Dysgenesis of cephalic neural crest derivatives in Pax7^−/− mutant mice

The mouse Pax gene family consists of nine members (see Gruss and Walther, 1992 for review) which have been isolated on the basis of sequence homology to Drosophila segmentation genes (Bopp et al., 1986; Coté et al., 1987; Baumgartner et al., 1987). They all contain the paired box, a DNA binding domain of 128 amino acids, which is located close to the amino terminus. The paired box has been highly conserved during evolution and paired box containing genes have been cloned from different organisms such as mouse (Gruss and Walther, 1992), zebrafish (Krauss et al., 1991), chick (Goulding et al., 1993a) and man (Buri et al., 1989). In the mouse, Pax genes are expressed in very restricted domains (Deutsch et al., 1988; Dressler et al., 1990; Plachov et al., 1990; Jostes et al., 1991; Walther and Gruss, 1991) and with the exception of Pax1 and Pax9 (Deutsch et al., 1988, Neubüser et al., 1995; Wallin et al., 1993; Stapleton et al., 1993), they all exhibit discrete expression patterns in the developing and adult nervous system (Stoykova and Gruss, 1994).

The crucial role that Pax genes play in embryonic development, is documented by three mouse mutants with mutations in three corresponding Pax genes (for review see Hastie, 1991; Gruss and Walther, 1992; Chalepakis et al., 1993; Mansouri et al., 1994). Furthermore, Pax genes are also mutated in three human syndromes (Hill and Van Heyningen, 1992; Strachan and Reed, 1994; Sanyanusin et al., 1995) and Pax6 is mutated in Drosophila eyeless mutants (Quiring et al., 1994). These developmental defects described for the mouse mutants and human syndromes (for review see Mansouri et al., 1994; Strachan and Reed, 1994) correlate very well with the expression domains described for the respective Pax genes.

The mutation of Pax3 in Splotch mice (Epstein et al., 1991) leads to defects in neural crest derivatives and in limb muscle (Auerbach, 1954; Moase and Trasler, 1989; Franz, 1990). In man, Pax3 mutation leads to Waardenburg syndrome (Tassabehji et al., 1992, 1993; Baldwin et al., 1992; Morell et al., 1992). According to the genomic organisation and sequence similarities in the paired domain, Pax genes can be subdivided into subgroups which share common expression domains. Pax3 and Pax7 form such a paralogous group. Although the Pax3 function is revealed by the Splotch mutation, no spontaneous mutant is available which indicates the role of Pax7. Therefore, we wanted to mutate the Pax7 gene in mice. Using the approach of homologous recombination in embryonic stem cells, we generated mice with a null allele for Pax7. The heterozygous animals are normal and fertile, while homozygous offspring develop to term but die within three weeks after birth. The analysis of the Pax7 mutant mice revealed that in all mutant animals mainly facial skeletal structures are affected which could be related to a neural crest cell defect.

For the first targeting vector (Fig. 2B) an EcoRI fragment of 5.4 kb derived from a λ genomic clone obtained from a genomic library (Balb/c) was used. The PGKNeo plasmid kindly provided by P. Soriano (Soriano et al., 1991) was digested with XhoI and the 1.7 kb neo insert containing the PGK promoter and bovine growth hormone Poly(A) was ligated into the unique SalI site in the first exon of the paired box. The PGKneo and the Pax7 gene have the same transcriptional orientation. At the 5′ end of the construct, 1 kb upstream of the neomycin insertion the HSV-Tk (Mansour et al., 1988) was inserted into a blunt-ended KpnI site. The construct contains 1 kb on the 5′ end and 3.5 kb of homologous sequences on the 3′ end (Fig. 2B).

The second construct is identical to the one described for Balb/c DNA except that the DNA is of isogenic origin isolated from a λ clone obtained from a 129Sv genomic library (kindly provided by T. Doetschman). The third construct (Fig. 2C) was also derived from 129Sv origin. In this construct the amount of homology was extended to 15.3 kb on the 5′ end while at the 3′ end the homology is reduced to 1.1kb. In the third construct no HSV-Tk gene was included. All cloning procedures were done in bluescript KSII⁺ using standard protocols.

The construct containing β-galactosidase (data not shown) is identical to the construct in Fig. 2C with the exception of the insertion of β-galactosidase in front of the neomycin gene and in frame with Pax7 sequences.

ES cell line, D3, was kindly provided by R. Kemler and A. Gossler and R1 cells by A. Nagy. ES cells were cultured as described by Robertson (1987). 10⁷ ES cells were mixed with 25 μg of linearized targeting construct in a volume of 800μl PBS and electroporated using the Bio-Rad Gene Pulser (250 Volt, 500 μF, electrode distance 0.4 cm). ES cells were seeded 5 minutes later on 6×10 cm dishes containing embryonic fibroblasts which had been freshly treated with MitomycinC. Drug selection (G418: 300 μg/ml) or (G418 + gancyclovir 2 μM) was started 24 hours later, and after 8 days resistant colonies were picked and grown on 24-well plates. After a few days confluent colonies were trypsinized and three-quarters of each colony was frozen in 24-well plates at –70°C and the remaining quarter was grown on gelatinized 24-well plates until confluent. Genomic DNA was made from cells in the 24 wells and used for further analysis by genomic Southern blot. For analysis, genomic DNA was digested by EcoRI and after agarose gel electrophoresis and blotting, hybridization was performed using an external probe (Fig. 2E). Positive clones were confirmed by using internal (Fig. 2F) and neo probes (data not shown).

Embryos for histological analysis were fixed in Bouin’s solution. Embryos for in situ hybridization were fixed o.n. at 4°C in 4% paraformaldehyde in PBS. Bouin-fixed embryos were washed several times in 70% ethanol, dehydrated through graded alcohols, cleared in toluene, embedded in paraffin, sectioned at 12 μm and stained with hematoxylin-eosin. Embryos for whole-mount in situ hybridization were washed after fixation twice in PBT (PBS, 1% Tween 20) dehydrated through 25%, 50%, 75% and 100% methanol and kept at –20°C. Once enough homozygous embryos have been collected, embryos were hydrated through graded methanol/PBT and processed for whole-mount in situ hybridization as described by Wilkinson (1992). Radioactive in situ hybridization was performed as described previously (Stoykova and Gruss, 1994). Whole-mount β-galactosidase staining were performed as described by Allen et al. (1988).

Whole foetal skeletons were dissected and stained with alizarin red and alcian blue as described previously (Kessel and Gruss, 1991).

Pax7 is detected at 8.5 days post coitum (d.p.c.) in all brain vesicles and this expression is later (11.5 d.p.c.) retracted to the mesencephalon with an anterior boundary at the posterior commisure (Jostes et al., 1991; Stoykova and Gruss, 1994). In the neural tube, Pax7 is confined to the alar plate and Pax7 mRNA is first detected after the closure of the neural epithelium (Jostes et al., 1991). The onset of Pax3 expression in the neural tube is earlier than that of Pax7 and starts before closure of the neuroepithelium (Goulding et al., 1991).

In the somites, Pax7 is first detected in the dermomyotome and later in development is confined to the intercostal muscle (Jostes et al., 1991; Mansouri and Gruss, unpublished data) whereas Pax3 is already down regulated upon activation of the myogenic markers (Goulding et al., 1991, 1994; Williams and Ordahl, 1994).

Pax7 has been previously detected in the nasal pit and nasal neuroepithelium (Jostes et al., 1991). We performed comparative in situ analysis of the expression of the paralogous Pax7 and Pax3 genes on adjacent sagittal and transverse sections of the head of 13.5 d.p.c. embryos. The results are shown in Fig. 1 The expression of Pax7 is confined to the medial region of the frontonasal mass (Fig. 1E,H, white arrow) including the nasal capsule, while Pax3 is detected in both the medial and lateral parts of this region (Fig. 1F,I, white arrowhead). Furthermore, Pax3 is expressed in the mandible (Fig. 1C,I).

We wanted to follow the expression of Pax7 in early developmental stages in the head region in order to clarify its presumptive involvement in the patterning of the head neural crest. This was done by inserting the gene for β-galactosidase in front of the neomycin gene (shown in the third construct in Fig. 2C, see legend for details). The β-galactosidase gene (lacZ) is inserted in-frame to the Pax7 sequences so that a fusion to the first exon of the paired box was generated (data not shown). Upon electroporation in ES cells, a targeted clone was obtained and the new Pax7 mutation containing lacZ was introduced into the mouse germline.

In animals heterozygous for this mutation, using wholemount staining for β-galactosidase, we could detect expression domains of Pax7 during embryogenesis which have not been described previously (Jostes et al., 1991). Pax7 is expressed in discrete domains in the rhombencephalon. As shown in Fig. 3, in 8.5 and 9.5 d.p.c. embryos, Pax7 is detected in presumptive rhombomere 1, 3 and 5. This expression pattern is first detected at 8.5 d.p.c. (Fig. 3B) and persists through 9.5 d.p.c. (Fig. 3C,D) until 10 d.p.c. (data not shown) and was confirmed by whole-mount in situ hybridization (data not shown). Furthermore Pax7 has been found to be expressed in streams of cells at the level of the forebrain around the optic vesicle (Fig. 3A, see small arrow), at the level of the midbrain (Fig. 3A, see arrowhead) and rhombencephalon (Fig. 3B, arrow and Fig. 3C,D). These neural crest cells migrate rostrally and laterally as shown in Fig. 3 for 8.5 and 9.5 d.p.c. embryos. Some cells migrate caudally as shown in Fig. 3D for rhombomere 5 in a 9.5 d.p.c. embryo (see arrow). These migratory cells are cephalic neural crest cells (Serbedzija et al., 1992; OsumiYamashita et al., 1994; for review see Bronner-Fraser, 1993). Some of these β-galactosidase-positive cells also migrate into the first and second branchial arches (data not shown).

To inactivate the Pax7 gene in ES cells the neomycin gene was inserted into the first exon of the paired box (the paired box is encoded by three exons) (Fig. 2). This was expected to abolish the DNA-binding activity of the paired domain as has been demonstrated by Chalepakis et al. (1991, 1994). We used replacement targeting vectors (see legend to Fig. 2). Two different ES cell lines, D3 and R1, were used for electroporation and from each cell line two targeted clones have given germline transmission. Chimeras were generated by blastocyst injection and morulae aggregation. Heterozygous Pax7 mice appeared healthy and fertile. RT-PCR performed on RNA from 12.5-day embryos, using primers from the first and the second exon of the paired box (see legend to Fig. 2), revealed no Pax7 RNA in Pax7^−/− mice (data not shown) which demonstrates that null mutant mice were probably generated.

Pax7^–/– mice develop to term and appear normal. They represent 23% of the offspring. After a few days, homozygous embryos can already be recognized by their growth retardation and frequent lethality. Most of the Pax7^–/– mice (97%) die within 3 weeks after birth, without an obvious reason for premature death. The few Pax7^–/– animals that survived to adulthood appeared very unhealthy. They were usually killed and examined and we found that many of them exhibited dilatations in the small intestine and appendix.

New born animals were histologically analyzed. Examinations of eosin-hematoxylin stainined sagittal and transverse sections of whole embryos revealed no obvious defect. Domains where Pax7 has been found to be strongly expressed, namely the mesencephalon, hindbrain, neural tube and adult brain (Jostes et al., 1991; Stoykova et al., 1994), appear morphologically normal. No obvious abnormality could be detected in neuronal derivatives of cephalic neural crest.

Homozygous animals were also analyzed by whole mount in situ hybridization using three myogenic markers (Myf5, MyoD and Myogenin; Weintraub, 1993) to examine the dermomyotome and the myotome where Pax7 has also been shown to be expressed. Analysis was done using embryos between 9 and 11.5 d.p.c. Examination of whole-mount staining for the RNA of these myogenic markers revealed no obvious changes in their pattern of expression in mutant embryos (data not shown). This indicates that the morphology of these somitic structures is not affected in Pax7^–/– mice. At later stages of development and in new born mice histological analysis revealed no obvious changes in the intercostal muscles.

However, skeletal preparations from new born Pax7^–/–?animals analyzed using alizarin red and alcian blue staining (shown in Fig. 4) revealed that Pax7^–/– mice have reduced maxilla. The maxilla is shortened in anterior/posterior direction (Fig. 4, arrowhead). This is more obvious in a lateral view of the same specimen (Fig. 4C,D). This skeletal phenotype has been found in all analyzed homozygotes. Transverse sections of the head of new born animals at the level of the nose also revealed some phenotypic changes. As shown in Fig. 5B, the tubules of serous glands, which are associated with the lateral wall of the middle meatus (see arrowhead) and those associated with the nasal septum (see arrow) are very reduced in number as compared to wild-type animals.

In further transverse sections, the inferior lateral part of the nasal capsule was not formed in Pax7^–/– mice. This missing structure is normally associated with the cartilage which lines the anterior nasal cavity. Therefore, homozygous animals appear to have a pointed snout which distinguishes them phenotypically from the wild-type and heterozygous animals. This is shown in Fig. 6B (see arrow in Fig. 6A). Analysis of embryos at 14.5 d.p.c. revealed that this part of the nasal capsule is not formed in Pax7^–/–.

The analysis of Pax7^–/– mice revealed defects in the maxilla, nasal tubules of serous glands and in the anterior part of the nasal capsule. The neural tube, the brain and the skeletal muscle are not obviously affected.

In the developing neural tube, Pax7 and Pax3 expression overlaps in the alar plate. At this level, Pax3 is already detected at the neural plate stage and precedes that of Pax7 which starts after neural tube closure (Goulding et al., 1991; Jostes et al., 1991). Therefore, Pax3 could complement the function of Pax7 in the neural tube of Pax7^–/– mice. In contrast, normal Pax7 expression does not compensate for all the functions of Pax3 in the neural tube of Splotch mice since they exhibit an open posterior neuropore. However, this latter phenotype involves only the most dorsal neural tube where Pax7 is not expressed and thus cannot compensate for the absence of Pax3. Alternatively, Pax7 may not rescue the lack of Pax3 because it is expressed slightly later.

The dermomyotome of Pax7^−/− mice is normal morpho-logically, as demonstrated by the expression of myogenic markers. This clearly indicates that the Pax7 mutation does not affect this somitic derivative. In Splotch mice, the development of the limb muscle and the associated shoulder is affected, the dermomyotome is disorganized and the limbs devoid of Pax3expressing cells (Franz et al., 1993; Bober et al., 1994; Goulding et al., 1994). Again Pax3 and Pax7 share extensive overlapping expression domains in the dermomyotome, and Pax7 cannot substitute for the function of Pax3 in the limb formation of Splotch mice. This may be the consequence of the slightly different dynamics of expression of both genes in the dermomyotome. In the most lateral part of the dermomyotome, which gives rise to the precursors of the limb muscles (Ordahl and Le Douarin, 1992), Pax3 is more strongly expressed, while Pax7 is more prominent in the medial part. Therefore partial redundancy may explain the lack of phenotype in the skeletal muscles of Pax7^−/− mice and in the axial muscles of the Pax3^−/− mice. However, while Pax3 expression is down regulated upon activation of the myogenic markers, Pax7 is still strongly expressed (Jostes et al., 1991; Goulding et al., 1991, Williams and Ordahl, 1994). The function of Pax7 in the myotome still remains unclear.

Partial redundancy has been reported recently for a number of genes. Only the generation of compound mutations revealed the phenotype, as anticipated by expression studies. For the MyoD family, the inactivation of MyoD, Myf5 and myogenin individually does not affect muscle development (Braun et al., 1992; Rudnicki et al., 1992). However, when mice carry null alleles for both the MyoD and myf5 genes they have no muscle, and myogenin is not transcribed (Rudnicki et al., 1993). Redundancy has also been described for Hox (Kostic and Capecchi, 1994; Horan et al., 1995) and retinoic acid receptor genes (Lohnes et al., 1994; Mendelsohn et al., 1994). Recently functional redundancy for engrailed 1 (En-1) and engrailed 2 (En-2) in the mouse has been very elegantly demonstrated by expressing En-2 under the promoter of En-1 by using the knock-in technique to rescue En-1 mutants (Hanks et al., 1995). Accordingly, we propose that the function of Pax3 and Pax7 may be in part redundant.

The structures affected by the inactivation of Pax7 are neural crest cell derivatives. This has been demonstrated in birds (Le Lievre, 1978; Le Douarin, 1982; Couly and Le Douarin, 1985; Couly et al., 1993; Noden, 1978, 1983), mouse (Serbedzija et al., 1992; Osumi et al., 1994) and rat (Tan and Morriss-Kay, 1986; Matsuo et al., 1993). Does functional redundancy also occur in the cephalic neural crest in Pax7^−/− mice? Pax3 is expressed in neural crest, and Splotch mice exhibit defects in neural crest-derived structures such as dorsal root and cranial ganglia and Schwann cells (Moase and Trasler, 1989; Franz, 1990; Tremblay et al., 1995). These defects cannot be compensated for by Pax7, as it does not appear to be expressed in caudal neural crest (Mansouri and Gruss, unpublished data). The origin of the nasal serous glands is not documented.

In the hindbrain, Pax7 expression is confined to rhombomeres 1, 3 and 5. Pax3 is found in all of the hindbrain except at the most rostral level, and in rhombomere 5 (the area facing the otic vesicle), where the expression is lower (Fig. 7, see arrow and arrowhead). The neural crest from the most rostral part of the rhombencephalon also contributes to the formation of the maxilla and the frontonasal mass (Lumsden et al., 1991; Osumi-Yamashita et al., (1994). Therefore this defect observed in Pax7^−/− mice cannot be overcome by the expression of Pax3, as it is is not present in the most rostral hindbrain. In other words, as in the dermomyotome and possibly the neural tube, in the hindbrain Pax3 and Pax7 exhibit different expression patterns which could lead to the observed phenotypes. In this context, the possible significance of the lower expression of Pax3 in rhombomere 5 remains unclear. Taken together these results strongly suggest that the development of a sub population of cephalic neural crest cells is regulated by Pax7. Further experiments are needed to elucidate the respective functions of Pax3 and Pax7 during this process. Both genes may, directly or indirectly, influence the specification of these cells.

We would like to thank Jens Krull for excellent technical assistance, Hans-Peter Geithe for oligonucleotide synthesis and Ralf Altschäffel for excellent photographs. We also thank Thomas Braun for the Myf5, MyoD and Myogenin probes for in situ hybridization and Rainer Libal and the animal house crew for constant help. We would also like to thank P. Bonaldo, M. Kessel, R. Fritsch, G. Proetzel, Y. Yokota and especially M. Hallonet for valuable discussions and David N. Cooper, Luc St-Onge and Eduard N. Stewart for critically reading the manuscript. This work was supported by the Max-Planck Gesellschaft.