The effects of hypermuscularity on shoulder morphology in myostatin-deficient mice


  • David J. Green,

    1. Center for the Advanced Study of Hominid Paleobiology, Department of Anthropology, The George Washington University, Washington, D.C., USA
    Search for more papers by this author
  • Mark W. Hamrick,

    1. Department of Cellular Biology and Anatomy, Medical College of Georgia, Augusta, GA, USA
    Search for more papers by this author
  • Brian G. Richmond

    1. Center for the Advanced Study of Hominid Paleobiology, Department of Anthropology, The George Washington University, Washington, D.C., USA
    2. Human Origins Program, National Museum of Natural History, Smithsonian Institution, Washington, D.C., USA
    Search for more papers by this author

David J. Green, Department of Anthropology, The George Washington University, 2110 G St., NW, Washington, DC 20052, USA. T: + 1 202 9947152; F: + 1 202 9946097;


Mechanical loads, particularly those generated by skeletal muscle, play a significant role in determining long-bone shape and strength, but it is less clear how these loads influence the morphology of flat bones like the scapula. While scapular morphology has been shown to vary with locomotor mode in mammals, this study seeks to better understand whether genetically modified muscle size can influence scapular shape in the absence of significant locomotor differences. The soft- and hard-tissue morphological characteristics were examined in 11 hypermuscular, mutant (myostatin-deficient), 20 heterozygote, and 15 wild-type mouse shoulders. Body mass did not significantly differ among the genotype groups, but homozygous mutant and heterozygote mice had significantly larger shoulder muscles than wild-type mice. Mutant mice also differed significantly from the wild-type controls in several aspects of scapular size and shape, including glenohumeral joint orientation, total scapular length, superior border length, and supraspinous and infraspinous fossa length. Conversely, several traits describing superoinferior scapular breadth measures (e.g. total breadth and dorsal scapular fossa breadth) did not significantly differ between mutant and wild-type mice. Since the intrinsic musculature of the scapula is oriented in a mediolateral fashion, it follows that mediolaterally configured hard-tissue features like scapular length were most distinct among genotype groups. As had been noted previously with long bones, this study demonstrates that genetically enhanced muscle size has marked effects on the morphological characteristics of the shoulder.


The mechanical loads imparted on bone emanating from muscle can dramatically influence its morphology (Robling et al. 2006; Ruff et al. 2006; Robling, 2009). Likewise, developmental variation in muscle size, morphology, and volume can also influence aspects of bone shape such as the size of muscle attachment sites and long bone curvatures (Swartz, 1990; Hamrick, 2003). While some studies have attempted to better define the role of muscle in bone development using models of muscle atrophy or injury (Lanyon, 1980; Rodriguez et al. 1988; Turner, 2007; Robling, 2009; Gross et al. 2010), other studies have explored how bone strength and cross-sectional properties can be affected by genetically enhancing muscle size (McPherron et al. 1997; Hamrick, 2003; Montgomery et al. 2005; Hamrick et al. 2006a; Ravosa et al. 2007; Vecchione et al. 2007; Elkasrawy & Hamrick, 2010). A significant amount of this work has focused on long bones as the primary weight-bearing bones of the skeleton (Lanyon, 1980; Lanyon & Rubin, 1984; Robling et al. 2006; Ruff et al. 2006). This study, however, tests the effects of genetically enhanced musculature on the hard-tissue anatomy of the shoulder.

Scapular morphology has been closely linked with locomotor differences among extant primate taxa (Inman et al. 1944; Oxnard, 1963, 1967; Ashton & Oxnard, 1964; Roberts, 1974; Hunt, 1991; Larson, 1995; Young, 2002), and so it is of particular interest for physical anthropologists and functional anatomists in general. As the scapula is tethered to the body via musculotendinous attachments and the relatively weak acromioclavicular joint, increasing the size of surrounding musculature ought to have significant morphological effects. Roberts (1974) noted that normal scapular form is contingent upon the regular stresses induced by the surrounding musculature. To further test this observation, we used a mouse knockout model lacking the growth/differentiation factor-8 gene (GDF-8, also referred to as myostatin), a member of the transforming growth factor-β superfamily. GDF-8 is a negative regulator of muscle growth, such that myostatin-deficient mice possess significantly larger muscles than control mice while remaining completely viable (McPherron et al. 1997). By targeting this gene and disrupting its pathway, McPherron et al. (1997) found that myostatin-deficient mice were larger in body size and also possessed muscles that were two to three times larger than control mice. McPherron & Lee (1997) also identified mutations in the myostatin sequences of two hypermuscular domestic cattle breeds, further confirming the role of this gene in determining muscle growth and size.

Following the discovery of the GDF-8 role in muscle development, several researchers utilized myostatin-deficient strains to test how bone responds to the increased mechanical loads of muscle hypertrophy. Mutant and control individuals usually have similar body masses (Hamrick et al. 2000, 2006a; Hamrick, 2003), and as a result, mutants do not experience elevated ground reaction forces that might otherwise contribute to hypertrophic skeletal characteristics (Schmitt et al. 2010). At the same time, Byron et al. (2004, 2006) linked increased bite forces and more complex cranial suture morphology with the significantly larger temporalis muscles in myostatin-deficient mice. Similarly, Hamrick (2003) found that mutant mice possessed exaggerated osteological features in the proximal femur and also had increased bone mineral density (BMD) and cross-sectional strength properties. Montgomery et al. (2005) also noted that myostatin-deficient mice had enlarged third femoral trochanters, as well as greater BMD in the femur and spinal column.

In addition to increasing muscle size, the loss of myostatin has been shown to directly impact bone development. Hamrick et al. (2007) found that myostatin-deficiency increased differentiation of bone marrow-derived mesenchymal stem cells, which can influence bone density and strength. Hamrick et al. (2007) also determined that these effects are contingent upon loading, and related work has examined bone plasticity in response to the increased biomechanical stresses stemming from enlarged muscle sizes, specifically in the masticatory apparatus, which resulted in significant craniofacial modifications (Ravosa et al. 2007; Vecchione et al. 2007, 2010). Further, the masticatory changes brought about from myostatin-deficiency were shown to be similar to those seen in mammals fed a harder and tougher diet (Ravosa et al. 2008).

These studies highlight that bone development can be influenced by increased muscle size, much in the way that it responds to greater biomechanical demands. However, these principles have primarily been tested in long bones, with less attention being paid to bones like the scapula, whose morphology has been shown to be closely related to locomotor habits (Ashton & Oxnard, 1964; Oxnard, 1967; Roberts, 1974; Larson, 1995; Young, 2002, 2006, 2008). To better understand the various mechanisms that influence the hard tissue morphology of the scapula, we examined the shoulders of wild-type, heterozygote, and full-mutant mice (lacking both GDF-8 alleles). In evaluating how scapular shape changes in response to the mechanical loads borne from increased muscle mass, the following questions were considered:

  •  Do mutant and heterozygote mice possess enlarged shoulder muscles relative to control mice?
  •  What is the relationship of intrinsic shoulder muscle mass and scapular fossa size and does hypermuscularity alter this relationship?
  •  How are scapular features influenced by hypermuscularity?

As the glenoid and medial border regions develop from secondary endochondral ossification centers (Fig. 1), features related to joint orientation, scapular breadth and length, spine and superior border length, and relative size and shape of the scapular fossae will be of primary interest. The increased muscle forces resulting from hypermuscularity could stimulate additional bone and/or cartilage growth in these regions, resulting in morphological differences among the genotype groups (Hamrick, 1999).

Figure 1.

 Whole-mount skeletal stains (alcian blue staining indicates cartilage and alizarin red staining indicates bone) of mice at different developmental stages highlighting regions of secondary endochondral ossification, particularly the glenoid region and medial border. (a) Medial view of a forelimb of a 13.5-day-old embryo (E13.5), (b) dorsal view of a scapula, and (c) articulated view of E16.5 embryos (scapular blade is indicated with the red arrow), and (d) lateral view of a newborn forelimb. Photos are not to scale; (a–c) are modified from Kuijper et al. (2005) and (d) is modified from Di-Poï et al. (2010).

Materials and methods


Mouse forelimb samples consisted of 46 individuals: 15 outbred, wild-type (five male), 20 heterozygote (10 male), and 11 full mutant (three male) mice, all on a CD-1 background. The mutation was originally present in C57BL6 mice (McPherron et al. 1997), but was crossed into the CD-1 line by Se-Jin Lee and Alexandra McPherron in 2000 and several breeding pairs were loaned to one of us (M.W.H.). The forelimb samples for this study were derived from mice that had been used in another experiment to test the effects of muscle hypertrophy in contributing to bone (fibula) fracture repair (Kellum et al. 2009). All mice were housed in standard cages, underwent the same surgical procedure, and moved normally following osteotomy (M. W. Hamrick, personal observation). Animals ranged in age from just over 4 to 10 months old at the time they were euthanized. There were very few age-related morphological differences within genotype groups; for instance, there were no significant differences between three 4-month and five 10-month-old mutant female individuals. The Institutional Animal Care and Use Committee of the Medical College of Georgia approved all animal procedures and mouse strains included in this study.

Measurements and analyses

Mice were weighed immediately after euthanization, the forelimbs removed, and stored at −20 °C. The intrinsic shoulder muscles – supraspinatus, teres major, infraspinatus (teres minor was removed and weighed along with infraspinatus), and subscapularis – were then removed and weighed with a Denver Instrument APX-200 analytical balance (accurate to 0.1 mg). Following this, forelimbs were skeletonized by placing them in simmering water with a small amount of washing soda. The water was allowed to simmer briefly, after which the heat was turned off and the solution left to cool overnight; these steps were repeated once or twice until the soft tissue was completely removed. Following skeletonization, the scapulae were photographed and osteological measurements were made with image j software (Rasband, 1997–2007). These measurements, listed in Table 1, are the distance and angle measurements between the landmark points depicted in Fig. 2 and in subsequent figures.

Table 1.   Measurements used in this study.
Measurement nameDescriptionReference figure
  1. See also Fig. 2.

Supraspinous area (size)The total polygonal area of the supraspinous fossaFig. 4a
Infraspinous areaThe total polygonal area of the infraspinous fossaFig. 4b
Subscapularis areaThe total polygonal area of the subscapularis fossaFig. 4c
Axillary/medial border angle (AMB)The angle formed by the medial [between landmarks 2 (superior angle) and 4 (inferior angle)] and axillary borders [between landmarks 4 and 5 (infraglenoid tubercle)]Figs 2b and 7a
Axillary/infraspinous medial border angle (AIB)The angle formed by the medial border and infraspinous breadth line [between landmarks 3 (point where the spine meets the vertebral border) and 4]Fig. 2c
Axillary border/glenoid angle (ABG)The angle formed by the line between landmarks 4 and 5 and the glenoid fossa (line between the superior and inferior points on the glenoid – landmarks 7–8)Figs 2d and 7b
Total lengthThe distance between landmark 3 and the center of the glenoid fossa 
Total breadthThe distance between landmarks 2 and 4 
Medial border/spine angle (MBS)The angle formed by the medial border and the spine [between landmarks 3 and 6 (spinoglenoid notch)]Figs 2e and 7c
Axillary border/spine angle (ABS)The angle formed by the axillary border and the spine 
Spine lengthThe distance between landmarks 3 and 9 (distal-most point of the acromion) 
Superior border lengthThe distance between landmarks 2 and 9Fig. 8a
Infraspinous breadthThe distance between landmarks 3 and 4Fig. 8b
Infraspinous lengthThe distance between landmarks 3 and 5Fig. 8b
Supraspinous breadthThe distance between landmarks 2 and 3Fig. 8c
Supraspinous lengthThe distance between landmarks 1 (suprascapular notch) and 3Fig. 8c
Figure 2.

 Landmarks describing the scapular variables and angles used in this study. (a) landmarks, (b) axillary/medial border angle, (c) axillary/infraspinous medial border angle, (d) axillary border/glenoid angle, and (e) medial border/spine angle. See Table 1 for a full list and description of features considered.

Principal components (PCA) and canonical variates analyses (CVA) were performed in statistica (version 7.1) to assess gross multivariate differences across genotype groups. These two approaches are similar in that they incorporate and reduce multiple variables for the purpose of visualizing differences in two dimensions. However, while PCA does not take into account prior grouping by the observer in generating factor scores, CVA finds the maximum variation among a priori groups. Finally, a discriminant function analysis (DFA) was used to test the probability that a given individual was (or was not) properly assigned to their genotype condition in the CVA. In this way, CVA and DFA can evaluate the variables that best distinguish groups from one another, and the extent to which groups differ.

Most of the analyses that follow, however, are bivariate comparisons (e.g. supraspinous vs. infraspinous breadths). In these comparisons, reduced major axis (RMA) regression slopes were calculated. Unlike ordinary least squares (OLS) regression analyses, which only seek to minimize variation in the y-axis variable, RMA regressions minimize variance in both x- and y-axis variables. This was the preferred method, as both x- and y-axis variables were known to vary in this study, whereas OLS assumes the x-variable to be fixed (Sokal & Rohlf, 1995). The significance of differences among genotype groups was assessed using Mann–Whitney U-tests in statistica (version 7.1). Furthermore, given the number of comparisons for each group of scapular variables, the Dunn–Šidák method was used to adjust the significance level from α = 0.05 to decrease the likelihood of type I statistical errors (rejecting a true null hypothesis; Sokal & Rohlf, 1995). A total of 45 tests comprised the body mass, muscle and fossa size comparisons (P-values < 0.002 were considered significant at the adjusted Dunn–Šidák level for these comparisons), 18 for broad scapular shape characteristics (< 0.003), 18 for scapular spine and superior border (P < 0.003), and 21 for supra-spinous and infraspinous fossae characteristics (< 0.002). However, all P-values are reported below to highlight comparisons that were significant at the original α = 0.05 level as well as the adjusted alpha level, since the Dunn–Šidák method is a highly conservative test of significance.


Body mass, muscle size, and scapular fossa size

Body mass did not differ significantly among the three genotype groups, which accords with several previous studies (Hamrick et al. 2000, 2006a; Hamrick, 2003; but see Hamrick et al. 2006b); however, mutant mice had significantly more massive shoulder muscles than the wild-type mice (Table 2). The effect was dose-dependent: the teres major and subscapularis were significantly larger in full mutants than in the heterozygotes (at the ≤ 0.007 level), whereas heterozygote muscle masses were significantly greater than wild-type mice in all muscle size comparisons, save for teres major (≤ 0.001; Table 2). Relative to body mass, mutant mice again had significantly larger muscle masses than wild-type mice and had significantly greater teres major and subscapularis muscles than the heterozygotes. As with absolute muscle size, heterozygote mice had significantly larger muscle masses than wild-type mice, relative to body mass, in all comparisons but teres major (Table 2).

Table 2.   Scapular characteristics; mean measurement value (SD).
MeasurementGenotypePairwise comparison (P-value)
Wild-type (W)Heterozygote (H)Mutant (M)W-HW-MH-M
  1. P-values of each pairwise t-test are shown; values with an * were significant at the α = 0.05 level.

  2. Given multiple tests, comparison-wide significance based on P-values determined by the Dunn–Šidák method are also presented: a value significant at the < 0.002 level for all body mass, muscle mass, and scapular fossa comparisons (45 comparisons); a value significant at the < 0.003 level for gross scapular shape characteristics (18); §a value significant at the < 0.003 level for scapular spine and superior border characteristics (18); and a value significant at the < 0.002 level for supraspinous and infraspinous fossae characteristics (21).

Body mass (g)34.7 (4.7)37.2 (7.6)36.6 (5.0)0.540.150.90
Absolute supraspinous mass (mg)42.7 (8.1)56.0 (13.0)65.0 (12.7)0.001*< 0.0010.06
Relative supraspinous mass0.0012 (0.0002)0.0015 (0.0002)0.0018 (0.0002)< 0.001< 0.0010.01*
Absolute supraspinous fossa size (area mm2)13.2 (1.5)16.1 (1.7)15.2 (1.7)< 0.0010.007*0.20
Relative supraspinous fossa size1.11 (0.1)1.21 (0.1)1.17 (0.1)< 0.0010.03*0.26
Absolute teres major mass (mg)42.3 (9.0)48.7 (12.5)68.5 (16.0)0.08< 0.0010.002*
Relative teres major mass0.0012 (0.0002)0.0013 (0.0002)0.0019 (0.0002)0.17< 0.001< 0.001
Absolute infraspinous mass (mg)36.3 (4.6)47.7 (10.0)53.3 (13.4)< 0.001< 0.0010.28
Relative infraspinous mass0.0010 (0.0001)0.0013 (0.0002)0.0014 (0.0003)< 0.001< 0.0010.11
Absolute infraspinous fossa size (area mm2)23.9 (1.7)25.6 (2.7)26.3 (2.5)0.04*0.01*0.43
Relative infraspinous fossa size1.50 (0.1)1.52 (0.1)1.55 (0.1)0.330.190.51
Absolute subscapularis mass (mg)60.2 (7.2)76.5 (12.7)94.4 (19.3)< 0.001< 0.0010.007*
Relative subscapularis mass0.0017 (0.0002)0.0021 (0.0002)0.0026 (0.0003)< 0.001< 0.001< 0.001
Absolute subscapularis fossa size (area mm2)40.1 (2.9)46.0 (4.0)45.3 (2.9)< 0.001< 0.0010.77
Relative subscapularis fossa size1.94 (0.1)2.04 (0.1)2.03 (0.1)0.009*0.03*0.69
Axillary/medial border angle (°)48.3 (2.2)50.5 (2.0)51.7 (2.6)0.01*0.0020.17
Axillary/infraspinous medial border angle (°)75.2 (1.8)76.4 (3.4)76.9 (2.5)
Axillary border/glenoid angle (°)138.0 (2.4)133.4 (4.0)132.9 (2.4)< 0.001< 0.0010.74
Breadth/(body mass0.33)2.54 (0.1)2.622 (0.1)2.63 (0.1)
Length/(body mass0.33)3.07 (0.1)3.15 (0.2)3.20 (0.1)0.170.04*0.39
Breadth/length0.83 (0.03)0.83 (0.03)0.82 (0.03)0.910.620.46
Medial border/spine angle (°)105.8 (3.4)104.0 (3.1)102.5 (2.7)0.120.03*0.23
Axillary border/spine angle (°)26.4 (2.5)25.7 (1.8)25.7 (1.8)
Spine length/(body mass0.33)3.59 (0.2)3.67 (0.2)3.73 (0.2)0.290.04*0.39
Spine length/total length1.17 (0.01)1.16 (0.01)1.17 (0.02)0.430.530.63
Superior border length/(body mass0.33)2.71 (0.2)2.86 (0.2)2.98 (0.1)0.01*< 0.001§0.19
Superior border length/total length0.88 (0.04)0.91 (0.04)0.93 (0.03)0.04*0.006*0.10
Infraspinous breadth/(body mass0.33)1.59 (0.1)1.60 (0.1)1.62 (0.1)0.650.340.50
Infraspinous length/(body mass0.33)2.82 (0.1)2.87 (0.1)2.93 (0.1)0.380.04*0.26
Infraspinous breadth/length0.56 (0.03)0.56 (0.03)0.55 (0.02)0.480.220.82
Supraspinous breadth/(body mass0.33)1.35 (0.1)1.37 (0.1)1.32 (0.1)0.960.360.19
Supraspinous length/(body mass0.33)2.57 (0.1)2.64 (0.1)2.70 (0.1)0.170.03*0.16
Supraspinous breadth/length0.53 (0.02)0.52 (0.03)0.49 (0.02)0.210.0020.008*
Supraspinous/infraspinous breadth0.86 (0.1)0.86 (0.1)0.82 (0.05)0.880.150.18

Intrinsic shoulder muscles did not scale in a way that significantly differed from isometry in mutant mice – all muscles were enlarged to a similar degree as a result of the mutation (Fig. 3; Table 3). In heterozygote mice, the subscapularis muscle was negatively allometric with respect to both teres major and infraspinatus (e.g. if the teres major muscle of individual A was larger compared with individual B, the subscapularis muscle of individual A was also larger than that of individual B, but not to the same degree as in the teres major comparison). However, none of the other heterozygote muscle comparisons differed significantly from isometry (Fig. 3; Table 3). In wild-type mice, the subscapularis muscle was negatively allometric with respect to supraspinatus and teres major, and infraspinatus was negatively allometric with respect to teres major. Only comparisons of supraspinatus with teres major and infraspinatus produced RMA slopes that did not significantly differ from isometry in wild-type mice (Fig. 3, Table 3).

Figure 3.

 Bivariate plots of intrinsic muscles depicting reduced major axis comparisons: (a) teres major – infraspinatus, (b) teres major – subscapularis, (c) supraspinatus – teres major, (d) supraspinatus – infraspinatus, (e) supraspinatus – subscapularis, and (f) infraspinatus – subscapularis. Solid RMA slopes did not differ significantly from isometry, while dashed lines were all negatively allometric. Refer to Table 3 for details.

Table 3.   Reduced major axis comparisons of intrinsic muscle groups.
ComparisonGenotypeRMA slope−99%−95%+95%+99%
  1. Comparisons considered natural logarithms of muscle mass. This table represents multiple comparisons and so both 95% and 99% CI are presented. Bold RMA slopes are those that differed significantly from isometry, that is, 95% (*) and/or 99% (**) CI, did not include the slope of 1.0 (e.g. wild-type teres major muscle mass was negatively allometric with respect to subscapularis muscle mass). See also Fig. 3.

Teres major, infraspinatusWild-type0.61*
Teres major, subscapularisWild-type0.56**0.250.340.790.88
Supraspinatus, teres majorWild-type1.090.180.441.741.99
Supraspinatus, infraspinatusWild-type0.660.160.301.021.16
Supraspinatus, subscapularisWild-type0.62*
Infraspinatus, subscapularisWild-type0.930.270.461.411.59

Two-dimensional fossa size (area) measures were positively correlated with muscle size in some, but not all, cases. In keeping with the muscle size results, heterozygote and mutant fossae were significantly larger than those of the wild-type mice. Heterozygote supraspinous and infraspinous fossae averages were greater than those of mutants, but these differences were not significant (Table 2). None of the genotype groups differed significantly in relative infraspinous fossa size, but both the mutant and heterozygote mice had significantly larger supraspinous and subscapularis fossa sizes than the wild-type mice (≤ 0.03; Table 2). Both heterozygote and mutant individuals showed significant positive relationships among fossa and respective muscle size; however, save for subscapularis, wild-type individuals actually displayed a significant negative relationship between fossa size and muscle mass (Fig. 4, Table 4).

Figure 4.

 Bivariate plots of muscle mass and fossa size comparisons for the (a) supraspinatus, (b) infraspinatus, and (c) subscapularis regions. As expected, reduced major axis slopes were positive among heterozygote and mutant individuals, but wild-type individuals showed a negative slope that differed significantly from zero in the supraspinatus and infraspinatus comparisons. Refer to Table 4 for details.

Table 4.   Reduced major axis comparisons of intrinsic muscle groups and associated fossae.
Fossa – muscleGenotypeRMA slope−95%+95%
  1. Comparisons considered natural logarithms of the cube root of muscle mass and the square root of fossa area. This table represents multiple comparisons and, for the most part, CI of muscle–fossa comparisons did not differ significantly from an isometric slope of 1.0. Wild-type supraspinous–supraspinatus and infraspinous–infraspinatus comparisons (shown in bold) displayed a negative relationship, and differed significantly from positive isometry at the 0.05 level. See also Fig. 4.

Supraspinous, supraspinatusWild-type−1.12−1.79−0.45
Infraspinous, infraspinatusWild-type−1.17−1.86−0.48
Subscapularis, subscapularisWild-type1.090.451.72

Multivariate scapular shape characteristics

Two multivariate approaches were used to examine scapular shape differences among genotype groups. These analyses considered 12 linear and angular scapular measures (Tables 1, 5 and 6). All linear values were size-corrected by body mass to better represent shape differences in the first two principal component (or canonical variate) axes. In the PCA, heterozygote and mutant individuals did not clearly separate from wild-type mice along the first principal component axis (factor 1), which accounted for 50.3% of the variance (Fig. 5a, Table 5). Alternatively, genotype groups were more easily distinguishable along the second principal component axis (19.7% of the variance; Fig. 5a). Mutant and heterozygote individuals occupied more of the positive region of this axis relative to the wild-type individuals. Along the second factor axis, the axillary/medial border (AMB) angle, axillary/infraspinous medial border (AIB) angle, and superior border length had the most positive factor loadings, while the medial border/spine (MBS) angle, supraspinous breadth, and axillary border/glenoid (ABG) angle had the most negative factor loadings in driving this separation between the genotype groups (Table 5).

Table 5.   List of measurements used and the first two factor loadings in ascending order for the PCA depicted in Fig. 5a.
Factor 1Factor 2
  1. See Figs 2, 7, and 8 for depictions of listed measures.

Spine length−0.967−0.823Medial border/spine angle
Total length−0.963−0.547Supraspinous breadth
Infraspinous length−0.953−0.371Axillary border/glenoid angle
Supraspinous length−0.922−0.095Total breadth
Total breadth−0.8990.026Supraspinous length
Superior border length−0.7590.037Infraspinous length
Infraspinous breath−0.7330.060Spine length
Supraspinous breadth−0.5910.081Total length
Medial border/spine angle0.0040.134Infraspinous breath
Axillary/medial border angle0.1530.461Superior border length
Axillary border/glenoid angle0.1560.546Axillary/infraspinous medial border angle
Axillary/infraspinous medial border angle0.3070.836Axillary/medial border angle
Table 6.   List of measurements used and the first two root loadings in ascending order for the CVA depicted in Fig. 5b.
Root 1Root 2
  1. See Figs 2, 7, and 8 for depictions of listed measures.

Axillary border/glenoid angle−0.465−0.363Supraspinous breadth
Medial border/spine angle−0.252−0.221Medial border/spine angle
Supraspinous breadth−0.028−0.074Total breadth
Infraspinous breath0.0810.020Axillary/infraspinous medial border angle
Axillary/infraspinous medial border angle0.1650.119Total length
Total breadth0.1750.140Infraspinous breath
Infraspinous length0.1880.160Spine length
Spine length0.1940.164Axillary/medial border angle
Total length0.2020.168Supraspinous length
Supraspinous length0.2180.198Axillary border/glenoid angle
Superior border length0.3760.224Infraspinous length
Axillary/medial border angle0.3900.302Superior border length
Figure 5.

 (a) Principal components and (b) canonical variates analysis plots.

Secondly, CVA and DFA tested how well scapular traits distinguished genotype group association. There was a clear separation of the heterozygote and mutant individuals from wild-type individuals, positively along the first variate axis (root 1) (90.2% of the variance; Fig. 5b). Along the second root, there was a slight separation of the mutant and heterozygote groups, but this was not significant (9.8% of the variance; Fig. 5b). AMB, superior border length, and supraspinous length had the most positive coefficient scores, while ABG and MBS showed the strongest negative coefficient scores in driving the separation between the groups along the first root (Table 6). The more medial position of the mutant superior angle is evident in Fig. 6 and likely resulted in differences in AMB angle and superior border length (see below). Along the second root, superior border length, infraspinous length, and ABG had the most positive coefficient scores, while supraspinous breadth and MBS had the most negative coefficient scores in driving the slight separation between heterozygotes and mutants (Table 6). In the DFA, Mahalanobis D2 distances between wild-type and both mutant and heterozygote mice were significantly greater, such that there were a few instances where a ‘wild-type’ individual was misclassified as ‘mutant’ or ‘heterozygote’ and vice versa. In contrast, there were far more instances where ‘mutants’ were misclassified as ‘heterozygotes’ and vice versa, such that the difference between the two was not significant (Table 7).

Figure 6.

 Examples of scapulae from each genotype: (a) and (b) are dorsal and ventral views of a male wild-type individual, respectively, (c) and (d) a heterozygote individual, and (e) and (f) a mutant individual. Red arrows above the wild-type and mutant individuals highlight the more medial position of the mutant superior angle.

Table 7.   Pairwise comparisons of genotype groups based on the CVA described in Fig. 5b and Table 6.
  1. Values above the diagonal are Mahalanobis D2 distances; values below the diagonal are P-values. Differences significant at α ≤ 0.05 are indicated in bold.

Heterozygote< 0.0012.25
Mutant< 0.0010.48

Bivariate scapular shape characteristics

Scapular blade shape

Heterozygote and mutant mice did not differ significantly in overall scapular blade shape as determined by the AMB angle, but both had significantly broader scapular blades compared with wild-type mice (heterozygotes differed from wild-type mice at the α = 0.01 level; Fig. 7a, Table 2). Alternatively, none of the genotype groups differed significantly in AIB angle (Table 2). These contrasting results could be attributed to the relative position of the superior angle, particularly in full mutant individuals. A more medially positioned superior angle would promote a larger AMB angle, but does not influence AIB angle, as its value is contingent on the point where the scapular spine meets the vertebral border, not the superior angle (Figs 2, 6, and 7a; Table 1).

Figure 7.

 Box plots of (a) axillary/medial border, (b) axillary border/glenoid angle, and (c) medial border/spine angle.

Glenohumeral (GH) joint orientation

The ABG angle was used to estimate shoulder joint orientation. Heterozygote and mutant mice did not differ significantly from one another, but both had significantly more acute angles than wild-type mice, such that individuals lacking at least one GDF-8 allele had more superiorly oriented GH joints than wild-type mice (Fig. 7b, Table 2).

Total scapular breadth and length

Mutant mice had slightly longer scapulae compared with wild-type mice, relative to body mass (P = 0.04); heterozygotes did not differ significantly from either group. The genotype groups did not differ significantly from one another in relative breadth or total scapular breadth : length ratios, despite a significant difference in total length between mutant and wild-type mice (Table 2).

Scapular spine and superior border characteristics

Scapular spine orientation

The orientation of the scapular spine was estimated by the MBS and the axillary border/spine (ABS) angles. Only mutant mice had a slightly more acute spine orientation with respect to wild-type mice as determined by MBS angle (= 0.03); heterozygote mice were not significantly different from either of the genotype groups (Fig. 7c, Table 2). Unlike the MBS results, none of the genotype groups significantly differed in the ABS angle comparisons.

Scapular spine length

By and large, scapular spine length was not significantly different among genotype groups, although mutant mice had minimally longer spines compared with wild-type mice, relative to body mass (= 0.04; Table 2). None of the genotype groups significantly differed when spine length was compared to total length, indicating that spine length scaled closely with overall scapular length (Table 2).

Superior border length

Heterozygote and mutant mice did not differ significantly in superior border length relative to body mass, but mutants were significantly longer than wild-type mice (Fig. 8a, Table 2). Mutant mouse superior borders were about 93% of total length, which was significantly longer than those of wild-type mice at the = 0.006 level (Table 2). Again, this highlights the relatively medial position of the superior angle in the mutant mice, which also resulted in significantly larger AMB and more acute MBS angles (Fig. 6).

Figure 8.

 Bivariate plots of (a) superior border length and the cubed root of body mass, (b) infraspinous length and breadth, and (c) supraspinous length and breadth. Reduced major axis regression lines are depicted.

Infraspinous and supraspinous fossae characteristics

Infraspinous fossa breadth and length

Genotype groups did not significantly differ in relative infraspinous breadth, but mutant mice had moderately longer infraspinous fossae with respect to wild-type mice (= 0.04; Table 2). Infraspinous breadth : length ratios did not differ significantly among the genotype groups (Fig. 8b, Table 2).

Supraspinous fossa breadth and length

Genotype groups did not significantly differ in relative supraspinous breadth but, again, mutants had slightly longer supraspinous fossae compared with wild-type mice (= 0.03; Table 2). This resulted in supraspinous breadth : length ratios that were significantly lower than both heterozygote (= 0.008) and wild-type mice (= 0.002), but the latter two groups were not significantly different from one another (Fig. 8c, Table 2).

Supraspinous and infraspinous breadth

Just as was the case when relative infraspinous and supraspinous breadth were considered separately, none of the genotype groups differed in relative supraspinous : infraspinous breadth ratios (Table 2).


This study used a myostatin-knockout mouse model to test the hypothesis that heritable variation in shoulder muscle size can yield significant variation in scapular form. Since the discovery of the role of myostatin in regulating muscle development and size, a great deal of research has been devoted to understanding how the skeleton responds to such a dramatic change in muscle mass (e.g. McPherron et al. 1997; Hamrick, 2003; Byron et al. 2004; Montgomery et al. 2005; Hamrick et al. 2006a; Ravosa et al. 2007; Vecchione et al. 2007; Robling, 2009). Many of these studies focused on long bones, as have a great many related studies investigating how bones respond to mechanical loads borne from other external forces (e.g. an experimental load or elevated exercise levels) (Lanyon, 1980; Kimes et al. 1981; Lanyon & Rubin, 1984; Heinonen et al. 2000; Notomi et al. 2001; Mori et al. 2003; Lieberman et al. 2004; Hamrick et al. 2006b,c; Robling et al. 2006).

The results of this study show that hypermuscularity significantly influenced both the soft- and hard-tissue characteristics of the shoulder. As predicted, the effect was more considerable in the full mutant mice (missing both GDF-8 alleles) than heterozygotes. The individuals in this study did not differ significantly in body mass (Table 2), as reported in related studies of myostatin-deficient mice (Hamrick, 2003; Montgomery et al. 2005; Hamrick et al. 2006a), although significant differences in body mass between mutant and control mice have been demonstrated (McPherron et al. 1997; Hamrick et al. 2006b). Previous work with myostatin-deficient mice has evaluated larger muscle groups (e.g. triceps brachii and quadriceps femoris), but less attention has been devoted to smaller regions like the shoulder and its intrinsic musculature. A major function of these deeper muscles is shoulder joint stabilization (Larson & Stern, 1986), and as a result, these muscle groups tend to have a higher proportion of slow-twitch, fatigue-resistant muscle fibers (Schmidt & Schilling, 2007). This is an important consideration, as GDF-8 has been shown to disproportionately influence fast-twitch muscle fibers (Girgenrath et al. 2005; Byron et al. 2006; Hamrick et al. 2006b). Nonetheless, these muscles did show significant increases in mass as a result of the mutation. Specifically, full mutants possessed significantly greater muscle sizes than wild-type mice, and had larger teres major and subscapularis muscles with respect to heterozygotes; heterozygotes were significantly greater than wild-type mice for all muscles, save for teres major (Table 2). Thus, while mutant muscle mass was always greater than heterozygote muscle mass, as with heterozygotes relative to wild-type mice, these differences were not always significant. It is possible, then, that a greater proportion of slow-twitch fibers predominating in these deeper muscles were not as affected by the lack of myostatin, and did not contribute to a statistically significant difference in some comparisons.

Following with the muscle size results, wild-type mice had the smallest absolute scapular fossa sizes for all three regions (Table 2). However, mutant and heterozygote mice did not significantly differ in absolute or relative fossa size, contrasting slightly with the muscle mass results. In addition, wild-type supraspinous and infraspinous fossae scaled negatively with muscle size (Fig. 4, Table 4). Put together, these results suggest that the relationship between intrinsic scapular muscle and fossa size is complex, a finding consistent with previous comparisons of muscle and origin/attachment site size in experimentally exercised sheep (Zumwalt, 2006). The wild-type plots in Fig. 4a and b both have low correlation coefficients and ordinary least-squares regression slopes did not significantly differ from zero, unlike the RMA results. To this end, a more conservative interpretation of these results might be that wild-type supraspinous and infraspinous fossae have no relationship with the muscles they house, rather than a negative one. Moreover, an exercise study utilizing a larger sample of wild-type mice (Green, 2010; Green et al. 2010) did find significant positive relationships for all the fossa/muscle size results presented here, indicating that smaller sample sizes may have also been a factor in the negative RMA results.

Scapular shape characteristics

In the principal components and canonical variates analyses, the heterozygote and mutant mouse data scatters separated from the wild-type individuals along the second variate axis in the PCA and the first root in the CVA (Fig. 5). AMB, ABG, and MBS angles as well as superior border length, and supraspinous fossa breadth and length had the strongest loading scores in driving the genotype differences in the multivariate analyses (Tables 5 and 6). Not surprisingly, these scapular features were all important in distinguishing the genotype groups in the bivariate comparisons.

Although there was some separation between the heterozygote and mutant mice along the second CV axis, these differences were not significant. Likewise, the characteristics driving the slight separation between the heterozygotes and mutant were not significantly different in bivariate comparisons either. Moreover, there were few significant differences between the two hypermuscular groups – only the supraspinous breadth : length ratio comparison produced a small but significant difference between the two (= 0.008; Table 2). On the other hand, many more mutant/wild-type comparisons resulted in significant differences than heterozygote/wild-type comparisons. Thus, although the mutant and heterozygote individuals did not dramatically differ from one another, full mutant mice possessed more extreme hard-tissue characteristics relative to the wild-type control.

It is also notable that enlarged muscle size produced differences in several angles describing shoulder joint and blade orientation characteristics: AMB, ABG, and MBS. The significant differences in GH joint orientation are of particular interest for physical anthropologists, as this trait distinguishes the shoulders of suspensory climbing and terrestrially quadrupedal primates (Ashton & Oxnard, 1964; Oxnard, 1967; Roberts, 1974; Hunt, 1991). Furthermore, the cranial orientation of the shoulder joint of some early hominin fossils has been contentiously cited as evidence for arboreal capabilities in Australopithecus afarensis (Stern & Susman, 1983; Susman et al. 1984; Susman & Stern, 1991; Stern, 2000; Alemseged et al. 2006; but see Lovejoy, 1988; Latimer, 1991; Inouye & Shea, 1997). Suspensory apes have significantly larger forelimb : hindlimb ratios than modern humans, and australopiths have also been shown to be more forelimb-dominant relative to modern humans (Jungers, 1985, 1988; McHenry, 1992; McHenry & Berger, 1998; Richmond et al. 2002; Green et al. 2007). It may be that increased forelimb musculature is related to the cranial orientation of suspensory ape shoulders, though the forelimb-dominant mode of locomotion is likely the predominant contributor to both their muscularity and unique hard-tissue characters. While it is difficult to distinguish forelimb muscle size and locomotor habits as the principal determinants of ape shoulder morphology, it is noteworthy that increased muscle size, in the absence of other locomotor differences, also resulted in statistically significant differences in mouse GH joint orientation.

The apparent position of the superior angle relative to the scapular blade in the hypermuscular mice also contributed to several differences among the genotype groups. Both of the angular measurement comparisons that included the medial border – AMB and MBS – produced significant differences between the mutant and wild-type mice (≤ 0.03; Table 2). Additionally, the mutant superior border was significantly longer than that of wild-type mice, both when considered relative to body mass and total length (= 0.006). In contrast, spine length only differed between mutant and wild-type mice when considered relative to body mass. This would indicate that the relative position of the superior angle, as opposed to the distal end of the acromion, was the factor contributing most to the superior border differences (Fig. 6, Table 2). With regard to the difference in MBS between the wild-type and mutant mice, the lack of significant difference among genotype groups in both supraspinous and infraspinous breadth indicates that the difference in this angle measurement was similarly related to a unique orientation of the hypermuscular medial border, and not the scapular spine. ABS angle was also considered as an alternative measure of spine orientation and it did not significantly differ among the genotype groups (Table 2). This is further evidence that superior angle position in the hypermuscular mice (and the resulting orientation of the medial border) was an important factor driving the morphological differences in the MBS comparisons (Fig. 6). It is also plausible that the more medial position of the superior angle could be a structural modification to accommodate an enlarged supraspinatus muscle (Table 2).


This study demonstrated that genetically enhanced muscle mass can have dramatic effects on the hard and soft tissue anatomy of the shoulder. This was not an unexpected result, given previous studies of long bones and the fact that the shoulder girdle is mainly connected to the torso via musculotendinous attachments. At the same time, it also highlights that the scapula is developmentally plastic and susceptible to modification in response to significant changes in muscle size. Several regions of the scapula were affected as a result of the muscle size differences, ranging from relative fossa length to more general characteristics relating to scapular orientation, like medial border and GH joint configuration. In addition, nearly all of the characters that significantly differed among the genotype groups were those that described mediolateral scapular shape parameters, while superoinferior scapular measures – total breadth, supraspinous and infraspinous breadth – did not differ significantly among the genotype groups. With enlarged scapular muscles that are generally oriented in a mediolateral fashion, it follows that these characters would be the most likely to change in response and to accommodate the larger muscles (Hamrick, 2003; Hamrick et al. 2006b; Robling, 2009). Since the scapula has been shown to be a reliable predictor of an organisms’ preferred locomotor pattern (Inman et al. 1944; Oxnard, 1963, 1967; Ashton & Oxnard, 1964), it is noteworthy that increased muscle size significantly influences scapular development (Roberts, 1974; Robling et al. 2006; Robling, 2009). Further, this study provides evidence that experimentally enlarged muscle size markedly changes shoulder morphological characteristics – even in the absence of significant locomotor differences.


We wish to thank Drs. Bernard Wood, Robin Bernstein, L. Patricia Hernandez, Stefan Milz, and three anonymous reviewers for critical comments on an earlier draft of this manuscript and the following funding sources: National Science Foundation IGERT grant (9987590), NSF Doctoral Dissertation Improvement Grant (BCS-0824552), and the National Institutes of Health (AR049717).