Mutations in the Notch Pathway Alter the Patterning of Multifidus


  • Rebecca E. Fisher,

    Corresponding author
    1. Department of Basic Medical Sciences, The University of Arizona College of Medicine–Phoenix, Phoenix, Arizona
    2. School of Life Sciences, Arizona State University, Tempe, Arizona
    • Department of Basic Medical Sciences, The University of Arizona College of Medicine-Phoenix, 550 East Van Buren Street, Phoenix, AZ 85004
    Search for more papers by this author
    • Fax: 602 827-2212

  • Heather F. Smith,

    1. Department of Basic Medical Sciences, The University of Arizona College of Medicine–Phoenix, Phoenix, Arizona
    Current affiliation:
    1. Department of Anatomy, Arizona College of Osteopathic Medicine, Midwestern University, Glendale, AZ 85308
    Search for more papers by this author
    • Rebecca E. Fisher and Heather F. Smith contributed equally to this work.

  • Kenro Kusumi,

    1. School of Life Sciences, Arizona State University, Tempe, Arizona
    Search for more papers by this author
  • Erica E. Tassone,

    1. School of Life Sciences, Arizona State University, Tempe, Arizona
    Search for more papers by this author
  • Alan Rawls,

    1. School of Life Sciences, Arizona State University, Tempe, Arizona
    Search for more papers by this author
  • Jeanne Wilson-Rawls

    1. School of Life Sciences, Arizona State University, Tempe, Arizona
    Search for more papers by this author


Clinical studies have suggested that defects in the epaxial muscles, particularly multifidus, may contribute to the etiology of idiopathic scoliosis. While the epaxial muscles and the vertebrae derive from the same embryonic segmentation process, the mechanisms that pattern the multisegmental back muscles are still unclear. The process of segmentation is regulated by the Notch signaling pathway, and mutations in the modulators delta-like 3 (Dll3) and lunatic fringe (Lfng) are genetic models for spinal disorders such as scoliosis. Osteological defects have been characterized in these genetic models, but myological phenotypes have not previously been studied. We analyzed the multifidus muscle in the mouse (Mus musculus) and observed intriguing changes in the cranio-caudal borders of multifidus in Dll3 and Lfng models. Statistical analysis did not find a significant association between the majority of the multifidus anomalies and the vertebral defects, suggesting a previously unappreciated role for Notch signaling in patterning epaxial muscle groups. These findings indicate an additional mechanism by which DLL3 and LFNG may play a role in the etiology of human idiopathic scoliosis. Anat Rec, 2012. © 2011 Wiley Periodicals, Inc.

The vertebral column is composed of alternating vertebrae and intervertebral discs that are supported by spinal ligaments and the paravertebral musculature. All of these tissues are derived from somites, clusters of paraxial mesoderm that are deposited in a metameric pattern on either side of the neural tube early in development. Each somite differentiates into four cell lineage-specific compartments, including the sclerotome (vertebrae and ribs), syndetome (axial tendons), myotome (axial skeletal muscle), and dermomyotome (dermis and skeletal muscle progenitor cells). As each of these compartments give rise to key elements of the spine, disruptions of somite differentiation may result in debilitating spinal defects.

The importance of properly coordinated spinal development is underscored in patients with scoliosis. Scoliosis affects up to 3% of the general population and is defined by the presence of a lateral curvature of greater than 10° (reviewed in Kusumi and Turnpenny, 2007; Alman, 2010). In actuality, scoliosis is often a complex, three-dimensional defect with deformation occurring in the frontal and sagittal planes along with a rotational component (Meier et al., 1997). Congenital scoliosis is difficult to manage clinically, as the curves are resistant to correction and tend to progress into debilitating deformities (McMaster, 2001). Malformation of the vertebrae can produce the lateral curvatures of congenital scoliosis (Erol et al., 2004). However, the majority of scoliosis deformities are classified as “idiopathic,” with an unknown etiology. Gao et al. (2007) found that 2%–3% of school-aged children have idiopathic scoliosis. In these cases, it is possible that disruptions of the paravertebral musculature could result in decreased stability of the vertebral column, and increased lateral flexion and rotation of the spine.

Stabilization and movements of the vertebral column are dependent on the precise patterning of paravertebral muscles and their attachments to skeletal elements through tendons. In tetrapods, the paravertebral musculature is arranged in a series of longitudinal columns, innervated by the dorsal rami of spinal nerves. These muscles are divided into four layers in humans, including a superficial splenius layer, the erector spinae, the transversospinalis, and a deep layer including the levatores costarum, intertransversarii, and interspinales (Fig. 1). The transversospinalis group is itself comprised of three subdivisions: the semispinalis, rotatores, and multifidus (Fig. 1B). In cases of idiopathic scoliosis, structural and electromyographic asymmetry of the paravertebral muscles has been documented (Riddle and Roaf, 1955; Zuk, 1962; Butterworth and James, 1969; Spencer and Zorab, 1976; Alexander and Season, 1978; Yarom and Robin, 1979; Reuber et al., 1983; Sahgal et al., 1983; Zetterberg et al., 1983; Mannion et al., 1998), and a number of studies have reported asymmetry in the multifidus muscle in particular (e.g., fiber type distribution, hypertrophy, atrophy, centralization of nuclei, and disruption of sarcotubular and myofibrillar elements; Fidler and Jowett, 1976; Khosla et al., 1980; Ford et al., 1984; Bylund et al., 1987; Meier et al., 1997; Chan et al., 1999). Whether this asymmetry is responsible for the initiation of idiopathic scoliosis, its progression, or both, remains unclear.

Figure 1.

The epaxial muscles of the back. (A) On the left, the splenius muscles; on the right, the erector spinae muscles, including iliocostalis, longissimus, and spinalis. (B) On the left, the transversospinalis muscles, including semispinalis, multifidus, and rotatores; on the right, the levatores costarum, intertransversarii, and interspinales muscles. Drawing by Brent Adrian.

Currently, little is known about the signals that direct the patterning of the paravertebral muscles (reviewed in Rawls and Fisher, 2010). Embryonic muscle masses are responsive to signals that establish the general body plan. The best described are the Hox genes, which impose regional identity along the cranio-caudal axis of both the body and limbs (Hashimoto et al., 1999; Alvares et al., 2003). At the level of the limbs, migrating myogenic cells respond to Hox signaling from the surrounding lateral plate mesoderm. In other regions, there is evidence that Hox genes are able to instruct myoblasts in a cell autonomous manner (Alvares et al., 2003). The interplay between signaling factors contributing to dorsoventral and proximodistal pattering in the limb bud, including FGFs, SHH, and BMPs, also impose identity on muscle masses (Riddle et al., 1995; Vogel et al., 1995). The further segregation of muscle masses into anatomically distinct muscles is dependent on signals from neighboring tissue. In the limb, muscle patterning requires reciprocal interactions with the adjacent tendon primordial (Kardon, 1998). Additional signals from the surrounding mesoderm have been described that are dependent on the expression of TCF4 (Grim and Wachtler, 1991; Kardon et al., 2003).

The Notch signaling pathway is a candidate for regulating skeletal muscle patterning. This pathway has been associated with the regulation of the segmental clock and the differentiation of skeletal muscle. Genetic disruptions in the modulators of Notch signaling, lunatic fringe (Lfng) and delta-like 3 (Dll3), have been shown to produce severe segmental defects affecting the vertebrae and ribs in both mouse models and in the autosomal recessive congenital vertebral disorder, spondylocostal dysostosis (Fig. 2; Evrard et al., 1998; Kusumi et al., 1998; Zhang and Gridley, 1998; Bulman et al., 2000; Dunwoodie et al., 2002; Sparrow et al., 2006). While both mutations produce segmental defects, the effects of segmentation and vertebral malformations are distinct due to their differing roles in Notch regulation (reviewed in Turnpenny, 2010). Both Lfng and Dll3 continue to be expressed in the rostral and caudal compartments of somites, respectively, after segmentation, and the significance of this somitic expression is not clear (Dunwoodie et al., 1997; Aulehla and Johnson, 1999). While studies have described the osteological malformations due to Lfng and Dll3 mutations, the effects on the paravertebral musculature have not been reported. We analyzed the multifidus muscle in the mouse (Mus musculus) and observed defects due to null mutations in two different genes in the Notch signaling pathway, Lfng and Dll3. The observation of global changes in multifidus in both Lfng and Dll3 mutants suggests a previously unappreciated role for Notch signaling in the patterning of epaxial muscles.

Figure 2.

Lfng and Dll3 mutants display severe vertebral and costal defects.



Mouse Dll3pu, Dll3tm1Rbe (referred to as Dll3neo), Lfngtm1Rjo (null animals referred to as Lfng–/–), C57BL/6J, and CD1 lines were used for analysis. All animals were maintained in an AALAC certified facility with standard light cycles and an ad libitum food and water regimen under protocols approved by the Institutional Animal Care and Use Committee at Arizona State University. Dll3neo and Lfngtm1Rjo mutations were maintained on a C57BL/6J background, having been backcrossed 10 generations or greater onto this strain. Since homozygous mutants are not viable neonatally on a C57BL/6J background, outcrosses to the CD1 line were required to produce viable adult mutants. Lfngtm1Rjo were also maintained by intercross with CD1 for viability. Dll3pu animals were maintained by intercrosses of heterozygous animals for this inbred line. Homozygous animals were generated by an intercross of heterozygous carriers. Euthanasia was carried out by carbon dioxide inhalation. Genotype was determined using PCR-based assays, as described previously (Evrard et al., 1998; Kusumi et al., 1998; Dunwoodie et al., 2002).

Analysis of Multifidus

While the myology of the laboratory rat (Rattus norvegicus) is well documented (e.g., Hebel and Stromberg, 1976; Brink and Pfaff, 1980; Wingerd, 1988; Chiasson 1994; Walker and Homberger, 1997), detailed muscle descriptions are largely lacking for the laboratory mouse (Mus musculus; e.g., Hummel et al., 1975; Cook, 1976; Feldman and Seely, 1988; Popesko et al., 1992; Komárek, 2004). Origin and insertion data are available for a limited number of paravertebral muscles, including multifidus, but these data are confined to the T12 to L6 vertebral levels and the number of specimens sampled is relatively small (Cornwall et al., 2010, N = 3; Hesse et al., 2010, N = 5). As a result, we first examined the multifidus muscles in 12 wild-type specimens from our colony of C57BL/6J-CD1 hybrids. We then analyzed littermates who were homozygous null for Lfng and Dll3. All dissections were carried out bilaterally on adult animals (6–8 weeks) using a Nikon SMZ800 stereodissecting microscope and photodocumented using a Nikon Coolpix 4500 digital camera. For each specimen, we noted both the origins and insertions of multifidus, or its unilateral or bilateral absence.

Analysis of Vertebrae

After the analysis of multifidus was complete, specimens were prepared for osteological analysis by fixation in ethanol followed by staining with Alizarin Red/Alcian Blue, as described previously (McLeod, 1980; Kusumi et al., 1998). In wild-type and mutant specimens, vertebral malformations were assessed at muscle origin and insertion sites, including the spinous processes, transverse processes, mammillary processes, and cranial and caudal articular processes. The cervical, thoracic, and lumbar vertebrae were disarticulated from each other, and from the cranium and sacrum. The processes of each vertebra were then visually inspected, and any defects were recorded. Vertebral malformations were treated as a dichotomous variable in which defects at each attachment point were scored as present or absent for each vertebral level.

Statistical Analyses

A Yates corrected chi-squared (χ2) test with the α level set at 0.05 was conducted to determine whether statistically significant differences exist between the frequencies of multifidus defects in mutant versus wild-type littermates. This analysis was also conducted separately for Dll3 and Lfng mutants. Although it has been reported previously that mutations in the Notch signaling pathway lead to osteological segmentation defects (Evrard et al., 1998; Kusumi et al., 1998; Zhang and Gridley, 1998; Dunwoodie et al., 2002), a second chi-squared test was performed between the frequencies of vertebral defects in the mutant and wild-type specimens to statistically test these observations in the current sample.

Several statistical tests were conducted using Statistica 4 to assess whether a significant association exists between the multifidus defects and the vertebral defects. Two-way analysis of variance (ANOVA) and subsequent Tukey post hoc comparisons were used to test whether differences in the levels of vertebral defects exist between genotypes or multifidus defect category. In this analysis, multifidus category was treated as a categorical variable with four possible states: normal, reduced, expanded, or absent. Regression Analyses were also conducted for the cranio-caudal extents of vertebral and multifidus defects. Significant correlation coefficients (P < 0.005) were interpreted as evidence of a relationship between the levels of vertebral defects and the vertebral levels of defects in multifidus.


In the mouse, the multifidus muscles originate on a mammillary (lumbar region) or transverse process (thoracic region) and insert onto a spinous process two to four vertebrae craniad to the site of origin. These muscles span from the L5 to T2 vertebral levels. In Lfng mutants, 13/15 specimens displayed multifidus anomalies. Anterograde shifts were observed in both the cranial and caudal borders of this muscle group (Fig. 3). The most common anomaly was a unilateral or bilateral anterograde shift of the cranial multifidus border to C2 or C1 rather than T2 (N = 6/15; Figs. 3 and 4). This anterograde shift was associated with the addition of new muscle segments rather than an increase in muscle fiber length with novel insertion sites craniad to T2. In addition, the presence of multifidus in the cervical region was not associated with a reduction or loss of the epaxial muscles that normally populate this region.

Figure 3.

Lfng and Dll3 mutants display anomalous cranial and caudal borders of the multifidus muscle. Multifidus muscles are shown as left-right paired columns indicating vertebral level. Observations from each specimen are shown. The arrows indicate the location of the cranial and caudal borders of multifidus in the wild-type condition. Vertebral levels: C, cervical; T, thoracic; L, lumbar.

Figure 4.

Comparison of the cranial border of the multifidus muscle (M) in a wild-type versus mutant Lfng mouse. Note the anterograde shift to a C2 insertion, compared to the T2 insertion in the wild-type. All of the other epaxial muscles have been removed from the neck in these views. Scale bar is 1 mm.

Posterograde shifts of the cranial multifidus border were also observed bilaterally or unilaterally, with a cranial insertion point at only T6 or T9 rather than T2 (N = 3/15; Fig. 3). In three specimens, both cranial and caudal borders were shifted in an anterograde fashion, with cranial borders at C2 and caudal borders at T6, T12, or L1 (N = 3/15; Fig. 3). In one Lfng mutant specimen, there was a complete bilateral absence of the multifidus muscle (N = 1/15; Fig. 3). Yates corrected χ2 analysis revealed highly significant differences in the frequency of multifidus defects in Lfng mutants (χ2 = 19.55, P < 0.0001) compared to their wild-type littermates. Significant changes in the cranial border of multifidus were also observed in Dll3 mutants (Fig. 3; N = 9/12; χ2 = 13.67, P < 0.0001).

While the developmental processes regulating muscle and bone formation are clearly distinct, defects in underlying osteological structures could lead to alterations in muscle attachment sites and patterning. To distinguish between muscle defects that are secondary to skeletal anomalies versus those that may arise from a role for Notch regulation in muscle patterning, we carried out an osteological analysis on all specimens. Mouse mutations of Lfng and Dll3 have been previously described to display widespread vertebral and rib defects (Grüneberg, 1961; Evrard et al., 1998; Kusumi et al., 1998; Zhang and Gridley 1998; Dunwoodie et al., 2002). We confirmed that defects occur at almost all vertebral levels in Dll3 and Lfng homozygous mutants (Fig. 2). In addition, the Lfng and Dll3 mutant mice exhibited rib fusions and bifurcations that led to highly variable intercostal distances. Osteological defects were not observed in any of 12 wild-type littermates analyzed.

In mice, the multifidus muscles course between a transverse or mammillary process and a spinous process. If the multifidus border shifts observed in the Lfng and DII3 mutant mice are secondary to vertebral defects, then defects in these processes should correspond with the vertebral levels of myological interruption. The vertebral defects were much more severe in the lumbar region than in the thoracic and cervical regions in both the Lfng and Dll3 mutant mice. Though fusions were observed between vertebrae, the axial level of individual transverse and spinous processes could be distinguished in the cervical and thoracic regions. Importantly, the spinous process at T2 was morphologically distinct and consistently present in both mutant strains. A two-way ANOVA of the vertebral levels of skeletal defects by both genotype (wild-type, Lfng−/−, Dll3−/−) and muscular category (normal, reduced, expanded) revealed no significant differences overall in the level of vertebral defects among normal, reduced and expanded muscles (P = 0.486). However, breaking down the analysis by side and cranio-caudal extent revealed that for the right caudal-most extent, muscular defect categories (normal, reduced, expanded, and absent) were significantly different overall (Table 1). Additionally, a regression analysis between the cranio-caudal extents of the multifidus muscle and vertebral defects revealed a significant correlation between the caudal-most osteological defect and the caudal-most myological defect in the DII3 mutant mice (Table 2). However, for the majority of cases, there was no statistical association between the cranio-caudal borders of the multifidus muscle and defects in its attachment points on the vertebrae.

Table 1. Results of two-way ANOVA and Tukey post hoc tests of the vertebral levels of relevant osteological defects by both genotype and multifidus defect category
 Right cranial-most defectsLeft cranial-most defectsRight caudal-most defectsLeft caudal-most defects
  1. Significant correlations are indicated in bold.

Multifidus defect categoriesF = 1.482 (P = 0.253)F = 0.9553 (P = 0.403)F = 5.70 (P = 0.027)F = 2.61 (P = 0.122)
GenotypesF = 6.078 (P = 0.240)F = 3.275 (P = 0.087)F = 2.06 (P = 0.168)F = 1.66 (P = 0.213)
Significant Tukey testsNoneNoneReduced vs. normal (P = 0.0138)None
Table 2. Results of regression analyses between cranio-caudal levels of relevant vertebral defects and multifidus border defects
 Right cranial-most defectsLeft cranial-most defectsRight caudal-most defectsLeft caudal-most defects
  1. Significant correlations are indicated in bold. NSD, no significant difference.

LfngR = 0.450 (P = NSD)R = 0.298 (P = NSD)R = 0.180 (P = NSD)R = 0.210 (P = NSD)
Dll3R = 0.0286 (P = NSD)R = 0.264 (P = NSD)R = 1.000 (P < 0.05)R = 1.000 (P < 0.05)
OverallR = 0.318 (P = NSD)R = 0.201 (P = NSD)R = 0.296 (P = NSD)R = 0.325 (P = NSD)


The epaxial muscle groups that span adjacent vertebrae (e.g., rotatores) are derived from individual somite myotomal compartments. However, when epaxial muscles span multiple vertebral levels, the cells from multiple myotomes contribute to individual muscle fibers (Huang et al., 2000). Studies performed in rat embryos indicate that these transversospinalis muscles first attach at the craniad insertion point and extend caudally (Deries et al., 2010). This requires large scale patterning cues, which are still largely unknown, that delimit the cranial and caudal borders of these muscles. Here we report abnormalities in the cranial and caudal boundaries of the multifidus muscles in adult mice deficient for either Lfng or Dll3. Although osteological defects are observed in the ribs and vertebrae of these mice, alterations in the muscle borders do not simply map onto these deficits. These findings reveal a previously unappreciated role of the Notch pathway in the proper patterning of epaxial muscle groups.

The Notch pathway has been implicated in regulating the embryonic development of the musculoskeletal system. It functions as a key component of the segmental clock that controls the periodicity of somite formation and ultimately the segmental organization of the axial skeleton (reviewed in Kusumi et al., 2010). Notch1 and Dll1 have been implicated in the inhibition of muscle differentiation of myogenic progenitor cells during embryonic development (Hirsinger et al., 2001; Schuster-Gosser et al., 2007). Inactivation of Dll1 results in precocious muscle differentiation and hypomorphic fetal muscles associated with depletion of progenitor cells (Schuster-Gosser et al., 2007). This is in contrast to the muscle anomalies observed in the Dll3 and Lfng deficient mice, where the anomalies were restricted to a small number of muscle groups and led to either the gain or loss of muscle. This predicts that Notch signaling can assume multiple roles in myogenesis, as demonstrated by different phenotypes from disruptions of Notch activation (Dll1 and Notch1) versus mutations in Notch inhibitors (Dll3 and Lfng).

Specification along the cranial-caudal axis has typically been associated with Hox genes, both in the development of the hindbrain as well as specifying vertebral identity (reviewed in Alexander et al., 2009). We observed that mutations in the Notch pathway genes Dll3 and Lfng led to a cranial shift in the normal multifidus T2 boundary towards C1–C2. In comparing these findings with Hox mutation phenotypes, single gene mutations in the Group 4 genes (Hoxa4, Hoxb4, Hoxd4) produced axial phenotypes in C2–C3 levels while mutations in Group 5 genes (Hoxa5, Hoxb5) produced phenotypes in C6–C7 levels (Fig. 5). Disruption of the entire Hox paralogous Group 5 set of genes led to axial disruptions ranging from C3 to T2 levels. Thus, the extension of the cranial multifidus border in Notch mutants corresponds to a region where Hox Group 5 genes play a key functional role in the paraxial mesoderm. Similarly, the caudal border of the multifidus is shifted cranially in Lfng mutant animals, in a zone corresponding to the lumbar region where Hox paralogous Group 9 plays a functional role (Fig. 5).

Figure 5.

Summary of mouse axial phenotypes characterized in Hox single gene and paralogous group disruptions and observations presented from analysis of Lfng and Dll3 mutant animals. The vertebral regions altered by deletion of individual Hox genes and paralogs are presented on the left side of the figure (reviewed in Alexander et al., 2009). The presence of multifidus abnormalities observed in Lfng and Dll3 mutants are presented as a HEAT map, considering axial halves independently for analysis. Vertebral levels: C, cervical; T, thoracic; L, lumbar.

Genetic interactions between the Hox and Notch pathway genes have been investigated. In the paraxial mesoderm, Notch signaling regulates Hoxd1 cyclical expression (Zákány et al., 2001) leading to changes in Hox gene expression and vertebral identity (Cordes et al., 2004). Evidence of converse Hox regulation of Notch pathway genes has been found in hindbrain rhombomeres, where Hoxb1 activates Notch regulation of neural stem cells (Gouti and Gavalas, 2008). Further analysis will be required to determine the mechanism by which Notch signaling regulates the cranial-caudal borders of epaxial muscles such as multifidus. These investigations may also help elucidate the etiology of idiopathic scoliosis, as disruptions of the normal development and patterning of these muscles could result in decreased stability of the vertebral column and increased lateral flexion and rotation of the spine. Evidence that Notch signaling plays a role in patterning the multifidus muscle indicates that the genes DLL3 and LFNG may play a role in the etiology of idiopathic scoliosis.


The authors would like to thank Brent Adrian, Brian Beres, Rajani George, and Allanceson Smith for technical assistance. We are also grateful to Brent Adrian for producing the illustrations in Fig. 1. The authors would like to thank Sally Dunwoodie, Randy Johnson, and Susan Cole for their gift of the Dll3tm1Rbe and Lfngtm1Rjo mouse mutations, respectively. In addition, we would like to thank Stephen Pratt for reviewing the manuscript.