted disruption of the hoxb-2 gene were identified. Additional restriction enzyme digests, probed with both 5 ′ and 3 ′ flanking probes, as well as an internal probe, showed that these cell lines contained no modifications other than the desired neo insertion into the hoxb-2 homeobox (data not shown). Two of these cell lines were used to generate germ-line chimeras. The two colonies of mice derived from these cell lines were found to have indistinguishable pheno-types. Most of the analysis described herein involved the colony of mice derived from the 2f-4 ES cell line.
Viability and fertility of hoxb-2 mutant mice
Mice heterozygous for the hoxb-2 neo mutation appeared outwardly normal and were viable and fertile. Of the mutant homozygotes, 88% died by 3-3.5 weeks of age in two stages: 75% died within 24 hours of birth, and the remaining 13% died between 21 and 25 days after birth. The eleven mutant homozygotes that survived to 3 weeks of age were readily distinguishable from their wild-type and heterozygous littermates. Most (10/11) were runted, all had narrow faces; (Fig. 2A,B) and all had receded lower lips; (Fig. 2C,D). In addition, these mutant mice had severe paralysis of the muscles of the face. They showed no movements of the whiskers and nose, did not close their eyes in response to touch and failed to move their ears in response to touch or noise. Although the facial defects closely resemble those observed in hoxb-1 mutant homozy-gotes (Goddard et al., 1996), the defects in hoxb-2 mutant mice are uniformly more severe. The narrowing of the face is first apparent 10 days following birth and becomes progressively more severe with age. At 3 weeks of age, approximately half of the remaining mutant homozygotes become weak and die shortly thereafter (21-25 days).
Those animals that survived past 25 days (2 males and 3 females) were fertile. However, the females had difficulty raising their litters, irrespective of the genotype of their pups. Two of the females had no surviving pups 24 hours after birth. The third lost 5/7 pups within 24 hours of birth.
As mentioned above, 75% of the hoxb-2 mutant homozy-gotes died within 24 hours of birth. Genotype analysis of newborns from heterozygous intercrosses showed that the distribution of wild-type and mutant alleles did not deviate significantly from the expected Mendelian ratio (Table 1), showing that homozygous mutant animals were not dying in utero. An additional 32 mutant homozygotes derived from hoxb-2 heterozygous or from homozygous by heterozygous intercrosses were observed within 24 hours of birth and assessed for viability. 24 of the 32 mutant homozygotes (75%) were found dead or appeared to be dying from respiratory distress. Most of the dead or dying mutants had severe sternal defects, which were readily apparent from gross examination of the newborn mice (Table 2). A smaller number appeared normal with respect to formation of the sternum, but were unable to nurse (i.e., did not have milk in their stomachs). Extensive facial paralysis would be expected to interfere with nursing since the orbicularis oris muscle, which surrounds the mouth, is controlled by the VIIth nerve.
Skeletal defects in hoxb-2 mutant homozygotes
The skeletons of 29 mutant homozygous and 39 heterozygous newborn animals were examined. Three different classes of skeletal defects were observed in the mutant homozygote.
The first class, which comprised the majority of the mutant homozygotes (24/29; 83%), had a split sternum. In spite of the high penetrance of this defect, the expressivity was variable. For example, 38% (11/29) had a severely split sternum (Fig. 3E) whereas another 34% (10/29) had a milder form of this defect in which the two sternal bands were not fused but were in closer proximity to one another (Fig. 3D). Three of the 29 showed partially fused sternums with either the top or the bottom being unfused (Fig. 3C). The five remaining newborn homozygotes had completely fused sternums; however, in one of these animals the sternebrae were not aligned properly along the ventral midline (crankshaft phenotype). None of the viable newborn mutant animals exhibited the most severe sternal abnormalities, which suggests that this defect was a major contributor to morbidity.
None of the hoxb-2 mutant heterozygotes had a split sternum; however, 2/39 (5%) had a crankshaft phenotype (Fig. 3B). We have never observed this defect in numerous skeletons from wild-type mice.
Very similar sternal defects were reported for mice homozygous for a targeted disruption of the first exon of the hoxb-4 gene (Ramirez-Solis et al., 1993). These authors reported 50% neonatal lethality and attributed all of the deaths to sternal defects. We also observed that ∼50% (14/29) of the hoxb-2 mutant homozygotes appeared to die from severe or moderate sternal defects (Table 2) but that an additional 25% (7/29) of these mutant newborn animals died of other causes (i.e., severe facial paralysis).
In the second class of skeletal defects, hoxb-2 heterozygous and homozygous mice showed anterior transformations of the axis (C2) such that it more closely resembled the atlas (C1). The defect consists of a broadening of the neural arch of the axis (Fig. 4). In addition, a small number of mutants with a broadened neural arch also had an ectopic anterior arch on the ventral side of the axis, a feature normally associated only with the atlas (Fig. 4). The penetrance of the cervical vertebral transformations was 55% (16/29) in the homozygous mutants and 13% (5/39) in heterozygotes. The presence of the cervical vertebral transformations did not correlate with the presence of the sternal defects, suggesting that the two sets of abnormalities are of independent developmental origin. These cervical transformations are also observed in hoxb-4 mutant mice with similar penetrance (at least for the hoxb-4s allele; see Ramirez-Solis et al., 1993 and below). Unlike the sternal defect, the cervical vertebral transformations do not appear to compromise viability.
In the third class, both hoxb-2 +/− and −/− animals had a change in the orientation of the anterior arch of the atlas compared with wild-type control animals. In wild-type animals, the anterior arch protrudes ventrally, perpendicular to the vertebral column (Fig. 4). In 41% of +/− animals (16/39) and 76% of −/− animals (22/29), the anterior arch was slanted rostrally. The severity of the phenotype was more pronounced in homozygotes than it was in heterozygotes. For example, in 11/22 of the homozygous mutants possessing this defect, the anterior arch was almost parallel to the axial skeleton (data not shown) whereas in the most severe heterozygous animals, it was tilted at only a 45° angle with respect to the vertebral column (similar to that shown in Fig. 4B, which depicts a mutant homozygote). No wild-type mice appeared to have this defect. As was the case with the transformations of the cervical vertebrae, this defect was not associated with death.
Second branchial arch skeletal structures are unaffected in hoxb-2 mutant mice
Hoxb-2 is strongly expressed in the second branchial arch and could therefore participate in the specification of skeletal structures derived from this arch, namely the lesser horn of the hyoid, the stapes and the stylohyoid ligament. None of the 29 mutant skeletons analyzed had defects in these structures. This is in sharp contrast to mice with mutations in the other member of this paralogous family, hoxa-2, in which second arch skeletal structures are not formed (Gendron-Maguire et al., 1993; Rijli et al., 1993).
Expression of hindbrain markers appears to be normal in hoxb-2 mutant mice
The integrity of the rhombomere pattern of the early hindbrain in hoxb-2 mutant homozygous embryos was examined by the analysis of the patterns of a series of molecular markers specifically expressed in the hindbrain. Mutant embryos (E8.5 and E9.5) along with heterozygous and wild-type controls were subjected to whole-mount RNA in situ hybridization analysis with hoxa-2, hoxb-3 or hoxb-4 RNA antisense probes, or to immunohistochemistry with a krox-20 antibody. For each gene, the boundary and level of expression in the hindbrain of hoxb-2 mutant homozygotes were indistinguishable from those in control embryos. This analysis showed that there were no gross hindbrain abnormalities in these mutant embryos and that the pattern of rhombomeres (r) from r2 through r7 was normal (data not shown).
Hoxb-2 mutants have defects in the VIIth motor nucleus and VIIth nerve
Histological sections of six homozygous hoxb-2 mutant and two heterozygous control newborn mice were examined for defects in internal structures and organs of hoxb-2 mutant homozygous mice. In all six mutants, the VIIth motor nucleus in the pons was undetectable (Fig. 5B). This is similar to, but more extreme than, the defect seen in hoxb-1 mutants (Goddard et al, 1996). In addition, the VIIth (facial) nerve, which consists primarily of axons from the VIIth motor nucleus, was reduced in five of the mutants (Fig. 5D,F) and not detected in the sixth (data not shown). Fig. 5D represents the mildest VIIth nerve defect, whereas the hypotrophy of the VIIth nerve shown in Fig. 5F is more typical (4/6 mutants). The facial paralysis of hoxb-2 mutant homozygous mice, as mentioned above, can be attributed to the reduction of the motor components of the VIIth nerve, which normally control the muscles of facial expression. These defects include a retracted lower lip, a narrowed face and an inability to close the eyelids or move the whiskers and ears. Since the bones of the skull in hoxb-2 mutant homozygotes appeared normal in size and shape, the progressive narrowing of the face with age suggests that this emaciated look results from progressive atrophy of the facial muscles. Indeed, in a surgical examination of hoxb-2 homozy-gotes at 3 weeks of age, the levator nasolabialis and levator labii maxillaris could not be detected and the remaining muscles of facial expression were reduced approximately two-fold in mass relative to control littermates (see also Goddard et al., 1996). In normal 3-week-old mice, the peripheral branches of the VIIth nerve are readily visible through a dissecting microscope following exposure by surgical dissection of the facial muscles. In hoxb-2 mutant animals, only the zygomatic branch of the VIIth nerve could be detected and it was reduced two to three-fold in diameter.
In contrast to the VIIth nerve, the VIIth ganglion, which is largely made up of sensory nuclei, appeared to be normal in the hoxb-2 mutant homozygotes (Fig. 5E,F). Although the VIIth ganglion shown in Fig. 5F appears somewhat smaller than the control (Fig. 5E), examination of adjacent sections showed that the ganglion is of normal size (data not shown). The fact that the VIIth ganglion is unaffected is corroborated by data obtained from immunohistochemical analyses of E10.5 and E11.5 embryos with an anti-neurofilament antibody. In hoxb-2 mutant homozygous embryos, no difference in the VIIth or other cranial ganglia was apparent (data not shown). Therefore, histological and immunohistochemical data suggest that the sensory component of the VIIth cranial nerve is not grossly affected by mutations of hoxb-2.
Expression of hoxb-1 in hoxb-2 mutants
The similarity of the facial defects in hoxb-2 homozygous mutant mice to those observed in hoxb-1 mutant mice (Goddard et al., 1996) suggests either that both genes have overlapping functions in specifying the motor components of the VIIth nerve, or that expression of one of the genes is altered by the mutation in the other, or both. Goddard et al. (1996) have shown that hoxb-2 expression is normal and hoxb-1 protein is not detected in hoxb-1 mutant homozygous embryos. Thus the facial defects in these mutants appear to be due solely to the absence of functional hoxb-1 protein. Hoxb-1 expression in hoxb-2 mutant embryos was examined by immunohisto-chemistry with a hoxb-1 antibody (Manley and Capecchi, 1995; Goddard et al., 1996). At E7.5 the hoxb-2 mutant homozygous and control embryos could not be distinguished on the basis of the level and distribution of hoxb-1 protein (Fig. 6A,B). Thus, initiation of hoxb-1 expression in the primitive streak of gastrulating embryos appears to occur normally in hoxb-2 mutants. At E8.0 (2-4 somites), when restricted hoxb-1 expression in rhombomere 4 is just becoming apparent, mutant and control embryos were still indistinguishable (data not shown). However, by the 6- to 9-somite stage (E8.5), clear differences in hoxb-1 expression were apparent (Fig. 6C,D). Hoxb-1 protein was not detected in r4 of hoxb-2 mutant homozygous embryos and expression in the caudal region of the embryo was greatly reduced. By E9.5 hoxb-1 protein was not detected anywhere in hoxb-2 mutant homozygous embryos (Fig. 6E,F). The same results were obtained when hoxb-1 RNA patterns were examined by in situ RNA hybridization in whole embryos, demonstrating that the effect of the hoxb-2 mutation on hoxb-1 expression is at the level of transcription (data not shown). The absence of hoxb-1 protein in rhombomere 4 of hoxb-2 mutant homozygotes (i.e., from E8.5 onward) provides a molecular explanation of why hoxb-2 mutants have facial defects similar to those of hoxb-1 mutant mice. However, the greater severity of the facial defects in hoxb-2 mutant homozygotes suggests that hoxb-2 also participates in the specification of the motor component of the VIIth nerve.
The neo insertion in hoxb-2 disrupts hoxb-1 expression in cis
The absence of hoxb-1 expression in hoxb-2 mutant embryos could be due either to its being a downstream target of hoxb-2 (regulation in trans) or to disruption of expression of the linked hoxb-1 gene in cis by the neo insertion in hoxb-2. To distinguish between these two possibilities, hoxb-1/hoxb-2 trans-heterozygotes were generated and expression of the hoxb-1 gene in the trans-heterozygote was determined by immunohistochemistry with the hoxb-1 antibody. If hoxb-2 regulates hoxb-1 in trans, then one would expect normal hoxb-1 expression in the trans-heterozygote (Fig. 7B), whereas if the hoxb-2 neo insertion disrupts hoxb-1 expression in cis (Fig. 7A), then hoxb-1 protein should not be synthesized (from E8.5 onward). Hoxb-1 expression was not detected in any of the trans-hetrozygotes analyzed at E9.5 (Fig. 7C). Thus, it appears that the insertion of neo into the hoxb-2 homeobox disrupts the expression of the linked hoxb-1 gene in cis.
Phenotypic analysis corroborated the molecular results. Five adult trans-heterozygotes were obtained from hoxb-1/hoxb-2 heterozygous intercrosses. None of these had sternal defects, nor did they have facial phenotypes as severe as those of the hoxb-2 mutants, but instead had facial defects that closely resembled those observed in hoxb-1 homozygous mutants. As a consequence of the lack of sternal defects and the milder facial defects, these trans-heterozygotes are viable. Therefore, as predicted, in the absence of hoxb-1 expression, one functional copy of hoxb-2 is sufficient for normal sternal development and ameliorating the facial defects. These results are consistent with the interpretation that hoxb-2 functions together with hoxb-1 in specifying the VIIth motor neurons.
hoxb-2/hoxb-4 trans-heterozygotes exhibit a split sternum and cervical transformation
As neo insertions into either hoxb-2 or hoxb-4 cause malformations of the sternum and cervical transformations in mutant mice, it was important to determine the molecular basis for this overlap in phenotype. Our laboratory has generated mice with a KT3NP4 neo insertion in the first exon of hoxb-4 (Manley, Barrow and Capecchi, unpublished results). Although the neo cassette was inserted at precisely the same SalI site as in experiments reported by Ramirez-Solis et al. (1993), the sternum phenotype of our mice in a Bl/6 genetic background is completely penetrant, rather than 50% penetrant, and causes 100% lethality. The difference in phenotype in the same genetic background may reflect the use of a different neo cassette. The completely penetrant sternum phenotype and lethality in our mutant resembles the results reported by Ramirez-Solis et al. (1993) when their mutation was crossed into a pure SVJ129 background. In hoxb-2 mutant mice, the rib cage, including the split sternum, was covered with the pericardial fold, a thin transparent membrane that aids in holding the two sternal bands together. This membrane was not present in our hoxb-4 mutant homozygotes, so that the internal organs were exposed after removal of the skin (Manley, Barrow and Capecchi, unpublished results). Nine hoxb-2/hoxb-4 trans-heterozygous newborn mice were generated by intercrosses between hoxb-2 and hoxb-4 heterozygous mice (see Table 3). Six died at birth, the other three were in respiratory distress and were killed. The sternal defects of eight of the trans-heterozy-gotes appeared to be intermediate between those of the hoxb-2 and hoxb-4 mutant homozygotes. Three of the nine lacked the pericardial fold covering the gap between the two sternal bands. Eight of the newborns were processed for skeletons, and seven were found to have a transformation of the axis to atlas as seen in each of the parental mutant backgrounds. The skeletal defects in the trans-heterozy-gotes could be interpreted in two ways: (1) there is nonallelic noncomplementation between the two genes (i.e. the genes have overlapping function, or the gene products function in a common complex) and/or (2) the expression of one (or both) of the genes is altered in the other mutant background.
hoxb-4 is not expressed in the ventral body wall of hoxb-2 mutant homozygous embryos
To determine whether hoxb-4 was expressed normally in hoxb-2 mutant homozygotes, whole-mount hoxb-4 RNA in situ hybridization analysis was carried out on E9.5 and E12.5 hoxb-2 mutant and control embryos. At E9.5 no differences in hoxb-4 expression were seen between mutant and control embryos (data not shown). However, a significant change in the pattern of hoxb-4 expression in hoxb-2 mutants relative to controls was observed at E12.5. The hoxb-4 hybridization signal observed in cells extending from the neural tube to the ventral body wall in normal embryos (Fig. 8A-D) was absent in hoxb-2− homozy-gotes (Fig. 8E,F). It is not certain whether this change in expression contributes to the sternal defects observed in hoxb-2 mutant homozygous mice. However, our hoxb-4 mutant mice showed an embryonic phenotype not reported by Ramirez-Solis et al. (1993). Homozygous hoxb-4 mutant embryos at E13.5 and later, showed a failure of ventral body wall closure. This failure could underlie the formation of a split sternum in hoxb-2 and hoxb-4 mutant mice (Manley, Barrow and Capecchi, unpublished results). That is, closure of the ventral body wall may be a necessary prerequisite for bringing together the two separate sternal bands, which then allows proper fusion to occur. Thus, the lack of hoxb-4 expression in the ventral body wall in hoxb-2 mutant embryos could well contribute to the sternal defects observed in these mutant mice. In addition, we found an overall decrease in the levels of hoxb-4 expression in the cervical prevertebrae that could contribute to the cervical transformations in hoxb-2 mutant homozygotes. As was the case with hoxb-1 and hoxb-2, it is important to determine whether the changes in hoxb-4 expression in the hoxb-2 mutant homozygotes are due to an effect of the hoxb-2 mutation in cis or if hoxb-4 is a downstream target of hoxb-2 (i.e., a trans effect). Unfortunately, a cis/trans test which uses hoxb-2/hoxb-4 trans-heterozygotes cannot be carried out by the use of RNA in situ hybridization, because hoxb-4 mutant homozygotes generate abortive transcripts initiated from within the neomycin resistance gene that interfere with this analysis. We are currently trying to generate a hoxb-4 antibody, which will permit this analysis.
hoxb-2 is down regulated in rhombomere 6 in hoxb-4 mutant embryos
Since the sternal defects seen in hoxb-4 mutants could be due in part to disrupted expression of hoxb-2, it was also important to determine whether hoxb-2 expression is normal in a hoxb-4 mutant background. RNA in situ hybridization analysis uncovered a subtle but reproducible difference of hoxb-2 expression between hoxb-4 mutants and control embryos at E9.5 (Figs. 9A-C). In normal embryos, hoxb-2 is expressed strongly in r3-r5, slightly reduced in r6 and then at a basal level in the rest of the neural tube (Sham et al., 1993; and this work, Fig. 9A). In addition, it is expressed in neural crest emanating from r4 and to a lesser extent in neural crest cells migrating from r6 (Fig. 9A). In hoxb-4 mutant homozygotes, the hoxb-2 expression pattern was similar to controls except that, in r6, the level of expression was reproducibly lowered. In addition, we were not able to detect hoxb-2 expression in neural crest cells emanating from r6 (Figs 9B,C). Expression of hoxb-2 in hoxb-4 mutant homozygous embryos was also examined at E12.5 but no difference in expression pattern relative to control embryos was apparent.
Hoxb-2 expression was also examined in hoxb-2/hoxb-4 trans-heterozygotes in order to determine whether the changes in hoxb-2 expression in the hoxb-4 mutant embryos were due to a cis or trans effect. The same changes in hoxb-2 expression that were observed in the hoxb-4 mutant homozygous embryos were observed in the hoxb-2/hoxb-4 trans-heterozygous embryos (Fig. 9B-D), demonstrating that the hoxb-4 mutation affects hoxb-2 expression in cis.
DISCUSSION
Hoxb-2 mutant homozygotes have a complex phenotype that includes atrophy and paralysis of the facial muscles of expression, defects in the formation of the sternum and trans-formations of cervical vertebrae. The facial defects are similar to the clinical profile of humans born with Moebius Syndrome (Moebius, 1888). Both the human and mouse disorders are associated with a failure to form the motor component of the VIIth (facial) nerve. The facial defects common to hoxb-2 mutant homozygotes are also very similar to, but more severe than, those seen in hoxb-1 mutant mice (Goddard et al., 1996). In one of the hoxb-1 mutant alleles that we generated, hoxb-1neoB, the coding sequences in both the first and second exons were disrupted. This mutant allele, therefore, is likely to represent a null mutation. However, since hoxb-2 mutant homozygotes show a more severe facial phenotype than hoxb-1neoB homozygotes, it is likely that the loss of hoxb-2 protein contributes directly to the facial defects, above and beyond the effect of the neo insertion in the hoxb-2 gene on hoxb-1 expression.
The co-participation of hoxb-2 with hoxb-1 in specifying the formation of the motor component of the VIIth nerve is apparent from three perspectives. Particular features of the phenotype, such as the narrowness of the face, the receding lower lip, or the level of lethality at birth, that are attributable to deficiences of the VIIth nerve, are all more pronounced in hoxb-2 than in hoxb-1 mutants. A second argument suggesting direct interactions between hoxb-1 and hoxb-2 in specifying the motor component of the VIIth nerve centers around the phenotype of hoxb-1/hoxb- 2 trans-heterozygotes. These mice, in effect, have no functional copies of the hoxb-1 gene, but one functional copy of the hoxb-2 gene. These mice, as predicted if there is a direct contribution of hoxb-2 to the formation of the VIIth nerve, have less severe facial defects than those in hoxb-2 mutant homozygotes (which have no functional copies of either the hoxb-1 or hoxb-2 genes). Third, the variability in the expressivity of the facial defects is greater in hoxb-1 than in hoxb-2 mutant mice. For example, in hoxb-1 mutant homozygotes, it is not uncommon to observe complete paralysis in the facial muscles on one side of the face, but only mild paralysis on the contralateral side. In hoxb-2 mutant homozygotes, uniform paralysis of both sides of the face is observed. Variability of expressivity of a mutant phenotype often suggests stochastic compensation for the mutant gene product by another gene product. Conversely, reduction in the variability of expressivity suggests reduced capacity for this type of compensation. If hoxb-1 and hoxb-2 compensate for each other in specifying the VIIth nerve then, in the absence of both gene functions, we would anticipate reduction in the variability of expressivity in this defect.
Although hoxb-2 mutant homozygous mice show less variability in the expressivity of the facial defects relative to hoxb-1 mutants, they nevertheless still show some variability in this phenotype. For example, hoxb-2 mutant homozygotes show three classes of viability attributable to VIIth nerve defects: those that die within 24 hours of birth with normal sternums, those that die between 21 and 25 days of age, and those that have a normal lifespan. This variability in expressivity is probably due to compensation by yet a third Hox gene, hoxa-1. Indeed, mice doubly mutant for hoxa-1 and hoxb-1 show severe exacerbation of the hoxb-1 mutant phenotype (Rossel and Capecchi, unpublished results). Thus, at least three Hox genes appear to participate in the specification of the motor component of the VIIth nerve, hoxb-1, hoxb-2 and hoxa-1.
Besides its apparent direct involvement in VIIth nerve formation, the hoxb-2 mutation clearly disrupts expression of hoxb-1. Interestingly, hoxb-1 transcription and protein synthesis in the primitive streak of the gastrulating mouse embryo appears to be initiated normally in hoxb-2 mutants. However, the sustained, high level of hoxb-1 expression in rhombomere 4 is eliminated. This was an unexpected result because Marshall et al. (1994), have demonstrated that cis regulatory elements in close proximity to the hoxb-1 gene were sufficient to recapitulate normal hoxb-1 r4 expression in transgenic mice. Further, this expression was sensitive to autoregulation (Pöpperl et al., 1995). The hoxb-2 mutation is ∼15 kb from the 5 ′ end of hoxb-1, yet it is capable of eliminating hoxb-1 r4-specific expression. Cis/trans analysis with the use of hoxb-1/hoxb-2-trans- heterozygotes, showed that the hoxb-2 mutation affects hoxb-1 expression in cis. It has not been determined whether the neo insertion into the hoxb-2 homeobox directly disrupts a critical hoxb-1 cis regulatory element, or whether the neo gene itself interferes with hoxb-1 expression. We may be able to distinguish between these two possibilities by generating a new hoxb-2 mutant allele, in which the neo gene is flanked by bacterio-phage loxP/CRE recognition sites, to allow removal of the neo gene with CRE recombinase (Gu et al., 1993). In either case, the above results highlight the complexity of the regulatory networks used to control Hox gene expression within the complex and emphasize the possibility that mutations distal to a gene can have profound effects on its expression. Indeed, the interspersal and sharing of cis regulatory elements among Hox genes in the same linkage group may be a major constraint to the dispersal of Hox genes from their respective linkage groups during evolution (van der Hoeven et al., 1996).
Hoxb-2 mutant homozygotes are also characterized by severe defects in the formation of the sternum and by apparent homeotic transformations of the second cervical vertebra towards the identity of C1. Both of these defects are also seen in hoxb-4 mutant homozygotes (Ramirez-Solis et al., 1993). Ramirez-Solis et al. (1993) generated two mutant alleles of hoxb-4, one containing a neo insertion in the first exon, the second with a frameshift mutation in the second exon, which encodes the homeodomain. Only mice homozygous for the neo insertion in the first exon showed defects in the formation of the sternum. The two sternal bands are produced in these mutant mice, but they fail to come together and fuse. We have also generated mice with a neo insertion in the first exon of hoxb-4 (Manley, Barrow and Capecchi, unpublished results). The sternum defects in these mutant mice are similar to those reported by Ramirez-Solis et al. (1993) except that they are more severe (i.e., these mutant mice show 100% rather than 50% penetrance of the sternal defects in a BL/6 genetic background) such that all of the homozygous mutant neonates die within 24 hours of birth. We have noted that these mutant mice have defects in ventral body wall closure. It is reasonable to suggest that a failure in ventral body wall closure would prevent the two sternal bands from coming together and fusing. Interestingly, hoxb-4 is expressed during embryogenesis in ventral body wall cells that are presumed to contribute to the pericardial fold used for closure. In hoxb-2 mutant mice, this component of hoxb-4 expression is disrupted. Reciprocally, the neo insertion in the first exon of hoxb-4 appears to eliminate hoxb-2 expression in neural crest cells emanating from rhombomere 6. In the chick, neural crest cells from r6-r8 contribute to the pericardial fold used for ventral body wall closure (Männer et al., 1995; Waldo et al., 1996). Based on these observations, we propose as a working model that the sternal defects observed in hoxb-2 and hoxb-4 mutant mice are secondary to the defects in ventral body wall closure and that both hoxb-2 and hoxb-4 are required for this morphogenetic process. Thus, only mutations that affect the expression of both hoxb-2 and hoxb-4 show defects in ventral body wall closure and sternum formation. For example, mice with a frameshift mutation in the second exon of hoxb-4, which certainly compromises the ability of the hoxb-4 protein to bind to specific DNA sequences, do not have sternum defects (Ramirez-Solis et al., 1993). It is presumed that hoxb-2 expression is not affected in these mutant mice. On the contrary, mice homozygous for either the hoxb-2 mutation or the hoxb-4 mutation (neo insertion in the first exon) or mice trans-heterozygous for the hoxb-2 and hoxb-4 mutation have problems with ventral body wall closure and sternum defects. In these three latter genotypes, expression of both hoxb-2 and hoxb-4 was shown to be compromised. Since mice with hoxb-4 mutations that only affect hoxb-4 protein function do not have defects in the formation of the sternum, it could be argued that only hoxb-2 is critical for this function. However, this does not appear to be the case since hoxb-4 mutations that do affect the function of both hoxb-2 and hoxb-4 have more severe sternum defects than mice with the hoxb-2 mutation.
The transformation of the second cervical vertebra (axis) to a form resembling the first cervical vertebra (atlas) is similar in hoxb-2 mutant mice and in mice homozygous for either of the hoxb-4 mutant alleles (Ramirez-Solis et al., 1993, and herein).
The penetrance of this defect is also similar in hoxb-2 and hoxb-4 genotypes. We interpret these results to indicate that the C2 cervical transformation is primarily caused by a loss of hoxb-4 function. Consistent with this interpretation, a general reduction in hoxb-4 expression is observed in the prevertebrae of hoxb-2 mutant embryos relative to normal controls.
Hoxb-3 is located between hoxb-2 and hoxb-4 on the mouse chromosome. It is curious that the hoxb-2 mutation was found to affect the expression of hoxb-4 but not hoxb-3. The selective effects of the hoxb-2 mutation on neighboring Hox gene expression again emphasizes the intricacies in the distribution of the regulatory elements controlling neighboring Hox genes.
A majority of the defects observed in hoxb-2 mutant mice can be attributed to a combination of alteration in the expression of neighboring Hox genes, hoxb-1 or hoxb-4, and the loss of hoxb-2 function. However, there are two defects seen in hoxb-2 mutant mice not observed in either hoxb-1 or hoxb-4 mutant mice. First, hoxb-2 homozygous females show a marked loss of ability to raise pups regardless of the pups’ genotype. The cause of this defect is not known. Second, the orientation of the anterior arch of the atlas relative to the vertebral column is altered in hoxb-2 mutant homozygotes. These phenotypes thus seem to arise directly from the loss of hoxb-2 function.
Surprisingly, the hoxb-2 mutant phenotype does not overlap significantly with the hoxa-2 phenotype (Gendron-Maguire et al., 1993; Rijli et al., 1993). Specifically, hoxb-2 mutant homozygotes do not have defects in skeletal elements derived from the second branchial arch. This situation is similar to what is observed in mice individually mutant for hoxa-3 and hoxd-3 (Chisaka and Capecchi, 1991; Condie and Capecchi, 1993). Although mice homozygous for mutations in these individual genes did not share overlapping phenotypes, double mutant mice showed exacerbation of both hoxa-3 and hoxd-3 defects (Condie and Capecchi, 1994). Thus, although these two genes individually have separate functions, in the double mutants it is apparent that they also strongly interact. We anticipate that similar interactions between hoxa-2 and hoxb-2 will be evident in hoxa-2, hoxb-2 double mutants.
In summary, hoxb-2 mutant homozygous mice with a neo insertion in the homeobox have a complex phenotype. The complexity of the phenotype results, to a large extent, from alterations in subdomains of the hoxb-1 and hoxb-4 expression patterns. However, hoxb-2 itself also appears to be a participant with hoxb-1 and hoxb-4 in specification of the motor component of the VIIth nerve and in the formation of an intact sternum, respectively. Ironically, this participation would not have surfaced if the hoxb-2 mutation had not concomitantly disrupted the expression of neighboring Hox genes as well as its own function.
ACKNOWLEGMENTS
We wish to thank N. Manley for helpful discussion concerning the hoxb-4 phenotype and for providing critical reagents. Advice from J. Goddard and M. Rossel on facial nerve analysis is gratefully acknowledged. We would also like to thank C. Lenz, M. Allen, G. Peterson, E. Nakashima and S. Barnett for excellent technical assistance and L. Oswald for preparation of the manuscript. J. R. B. is supported by an NIH developmental training grant.