Alveolar rhabdomyosarcomas in conditional Pax3:Fkhr mice: cooperativity of Ink4a/ARF and Trp53 loss of function

Abstract

Alveolar rhabdomyosarcoma is an aggressive childhood muscle cancer for which outcomes are poor when the disease is advanced. Although well-developed mouse models exist for embryonal and pleomorphic rhabdomyosarcomas, neither a spontaneous nor a transgenic mouse model of alveolar rhabdomyosarcoma has yet been reported. We report the first mouse model of alveolar rhabdomyosarcoma using a conditional Pax3:Fkhr knock-in allele whose activation in late embryogenesis and postnatally is targeted to terminally differentiating Myf6-expressing skeletal muscle. In these mice, alveolar rhabdomyosarcomas occur but at low frequency, and Fkhr haploinsufficiency does not appear to accelerate tumorigenesis. However, Pax3:Fkhr homozygosity with accompanying Ink4a/ARF or Trp53 pathway disruption, by means of conditional Trp53 or Ink4a/ARF loss of function, substantially increases the frequencies of tumor formation. These results of successful tumor generation postnatally from a target pool of differentiating myofibers are in sharp contrast to the birth defects and lack of tumors for mice with prenatal and postnatal satellite cell triggering of Pax3:Fkhr. Furthermore, these murine alveolar rhabdomyosarcomas have an immunohistochemical profile similar to human alveolar rhabdomyosarcoma, suggesting that this conditional mouse model will be relevant to study of the disease and will be useful for preclinical therapeutic testing.

Rhabdomyosarcomas are a substantial problem in pediatric oncology. The most difficult patients to treat are those with unresectable local disease or metastases (Wexler and Helman 1997; Arndt and Crist 1999). Rhabdomyosarcomas are divided into two major histological subtypes, alveolar and embryonal (Arndt and Crist 1999). Alveolar histology is an independent predictor of worsened outcome, and children with metastatic alveolar rhabdomyosarcomas have a five-year survival rate <30% (Wexler and Helman 1997; Arndt and Crist 1999). The unique histological appearance and significantly worse prognosis for alveolar rhabdomyosarcomas compared with embryonal rhabdomyosarcoma suggest that alveolar rhabdomyosarcoma is biologically distinct from non-alveolar subtypes (e.g., embryonal). Furthermore, despite collaborative national trials of multimodal therapy and attempts at chemotherapy intensification, the outcome for patients with advanced stage alveolar rhabdomyosarcomas has not improved in more than two decades (Wexler and Helman 1997; Arndt and Crist 1999). This plateau in conventional treatments suggests that novel molecularly targeted therapy may be necessary to treat this tumor effectively.

The cell (or cells) of origin for alveolar rhabdomyosarcomas is unknown. The genes and proteins expressed by tumors are characteristic of both immature and well-differentiated embryonic and postembryonic muscle lineages (Tonin et al. 1991). Although congenital cases of alveolar rhabdomyosarcomas have been reported (Grundy et al. 2001), 84% of pediatric rhabdomyosarcomas arise after the first year of life (Gurney et al. 1999), suggesting either a long latency of an embryonically derived tumor precursor or, more likely, a postembryonic origin of the tumors. In the accompanying paper (Keller et al. 2004), we show that initiation of alveolar rhabdomyosarcoma from somite-derived myogenic precursors, the embryonic satellite cell pool, and the postnatal satellite cell pool does not occur in a mouse model. Therefore, our experimental design for this study was influenced strongly by the observation that the immunohistochemical profile of rhabdomyosarcomas is similar to differentiated, multinucleated skeletal muscle, and that the signature cell of rhabdomyosarcoma is the multinucleated rhabdomyoblast (Qualman et al. 1998).

Most alveolar rhabdomyosarcomas share a common genetic feature, the translocation-mediated fusion of a Pax gene to a Fkhr gene (Barr et al. 1993). The t(2;13)-mediated Pax3:Fkhr fusion gene is believed to play a key role in 55%-75% of alveolar rhabdomyosarcomas, whereas the t(1;13)-mediated Pax7:Fkhr fusion gene is believed to play a key role in 10%-22% of alveolar rhabdomyosarcomas (Sorensen et al. 2002). Both of these chimeric genes results in production of a novel transcription factor that has the DNA-binding domain from Pax3 or Pax7, important regulators of muscle development, fused to the potent transactivation domain of the Fkhr tumor suppressor (Arndt and Crist 1999). These chimeric Pax:Fkhr genes are believed to result in a molecular gain of function that may in part lead to the initiation of an inappropriate muscle development program in the affected cells.

Our Pax3:Fkhr conditional knock-in allele builds on but is distinguished from previous informative germ-line Pax3:Fkhr models that do not generate tumors, because our model more carefully recapitulates the translocation-mediated genomic DNA rearrangement found in human tumors. Specifically, our model exchanges potentially important gene regulatory elements in the 3′ genomic region of Pax3 for those from the 3′ genomic DNA of Fkhr. A Cre/LoxP-mediated translocation across heterologous chromosomes would represent the ideal method for modeling this cancer, and this approach has been successfully demonstrated for a leukemia model (Forster et al. 2003). However, because the Pax3:Fkhr fusion gene is not a robust oncogene in previous germ-line Pax3:Fkhr mice (Anderson et al. 2001; Lagutina et al. 2002; Relaix et al. 2003), other additional genetic events are likely required for alveolar rhabdomyosarcoma tumor progression. The much greater efficiency of the knock-in approach versus the chromosomal translocation approach (50% vs. 10^-6) permits generation of the fusion gene in a sufficient number of cells to allow for spontaneous occurrence of secondary genetic events at a reasonable frequency. Our model addresses these issues.

Conditional modeling of alveolar rhabdomyosarcoma in the mouse is especially attractive because of the strong association between Pax3:Fkhr and this somatic tumor. We have generated a Cre/LoxP-mediated conditional Pax3:Fkhr knock-in allele that can be triggered in distinct myogenic lineages by breeding to muscle-specific Cre drivers. In combination with a conditional Fkhr knockout allele generated for these studies, the Cre driver acts to recapitulate the t(2;13) human translocation by both activating Pax3:Fkhr and inactivating single normal alleles of Pax3 and Fkhr. Other putative initiating events can also be triggered in a muscle-lineage-specific manner by breeding these alleles with mouse lines harboring conditional tumor suppressors or oncogenes.

Disruption of Ink4a/ARF and Trp53 pathways has been implicated as a cooperative effect for alveolar rhabdomyosarcomagenesis in humans (Felix et al. 1992; Takahashi et al. 2004) and for non-alveolar rhabdomyosarcomagenesis for humans with germ-line Trp53 mutations (Li-Fraumeni Syndrome) or sporadic tumors (Li and Fraumeni 1969) as well as for transgenic mouse models of embryonal and pleomorphic rhabdomyosarcoma (Jacks et al. 1994; Sharp et al. 2002; Fleischmann et al. 2003; Nanni et al. 2003). While Trp53 contributes to many cellular functions such as differentiation, angio-genesis, cellular senescence, and apoptosis (Kirkwood 2002), its overlapping role with the Ink4a/ARF locus controlling cell cycle withdrawal via the G1-S checkpoint is especially relevant when considering the potential cell of origin for rhabdomyosarcomas. Loss of replicative checkpoint would be advantageous not only to the expansion of a myogenic precursor transforming into rhabdomyosarcoma, but it would be especially critical to rhabdomyosarcomagenesis from terminally differentiating myofibers in quiescence.

In this study we report the first mouse model of alveolar rhabdomyosarcoma, which has been generated by conditional activation of Pax3:Fkhr and conditional inactivation of the Ink4a/Arf and Trp53 pathways in a target pool of terminally differentiating, Myf6-expressing skeletal muscle. These tumors closely recapitulate the histological features and immunohistochemical profile of human alveolar rhabdomyosarcomas. This report provides an important starting point for understanding the initiation and progression of this tumor and its potential cell(s) of origin.

Results

Design and generation of targeted mouse lines

The conditional Pax3:Fkhr (P3Fm) knock-in allele was designed to allow normal Pax3 expression of that allele until Cre is present (Fig. 1A). With expression of Cre, this normally functioning Pax3 allele is converted to a Pax3:Fkhr (P3Fa) allele, fusing exons 1-7 of Pax3 to a 9.3-kb 3′ genomic region of the Fkhr locus containing Fkhr exon 2, exon 3, and 6.5 kb of the untranslated region and genomic DNA. Simultaneously, the eYfp fluorescent marker gene is activated as a second cistron downstream from an encephalomyocarditis virus-derived internal ribosome entry site (IRES). This design mimics the disease-associated translocation with respect to Pax3:Fkhr expression being directed by the Pax3 promoter, and being potentially influenced by 3′ Fkhr cis-regulatory elements knocked into the Pax3 locus. Germ-line mice were established carrying the conditional Neo-containing Pax3:Fkhr allele (Pax3^P3Fp/wt) (data not shown). After removing the positive selection cassette by Flpe-mediated recombination, Neo-excised Pax3^P3Fm/wt and Pax3^P3Fm/P3Fm mice prove to be viable and fertile, breeding normally. Whereas mice and humans heterozygous for Pax3 loss-of-function mutations are readily discernable because of their deficiencies of melanocyte production (Goulding et al. 1993; Tassabehji et al. 1993), Pax3^P3Fm/wt mice without Cre resemble wild-type mice and have normal coat color pigmentation. It therefore appears that, in the absence of Cre, the targeted modifications introduced into the Pax3 locus do not affect normal Pax3 function.

Figure 1.

Structure of targeted alleles. (A) For the Pax3:Fkhr-targeted knock-in allele, a targeting vector was designed for the insertion of a complex cassette into the 3′-region of the Pax3 gene, thereby modifying the last three of 10 exons. FRT sites flank the neomycin resistance gene (Neo) and allow excision of this gene upon breeding to an Flp-e mouse (Rodriguez et al. 2000). The P3Fp allele indicates the presence of the neomycin resistance gene (Neo⁺), whereas the P3Fm allele indicated the absence of the neomycin resistance gene (Neo^-). LoxP sites flank exons 8-10 of Pax3 as well as an inserted polyadenylation/stop signal from the Complement2 (C2) gene. Prior to activation, the P3Fm allele will transcribe a full-length Pax3 transcript. Upon activation by CRE, exons 8-10 and the stop signal are removed and exon 7 of Pax3 is juxtaposed to exons 2 and 3 of Fkhr, thereby creating an activated Pax3:Fkhr fusion allele (P3Fa). An internal ribosomal entry site (IRES) downstream from Fkhr exon 3 allows for bicistronic expression of eYFP. The positions of PCR genotyping primers ck82, ck83, and ck10 are shown. (TK1) Thymidine kinase; (FRT) Flp-e recombinase recognition sequence; (RI) EcoRI restriction site; (RV) EcoRV restriction site; (UTR) untranslated region. (B) For the Fkhr conditional mouse line, a targeting vector was designed for which exons 2 and 3 were flanked by Lox511 sites. The F2-3p allele indicates the presence of the neomycin resistance gene (Neo⁺), whereas the F2-3m allele indicates the absence of the neomycin resistance gene (Neo^-). In the presence of CRE, the inactivated DEL2-3 allele is generated. The positions of PCR genotyping primers ck142, ck143, and ck30 are shown. (C) For the Myf6-Cre allele, a targeting vector was designed for the insertion of a complex cassette into the 3′-region of the Myf6 gene. At an engineered XhoI site in the 3′-UTR, an IRES-Cre was inserted to allow bicistronic expression of Myf6 and Cre. An FRT-Neo-FRT cassette was inserted 3′ to IRES-Cre. The ICNp allele indicates the presence of the neomycin resistance gene (Neo⁺), whereas the ICNm allele indicates its absence. The positions of PCR genotyping primers ck219, ba88, and ba86 are shown.

The Fkhr conditional knockout allele was designed to allow normal expression of that allele until Cre is present (Fig. 1B). Upon Cre expression, this functionally normal Fkhr allele is inactivated by excision of exons 2 and 3 removing the DNA-binding domain and the full transactivation domain of the Fkhr gene. Germ-line mice were established and confirmed by Southern analysis to carry the conditional Neo-containing allele (Fkhr^F2-3p/WT) (data not shown). The Neo-excised Fkhr^F2-3m/WT and Fkhr^F2-3m/F2-3m mice are viable and fertile, breeding normally. Homozygous Fkhr^{DEL2-3/DEL2-3} mice with germ-line absence of both Fkhr alleles die by embryonic day 10 (E10) (data not shown), consistent with previous reports for nonconditional Fkhr alleles (Hosaka et al. 2004).

To allow Cre expression in a target pool of differentiating skeletal muscle following the endogenous expression pattern of Myf6 (Fig. 1C), we designed an Myf6-Cre allele. This allele expresses Cre as a second cistron from an IRES downstream from the Myf6 gene. Germ-line mice carrying this allele were established (data not shown). Unlike previously reported mice with 3′ Myf6 modifications that alter the normal expression of Myf6 and/or adjacent Myf5 gene (Olson et al. 1996), our Myf6^ICNm/WT and Myf6^ICNm/ICNm mice are viable and fertile as both heterozygotes and homozygotes, breeding normally. By small animal computed tomography, these animals also lacked the rib abnormalities (data not shown) previously seen in Myf6/Myf5 loss-of-function alleles (Braun and Arnold 1995).

To test the muscle-specific pattern of expression of our Myf6-Cre alleles, we crossed mice carrying the Myf6-Cre allele to mice carrying the Rosa26 LacZ reporter (Sorriano 1999). In the latter mouse, production of LacZ mRNA is blocked by a strong polyadenylation signal flanked by LoxP sites. In the presence of Cre, the blocking cassette is excised and LacZ mRNA and β-galactosidase protein are synthesized. Shown in Figure 2A is whole-mount β-galactosidase staining of an E15.5 embryo heterozygous for both alleles (Myf6^ICNm/WT Gtrosa26^tm1Sor/WT). Embryos of this genotype exhibit a skeletal-muscle-specific pattern of Cre expression. To confirm that the Myf6-Cre allele was expressed in the target pool of terminally differentiating myofibers, and not in satellite cells, we looked for coexpression of nuclear proteins Cre and Pax7 in Myf6^ICNm/WT mice by immunohistochemistry in cross-section of skeletal muscle of 3-week-old mice. Non-overlap of these markers in Figure 2B demonstrates that Cre is not present in Pax7-expressing satellite cells; however, immunohistochemistry for Cre and myogenic transcription factor MyoD in Figure 2C demonstrates coexpression of these markers in the same nuclei as has been previously shown for postnatal terminally differentiating myofibers (Cornelison and Wold 1997).

Figure 2.

Characterization of the Pax3:Fkhr, Myf6-Cre experimental system. (A) The Myf6-Cre allele demonstrates a skeletal-muscle-specific pattern of expression in an E15.5 fetus carrying the ROSA26 Cre-activatable LacZ marker. (B) Nuclear-localized Cre (green) and the nuclear satellite cell marker Pax7 (red) are not coexpressed in cross-sectioned skeletal muscle of Myf6-Cre mice. (C) Cre and the post-satellite cell myogenic transcription factor MyoD are coexpressed in skeletal muscle of Myf6-Cre mice. (D) When crossed to the Pax3:Fkhr conditional knock-in allele, the Myf6-Cre allele causes heterogeneous expression of the eYfp marker of Pax3:Fkhr activation in cross-sectioned skeletal muscle of a 6-week-old mouse (right), with little autofluorescence in tissue of a control mouse (left). (E) Genomic DNA PCR of the Pax3:Fkhr locus from a 6-week-old mouse carrying the Myf6-Cre allele confirms heterogeneous activation of Pax3:Fkhr within each skeletal muscle sample from varying epaxial and hypaxial sites. Activation of Pax3:Fkhr in testes is noted. (F) RT-PCR of skeletal muscle tissue from a 6-week-old mouse carrying both the Pax3:Fkhr knock-in allele and the Myf6-Cre allele demonstrates Pax3:Fkhr transcript expression for the tissues shown to undergo genomic rearrangement of the Pax3:Fkhr knock-in allele.

Conditional production of the Pax3:Fkhr fusion protein in skeletal muscle

To generate the Pax3:Fkhr protein in skeletal muscle, we crossed mice carrying the Pax3:Fkhr knock-in allele to mice carrying the Myf6-Cre allele. Pax3^P3Fa/WT Myf6^ICNm/WT offspring were viable, fertile, and present in Mendelian ratios. To visualize the extent of Cre-induced formation of the Pax3:Fkhr fusion gene in skeletal muscle, we harvested the vastus lateralis muscle from these mice as well as from control mice and examined the expression of the Pax3:Fkhr reporter, eYFP. Epifluorescence microscopy of these muscles in cross-section showed eYFP expression in ∼50% of the cells, suggesting that the Cre-mediated rearrangement of the Pax3:Fkhr locus was incomplete among mature muscle fibers (Fig. 2D), which is advantageous for demonstrating that tumors arise from Pax3:Fkhr-expressing cells.

To verify that the Pax3:Fkhr knock-in allele was being partially rearranged by the Myf6-Cre driver at the genomic level in a skeletal-muscle-specific manner, we performed PCR of genomic DNA from a variety of tissues from Pax3^P3Fa/WT Myf6^ICNm/WT animals. Pax3:Fkhr rearrangements (Fig. 2E) were present not only in hypaxial skeletal muscle (lower leg, lower arm) but also in epaxial skeletal muscle (diaphragm, tongue). The Pax3:Fkhr allele was not rearranged in cardiac muscle, but a trace rearrangement was present in testes. However, the band for the unrearranged Pax3:Fkhr knock-in allele was still present in each skeletal muscle tissue tested, suggesting, as shown by eYfp microscopy of skeletal muscle, that the action of Cre occurred in ∼50% of the differentiating skeletal muscle myofibers.

To confirm the presence of the Pax3:Fkhr transcript in skeletal muscle tissues of 6-week-old Pax3^P3Fa/WT Myf6^ICNm/WT mice, we performed nested RT-PCR using 5′ Pax3 exon 6 and 3′ Fkhr exon 2 primers (Fig. 2F). Analysis revealed the expected pattern of hypaxial (upper and lower limb) and epaxial (tongue, back, diaphragm, and pectoralis) skeletal-muscle-specific Pax3:Fkhr expression, as well as trace expression in the testes. Nonspecific expression in cardiac muscle or liver was absent.

Mice expressing Pax3:Fkhr develop alveolar rhabdomyosarcoma

Mice carrying the Pax3:Fkhr knock-in allele and a Myf6-specific Cre driver were followed for up to 29 mo (Table 1). A tumor occurred in one of 228 animals at 12 mo of age. This tumor is shown in Figure 3A. Microscopic computed tomography (microCT) of the animal revealed the presence of a pedunculating tumor arising from the pectoralis major muscle, with stellate internal calcification suggestive of necrosis (Fig. 3B). On necropsy the tumor was highly vascular (Fig. 3C). Histology of the tumor at low magnification revealed the attachment of normal skeletal muscle at the tumor base (Fig. 3D). Higher power magnification revealed densely packed, small, round, blue cells with scant cytoplasm and corrugating vessels consistent with the solid variant of alveolar rhabdomyosarcoma (Fig. 3E). An example of the solid variant of alveolar rhabdomyosarcoma from human is shown in Figure 3F for comparison. The diagnosis was confirmed by positive immunohistochemistry for myogenin (Fig. 3G). Immunohistochemistry for smooth muscle actin and muscle-specific actin was also positive (data not shown). To provide supportive evidence that this tumor was nonsporadic, we genotyped two areas of the tumor, demonstrating that the conditional P3Fm allele had been clonally converted to the activated P3Fa allele (Fig. 3H), rather than exhibiting a nonactivated conditional allele or heterogeneity of nonactivated and activated alleles. Therefore, the tumor was derived from a Myf6-Cre-expressing cell.

Table 1.

Mice generated for this study

Genotype				(n)	Age (days) median, range	Tumor incidence	Onset (days)
Pax3:Fkhr heterozygote lines
Myf6ICNm/any	Pax3P3Fa/WT	—	—	228	348, 35-889	1	383
Myf6ICNm/any	Pax3P3Fa/WT	FkhrF2-3m/WT	—	96	332, 49-739	0	NA
Myf6ICNm/any	Pax3P3Fa/WT	FkhrF2-3m/WT	Trp53F2-10/WT	57	282, 76-406	0	NA
Myf6ICNm/any	Pax3P3Fa/WT	—	Trp53F2-10/WT	16	235, 89-406	0	NA
Myf6ICNm/any	Pax3P3Fa/WT	—	Trp53F2-10/F2-10	12	75, 65-295	0	NA
Myf6ICNm/any	Pax3P3Fa/WT	—	Ink4a/ARFF2-3/WT	68	235, 16-393	0	NA
Myf6ICNm/any	Pax3P3Fa/WT	—	Ink4a/ARFF2-3/F2-3	31	72, 47-295	0	NA
Control lines
Myf6ICNm/any	—	—	Trp53F2-10/WT	50	262, 41-399	0	NA
Myf6ICNm/any	—	—	Trp53F2-10/F2-10	5	72, 57-303	0	NA
Myf6ICNm/any	—	—	Ink4a/ARFF2-3/WT	57	269, 41-388	0	NA
Myf6ICNm/any	—	—	Ink4a/ARFF2-3/F2-3	18	62, 41-295	0	NA
Pax3:Fkhr homozygote lines
Myf6ICNm/any	Pax3P3Fa/P3Fa	—	—	36	219, 44-542	0	NA
Myf6ICNm/any	Pax3P3Fa/P3Fa	—	Trp53F2-10/WT	12	216, 78-406	1	202
Myf6ICNm/any	Pax3P3Fa/P3Fa	—	Trp53F2-10/F2-10	5	89, 72-295	2	75-91
Myf6ICNm/any	Pax3P3Fa/P3Fa	—	Ink4a/ARFF2-3/WT	13	195, 66-295	0	NA
Myf6ICNm/any	Pax3P3Fa/P3Fa	—	Ink4a/ARFF2-3/F2-3	14	72, 47-108	4	56-89

NA = not applicable.

any = ICNm or WT allele.

Figure 3.

Alveolar rhabdomyosarcoma in Pax3:Fkhr Myf6-Cre mice. (A) A 12-month-old female Pax3^P3Fa/WT Myf6^ICNm/WT mouse was found to have a large mass (black arrowhead) arising from the left chest wall near the axilla. (B) Computed tomography scan revealed a well-circumscribed 13.8 × 6.4 × 9-mm mass (white arrowhead) arising from a more infiltrative base. A stellate area of calcification was indicative of central necrosis within the tumor. (C) Gross dissection revealed a highly vascular, rubbery tumor. (D) Low-power magnification of the hematoxylinand eosin-stained tumor (blue) demonstrates adherent skeletal muscle (pink). (E) Higher-power magnification of the hematoxylin- and eosin-stained tumor shows densely packed cells with scant cytoplasm, consistent with the solid variant of alveolar rhabdomyosarcoma. (F) A human solid variant of alveolar rhabdomyosarcoma is shown for comparison. (G) Immunohistochemistry for myogenin is positive. (H) Genotyping of two areas of the tumor reveals clonal conversion of the conditional P3Fm allele to an activated P3Fa allele.

Haploinsufficiency of Fkhr does not significantly accelerate tumorigenesis

The reciprocal chromosome 2 and 13 translocations observed in human alveolar rhabdomyosarcomas result in four genetic consequences: formation of the Pax3:Fkhr fusion gene, formation of the reciprocal Fkhr:Pax3 fusion gene, loss of a single Pax3 allele, and perhaps most significantly, the loss of a single Fkhr allele. To test the possibility that Fkhr loss contributes to the pathogenesis of alveolar rhabdomyosarcoma, we generated a conditional allele of Fkhr and introduced it into Pax3^P3Fa/WT Myf6^ICNm/WT mice. We followed 96 Pax3^P3Fa/WT Fkhr^F2-3m/WT Myf6^ICNm/WT animals for up to 25 mo without the development of a tumor (Table 1). We therefore concluded that haploinsufficiency of Fkhr does not significantly accelerate tumorigenesis when Pax3:Fkhr is expressed in the Myf6 domain for terminally differentiating skeletal muscle.

Ink4a/ARF and Trp53 mutations accelerate tumorigenesis

Because human alveolar rhabdomyosarcomas frequently carry mutations in the Ink4a/ARF and Trp53 pathways, we wished to test whether mutations that prominently affect G1-S cell cycle re-entry could accelerate the formation of alveolar rhabdomyosarcomas from a target pool of quiescent, terminally differentiating Myf6-expressing skeletal muscle. For this goal, we introduced conditional knockout alleles for INK4a/ARF (Aguirre et al. 2003) or Trp53 (Jonkers et al. 2001) into our Pax3:Fkhr model. A total of 57 control mice with the genotype Ink4a/ARF^F2-3/WT Myf6^ICNm/WT and 50 mice with the genotype Trp53^F2-10/WT Myf6^ICNm/WT were followed for up to 13 mo without the development of alveolar rhabdomyosarcomas. We also followed 68 triple heterozygous Pax3^P3Fa/wt Ink4a/ARF^F2-3/WT Myf6^ICNm/WT mice and 16 triple heterozygous Pax3^P3Fa/wt Trp53^F2-10/WT Myf6^ICNm/WT animals for up to 13.5 mo without the development of a tumor (Table 1).

To increase the frequency of tumors, we began breeding for compound homozygotes. A total of 18 Ink4a/ARF^F2-3/F2-3 Myf6^ICNm/WT control mice and five Trp53^F2-10/F2-10 Myf6^ICNm/WT control mice remain cancer-free for 10 mo. However, in our first animal homozygous for the Pax3:Fkhr knock-in allele and homozygous for the INK4a/ARF knockout allele, we observed a soft tissue mass arising from the tongue base (Fig. 4A) at age 3 mo. MicroCT revealed a large neck mass encasing the trachea (Fig. 4B), which was also evident on necropsy (Fig. 4C). Histology of this tumor revealed classic alveolar rhabdomyosarcoma with entrapped normal skeletal muscle (Fig. 4D). Immunohistochemistry revealed the tumor to be myogenin positive, confirming the histological diagnosis. Three more mice with the identical genotype have also developed unifocal alveolar rhabdomyosarcomas of the masseter, neck, or anterior chest wall at age 2 mo (data not shown).

Figure 4.

Alveolar rhabdomyosarcoma in Pax3:Fkhr Myf6-Cre mice with cell cycle control mutations. Three animals with Trp53 or Ink4a/ARF mutations that have developed classical alveolar rhabdomyosarcomas. (A,F,K) Gross appearance of the tumors of the tongue/jaw, chest/arm, and upper and lower extremities, respectively, marked with black arrowheads. (B,G,L) MicroCT of the tumor(s) for each animal demarcated by white arrows. (C,H,M) Appearance of the tumors (white arrowheads) at necropsy. (D,I,N) Hematoxylin and eosin staining of the tumors demonstrating classical alveolar histology and entrapped skeletal muscle (D, black arrowhead), strap cell (D, inset), and multinucleated rhabdomyoblasts (N, inset). (E,J,O) Immunohistochemistry for myogenin was positive for each tumor.

Because the INK4a/ARF gene products influence both Trp53 and pRb pathways, we sought to determine which pathway was primarily responsible for rhabdomyosarcomagenesis. We began breeding for mice with compound conditional Pax3:Fkhr homozygous alleles and conditional Trp53 alleles. In these mice, we observed a soft tissue tumor arising from the right pectoralis major muscle of a 6.7-month-old Pax3^P3Fa/P3Fa Trp53^F2-10/wt Myf6^ICNm/WT animal (Fig. 4F-H). Histology of this tumor revealed classic alveolar rhabdomyosarcoma (Fig. 4I). As was the case for the prior tumor, numerous multinucleated rhabdomyoblasts were present. Immunohistochemistry revealed the tumor to be myogenin positive, confirming the histological diagnosis. In a second mouse that was homozygous instead of heterozygous for the conditional Trp53 allele, we observed earlier onset (age 3 mo) for four multifocal tumors arising from three extremities (Fig. 4K-M). These tumors also exhibited classic alveolar rhabdomyosarcoma histology and myogenin immunopositivity (Fig. 4N,O). An identical pattern of multifocal limb tumors was observed in a 2.5-month-old animal of the same genotype. By immunohistochemistry, Trp53 was absent from each of these alveolar rhabdomyosarcomas from Ink4a/ARF or Trp53 animals. We therefore suggest that Trp53 pathway disruption, either for its cell cycle control role or other roles, contributes to alveolar rhabdomyosarcoma formation in Pax3:Fkhr mice.

To answer the underlying question whether homozygosity for Pax3:Fkhr was a necessary and sufficient event for tumor formation in this model, irrespective of Trp53 or Ink4a/ARF mutations, we observed 36 double homozygote Pax3^P3Fa/P3Fa Myf6^ICNm/ICNm mice for up to 18 mo without the formation of tumors; however, a small number of mice with homozygous deletion of Ink4a/ARF or Trp53 and only a heterozygous Pax3:Fkhr allele have not developed tumors within the same period of time as animals with homozygous deletion of Ink4a/ARF or Trp53 and homozygous Pax3:Fkhr alleles, suggesting for this mouse model that two Pax3:Fkhr alleles may be required for tumorigenesis.

Murine alveolar rhabdomyosarcomas have an immunohistochemical profile similar to human alveolar rhabdomyosarcomas

To assess the similarity of this preclinical model of alveolar rhabdomyosarcoma to the human disease, we examined the immunohistochemical profile of the murine tumors. A summary is presented in Table 2, and representative staining for Pax3^P3Fa/wt Ink4a/ARF^F2-3/WT Myf6^ICNm/WT tumors is shown in Figure 5. The hepatocyte growth factor receptor, c-Met, is a known downstream target of Pax3:Fkhr expressed by alveolar rhabdomyosarcomas (Ginsberg et al. 1998) and was focally present in all murine alveolar rhabdomyosarcomas (Table 2; Fig. 5A). Pax7 expression was also present in murine alveolar rhabdomyosarcomas (Fig. 5B), as has been reported for many human alveolar rhabdomyosarcomas, albeit at low levels (Tiffin et al. 2003; Tomescu et al. 2004). In human alveolar rhabdomyosarcoma samples we tested, Pax7 expression comparable or greater than our murine tumors was seen in 6 of 32 cases (see Supplementary Fig. S1A). Similarly, the human rhabdomyosarcoma marker MyoD (Tonin et al. 1991) was uniformly expressed in the murine tumors (Fig. 5C). Trp53 was expressed in the tumor without Trp53 or Ink4a/ARF conditional deletion, but a distinction could not be made between wild-type and mutated Trp53 (Table 2). In this tumor, murine p19ARF was not expressed (data not shown), suggesting that Trp53 function was intact. Nevertheless, the downstream effector of G1-S cell cycle rerentry, Cyclin D1 (Cavenee 2002), was strongly expressed in all murine alveolar rhabdomyosarcomas (Fig. 5D). C-MYC, a proliferative factor expressed in some human alveolar rhabdomyosarcomas (Kouraklis et al. 1999) that often accompanies an apoptotic defect (Nilsson and Cleveland 2003), was strongly expressed in murine alveolar rhabdomyosarcomas for all genotypes (Fig. 5E) at comparable levels to the human rhabdomyosarcomas we tested (see Supplementary Fig. S1B); furthermore, the Pax3:Fkhr target gene and antiapoptotic factor Bcl-XL (Margue et al. 2000) was uniformly expressed across all murine rhabdomyosarcomas (Fig. 5F). A less common marker of human alveolar rhabdomyosarcomas, Neu (Mark et al. 1998), was absent from all murine Pax3:Fkhr-expressing tumors. CD44, a favorable prognostic marker of human rhabdomyosarcomas that is typically absent from alveolar rhabdomyosarcomas (Humphrey et al. 1999), was also absent in the murine alveolar rhabdomyosarcomas.

Table 2.

Immunohistochemical profile of murine alveolar rhabdomyosarcomas

Genotype			c-Met	Pax7	MyoD	Trp53	Cyclin D1	C-MYC	Bcl-XL	Neu	CD44
Myf6ICNm/any	Pax3P3Fa/WT	—	+++	+	++	++	+++	++	++	0	0
Myf6ICNm/any	Pax3P3Fa/P3Fa	Trp53F2-10/WT	++	++	++	0	+++	++	++	0	0
Myf6ICNm/any	Pax3P3Fa/P3Fa	Trp53F2-10/F2-10	++	+	++	0	+++	++	++	0	0
Myf6ICNm/any	Pax3P3Fa/P3Fa	Ink4a/ARFF2-3/F2-3	++	++	++	0	+++	++	++	0	0

Semiquantitative scoring is based on the scale used by Green et al. (1993). In brief, 0 equals no staining, + equals weakly positive staining, ++ equals strongly positive staining equivalent to the control, and +++ equals very strongly positive staining.

Figure 5.

Representative immunohistochemical profile of alveolar rhabdomyosarcomas from Pax3^P3Fa/wt Ink4a/ARF^F2-3/WT Myf6^ICNm/WT mice. Tumors expressed myogenic markers common to human alveolar rhabdomyosarcomas, including c-Met (A), Pax7 (B), and MyoD (C). (D) The cell cycle re-entry marker, Cyclin D1, was strongly expressed. (E) Other markers common to human and mouse alveolar rhabdomyosarcomas included C-MYC. (F) The anti-apoptotic protein, Bcl-XL, was strongly expressed and appeared to localize to the nuclear membrane as has been previously reported (Hsu et al. 1997). (Insets) Higher magnification of positively staining cells (white arrowhead).

Discussion

The alveolar subtype of rhabdomyosarcoma portends a poor prognosis when the disease is advanced (Wexler and Helman 1997). Although well-characterized embryonal and pleomorphic rhabdomyosarcomas mouse models exist, neither a spontaneous nor a transgenic mouse model of alveolar rhabdomyosarcoma has yet been reported. In this manuscript we report a faithful model of alveolar rhabdomyosarcoma built on the signature genetic change of this tumor. Using conditional mutagenesis to simulate events that occur as a result of a t(2;13) translocation associated with human alveolar rhabdomyosarcomas, we have been able to convert a functionally normal Pax3 allele to a Pax3:Fkhr allele. In this report, Pax3:Fkhr activation was restricted to a target pool of terminally differentiating Myf6-expressing skeletal muscle. In Pax3^P3Fa/WT Myf6^ICNm/WT mice, alveolar rhabdomyosarcomas are rare (1 in 228). However, Pax3:Fkhr homozygosity and Ink4a/ARF or Trp53 pathway loss of function significantly accelerates alveolar rhabdomyosarcoma formation from this target pool, raising the interesting possibility that some or most alveolar rhabdomyosarcomas arise from postnatal, terminally differentiating, Myf6-expressing myofibers. However, we cannot rule out the possibility that within the postnatal Myf6 cell population there is a yet-to-be-identified sub-population of primitive nondifferentiating myogenic cells, similar to those that transiently express Myf6 in the myotome of the rostral somite but that do not necessarily contribute to embryonic muscle mass (Bober et al. 1991; Summerbell et al. 2002).

These mouse alveolar rhabdomyosarcomas derived from a target pool of maturing myofibers have an immunohistochemical profile that closely parallels human alveolar rhabdomyosarcomas, including a markedly dis-regulated array of early myogenic marker expression (e.g., myogenin, MyoD, and Pax7), known Pax3:Fkhr targets (e.g., c-Met, Bcl-XL), and markers of proliferation (e.g., C-Myc and Cyclin D1). Because of this immunohistochemical similarity, the inherited immunocompetency of the conditional mouse model, and the ability to form tumors in situ, that is, from a normal muscle environment, this model promises to be valuable for studies of tumor biology as well as for preclinical therapeutic testing.

For the paradigm of accurately modeling translocation-mediated tumor initiation, there are four significant genetic changes to take into account with each fusion. In the case of the t(2;13) translocation in alveolar rhabdomyosarcomas, two fusion genes are created: Pax3:Fkhr and Fkhr:Pax3. Whereas Pax3:Fkhr is consistently transcribed in tumors and is composed of functional elements from both Pax3 and Fkhr (Arndt and Crist 1999), Fkhr:Pax3 is expressed in a fraction of tumors tested (Frascella et al. 1998) and consists of only a partial Fkhr DNA-binding domain and a partial Pax3 transactivation domain. The two less appreciated genetic changes associated with t(2;13) are the destruction of one allele of Pax3 and the destruction of one allele of Fkhr. Whereas Pax3 haploinsufficiency is known to cause Waardenburg Syndrome in humans (Tassabehji et al. 1993), a condition not associated with rhabdomyosarcomas or tumorigenesis, the effects of haploinsufficiency of the putative tumor suppressor Fkhr are unknown. To test the hypothesis that haploinsufficiency of the tumor suppressor Fkhr may, indeed, be associated with the initiation and/or progression of alveolar rhabdomyosarcoma, we introduced a conditional knockout allele of Fkhr into Pax3^P3Fa/WT Myf6^ICNm/WT mice. For this model, Fkhr haploinsufficiency does not appear to significantly accelerate tumorigenesis in Pax3^P3Fa/WT Fkhr^F2-3m/WT Myf6^ICNm/WT animals. However, we have experiments ongoing to determine whether Fkhr haploinsufficiency, as a downstream effector of PTEN, acts synergistically with Ink4a/ARF or Trp53 loss of function in the acceleration of rhabdomyosarcomagenesis.

In our model, tumors most often arise from mice homozygous for the Pax3:Fkhr allele (although the animal without Trp53 or Ink4a/ARF knockout alleles that developed a tumor at age 12 mo was heterozygous for Pax3:Fkhr). This observation raises the possibility that dosage of the Pax3:Fkhr knock-in allele is important for tumor initiation, and that in our construction of the knock-in allele, albeit as carefully as was possible, we were missing an important positive cis-regulatory element from the Fkhr 3′ genomic region, or we did not eliminate an important negative cis-regulatory element from the Pax3 3′ genomic region. Alternatively, it may simply be a result from species-specific levels of transcription from the Pax3 locus. This will be an area of further investigation.

It had been an open question whether alveolar rhabdomyosarcomas initiate before or after birth, and whether certain subsets of myogenic progenitors are more susceptible to give rise to this malignancy. Pediatric cancers are frequently associated with translocations, and these translocations are often asymptomatic at birth (Greaves and Wiemels 2003). In the accompanying manuscript (Keller et al. 2004), we have described the embryonic and postnatal consequences of Pax3:Fkhr expression within subsets of the Pax3 expression domain. These results suggest a framework for the molecular gains of function for Pax3:Fkhr, as well as the potential cell(s) of origin for alveolar rhabdomyosarcomas. The severity of birth defects in Pax3^P3Fa/wt RajCreTg^Cre/wt or Pax3^P3Fa/wt Pax7^ICNm/wt mice and associated lethality suggest that the translocation event in humans is unlikely to occur either as a germ-line mutation or as a mosaic in a broad range of embryonic myogenic precursors. Based on their normal capacity for self-renewal and proliferation during postnatal myogenesis and muscle repair, postnatal multipotent stem cells in muscle (e.g., side population cells) and committed muscle stem cells (i.e., satellite cells) are theoretically the best candidates for being the cells of origin for alveolar rhabdomyosarcomas, but until now the theory has not been tested.

Results from our accompanying paper (Keller et al. 2004) go on to demonstrate that Pax3^P3Fa/wt Pax7^ICNm/wt mice carrying the activated Pax3:Fkhr allele in the Pax7+ prenatal and postnatal satellite cell pools do not develop tumors even up to 3.5 mo of age, well into the young adulthood of these mice. By inference from this experiment for which Pax3^P3Fa/wt Pax7^ICNm/wt mice have a severely diminished satellite cell pool and inhibited Pax7 target gene activation, formation of the human 2;13 chromosomal translocation should occur in embryonic or neonatal muscle lineages only after normal Pax7 expression is no longer critical for establishment or renewal of the satellite cell pool, such as during myoblast differentiation following the postnatal response to muscle injury—or the translocation should occur in only a small minority of Pax7+ satellite cells.

Our results in this report raise the interesting possibility that terminally differentiating skeletal muscle myofibers have the capacity to give rise to alveolar rhabdomyosarcomas, which, in turn, can express markers (myogenin, MyoD, Pax7) of earlier myogenic precursors. This result may not be surprising given that fully differentiated mature skeletal muscle has been demonstrated to have the capacity to dedifferentiate into earlier precursor myogenic forms or even into different lineages (Odelberg et al. 2000). One might expect that for a terminally differentiated cell to begin replicating, loss of cell cycle checkpoint would be necessary, consistent with the fact that a quarter of human rhabdomyosarcomas have aberrant or absent Trp53 (Felix et al. 1992) or Ink4a/ARF (Iolascon et al. 1996) expression. Furthermore, growing evidence from non-alveolar rhabdomyosarcoma mouse models suggests that mutations affecting cell cycle control significantly increase rhabdomyosarcoma frequency for mice with aberrant growth factor expression (see Supplementary Table 1). With regard to cell(s) of origin, the resulting loss of replicative checkpoint(s) would perhaps be advantageous to the expansion of a myogenic precursor transforming into rhabdomyosarcoma, but would be critical to rhabdomyosarcoma arising from terminally differentiating myofibers in quiescence. We therefore postulate that many but perhaps not all human Pax3:Fkhr-expressing alveolar rhabdomyosarcomas are derived from differentiating myofibers. These cells would presumably require cell cycle checkpoint loss of function.

In summary, we report a mouse model of alveolar rhabdomyosarcoma, which has been generated by conditional activation of Pax3:Fkhr and conditional inactivation of the Trp53 pathway in a target pool of terminally differentiating, Myf6-expressing skeletal muscle. This report provides the starting point for understanding the initiation and progression of this tumor and its potential cell(s) of origin. Furthermore, this conditional genetic model will be a valuable tool for understanding the unique biology of alveolar rhabdomyosarcomas and will serve as an exciting new preclinical platform for testing new therapeutic strategies.

Materials and methods

Genomic clones isolation

Genomic DNA clones were isolated from a λ bacteriophage library of mouse strain SvJ-129 (Stratagene). These clones consisted of a 14.2-kb clone spanning intron 7 to the 3′ untranslated region of Pax3, an 18.9-kb clone spanning intron 2 to the 3′ untranslated region of Fkhr, and a 12-kb clone spanning exon 8 and the 3′ untranslated region of Myf6.

Targeted mouse line production

For the Pax3:Fkhr conditional mouse line, a targeting vector was constructed for which LoxP sites flank exons 8-10 of Pax3 at AflII and BstXI sites, respectively. The compact polyadenylation signal and transcriptional pause elements from the Complement2 gene 3′ untranslated region (Ashfield et al. 1991; Moreira et al. 1995) is inserted upstream of the native Pax3 polyadenylation signal and the 3′ LoxP site. The 3′ LoxP site is followed by Fkhr genomic DNA containing 338 bp of intron 1, 2.9 kb containing exon 2/intron 2/exon 3, and 6.5 kb of the 3′ untranslated region and genomic DNA. An internal ribosome entry site from encephalomyocardial virus and the enhanced yellow fluorescent protein eYfp were inserted in exon 3 at a BglII site. This targeting vector was electroporated into R1 embryonic stem cells (Nagy et al. 1993), and the cells were subjected to positive and negative selection (Mansour et al. 1988). Genomic DNA from 144 colonies was analyzed by Southern hybridization using a 3′ external probe to identify 12 clones with a targeted modification of the Pax3 locus. Southern hybridization using an internal probe and EcoRI-digested genomic DNA was used to further identify three clones that contained the desired 5′ LoxP site from the targeting vector. Cells from these three recombinant clones were microinjected into C57BL/6 blastocysts to generate chimeric mice. Chimeric mice were mated to C57BL/6 females, and their Agouti offspring were tested by Southern hybridization to confirm germ-line transmission of the Pax3:Fkhr conditional allele. The FRT-flanked neomycin resistance gene (Neo) was removed by breeding Pax3^P3Fp/wt mice to transgenic mice expressing Flp-e (Rodriguez et al. 2000), thereby generating Pax3^P3Fm/wt mice.

For the Fkhr conditional knockout mouse line, a targeting vector was constructed for which Lox511 sites flank exons 2 and 3 of the Fkhr gene. This targeting vector was electroporated into R1 embryonic stem cells subjected to positive and negative selection. Genomic DNA from 71 colonies was analyzed by Southern hybridization using a 3′ external probe to identify 41 clones with a targeted modification of the Fkhr locus. Chimeric and germ-line mice were produced as discussed above.

For the Myf6-Cre mouse line, a targeting vector was constructed that placed an ires-Cre-FRT-Neo-FRT cassette at an engineered XhoI site within the untranslated region of the third exon of the Myf6 gene, 226 bp after the stop codon. This targeting vector shown in Figure 1C was electroporated into R1 embryonic stem cells subjected to positive and negative selection. Two of 70 clones were identified as positive by screening with AflII, NcoI DNA digests and Southern hybridization using a 5′ external probe, as demonstrated by a downshift from 11.2 to 7.3 kb (data not shown). Chimeric and germ-line mice were produced as discussed above.

Genotyping

Mice carrying targeted alleles were genotyped by PCR. For the Pax3:Fkhr mouse line, the 5′ and 3′ primers were ck82, 5′-GGATTGCATGGTTGGGATGTAGCTG-3′; ck83, 5′-CAAAGAAGCAACAGAGGGAGATG-3′; and ck10 5′-TGAGCATCCACCAAGAACTC-3′; respectively. DNA was extracted from tails, and 2 μL (5-20 ng) was used in the subsequent PCR reaction. Each 25 μL of PCR reaction contained 1× buffer, 2.5 mM MgCl₂, 200 μM deoxynucleotides, 0.6 μM ck82, 0.3 μM ck83, 0.3 μM ck10, and 0.75 U of Taq DNA polymerase (Promega). The cycling conditions were 5 min at 95°C, 27 cycles of 30 sec at 95°C/20 sec at 64°C/60 sec at 72°C, followed by 7 min at 72°C. The wild-type, P3Fm, or P3Fa alleles resulted in 196-bp, 250-bp, or 575-bp bands, respectively.

For genotyping Fkhr mouse lines, the 5′ and 3′ primers were ck142, 5′-CCTTCAGAGCTGCCAGGTGAATATG-3′ and ck143, 5′-TGTAGAGAGTATGCCGTCCAGAGTGG-3′; ck30, 5′-GTGTACAAACCAGCTGAGCAC-3′; respectively. Reaction conditions were as described above except that 0.4 μM each primer was used. Cycling conditions were 5 min at 95°C, 32 cycles of 30 sec at 95°C/20 sec at 64°C/60 sec at 72°C, followed by 7 min at 72°C. The wild-type, F2-3p, or F2-3m alleles resulted in 290-bp, 350-bp, or 661-bp bands, respectively.

For genotyping Myf6-Cre mouse lines, the 5′ and 3′ primers were ck219, 5′-CTTCAGCGCCTTTCTTCCATCG-3′; ba86, 5′-GGATAGTGAAACAGGGCAA-3′; and ba88, 5′-GAACTTCGTTTGCAACTGAC-3′; ba97, 5′-GATCTGGACGAAGAGCATCA-3′; respectively. Reaction conditions were as described above except that 0.8 μM each primer was used. Cycling conditions were the same as for Pax3:Fkhr. The wild-type, ICNm, and ICNp alleles resulted in 398-bp, 300-bp, or 550-bp bands, respectively.

Testing Myf6 driver expression

Whole-mount detection of the tissue-specific CRE expression for the Myf6-Cre mouse line was performed by crossing that line to ROSA26 LacZ reporter mice (Sorriano 1999). Embryos were harvested at E15.5 and processed as previously described (Arenkiel et al. 2003).

Detection of Pax3:Fkhr rearrangements in various tissues

Genotyping as described above was performed on various tissues from 6-week-old mice with the genotype Pax3^P3Fa/WT Myf6^ICNm/WT to determine the spectrum of Myf6-Cre-driven recombination for the Pax3:Fkhr knock-in allele. For each organ, 200 mg of tissue was diced and digested overnight in 330 μg/mL proteinase-K in 100 mM NaCl, 20 mM Tris-Cl (pH 7.6), 10 mM EDTA, and 0.5% SDS. The following day, samples were extracted with phenol, extracted with chloroform, and precipitated with 2.5 volumes of ethanol. Samples were resuspended in TE, and 50 ng was used for the Pax3:Fkhr PCR genotyping reaction described above.

Detection of Pax3:Fkhr transcripts and eYFP marker gene

For epifluorescence microscopy of skeletal muscle for Pax3:Fkhr mice, 6-week-old animals were killed by CO₂ asphyxiation and perfused with PBS. The vastus lateralis muscle was dissected and fixed in 4% paraformaldehyde in PBS for 3 h at 4°C. The tissue was washed for 10 min in cold PBS four times, then transitioned from 10% to 20%-30% sucrose in PBS at 60-min intervals. The tissue was embedded in OCT compound (Sakura Finetek U.S.A., Inc.) and snap-frozen on dry ice. Ten-micron-thick cross-sections were subsequently washed in 0.5% Triton X-100 in PBS and coverslipped using FluorMount-G mounting media (Southern Biotech). Slides were visualized with a Zeiss Axiovert 200M epifluorescence microscope using a yfp filter.

Histology and immunohistochemistry of tumors

Histology was performed as previously described (Boulet and Capecchi 2004). Immunohistochemistry for myogenin, smooth muscle actin, muscle-specific actin, Cyclin D1, C-Myc, and Bcl-XL was performed by ARUP Laboratories at the University of Utah on 3.5-μm sections. A rabbit anti-mouse F_AB2 secondary antibody (Dako Corp.) was used to avoid cross-reactivity with nonspecific mouse tissue antigens. For Pax7, MyoD, and myogenin immunohistochemistry on frozen skeletal muscle (prepared as described in the previous paragraph), staining was performed using the M.O.M. Immunodection Kit Staining Procedure (Vector Laboratories) following the manufacturer's instructions, without antigen unmasking. The Pax7 monoclonal primary antibody (Pax7; Developmental Hybridoma Studies Bank, Iowa City, IA) was used at a concentration of 1:10. The MyoD polyclonal primary antibody (PRB-106C; Covance Corp.) was used at a concentration of 1:3000. The myogenin monoclonal primary antibody (M3559; DAKO Corp.) was used at a concentration of 1:50. Slides were coverslipped with Fluoromount-G (Southern Biotechnology Associates) or DAPI mounting medium (Vector Laboratories) and visualized by standard light microscopy. Immunohistochemistry using rabbit polyclonal anti-Pax3 (Li et al. 1999) (a kind gift from J. Epstein, University of Pennsylvania, Philadelphia, PA), and rabbit polyclonal anti-cMet (Jeffers et al. 1997) (a kind gift from G. Vande-Woude, Van Andel Research Institute, Grand Rapids, MI) were performed as previously described.

Computed tomography scanning

Computed tomography of sacrificed mice was performed at 93 μm³ voxel resolution using an eXplore Locus Small Animal MicroCT Scanner (GE Healthcare). Images were reconstructed and visualized with the manufacturer's MicroView freeware.