Pax genes in development

The mechanisms that control embryonic development are highly conserved among different organisms such as Drosophila, mouse and nematodes and has led to the isolation of many developmental control genes, which are shared between different species including mammals.

Based on sequence homology to the Drosophila segmentation genes (Bopp et al., 1986; Frigerio et al., 1986; Coté et al., 1987; Baumgartner et al., 1987), a family of paired box-con-taining genes has been isolated in the mouse (Deutsch et al., 1988; Dressier et al., 1988; Plachov et al., 1990; Jostes et al., 1991; Asano and Gruss, 1992; Wallin et al., 1993; Stapleton et al., 1993; for review see Deutsch and Gruss, 1991; Fritsch and Gruss, 1993; Chalepakis et al., 1993; Gruss and Walther, 1992; Noll, 1993). The paired box has been highly conserved during evolution and paired box encoding genes have been found in different organisms such as zebrafish (Krauss et al., 1991), chicken (Goulding et al., 1993a) and man (Burri et al., 1989).

The Pax gene family now consists of nine members, referred to as Pax-1 to Pax-9. Unlike the Hox genes they are not clustered and are localised on different loci (Walther et al., 1991; Wallin et al., 1993; Stapleton et al., 1993).

Like several other developmental control genes, Pax proteins act as transcription factors, since they display sequence-specific DNA-binding activity (Chalepakis et al., 1991; Treisman et al., 1991; Dressier and Douglas, 1992; Zannini et al., 1992; Fickenscher et al., 1993; Czerny et al., 1993; Epstein et al., 1994) and can activate transcription. This is also in agreement with the nuclear localisation of Pax-1, Pax-2, Pax-5 and Pax-6 (Dressier and Douglas, 1992; Adams et al., 1992; R. Fritsch and P. Gruss, unpublished data).

Three conserved motifs exist in the Pax proteins (Fig. 1). A DNA-binding domain of 128 amino acids, the paired domain, which is located close to the amino terminus of the Pax proteins. A second DNA-binding domain of 61 amino acids, the paired-type homeodomain, is localised at the 3’ end of the paired box of Pax-3, Pax-4, Pax-6 and Pax-7. In contrast, Pax-2, Pax-5 and Pax-8 only contain the first helix of this domain, thus missing the whole helix-loop-helix part of the second DNA-binding motif, while Pax-1 and Pax-9 are missing the whole homeobox sequences. In addition (except for Pax-4 and Pax-6), all the Pax proteins share a conserved octapeptide, which is localised between the paired- and the homeodomains and whose function remains unknown. According to the genomic organisation and the paired domain sequence simi-larities, Pax genes can be subdivided into 5 subgroups. The paralogous genes within a subgroup share common intron-exon boundaries, similar protein structure and related expression pattern during development (Pax-1 and Pax-9’, Pax-2, Pax-5 and Pax-8’, Pax-3 and Pax-7).

In this article we will summarise the expression patterns of Pax genes during mouse development and correlate the expression domains with the phenotypes observed in some mouse mutants and in human diseases, where specific Pax genes are mutated. Finally we will discuss what we can learn from the developmental defects in the Pax mutants about the function of the Pax genes during development.

The potential importance of the Pax genes is reflected by their expression pattern during development. A common feature of all Pax genes, except for Pax-1 and Pax-9, is that they exhibit spatially and temporally restricted expression patterns in the developing nervous system. This strongly suggests that they may play a crucial role in the regionalisation of the neural tube and brain. (Fig. 2).

The first group of the early genes, which contain both the paired- and the homeodomain (Pax-3, Pax-6 and Pax-7) start to be expressed around day 8.5 pc., before the onset of cellular differentiation. Up to day 10.5 pc. their expression domains are confined to the ventricular zone of the entire developing CNS (epichordal and prechordal part). Thereafter, the expression of Pax-3 and Pax-7 is retracted from the telencephalon, while in all stages Pax-6 is not expressed in the mesencephalic roof and maintains a caudal limit of expression at the level of the posterior commissure (Walther and Gruss, 1991; Stoykova and Gruss, 1994; Goulding et al., 1991; Jostes et al., 1991).

The group of the late genes, which harber no paired-type homeodomain (Pax-2, Pax-5 and Pax-8) are first detected around day 9.5-10.5 pc. Their expression domains are confined only to postmitotic cells in the epichordal part of the neural tube with a rostral limit of expression at the midbrain-hindbrain boundary. It should be noted, however, that Pax-5 transcripts are detected in the posterior mesencephalic tegmentum (Asano and Gruss, 1992; Adams et al., 1992) and is strongly expressed on both sides of the fovea isthmi. The fact that the expression of all Pax genes in the epichordal part of the neural tube extends up to the midbrain hindbrain boundary, thus surpass-ing the most rostral expression limit of the Hox genes (Keynes et al., 1990), suggests a role for Pax genes in the regionalisation of the most rostral part of the hindbrain.

Recent evidence support the idea that Pax genes are involved in the specification of the midbrain-hindbrain boundary: injection of an antibody against Pax(zf-b) protein (probable homolog of Pax-2 in the mouse) results in a failure of the normal isthmus development (Krauss et al., 1992). Dis-ruption of Pax-2, Pax-5 and Pax-8 by homologous recombi-nation in embryonic stem cells will help to elucidate their role in the regionalisation of the midbrain-hindbrain area. Further-more their interaction with other genes already known to play such a role (En-1, En-2 and Wnt-1) (Davis et al., 1988; Davidson et al., 1988; Wilkenson et al., 1987) can also be studied.

Pax genes also show restricted expression domains along the dorso-ventral (D-V) axis of the developing neural tube. The specific location of the cells along the D-V axis appears to be a prerequisite for the commitment of the precursor cells to a distinct phenotype: motor neurons (in basal plate), commis-sural neurons and neurons involved in sensory pathway (in alar plate), interneurons and preganglionic sympathetic and parasympathetic neurons (intermediate plate). Inductive signals from the underlying notochord and floor plate determine the dorsoventral regionalisation of the spinal cord (van Straaten et al., 1988; Placzek et al., 1990). During devel-opment, the paralogbus genes Pax-3 and Pax-7 are expressed in the ventricular zone of the alar subdivision of the neural tube. Only the Pax-3 expression domain includes the roof plate, the neural crest and its derivatives (Goulding et al., 1991; Jostes et al., 1991).

In contrast, Pax-6 transcripts are detected in the ventricular zone of the intermediate and basal plate, partially overlapping with Pax-3 and Pax-7 at the upper region (Walther and Gruss, 1991). As shown in notochord transplantation experiments in chick the restriction of the Pax-6 and Pax-3 expression domain along the D-V axis is under the influence of the notochord (Goulding et al., 1993a). This support the notion that the expression of these genes is related to the D-V patterning of the developing spinal cord.

In the developing neural tube the paralogous genes Pax-2, Pax-5 and Pax-8 are expressed in two longitudinal columns of the intermediate gray on both sides of sulcus limitans (Nornes et al., 1990; Plachov et al., 1990; Asano and Gruss, 1992; Adams et al., 1992). Expression of the Pax-2 homologue in zebrafish is consistent with a role in the fate determination of the progenitor cells to the specific phenotype of the interneu-ron (Püschel et al., 1992; Mikkola et al., 1992).

Recent neuroanatomical and molecular analyses have shown that the segmental organisation of the hindbrain is determined by the differential expression of combinations of Hox genes (Lumsden, 1990; McG₁nnis and Krum1auf, 1992). In contrast, the molecular mechanisms that control the regionalisation and differentiation of the embryonic forebrain are still largely unknown. A large number of genes have been identified that are expressed in spatially restricted domains that outline lon-gitudinal and transverse domains in the developing forebrain. Consistent with the restricted expression pattern of numerous genes in the CNS, recently, new models for the neuromeric organisation of the forebrain have been proposed (Puelles and Rubenstein, 1993; Figdor and Stem, 1993).

The expression domains of several Pax genes during devel-opment were found to respect anatomical boundaries that relate to former neuromeric territories. Around day 10.5 pc, thus before morphological appearence of segmentation in the devel-oping diencephalon, the main expression domain of Pax-6 is confined to the ventral thalamus (the former parencephalon posterior), while Pax-3 and Pax-7 transcripts are detected in the epithalamus and entire pretectum (former synencephalon). Further, in a comparative in situ analysis in midgestation and adult brain, a similar transcript distribution along the anterior-posterior axis for Pax genes has been observed (Stoykova and Gruss, 1994). Similar to the situation in the embryonic brain (see Fig. 2), in the adult brain transcripts of six Pax genes are detected in distinct isthmic nuclei; in the midbrain three Pax genes (Pax-5, Pax-6 and Pax-7) are expressed, while in the telencephalon and anterior diencephalon only Pax-6 is detected. In many cases, a correlation exist between the expression of different Pax genes in defined brain structures and the respective domains of their origin in the embryonic brain. Therefore, the products of Pax genes may not only be important for the regionalisation of the developing nervous system, but also involved in the differentiation and/or mainte-nance of specific structures in the mature brain. The identifi-cation of Pax genes involved in specific mouse mutants and their corresponding human syndromes support this idea.

Some Pax genes show expression consistent with a role in the inductive process underlying the formation of sensory organs. The lens and cornea of the eye are formed by the inter-action of different cell layers. When the optic vesicle contacts a specific region of the head ectoderm, the ectoderm thickens into the lens placodes. All structures of developing eye (optic placode, optic vesicle, sulcus, retina, lens and cornea) express Pax-6; thus inductive and responsive tissues express Pax-6. (Walther and Gruss, 1991). Pax-6 is also expressed in the developing olfactory epithelium (Walther and Gruss, 1991).

The otic vesicle is derived from the otic placode, which invaginates towards the neural tube. Pax-2 has been detected in the otic vesicle of the mouse (Nornes et al., 1990) and zebrafish (Krauss et al., 1991; Püschel et al., 1992). Pax(zf-b) has been shown to exhibit the highest level of expression close to the neural tube surface. Therefore this gene and, in homology, Pax-2 have been suggested to be involved in the formation of the inner ear, in response to some secreted signal, from the neural tube (Püschel et al., 1992).

Some Pax genes are also expressed in the segmented mesoderm (Pax-1, Pax-9, Pax-3 and Pax-7;Deutsch et al., 1988; Dietrich et al., 1993; Wallin et al., 1994; Goulding et al., 1991; Williams and Ordahl, 1994; Jostes et al., 1991) and in the developing excretory system (Pax-2 and Pax-8;Dressier et al., 1990; Plachov et al., 1990).

Pax-1 expression is seen at day 8.5 pc. in the ventral half of the newly formed somite and mRNA is later detected in the sclerotome cells migrating to surround the notochord. This domain of expression later gives rise to a segmented structure called the intervertebral disc anlagen (Deutsch et al., 1988; Dietrich et al., 1993; Wallin et al., 1994).

During gastrulation Pax-3 is expressed in the primitive streak (Goulding et al., 1993a). At day 8.0 pc Pax-3 is detected in the undifferentiated somite before it looses its epithelial morphology. The expression of Pax-3 is later confined to the dorsal half of the somite. At day 9 pc Pax-3 and Pax-7 are expressed in the dermomyotome (Jostes et al., 1991). Here Pax-3 is expressed at higher levels in the dorsolateral than the dorsomedial half. At the early limb bud stage, Pax-3 expression is extended from the lateral edge of the dermomy-otome into the limb. Transplantation experiments in the chick-quail marking system (Williams and Ordahl, 1994) demon-strate that Pax-3 labels the migratory muscle precursor cells entering the limb to form skeletal muscle in this part of the body, suggesting that Pax-3 is probably necessary for the migratory process of these cells. (Williams and Ordahl, 1994; Bober et al., 1994). At later stages only Pax-7 is detected in the intercostal muscle, while Pax-3 is downregulated.

The paralogous genes Pax-2 and Pax-8 are expressed also in the developing kidney (Dressier et al., 1990; Plachov et al., 1990), while Pax-5 is detected in B lymphocytes and testis (Adams et al., 1992).

The Pax-2 mRNA is detected during kidney development namely in the pro- and mesonephros tubuli (Dressier et al., 1990). After induction of the metanephros Pax-2 protein is present in the mesenchymal condensations and its early epithe-lial derivatives. With the further differentiation of the epithe-lium, Pax-2 is downregulated. Therefore it has been suggested (Dressier et al., 1990) that Pax-2 may play a role in the con-version of the mesenchyme to epithelium during kidney organogenesis. This is also supported by experiments showing Pax-2 expression in Wilms tumors (Dressier and Douglas, 1992).

Direct evidence for the crucial role of the Pax genes during development has been demonstrated by the correlation of mouse developmental mutants with mutations in certain Pax genes. Three Pax genes have been correlated with mouse mutants (for review see Hastie, 1991; Chalepakis et al., 1993; Tremblay and Gruss, 1994; Stuart et al., 1994) and two with corresponding human syndromes (for review see Hill and Van Heyningen, 1992; Strachan and Reed, 1994).

Pax-1 has been found to be mutated in undulated (un) mice (Balling et al., 1988), a recessive mutation that exhibits skeletal abnormalities in the vertebral column and sternum (Griineberg, 1950, 1954). Their phenotype is clearly correlated with the expression pattern of Pax-1 in sclerotome (Deutsch et al., 1988; Koseki et al., 1993; Wallin et al., 1994; Dietrich et al., 1993). A point mutation (change from Gly to Ser) in the paired domain of Pax-1 results in a reduction of DNA-binding affinity (Chalepakis et al., 1991), interfering with the transcriptional activity of the Pax-1 protein. Two other undulated alleles are known: undulated extensive (un^ex) and undulated shorttail (un^s), where Pax-1 is also involved (Balling et al., 1992). In un^ex it has been shown that the RNA level of Pax-1 is highly reduced (Balling et al., 1992) and that a deletion of at least 28.2 kb is located within the 3 ′ end of the gene, which eliminates the terminal exon and the Poly A signal (Dietrich and Gruss, 1994, unpublished). In un^s a large chromosomal deletion has been detected (Balling et al., 1992) of at least 48.3 kb, which includes the Pax-1 gene thus removing the Pax-1 locus and 38 kb flanking sequences (Dietrich and Gruss, 1994, unpublished; Wallin et al., 1994). un^s exhibits the most severe phenotype and is semidominant, while un and un^ex are recessive.

Pax-3 has been shown to be mutated in the mouse mutant Splotch (Sp, spontaneous mutation; splice site defect; exon 4 skipped and truncation within paired domain; Russell, 1947; Epstein et al., 1991a,b, 1993; Goulding et al., 1993b). In humans, Pax-3 mutations have been correlated with Waarden-burg syndrome (Tassabehji et al., 1992, 1993; Baldwin et al., 1992; Morell et al., 1992). The Pax-3 mutation in Splotch mice is semidominant. The heterozygous mice have white spotting of the abdomen, tail and feet, probably due to the absence of melanocyte migration to these regions (Auerbach, 1954). The homozygous mice die at midgestation stage with severe defects in closure of the neural tube (exencephaly, meningocele and spina bifida) and multiple defects in structures of neural crest origin (schwann cells, spinal ganglia and melanocytes; Auerbach, 1954; Franz, 1989, 1990; Moase and Trasler, 1989). It is of interest to notice that neural tube defect and deficiency in neural crest-derived structures are independent from each other. In addition, the development of the muscle in the limb (fore- and hindlimb) and in the associated shoulder is impaired (Franz et al., 1993; Bober et al., 1994; Goulding et al., 1994). However the muscles of the neck, of the head and of the body wall are not affected (Franz et al., 1993) possibly due to the specific expression of the paralogous gene Pax-7 in this region (Jostes et al., 1991). Detailed analysis revealed that the der-momyotome is disorganised and the limbs are devoid of Pax-3 expressing cells (Franz, et al., 1993; Bober et al., 1994; Goulding et al., 1994). The reduction of axial muscles, which is more prominent in caudal segments, also follows the ros-trocaudal gradient observed in the neural crest defects.

The described defects in Splotch mice correlate very well with the expression domains of Pax-3 in neural crest cells, in the dermomyotome and in the neural tube. However, it is not clear whether the neural tube defect is due specifically to the absence of the Pax-3 protein or not, since other mouse mutants like curly tail also show this phenotype (Copp et al., 1988). Five other Splotch alleles have been described: Sp^2H (X-irra-diation-induced mutant of 32 bp deletion within exon 5; trun-cation within homeodomain; Epstein et al., 1991a), Sp^r (Splotch retarded; X-irradiation-induced mutation with a deletion of Pax-3 and several other genes; Epstein et al., 1991b), Sp^d (Splotch delayed; point mutation within the paired box: Gly to Arg; Vogan et al., 1993) and Sp^4H (Splotch 4H, X-irradiation-induced mutant with a deletion of Pax-3;Goulding et al., 1993b) and SplotchlH (SplH, undefined X-irradiation-induced mutant). The phenotype is almost identical in most splotch alleles known so far. The most severe phenotype is expressed by Sp^r, which die before implantation. (Evans et al., 1988), while Sp^d exhibit only spina bifida and die after birth (Dickie, 1964).

Waardenburg syndrome is an autosomal dominant disorder with incomplete penetrance (McKusick, 1992). It is clinically and genetically heterogenous. The Waardenburg syndrome consists of numerous defects of neural crest derivatives: deafness, pigmentation deficiency (heterochromia irides; white forelock and white skin patches) and lateral displacement of the inner comer of the eye. Based on the presence or absence of the lateral displacement of the inner comer of the eye the Waardenburg syndrome is classified into two categories: type 1 (WS1) and type 2 (WS2). In both WS1 and WS2 Pax-3 has been found to be mutated (Tassebehji et al., 1992, 1993). The cause of the deafness in Waardenburg syndrome patients is not clear. However, it has been often observed that frequently pig-mentation defects are associated with deafness in mice and man (Steel and Smith, 1992).

In the mouse small eye (sey) mutant and its corresponding human syndrome aniridia (Shaw et al., 1960), Pax-6 has been found to be mutated. The small eye homozygous mutant is recognised by complete loss of the eyes and the nasal cavities do not develop (Pritchard and Clayton, 1974). Like Splotch, small eye display a semidominant trait and heterozygous mice are characterized by smaller eyes. A number of different Sey alleles have been described (Sey, spontaneous mutation; Sey^H, X-irradiation mutant; Sey^Neu, ethylnitrosourea induced; Sey^Dey, spontaneous mutation). The most severe phenotype is exhibited by Sey^H where the homozygous animals die shortly after implantation and have a large chromosomal deletion including Pax-6 and possibly other genes (Hill et al., 1991). In the spontaneouly occurring Sey there is an in-frame stop codon before the homeobox of Pax-6 (Hill et al., 1991), while in Sey^Dey a deletion of the Pax-6 gene (Glaser et al., 1990) has been found. In Sey^Neu a splice defect in the Pax-6 gene results in a truncated protein lacking the last 115 amino acids but con-taining both the paired- and the homeobox (Hill et al., 1991). Most of the described mutations result in the same phenotype, suggesting a loss of function mutation.

Also in the rat there is a small eye mutation, where the Pax-6 gene is involved (Matsuo et al., 1993; Fujiwara et al., 1994) and the phenotype is similar to that described in the mouse.

Pax-6 mutations in aniridia range from single base pair mutations to large deletions (Ton et al., 1991; Hill et al., 1991; Jordan et al., 1992; Glaser et al., 1992). The patients suffer from complete or partial absence of the iris. Cornea, lens, retina and optic nerve are also affected. Recently, mutations at the Pax-6 locus in man have been described for Peter’s anomaly (heterogeneous anterior segment malformations; Hanson et al., 1994).

All the phenotypes described for the small eye in mouse and rat and aniridia in human correlate very well with the expression of the Pax-6 gene in the affected structures.

Finally, in human tumours PAX-3 has been found to be mutated (Barr et al., 1993). In alveolar rhabdomyosarcoma cell lines the PAX-3 gene is rearranged and translocated to a region of chromosome 13, where its 5 ′ end is fused to a gene of the forkhead family termed FKHR (Galili et al., 1993) or ALV (Shapiro et al., 1993). Recently also PAX-7 has been found to be rearranged in a variant of alveolar rhabdomyosarcoma. PAX-7 is involved in a translocation of t(1; 13)(p36; ql4) and the 5 ′ end of the gene is fused to the 3 ′ end of the FKHR region (Davis et al., 1994). This indicates that deregulation of Pax gene expression may be involved in tumorgenesis, which supports the previous observation that overexpression of Pax genes transforms fibroblasts and the resulting foci develop into tumours in nude mice (Maulbecker and Gruss, 1993).

The availability of Pax developmental mutants is a valuable tool to analyse the function of the Pax genes. The phenotypes characterized in Splotch, undulated, small eye and in the human syndromes Waardenburg and aniridia correlate very well with their expression pattern and highlights the fact that Pax genes are involved in the control of embryonic develop-ment in different cell types.

The described mutations Splotch and small eye seem to be loss of function mutations, since different mutations exhibit the same phenotype in most cases.

From all the mutants, except undulated, the Pax protein threshold seems to play an important role in Pax gene function. This is documented by the semidominance of small eye and Splotch phenotypes in the mouse and dominance of Warden-burg and aniridia diseases in man. Another example is given by the different undulated alleles. The phenotypes ranges from mild skeletal abnormalities to complete lack of vertebral bodies in the lumbar region and enhances the idea of phenotypic dependence on the Pax protein concentration (Hill and Hanson, 1992). DNA binding studies using undulated (Chalepakis et al., 1991), splotch or Wardenburg mutations (Chalepakis et al., 1994) could clearly show that these mutations affect DNA binding of the Pax proteins. In Waardenburg Brazil, where the invariant Pro residue is substituted by a Leu residue in the paired box, the DNA binding is even completely abolished (Chalepakis et al., 1994), suggesting that this invariant position in the paired domain is essential for DNA binding of the paired domain. The further biochemical analysis of such mutations will help to map the transcription activation domains essential for Pax function.

At the cellular level, Pax genes seem to be expressed in specific cell types: Pax-3, Pax-7 and Pax-6 are detected in mitotically active cells in the ventricular zone of developing neural tube and could play a role in cell growth. Pax-2 is specifically related to the differentiating spinal cord interneu-rons (Mikkola et al., 1992). In addition Pax-3 is possibly involved in the emigration of neural crest cells, which is delayed in Splotch mice (Moase and Trasler, 1990). It has been suggested that N-CAM is involved in this latter process (Moase and Trasler, 1991), however, it is not known if N-CAM is regulated by Pax-3. Using muscle-specific markers and mor-phological analysis, it has been shown that in splotch mice the development of the shoulder and limb muscle is disturbed, while body wall and axial muscle develop normally (Franz et al., 1993; Bober et al., 1994; Goulding et al., 1994). The mechanism described for the neural crest emigration defect could be identical in the migration process of muscle cell pre-cursors to the limb. However Pax-3 can also be involved in the differentiation and/or proliferation of these cells.

In vitro experiments using antisense oligonucleotides to inhibit Pax-5 protein synthesis demonstrate, that this gene could be involved in the proliferation of B lymphocytes (Wakatsuki et al., 1994).

In rat and mouse small eye the Pax-6 mutation seems also to interfere with the migration of the neural crest (Matsuo et al., 1993; Schmahl et al., 1993) cells. By using vital dye labelling, it was shown that anterior midbrain neural crest migration in the rat sey mutant is disturbed. Neural crest cells accumulate in areas adjacent to the anterior cardinal vein and around the optic vesicle and further migration to the fron-tonasal area is inhibited. Since Pax-6 is not expressed in neural crest cells (Walther et al., 1991), the impaired migration in the rat small eye mutant could be linked to the perturbation of the normal expression of a downstream target gene of Pax-6, involved in this process. This mechanism is similar to that described for the Splotch mice, where N-CAM is abnormally expressed (Moase and Trasler, 1991) and possibly also a down-stream gene of Pax-3.

Pax genes may also be involved in the determination of certain tissue structures. Thus Pax-6 could play a role in the induction of the lens during eye development. In vitro ablation experiments in the chick, however, suggest that Pax-6 may be necessary for the early determination of the lens-competent regions (Li et al., 1994).

Also Pax-2 has been shown to be involved in specifying certain tissues. Using antisense oligonucleotides to Pax-2 (Rothenpieler and Dressier, 1993) in kidney organ culture, it has been shown that Pax-2 is possibly required for the con-version of the mesenchyme to epithelium in kidney develop-ment.

Finally it should be kept in mind that Pax genes are paralo-gous genes, sharing many expression domains, which may implicate some redundancy. In Splotch mice axial muscle develops normally, which could be a compensation from Pax-7, the paralogous gene to Pax-3.

There is a need for more mutants to try to elucidate the function of the Pax genes. The homologous recombination approach is a powerful tool to mutate the whole Pax family in mice and to achieve double and triple mutants that will allow us to identify the redundant parts possibly present in each par-alogous group.

Furthermore, since the Pax protein concentration seems to play a crucial role for their appropriate function it is of interest to deregulate the Pax protein thresholds by generating gain of function mouse mutants by overexpressing Pax proteins ectopically and specifically in certain tissues. Also promoter swap experiments in transgenic mice by using homologous recombination approach in ES cells (for example expressing Pax-7 under the promoter of Pax-3) will be useful in order to distinguish the specific function of the paralogous Pax-3 and Pax-7 genes during development.

We thank Paolo Bonaldo and Edward Stuart for critically reading the manuscript and Ralf Altschaffel for the photographs. This work was supported by the Max-Planck Society.