Lbx1 is required for muscle precursor migration along a lateral pathway into the limb

The somitic mesoderm gives rise to multiple tissues in the developing embryo including bone, connective tissue and muscle. Somites form as an epithelial ball of cells that later segregate into sclerotome and dermomyotome in response to patterning signals that arise from adjacent tissues (Dietrich et al., 1993; Pourquie et al., 1993; Goulding et al., 1994; Fan and Tessier-Lavigne, 1994; Munsterberg and Lassar, 1995). Cells in the ventral half of the nascent somite undergo a transition from epithelium to mesenchyme, forming the sclerotome, that differentiates further to give rise to the axial skeleton. Cells in the dorsolateral half of the somite retain their epithelial morphology, forming the dermomyotome, which contains precursors for both the dermis and for skeletal muscles. Dermomyotome derived muscle precursors not only generate the epaxial muscles that attach to the vertebral column, but also the hypaxial musculature of the limb, tongue, diaphragm and ventral body wall (Christ and Ordahl, 1995). Whereas the epaxial muscles arise from cells in the medial dermomyotome, hypaxial muscles are derived from precursors in the lateral half of the dermomyotome (Ordahl and Le Douarin, 1992; Denetclaw et al., 1997). The precursors for hypaxial muscles exhibit markedly different morphogenetic behaviours at different axial levels of the embryo. Cells in the ventral dermomyotome at limb and cervical levels delaminate (Chevallier et al., 1977; Christ et al., 1977; Christ and Ordahl, 1995; Mackenzie et al., 1998) and undergo long range migration to form diaphragm, tongue muscles and appendicular muscles. In contrast, hypaxial muscle precursors at interlimb levels do not migrate. Instead, they retain their epithelial morphology, forming a bud that extends ventrally toward the midline to give rise to ventral body wall muscles (Parry, 1982; Christ et al., 1983).

A number of genes that control the development and migration of hypaxial muscles have been identified. Among these are the paired domain transcription factor Pax3, which is expressed throughout the dermomyotome, and c-Met, which is expressed in delaminating hypaxial muscle precursors (Bober et al., 1994; Goulding et al., 1994; Williams and Ordahl, 1994). Analysis of Splotch (Sp, Pax3⁻) mutants shows Pax3 is required for the normal development of all hypaxial muscles. Appendicular, tongue and diaphragm muscles are missing from homozygous Sp embryos, while the ventral body wall muscles are greatly reduced in size (Franz et al., 1993; Tajbakhsh et al., 1997). The loss of appendicular muscles in Pax3 mutant embryos is primarily due to a failure of muscle precursors to migrate into the limb (Daston et al., 1996). However, Pax3 also regulates muscle cell differentiation (Maroto et al., 1997; Tajbakhsh et al., 1997) and the dysgenesis of all hypaxial muscles in Sp embryos, including those that do not migrate, suggests Pax3 may also be required for the differentiation or survival of hypaxial muscle precursors. The c-Met receptor tyrosine kinase is also expressed in migratory muscle precursors, and mice lacking c-Met exhibit a muscle phenotype that is similar to the Sp muscle phenotype, except that the ventral body wall muscles are still present (Bladt et al., 1995). c-Met is expressed in the ventral lip of the dermomyotome as muscle precursors are delaminating (Daston et al., 1996; Bladt et al., 1995). In c-Met mutant mice, muscle precursors fail to delaminate, and the dermomyotomes remain elongated at limb levels (Dietrich et al., 1999). A similar phenotype is also seen in mice lacking Scatter Factor (SF), the ligand for c-Met (Bladt et al., 1995; Dietrich et al., 1999). Thus, c-Met activation is necessary for cells in the ventrolateral dermomyotome to undergo an epithelial to mesenchymal transition prior to migration.

Pax3 and c-Met are also expressed in non-migratory populations of dermomyotomal cells. Pax3 is expressed throughout the dermomyotome (Goulding et al., 1994; Daston et al., 1996), while c-Met is expressed in the ventral dermomyotome at interlimb levels and in cells located at the dorsal tips of the dermomyotome (Yang et al., 1996). Consequently, factors other than Pax3 and c-Met must specify and control the migratory behaviour of hypaxial muscle precursors. The Lbx1 gene is a candidate for regulating the migratory behaviour of these cells. Lbx1 encodes a homeodomain transcription factor that is expressed specifically in hypaxial muscle precursors that are destined to migrate from the ventrolateral dermomyotome at limb, cervical and occipital levels (Jagla et al., 1995). Subsequently, cells expressing Lbx1 leave the ventrolateral dermomyotomes and migrate into the limb buds and diaphragm at lower cervical and limb levels, and toward the pharynx at occipital levels (Mennerich et al., 1998; Dietrich et al., 1999). Thus, expression of Lbx1 is restricted to hypaxial muscle precursors that undergo long range cell migration. In addition, Lbx1 is not expressed in the dermomyotome in Sp embryos, demonstrating Lbx1 lies downstream of Pax3 and may therefore contribute to the loss of hypaxial muscles that occurs in Sp embryos (Mennerich et al., 1998; Dietrich et al., 1999; L. M. and M. G., unpublished results).

In this study, we show that Pax3 and Lbx1 are coexpressed in all migrating hypaxial muscle precursors. We have examined the function of Lbx1 in hypaxial muscle precursors by inactivating the Lbx1 gene in mice. Homozygous Lbx1^−/− mice lack most appendicular muscles; however, six forelimb flexors and two hindlimb extensors are still present in mutant newborns. In addition, diaphragm and tongue muscles still form, demonstrating Lbx1 is required only for limb muscle development. The extensive loss of limb muscles in Lbx1^−/− mice results from a cell migration defect. This defect is not due to the premature differentiation of limb muscle precursors or impaired cell motility. Rather, the observed changes in cell migration demonstrate that Lbx1 is required for muscle precursors to migrate laterally into the limbs. As a result, misplaced muscle precursors are found at the posterior-ventral and dorsal margins of the forelimbs and hindlimbs, respectively, thereby generating the residual hypaxial muscles that are present in Lbx1^−/− mice.

The Lbx1 targeting vector was assembled using the pKSloxPNT vector (provided by A. Joyner) that contains HSV TK and lox-P flanked Neomycin gene cassettes in a Bluescript KS backbone. Genomic sequences encompassing the mouse Lbx1 gene were isolated from a 129SV genomic phage library. The coding region of EGFP was cloned in frame into a NotI site at aa 62 of the mouse Lbx1 protein. A 4.1 kb XhoI fragment containing EGFP-pA (upstream arm) and a 3.6 kb NotI fragment (downstream arm) were cloned seperately into pKSloxPNT to generate the Lbx1 targeting vector. A frameshift in the Lbx1 homeobox was created by filling in the BglII site in exon 2 of the downstream arm prior to its insertion into the targeting vector.

W9.5 embryonic stem cells were maintained on primary fibroblast feeder layers supplemented with LIF. 2×10⁷ES cells were electroporated with 25 μg of the Lbx1 targeting vector after linearization with SalI. Fifty clones were screened by Southern analysis using an upstream external probe (Fig. 1). Two clones with a recombined Lbx1 allele were identified by Southern analysis and by PCR analysis using primers for Neo and the Lbx1-EGFP boundary. Both clones were injected into C57Bl6 blastocysts to generate chimeras. Germline founders and F₁ generations were generated on a C57Bl6 background. Mice and embryos were genotyped by PCR using tail or visceral yolk sac DNA. Primers MKG396 (CAGCTGCA-GAAGCCAGGACTG; 12 ng/μl), MKG321 (CCGGACACGCTGA-ACTTGTGG; 12 ng/μl), and MKG333 (ATGACTTCCAAGGAGG-ACGGCA; 24 ng/μl) were used in a 25 μl reaction containing Taq buffer (1.6 mM MgCl₂, 0.2 mM dNTPs, 10% DMSO) and 1.25 Units Taq polymerase (Perkin-Elmer). Amplification of mutant and wild-type Lbx1 alleles generated diagnostic bands of 315 and 445 bp, respectively.

Rat anti-Pax3 and rabbit anti-Lbx1 antibodies were generated against bacterial fusion proteins containing 122 aa of the Pax3 C terminus and 120 aa of the mouse Lbx1 protein that includes the homeodomain, respectively. A 365 bp PvuII fragment from the mouse Pax3 C terminus was inserted into SmaI/XhoI fill in vector pGEX4T-2 to generate GST/Pax3(CT). A fragment encoding His6-Pax3(CT) was inserted into the NheI/B cut vector pET11d. A BglII-EarI fragment from exon2 of the downstream mouse genomic clone was inserted into the BglII-EarI sites of the human LBX1 cDNA to create BS-Lbx1(C-mm). An Ecl136II/Xho fragment of BS-Lbx1(C-mm) was inserted into the SmaI/XhoI cut vector pGex4T2 to generate GST/Lbx(I120). Soluble GST fusion proteins were purified from BL21(DE3) bacteria according to standard methods (Pharmacia). GST/Pax3(CT) was also further purified from SDS-polyacrylamide gels using UV shadowing and elutrap (Schleicher and Schuell) elution. H6/Pax3(CT) was purified from BL21(DE3)plysS bacteria using a denaturing urea procedure (Qiagen) and was column renatured prior to elution with an imidazole gradient. Rats (anti-Pax3) were injected and boosted twice with SDS-treated GST-Pax3, boosted once with soluble GST/Pax3(CT) and boosted twice with soluble H6/Pax3(CT). Rabbits (anti-Lbx) were injected and boosted three times with soluble GST-fusion protein.

High titre sera were pooled and used for affinity purification. Soluble GST-Pax3(CT) and GST-LbxI120 were coupled to a 1:1 mixture of Affigel 10 and 15 according to manufacturer’s instructions (Biorad). Antibody was affinity purified according to the method of Harlow and Lane (1988), concentrated by ultrafiltration, adjusted to 10% glycerol and 1 mg/ml BSA then pre-absorbed with GST-whole cell extract beads, which were generated by coupling whole bacterial extracts of GST overproducing bacteria (Pharmacia) to Affigel 10 and 15.

Embryos were rinsed for 1-5 hours in PBS (4°C), fixed for 1 hour in 4% paraformaldehyde (4°C), and then rinsed overnight in PBS before being equilibrated in 25-30% sucrose (4°C) prior to embedding in OCT (Tissue-Tek). Cryostat sections (12-20 μm) were dried and treated as follows: PBS at room temperature (RT), 3× 5 minutes; methanol (−20°C), 5 minutes; air dry; 3SB (PBS, 0.3% BM blocker (Boehringer Mannheim) heat inactivated sera: 5% fetal calf, 5% goat, 1% chick) with 0.2% Triton X-100, 1 hour; primary antibody (rabbit anti-EGFP and rabbit anti-MyoD (Santa Cruz Biotechnology), mouse anti-Pax7 and mouse anti-Myogenin (Developmental Studies Hybridoma Bank) in 3SB/0.1-0.2% Triton X-100, overnight (4°C); PBST (PBS, 0.1% Tween 20), 3× 10 minutes (RT); secondary antibodies (Cy2-donkey anti-rat and Cy3-goat anti-rabbit (Jackson Laboratories), biotinylated goat anti-mouse (Vector Laboratories) in 3SB/0.1-0.2% triton, 2-3 hours (RT); PBST, 3× 10 minutes (RT); streptavidin-Cy5 (Jackson Laboratories) in PBST, 1-2 hours (RT); PBST, 3× 10 minutes (RT). Sections were dehydrated in an ethanol/xylene series before mounting with DPX (BDH). Images were recorded as three single tracks (red, blue, green) using a Zeiss LSM 510 laser scanning microscope. All figures were assembled using Adobe Photoshop.

Fixed newborns were dissected and muscle connections to bones were traced to identify muscles according to Gilbert (1968) and Gray’s Anatomy (Williams et al., 1989). For histology, newborn pups were killed and cut open along the abdominal midline and fixed for at least 2 days in 4% paraformaldehyde prior to dehydration in an ethanol series (8-12 hours each step). Embryos were transfered to Histoclear (8-12 hours) and then paraffin (2 hours and 12 hours) prior to embedding. 6 μm sections collected at regular intervals were stained with hematoxylin and eosin prior to mounting.

Previous RNA in situ analyses of Lbx1 expression during embryogenesis show Lbx1 is expressed in the ventral dermomyotome at cervical and limb levels in presumptive migratory muscle precursors and later in the tongue, diaphragm and limbs (Jagla et al., 1995; Mennerich et al., 1998; Dietrich et al., 1998; 1999). However, from these studies it was not clear whether Lbx1 and Pax3 marked different or identical populations of migrating limb muscle precursors, or if Lbx1 was expressed in all migrating muscle precursors. It was also not known if Lbx1 and Pax3 were coexpressed with muscle regulatory factors (MRFs) in differentiating muscle precursors, which would suggest a potential role in the differentiation or patterning of hypaxial muscles. Antibodies against mouse Pax3 and Lbx1 were therefore generated and used to analyze the expression of both proteins in developing embryos. The Pax3 antibody detected Pax3 expression in the dorsal neural tube, dorsal root ganglia, dermomyotome and in scattered cells in the limb, diaphragm and hypoglossal cord. Specific Lbx1 antibody staining was present in the marginal zone of the neural tube, as well as in migrating limb muscle precursors.

At forelimb levels in E9.5-E10 embryos, Lbx1 staining was observed in the ventral lip of the dermomyotome, in cells that are delaminating (Fig. 1A). Large numbers of Pax3⁺/Lbx1⁺ cells were scattered throughout the proximal region of the limb, confirming that Lbx1 is expressed in presumptive migrating limb muscle precursors (Fig. 1B). All Lbx1⁺ muscle precursors coexpressed Pax3 and all migratory Pax3⁺ cells expressed Lbx1. Pax3⁺/Lbx1⁺ cells were seen in both the ventral and dorsal muscle masses of the limb (Fig. 1B). Complete co-expression of Lbx1 and Pax3 was also seen in hypaxial muscle precursors that migrate ventrally into the diaphragm and tongue (Fig. 7, data not shown). Thus, co-expression of Pax3 and Lbx1 was observed in all migrating muscle precursors.

Myogenin expression was examined in sections through the forelimb that had been stained with antibodies to Pax3 and Lbx1 to determine at the cellular level whether limb myoblasts express Lbx1 and Pax3. Although Myogenin⁺ cells were clearly present in the developing epaxial dermomyotome (Fig. 1C,D) at E10.5, few Myogenin⁺ cells were present in the limb, consistent with previous studies showing that Myogenin expression in the limb begins at E10.5 (Ott et al., 1991; Yee and Rigby, 1993). In E11 embryos, Myogenin⁺ cells were found interspersed with Lbx1⁺/Pax3⁺ muscle precursors in both the dorsal and ventral muscle masses (Fig. 1D); however, very few, if any, cells were seen co-expressing Pax3 and Myogenin (Fig. 1D), or Lbx1 and Myogenin (data not shown). Similar results were obtained when Pax3, Lbx1 and MyoD expression was compared (data not shown). From E11.5 onwards, Pax3 and Lbx1 were expressed predominantly in undifferentiated cells that were located at the leading edge of migrating pools of limb muscle precursors. These results show that during limb muscle development muscle precursors that are initially Pax3⁺/Lbx1⁺/MRF⁻ become Pax3⁻/Lbx1⁻/MRF⁺. Together, these data demonstrate that Pax3 and Lbx1 are not expressed in limb myoblasts and indicate that their downregulation may be required for hypaxial muscle precursors to differentiate.

To examine the function of Lbx1 in hypaxial muscle precursors, we generated a targeted mutation in the Lbx1 gene by homologous recombination. An EGFP reporter cassette was fused in frame into the Lbx1 coding region at aa 62 to trace migrating muscle precursors in Lbx1^−/− mice. However, EGFP expression in both Lbx1^−/+ and Lbx1^−/− embryos was weak, requiring the use of an anti-EGFP antibody to detect cells that would normally express Lbx1. A frame shift mutation was also introduced into the homeobox of Lbx1 to ensure that no functional Lbx1 protein is produced (Fig. 2A). W9.5 ES cells were electroporated with the Lbx1 targeting vector. Two cell lines, 6C1(A) and 2B1(B), were isolated that exhibited homologous recombination at the Lbx1 locus (Fig. 2B). Both cell lines when injected into C57Bl6 blastocysts gave germline transmission, and offspring from the 6C1 and 2B1 ES cell lines exhibited identical mutant phenotypes. Heterozygous Lbx1^−/+ offspring were fertile and exhibited no gross developmental abnormalities. In contrast, homozygous Lbx1^−/− mutant mice died shortly after birth with striking defects in the organization of the appendicular musculature.

The cause of the perinatal lethality observed in Lbx1^−/− mice is not known, although the respiratory problems observed in one newborn mutant mice may be a major contributing factor. Homozygous Lbx1^−/− embryos did not express the Lbx1 protein, arguing that the Lbx1 mutation is a null allele. Nevertheless, it is possible that the first 62 amino acids of the Lbx1 protein are translated in frame with EGFP and could therefore exhibit some biological activity.

When heterozygous Lbx1 mice were bred, a number of newborn offspring with gross morphological defects were found dead in each litter. One live mutant pup was born and died within two hours. All abnormal mice were homozygous for the mutated Lbx1 allele. Lbx1^−/− mice exhibited a pronounced limb phenotype characterized by a severe reduction in muscle mass in the shoulders, pelvic girdle and the extremities of the limbs (Fig. 3). Hindlimbs were severely reduced in girth, while the forelimbs were hyperflexed and thinner than those of wild-type newborns (Fig. 3A,B).

Anatomical dissection of mutant and wild-type P0 mice revealed a number of differences in the organization of the skeletal musculature (Table 1). All hindlimb muscles were missing, with the exception of the gluteus medius and one other unidentified hypaxial muscle. Two hindlimb suspension muscles (quadratus lumborum, pubococcygeus) were retained, suggesting that all other suspension muscles are hypaxially derived. In the forelimbs, all extensor muscles were absent; however, a number of flexor muscles were still present (Table 1). Both extensors and flexors connecting phalanges, metacarpals, and carpals were not detected (Fig. 4C,D). Wrist flexors were all present but reduced in size. Two of three elbow flexors, the biceps brachii and coracobrachialis, were also present. A loss of extensors was also observed in muscles that surround the scapula. Muscles on the lateral (extensor) side of the scapula (supraspinatus, infraspinatus, spinodeltoid, teres major) which are derived from the dorsal muscle mass were absent, whereas the subscapularis muscle on the medial (flexor) side was only reduced in size. All muscle suspending the forelimb from the body with the exception of the clavobrachialis and latissimus dorsi muscles were present and normal, suggesting that their formation does not require Lbx1.

One difference that was seen in the organization of proximal forelimb flexors in Lbx1^−/− mice was the loss of the brachialis muscle, which is one of the two principal flexors of the elbow joint. Further differences were seen in the anatomy of the flexor muscles that connect the humerus, radius and ulna to the carpal bones. These flexor muscles, while still present and correctly attached, were truncated along the proximodistal axis of the limb. Muscle tissue was observed only at the proximal ends of both bones, giving way to tendon midway between the elbow and wrist. In wild-type P0 mice, skeletal muscle tissue extends three-quarters of the way from the elbow to the wrist (Fig. 4A,B). In contrast to the severe loss of appendicular muscles in Lbx1^−/− mice, the deep muscles of the back, intercostal muscles and ventral body wall muscles were similar to age matched wild-type and heterozygous littermates (data not shown).

To determine whether skeletal muscles develop in the limbs of Lbx1 mutant mice and are subsequently lost or whether these limb muscles never form, the pattern of differentiating muscles in E13.5 mouse embryos was analyzed. At this time, myoblasts in the limb have segregated into distinct populations that mark the developing muscle anlagen. A complete series of cross sections through forelimbs and hindlimbs of wild-type and Lbx1^−/− embryos was stained with an antibody to MyoD. MyoD⁺ cells were absent from the hindlimbs of Lbx1^−/− embryos, except for the anlage of the gluteus medius that was located lateral to the ileum and one other unidentified muscle. The number and size of MyoD⁺ muscle anlagen in the forelimb was compared at increasing distances from the junction of the humerus and ulna bones (Fig. 4E-I). In Lbx1^−/− embryos, extensor muscle anlagen were completely absent at this stage, while the developing flexor muscles showed a substantial reduction in their size midway between the wrist and elbow (Fig. 4G,H). In wild type embryos, MyoD expression extended further toward the wrist (Fig. 4I). Sections through E13.5 Lbx1^−/− embryos at the level of the elbow, show the anlagen for those muscles that are retained in mutants, i.e. biceps brachii, coracobrachialis and wrist-flexors, were located between the humerus/ulna and ribs (Fig. 4J). This shows that proximal forelimb flexor muscles form when the elbow joint is no longer separated from the body wall and suggests a model to explain the selective development of some flexor muscles in Lbx1^−/− mice.

The expression of Lbx1 in migrating muscle precursors and the loss of appendicular muscles in E13.5 Lbx1^−/− embryos led us to ask whether limb muscle precursors are impaired in their ability to migrate in Lbx1^−/− mice. The distribution of Pax3⁺ muscle precursors was analyzed in a complete series of sections through the limbs of E9.5 and E10.5 wild-type, Lbx1^−/+ and Lbx1^−/− embryos. At E9.5 a clear difference was seen in the distribution of Pax3⁺ cells in wild-type versus Lbx1 mutant embryos. Whereas in wild-type embryos, Pax3⁺ cells had already invaded the limb, in Lbx1 mutant embryos these cells were clustered in the trunk, just beneath the ventrolateral lip of the dermomyotome (Fig. 5A,B).

The differences observed in the migration of limb muscle precursors were even more pronounced at E10.5. In wild-type (not shown) and Lbx1^−/+ embryos (Fig. 5C,E) Pax3⁺ cells had already segregated into dorsal and ventral populations. In contrast, no Pax3⁺ cells were seen in the forelimbs of E10.5 Lbx1^−/− mutant embryos. Instead, an enlarged stream of Pax3⁺ cells was observed immediately adjacent to the ventral aspect of the limb at posterior forelimb levels (Fig. 5F arrow). These Pax3⁺ cells expressed EGFP, demonstrating that the cells that normally express Lbx1 fail to migrate into the limbs of Lbx1^−/− embryos. When Pax3 expression was examined in sections through the hindlimbs of E10.5 Lbx1^−/− embryos, a pool of delaminated Pax3⁺/EGFP⁺ cells was seen just ventral to the dermomyotome (Fig. 5H). In comparable sections through Lbx1^−/+ hindlimbs, Pax3⁺/EGFP⁺ cells were seen in the limb (Fig. 5G). In Lbx1^−/− embryos, hindlimb muscle precursors delaminate, but fail to migrate far from the ventral edge of the dermomyotome.

In mice lacking c-Met, limb muscle precursors fail to migrate into the limb, resulting in the complete loss of limb muscles (Bladt et al., 1995; Maina et al., 1996) and raising the possibility that c-Met is no longer expressed in Lbx1^−/− embryos. When c-Met expression was examined at E9.5, strong c-Met staining was observed in the ventral dermomyotome in both normal and Lbx1^−/− embryos (Fig. 5A,B). c-Met receptor expression was also seen in delaminating muscle precursors, demonstrating that the failure of limb muscle precursors to enter the limb in Lbx1^−/− embryos is not due to the loss of c-Met. This finding is consistent with the different morphologies of the ventral dermomyotome which we observe in Lbx1 mutant embryos compared to that reported for c-Met mutant embryos. In Lbx1^−/− embryos, muscle precursors delaminate from the dermomyotome, whereas in c-Met^−/− embryos these cells retain their epithelial morphology (Dietrich et al., 1999). Together, these results suggest that Lbx1 and c-Met regulate different morphogenetic events in migrating muscle precursors. Interestingly, we also noted that antibodies to c-Met did not detect large numbers of Pax3⁺ cells that co-express c-Met in the limb proper, suggesting that c-Met is downregulated once muscle precursors delaminate from the dermomyotome (Fig. 5A).

Myogenesis in the mouse begins at E9.5 in the dorsal dermomyotome (Ott et al., 1991); however, it is delayed in the ventral lip of the dermomyotome, especially at limb and cervical levels where Lbx1 is expressed. Expression of MyoD and Myogenin in the limb does not begin until after muscle precursor cells have finished migrating, suggesting myogenesis is not compatible with long-range cell migration. Furthermore, our results show that Myogenin is not expressed until after Lbx1 and Pax3 are downregulated (Fig. 1, data not shown). Therefore, Lbx1 may be required to delay myogenesis in migrating hypaxial muscle precursors, and premature muscle differentiation may cause the defects in Lbx1^−/− mice. To test this hypothesis, E9.5 Lbx1^−/− mutant embryos were examined for evidence of precocious myogenesis. Myogenin⁺ cells were observed beneath the dorsal dermomyotome in both Lbx1^−/+ and Lbx1^−/− embryos at E9.5, consistent with the early expression of Myogenin in epaxial myoblasts described elsewhere (Fig. 5A,B). However, no Myogenin⁺ cells were seen amongst the Pax3⁺/EGFP⁺ cells that had delaminated and failed to enter the limb in Lbx1 mutants. At E10.5, two populations of Myogenin⁺ cells were detected in the body wall adjacent to the limbs in both wild-type and mutant embryos (Fig. 5C,D, arrowhead). However, there was no increase in the number of Myogenin⁺ cells in Lbx1^−/− embryos. Furthermore, the EGFP⁺ population of cells did not express Myogenin (Fig. 5 compare D and F). These results argue that the cell migration defect in Lbx1 mutants is not caused by hypaxial muscle precursors differentiating prematurely.

Our observation that Pax3⁺/Lbx1⁺ cells migrate ventrally at forelimb levels (Fig. 5F), but not at hindlimb levels (Fig. 5H), led us to investigate whether this difference could account for the different patterns of muscle development that occur in the forelimbs and hindlimbs of Lbx1^−/− mutants. Expression of Pax3 and Myogenin was analyzed in a complete series of serial sections through E11.5 forelimbs and hindlimbs to identify populations of migrating muscle precursors (Pax3⁺) and myoblasts (Myogenin). In wild-type E11.5 embryos, large numbers of Pax3⁺ cells were seen within the limb as expected. These cells were located both dorsally and ventrally where they had begun to separate into muscle anlagen (Fig. 6A). When sections through the posterior forelimbs of Lbx1^−/− mice were analyzed, a large pool of Pax3⁺ cells was observed immediately adjacent to the ventral aspect of the limb (Fig. 6B). These muscle precursors (Fig. 6B arrow) did not appear to be migrating into the limb, and instead clustered at the junction of body wall and ventral limb. Myogenin⁺ cells were distributed throughout this posterior-ventral pool of Pax3⁺ cells, indicating limb muscle precursors had already begun to differentiate adjacent to the ventral limb. Furthermore, this pool of cells appear to be sufficiently close to the limb to receive normal muscle patterning signals, thereby giving rise to the biceps brachii and coracobrachialis flexor muscles, but not the brachialis. Since Pax3⁺ muscle precursors were only present adjacent to the posterior forelimb at E11.5, we conclude that all hypaxial muscle precursors in the anterior half of the forelimb, including the precursors that would normally give rise to extensors, instead migrate ventrally and enter the septum transversum. Consistent with this hypothesis, there is an increase in the number of Pax3⁺ cells in the diaphragm at E11.5 (Fig. 7).

We did not observe a similarly located pool of Pax3⁺ cells at hindlimb levels in E11.5 Lbx1^−/− embryos (Fig. 6C,D), confirming our observations that muscle precursors are unable to migrate at hindlimb levels (Fig. 5G,H). Nevertheless, an increase in the number of Myogenin⁺ cells at the ventral edge of the epaxial dermomyotome was noted in Lbx1^−/− embryos (Fig. 6E,F arrows) suggesting that cells that normally migrate laterally into the limb instead congregate just ventral to the epaxial dermomyotome. These cells may contribute to the gluteus medius muscle that is still present in the hindlimb and connects the ilium to the femur (see Table 1). This muscle is located lateral to the pelvic bone, indicating that it is an extensor and is derived from dorsal muscle precursors. In addition, the quadratus lumborum, a hindlimb suspension muscle, is enlarged in Lbx1^−/− mice, suggesting hypaxial muscle precursors that do not migrate contribute to this muscle.

Lbx1 and Pax3 are not only expressed in limb muscle precursors that migrate laterally into the limb buds, but also in muscle precursors that migrate ventrally into the thorax to generate the diaphragm muscles and into the branchial arches to form the intrinsic tongue muscles (Fig. 7; Dietrich et al., 1999; Mackenzie et al., 1998). To determine whether the ventral migration of these two populations of hypaxial muscle precursors is altered in Lbx1^−/− mice, E10.5 and E11.5 embryos were examined for the presence of migrating Pax3⁺ cells in the diaphragm and second branchial arch. In E11.5 wild-type and Lbx1^−/− embryos, Pax3⁺ muscle precursors were seen in the diaphragm. However, the diaphragms of Lbx1^−/− embryos at E11.5 consistently contained more Pax3⁺ cells (Fig. 7A,B), suggesting that the muscle precursors that normally enter the anterior forelimb migrate inappropriately into the septum transversum. The migration of tongue muscle precursors was also examined at E10.5 for evidence of changes in their migratory behaviour. A delay was seen in the migration of Pax3⁺ cells into the second branchial arch in E10.5 embryos (Fig. 7C,D). However, by E11.5 large numbers of Pax3⁺ cells were present within the tongue primordium of Lbx1^−/− mice (data not shown).

Anatomical and histological investigation of newborn Lbx1^−/− mice confirmed the presence of a muscular diaphragm in Lbx1^−/− mice (Fig. 7E,F). While small differences were noted in the distribution of muscle tissue within the diaphragm, i.e. there was slightly more muscle laterally in mutant than in wild-type newborn mice (data not shown), the overall organization of the diaphragm in Lbx1^−/− mutant mice was similar to newborn wild-type mice (Fig. 7 c.f. E,F). No significant difference was seen in the size or organization of the tongue muscles in Lbx1^−/− mice. The muscle fibres in the tongue were multinucleate and organized into the patterned bundles of muscle fibres that are characteristic for the intrinsic muscles of the tongue (Fig. 7G,H). Taken together, these results demonstrate that while muscle precursors are unable to migrate laterally into the limb, they still migrate along ventral routes into the septum transversum and branchial arches to form the muscles of the diaphragm and tongue, respectively.

We have examined the function of Lbx1 in the development of hypaxial muscles by inactivating the Lbx1 gene in mice. Our results show that the ablation of Lbx1 causes the widespread, but incomplete, loss of appendicular muscle and reveals a role for Lbx1 in directing the lateral migration of muscle precursors into the limbs. Our findings and a model for how these muscle defects arise in Lbx1^−/− mice are discussed below.

Inactivation of Lbx1 leads to the loss of most appendicular muscles, whereas diaphragm and tongue muscles are spared. While the loss of appendicular muscles that occurs in Lbx1 mutants is substantial, it is less severe than the muscle phenotypes seen in Pax3 and c-Met mutant mice. In Sp (Pax3⁻) mice, appendicular, tongue, diaphragm and ventral body wall muscles fail to form (Franz et al., 1993; Bober et al., 1994; Goulding et al., 1994; Tajbakhsh et al., 1997). Thus, Pax3 is necessary for the development of all hypaxial muscle precursors, including those populations that do not undergo long range cell migration. The defects in c-Met mutant embryos are less severe than those seen in Sp embryos. However, unlike Lbx1 mutants, the tongue and diaphragm muscles as well as all appendicular muscles are missing (Bladt et al., 1995; Maina et al., 1996).

Although Pax3 is required for the migration of limb muscle precursors (Daston et al., 1996), there is evidence that Pax3 also plays an early role in the differentiation of hypaxial muscle precursors (Maroto et al., 1997, Tajbakhsh et al., 1997). In Sp embryos the loss of hypaxial muscles is accompanied by a shortening of the dermomyotomes (Bober et al., 1994; Daston et al., 1996, Tajbakhsh et al., 1997) suggesting hypaxial muscle precursors may either degenerate or are not specified correctly in the absence of Pax3. Additional evidence that Pax3 controls the early specification of hypaxial muscle precursors comes from the demonstration that all trunk muscles are lost in Myf5/Pax3 mutant mice (Tajbakhsh et al., 1997). In c-Met mutant mice, the underlying cause of these muscle defects is the failure of cells in the ventrolateral dermomyotome to undergo an epithelial to mesenchymal transition (Dietrich et al., 1999). Unlike c-Met, Lbx1 is not required for the delamination of hypaxial muscle precursors. In Lbx1^−/− embryos, tongue and diaphragm muscle precursors migrate ventrally along their normal routes in the embryo, whereas appendicular muscle precursors delaminate, but fail to migrate laterally.

Further evidence that these genes control different steps in the development of hypaxial muscles comes from epistasis experiments. In Sp mutant embryos, the expression of c-Met and Lbx1 is severely reduced in the ventral dermomyotome, demonstrating that both genes lie downstream of Pax3 (Mennerich et al., 1998; Dietrich et al., 1999). However, it is not known if Pax3 directly regulates the expression of either gene, and while potential binding sites for Pax3 have been identified in the c-Met promoter (Epstein et al, 1996), these sites have not been shown to be functionally significant in vivo. c-Met does not function upstream of Lbx1, since Lbx1 is still expressed in mice lacking c-Met or its ligand, Scatter Factor (Dietrich et al., 1999). Furthermore, expression of c-Met and Pax3 also does not depend on Lbx1 as seen in Fig. 5. This is consistent with the different muscle phenotypes seen in the Lbx1^−/− mice compared to c-Met^−/− and Sp mice. Thus, Lbx1 and c-Met function independently downstream of Pax3 to control migration and delamination, respectively.

Lbx1 is expressed in all known migrating hypaxial muscle precursors, and contrary to expectations, some populations of hypaxial muscle precursors do migrate in Lbx1^−/− embryos. Thus, Lbx1 does not specify the general property of long range migration in hypaxial muscle precursors. However, Lbx1 is required for the migration of muscle precursors into the limb field. Three models could account for the loss of muscle precursors in the limbs of Lbx1^−/− mice. First, Lbx1 serves to maintain limb precursors in an undifferentiated state during migration, leading to the premature differentiation of migratory cells in Lbx1 mutants. This is unlikely since MRF expression was not observed in the proximal limb buds of Lbx1^−/− embryos between E9.5 and E10.5 (Fig. 5). Second, Lbx1 serves to activate cellular components required for cell motility. However, the presence of ventrally migrating EGFP⁺ cells in Lbx1^−/− embryos demonstrates that Lbx1 is not required for motility per se. Finally, Lbx1 could control the migration of hypaxial muscle precursors by allowing them to respond to guidance cues along their migratory routes. The observation that muscle precursors in Lbx1^−/− embryos can still migrate ventrally, but not laterally, argues for this model, and suggests that Lbx1 regulates responsiveness to a lateral migration cue emanating from the limb bud.

Flexors and extensor muscles in the limbs are thought to arise from the ventral and dorsal muscle masses, respectively. In Lbx1^−/− mice, six flexor muscles are retained in the forelimb and two extensors are still present in the hindlimb. This raises the question as to whether the defects in appendicular muscle formation in Lbx1^−/− mice are primarily due to changes in cell migration, or whether Lbx1 has an additional role in specifying subsets of appendicular muscle precursors. Our detailed examination of hypaxial muscle precursor migration, combined with an in depth analysis of the muscle defects in E13.5 and P0 mutants, explains how certain appendicular muscles persist in Lbx1 mutants without invoking an additional role for Lbx1 in muscle specification. A model that outlines the migration routes of muscle precursors in wild-type and Lbx1 mutant embryos is shown in Fig. 8. In this model, there is only a ventral migration route at occipital levels. Two overlapping migratory cues are present at anterior forelimb levels in mammals, and these direct cells to migrate laterally into the dorsal limb and ventrally into the septum transversum (see Fig. 8), while at posterior forelimb levels, cells migrate along a ventrolateral route. At hindlimb levels there is a lateral pathway, but no corresponding ventral route. With this model, we can explain the presence of all of the residual muscles in Lbx1^−/− mice by assuming that Lbx1 is required only for lateral migration, not ventral migration. Thus in Lbx1^−/− embryos, forelimb muscle precursors can migrate toward the ventral limb, but not enter it, whereas at hindlimb levels they are unable to migrate at all.

The absence of a ventral migratory pathway in the hindlimb is consistent with the accumulation of delaminated muscle precursors immediately below the dermomyotome in Lbx1^−/− embryos (Fig. 5H). EGFP⁺ cells are present in E10.5 Lbx1^−/− embryos at hindlimb levels (Fig. 5H), demonstrating hindlimb hypaxial muscles are still specified, and in E11.5 Lbx1^−/− embryos there are increased numbers of Myogenin⁺ cells near the ventral dermomyotome. In newborn Lbx1^−/− mice, the gluteus medius muscle is present, demonstrating that this hypaxial muscle still develops at hindlimb levels. The dorsal location of this muscle, argues that the precursors of the gluteus medius do not need to migrate extensively for it to form, hence its presence in Lbx1^−/− mice. Secondly, we observe a substantial increase in the size of the quadratus lumborum muscle, suggesting that non-migrating hypaxial muscle precursors may also be recruited to contribute to this muscle.

The constellation of flexors retained in the forelimbs of Lbx1^−/− mice is also very specific, and argues against a simple model where Lbx1 specifies dorsal forelimb muscle precursors (extensors), but not ventral forelimb precursors (flexors). We observe that in Lbx1^−/− embryos, Pax3⁺ muscle precursors still migrate ventrally at both forelimb levels. At anterior forelimb levels, these cells enter the septum transversum and contribute to the diaphragm (Fig. 7). At more posterior levels they are unable to do so, and instead form a pool of cells immediately adjacent to the posterior forelimb (Fig. 6). Thus, the loss of all dorsal appendicular muscles at forelimb levels most likely reflects a failure of these muscle precursors to migrate laterally into the limb. Not all forelimb flexor muscles are present in Lbx1 mutant mice. While the biceps and coracobrachialis are retained and are morphologically normal, the brachialis is missing. In wild-type E10 embryos, few Pax3⁺ cells enter the ventral half of the limb at anterior forelimb levels (Fig. 8). Conversely, many Pax3⁺ cells are present in ventral limb posteriorly (Fig. 8). However, by E10.5 the ventral muscle mass is evident both in the anterior and posterior forelimb, indicating that cells in the ventral forelimb relocate anteriorly. We propose that in Lbx1^−/− embryos, muscle precursors do not enter the posterior forelimb and are therefore unable to migrate anteriorly, and as a result the brachialis muscle never forms. Consistent with this, there is no pool of Pax3⁺ cells adjacent to the anterior forelimb in E11.5 Lbx1^−/− embryos (data not shown). Nevertheless, muscle precursors that are present at posterior forelimb levels can give rise to the biceps brachii and coracobrachialis without migrating into the limb. Thus, the loss of the lateral migration pathway is sufficient to account for the abnormal patterning of elbow flexors.

Previous studies have shown that elements of the skeletal primordia are the likely source of the signals that pattern skeletal muscles during development (see Kardon, 1998). This together with our observations, provides an explanation as to why some muscles form in Lbx1 mutant mice. At the time that the forelimb muscle anlagen are forming, the elbow joint is connected to the body wall where migrating Pax3⁺ muscle precursors have congregated in Lbx1^−/− embryos, thus allowing them to be patterned by the adjacent appendicular skeleton (Figs 4 and 6). Although hindlimb muscle precursors do not appear to migrate, they do detach from the dermomyotome and are positioned close to the dorsal pelvis. Thus, in all cases it appears that the muscles that form in Lbx1^−/− mice do so because they are located near elements of the appendicular skeleton that play an instructive role in patterning the developing muscle anlagen. The observation that abnormally migrating hypaxial muscle precursors still form anatomically correct muscles, indicates muscle precursors are naive and can be programmed to form different types of hypaxial muscles. Thus, it appears that the role of limb muscle precursor migration during embryogenesis is to position muscle precursors within the limb mesenchyme, so they can then respond to patterning signals derived from the appendicular skeleton.

This research was supported by grants from NIH (NS31978) and the March of Dimes to Martyn Goulding, and from the Human Frontier Science Program. Laura Moran-Rivard was supported by an NIH Predoctoral Training Fellowship (GM07420). The anti-Myogenin monoclonal antibody was obtained from the Developmental Studies Hybridoma Bank, University of Iowa. We would also like to thank Elise Lamar for her critical reading of the manuscript and Joseph Wu for technical assistance.