1 Department Sport and Movement Sciences, Faculty of Physical Education and Physiotherapy, Katholieke Universiteit Leuven, Maastricht 6229 ER, The Netherlands
2 Division of Genetics and Molecular Cell Biology, Universiteit Maastricht, Maastricht 6229 ER, The Netherlands
3 Center for Human Genetics, Faculty of Medicine, Katholieke Universiteit Leuven, Leuven 3000, Belgium
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
quantitative trait locus; hypertrophy; fat-free mass; human performance; candidate gene
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Myostatin (growth and differentiation factor 8 [GDF8]), is a member of the transforming growth factor-ß (TGF-ß) superfamily that acts as a negative regulator of skeletal muscle mass. Its structure and function is highly conserved during evolution across species (19, 20). Myostatin-deficient animals show dramatic increases in muscle growth. Mice completely lacking Gdf8 show a two- to threefold increase in skeletal muscle mass with both increases in myofiber size (hypertrophy) and myofiber number (hyperplasia) (20), and mutations in the GDF8 gene in cattle result in the double-muscling phenotype (11, 19). The negative regulation of myostatin on postnatal muscle growth is confirmed by a study where a myostatin blocker, in vivo, increased muscle size and strength in mice suffering from Duchenne muscular dystrophy (5). In addition, systemically administered myostatin inhibitors in adult mice show a postnatal positive effect on both muscle and fat loss (38).
However, whether myostatin regulates skeletal muscle mass in humans in the same way as in nonhuman species is unclear. A study on HIV-infected men demonstrated a strong association between muscle wasting and increased levels of myostatin in serum and muscle tissue (10). Ferrell et al. (8) identified five missense substitutions in the coding sequence of human myostatin, but only two of them were polymorphic (K153R, A55T). In their sample of Caucasian and African-American subjects, no significant association was found with differential muscle mass response to strength training. Yet, one allele (R153) was overrepresented in the nonresponder group, suggesting that this allele may play a role in other muscle phenotypes (8). Since muscle mass, especially cross-sectional areas of the muscle, and muscle strength are two strongly related characteristics, more muscle mass would implicate (in part) more strength. Indeed, two studies, both on women, suggest an association between strength (28) or gain in strength after a training program (13) with human myostatin variant K153R.
Here, briefly we provide a description of our selection of the candidate genes. As economic implications of myostatin on muscle growth for the cattle-breeding industry are obvious, many studies on animals have been performed and have unraveled the physiological pathway (16, 18, 26, 30, 31), although not all interactions with other muscle regulatory proteins are known. Myostatin pathway genes can directly or indirectly interact at each stage of the muscle development (Fig. 1). Mesoderm precursor cells transform into myoblasts under control of Myf5 and MyoD. The phosphorylation status of retinoblastoma (Rb) regulates the cell cycle (DNA synthesis) and therefore also the proliferation of myoblasts. After proliferation, committed myoblasts differentiate and fuse into mature myotubes under control of Myf6 and myogenin. Thomas et al. (31) propose the following model for myostatin (GDF8) in regulating muscle mass (Fig. 1A): myostatin signaling results in an upregulation of p21 (or Cdkn1a), which is an inhibitor of cyclin-dependent kinase 2 (Cdk2). This causes a hypophosphorylation of Rb and a cell cycle arrest (G1) in the proliferating myoblasts. Thus myoblast number and, hence, fiber number is regulated by GDF8. In addition, titin-cap (Tcap) can interact with myostatin and decreases the secretion of active myostatin, suggesting a possible regulatory effect on muscle development (22).
|
|
![]() |
MATERIALS AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Measurements
Body composition.
An extended set of skinfold, circumference, length, and width measurements were taken from all subjects by an experienced anthropometrist in standardized conditions to quantify body composition (for details, see Ref. 12). Cross-sectional muscle and bone area (MBA) of the thigh was calculated based on circumference of the midthigh corrected for skinfold thickness at the midthigh, and the muscle cross-sectional area of the quadriceps was estimated by the equation of Knapik et al. (14).
Muscle strength.
The Cybex NORM isokinetic dynamometer (Lumex, Ronkonkoma, NY) was used for concentric knee strength tests. After a 5- to 10-min warm-up on an ergometer cycle and light stretching exercises, subjects were positioned in the dynamometer, following the instructions of the manufacturer. Anatomical zero was set at full extension of the knee, and the rotation axis of the joint was aligned with the mechanical axis of the dynamometer.
Four concentric sub-maximal trials preceded the actual tests to get familiarized with the testing procedure (velocity, range of motion). Peak torque over the complete range of motion (090°) of knee extension and flexion was measured at 60°/s (3 repetitions), at 120°/s (25 repetitions), and at 240°/s (5 repetitions). During these contractions, torque at specific angles was also recorded. Following the force-strength relationship of a muscle, optimal strength is generated at longer muscle length, i.e., at an angle of 60° for knee extension (musculus quadriceps) and 30° for knee flexion (hamstring). Subjects were verbally encouraged to perform at their maximum effort, and visual feedback of their performance was presented after each test. For optimal comparison between subjects of different body size, "muscle quality" was calculated as the ratio of knee torque over muscle and bone cross-sectional area of the midthigh (N·m/cm2) and used in further linkage analysis.
Laboratory Methods
Genomic DNA was prepared from EDTA whole blood by the salting out method (21). Microsatellite markers (di- and tetranucleotide) were selected from the build 31 STS map (http://www.ncbi.nlm.nih.gov/genome/sts). Criteria for good markers were: location to the candidate gene (within 1 cM region), heterozygosity (>65%), and uniqueness of the primers. Positioning of the markers to the genes and heterozygosity is given in Table 1.
PCR conditions.
Five PCR protocols were used to amplify the 11 markers of the 10 candidate genes. A marker was selected in or near (<1 cM) the gene locus, and primers were, if necessary, engineered to fit in the multiplex PCR. Table 2 shows the different conditions of the three multiplex (PCR-A, PCR-B, and PCR-C) and two single-plex (PCR-D and PCR-E) PCRs together with the primer sequences, gene and marker names. The specifications of the total PCR reaction mix (15 µl) were as follows: 1.5 µl PCR buffer [200 mM Tris·HCl (pH 8.4) and 500 mM KCl; Invitrogen], 0.5 µl dNTPs, 10 mM each (Amersham Biosciences), and 0.1 µl of Taq DNA polymerase (5 U/µl; Invitrogen), 1 µl genomic DNA, 0.6 µl of 2 mM MgCl2 (or 0.675 µl of 2.25 mM MgCl2 in PCR-D), and primers (see Table 2 for details). Sterile water was added to this mixture to end up with 15 µl mix. Amplification in Biometra T1 thermal cyclers started with an initial denaturing of 5 min at 94°C and ended with a final extension of 7 min at 72°C. PCR-specific annealing temperatures are given in Table 2.
|
Statistical Analysis
Heritability estimations.
Upper-limit heritabilities of the traits (h2) were estimated by the variance components analysis procedure (VC) in QTDT (1). This estimate also includes common environmental variation in addition to the additive genetic component, because these two factors cannot be separated with sib-pairs only. Hence, it is called the upper-limit heritability.
Linkage analysis.
After a pedigree check that revealed no genotyping errors, single-point linkage analysis was performed using Merlin software ver. 0.9.11 (2). Genotypic and phenotypic information was analyzed with two different model-free linkage methods. The first is a linear regression for quantitative traits ("qtl" option in Merlin; Ref. 2) that uses the framework of Whittemore and Halpern (37) for calculating a mean deviate for each founder allele, and an LOD score is defined by the Kong and Cox (15) statistic (QTL). The second is the VC method that tests a model with unique environment (E) and additive genes (A) against a model with A, E, and a quantitative trait locus (Q). Test statistic is a 2 with 1df for comparison of the 2ln likelihoods of the two models. LOD scores are reported as (
2/4.6). Allele frequencies and heterozygosity of the markers were estimated by allele counting procedures, and a Hardy-Weinberg equilibrium test was performed in Merlin. Suggestive evidence for linkage was defined by P < 0.01.
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
Linkage Analyses
Heterozygosity of all markers was on average 72.5%. Genotype frequencies were tested for Hardy-Weinberg equilibrium and revealed no significant deviation from this equilibrium.
Single-point linkage analysis was performed between markers in or near the candidate genes from the myostatin pathway with muscle strength and mass phenotypes. Three markers showed consistent LOD scores on different strength phenotypes and are reported in Table 4. However, the two analysis methods (QTL and VC) did not always confirm each other. Marker D2S118 (GDF8, 2q32.2) had the strongest evidence for linkage with LOD scores between 2.63 and 1.24 (0.0002 < P < 0.008), except for extension at 120°/s and at 240°/s (P > 0.04). For QTL and VC, similar P values were observed in most cases, but large differences exist for extension at 60°/s (P = 0.0002 vs. P = 0.008) and flexion at 240°/s (0.007 vs. 0.2). D6S1051 (CDKN1A or p21, 6p21.2) and D11S4138 (MYOD1, 13q14.2) showed relative consistent linkage results with the VC method for most strength phenotypes (0.01 < P < 0.004), but the QTL method often failed to result in P values that were similar to the VC method. In addition, the number of significant LOD scores decreased with increasing velocity for the three loci. Moreover, only one significant LOD score (P = 0.007) was found with the highest velocity (flexion at 240°/s), i.e., with marker D2S118. The markers of the other candidate genes did not show a pattern of linkage with muscularity or knee strength. Suggestive linkage with estimated muscle cross-sectional area was only found with the marker for myostatin: MBA of the midthigh and muscle cross-sectional area of the quadriceps showed borderline significance (P = 0.01) with D2S118 applying the VC method.
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The maximum LOD score (QTL) of 2.63 (P = 0.0002) was observed for the ratio (N·m/cm2) of knee extension at 60°/s with marker D2S118, 680 kb near the myostatin gene (GDF8). However, the VC analysis revealed a lower LOD score (1.24), suggesting difference in power with QTL, although other phenotypes showed in general comparable LOD scores (Table 4). The pattern of LOD scores for this marker was consistent over the different muscle groups (musculus quadriceps and hamstrings for extension and flexion, respectively) and velocities with exception of knee flexion at 120°/s and 240°/s. On chromosome 6, marker D6S1051 was chosen close to CDKN1A (13 kb), and marker D11S4138 on chromosome 11 is located 14.6 kb near MYOD1, and the same pattern, although less strong, emerges with LOD scores ranging between 1.17 and 1.87 (0.02 < P < 0.002). Together with some borderline significant LOD scores (0.05 < P < 0.01) (Table 4), these data can be interpreted as an indication of linkage with these loci but not as strong evidence.
The myostatin pathway was chosen for its possible effects on muscle mass. Surprisingly, linkage was mainly found with strength phenotypes, but only marginal evidence was present for its effects on estimated muscle cross-sectional area. Two muscle cross-sectional areas had LOD scores that just reached P = 0.01 with D2S118 (myostatin) (Table 4). However, a recently revisited Haseman and Elston method, also implemented in Merlin ("merlin-regress" option) (29) showed confirmation of suggestive linkage with estimated muscle cross-sectional area (LOD 1.121.18). Unfortunately, this study lacks more accurate MRI- or CT scan-based measures of muscle mass. Maybe the anthropometric estimates of muscle mass cannot differentiate enough between levels of muscularity and therefore fail to detect higher allele sharing between sibs with similar muscle mass. However, previous studies have shown a good correlation between CT scan measures of muscle mass and estimates of muscle mass by anthropometric equations (32). Nevertheless, caution should be taken when interpreting results of estimated muscle cross-sectional area as muscle mass indicators.
Whether earlier reported polymorphisms in the myostatin gene A55T (exon 1) and BanII K153R, TaqI E164K, and BstNI P198A (all in exon 2) (8) are responsible for the weak signal as found in this linkage study is unclear. In an earlier study we found no variation in the A55T, E164K, and P198A genotypes of 57 strength athletes and 57 control individuals (34). Only one individual was heterozygous for the K153R polymorphism in both strength athletes and control group. Because of the very low predicted allele frequencies of the rare alleles in these polymorphisms, subjects were not genotyped for these sequence variants in the present linkage analysis.
When phenotypes are compared over the different markers, the slow contraction velocities tend to be more linked to the three candidate regions (GDF8, CDKN1A, and MYOD1) than the higher velocities. Since the myostatin mechanism described earlier in animals mainly affects muscle mass, and since muscle mass is more correlated with isometric strength, it is not surprising that we found more suggestive linkage signals with the slower velocities. Another explanation is given by Carlson et al. (7) and Wehling et al. (36). They studied the effects of modified muscle use (unloading) on myostatin expression in different fiber types and found higher concentration of myostatin mRNA in fast-twitch muscle than in slow-twitch muscle. This, probably more prominent, inhibitory role of myostatin in fast-twitch muscle could be caused by fiber-specific, posttranslational modifications in myostatin. But it is not clear why the two other candidate genes are also more linked to strength phenotypes at lower contraction speeds.
The linkage pattern between the three loci and maximal knee muscle strength becomes more apparent when torque is considered in a specific angle during the three different movement speeds. The angles were chosen in which optimal strength could be generated following the force-length relationship (30° for flexion and 60° for extension). These torque measurements represent the same biological trait as the concentric peak torque values, and if similar linkage results are found with these traits, power is added to the interpretation of (suggestive) linkage. Indeed, the LOD scores of the angle-specific torques replicate the peak torque results (data not shown).
As could be expected, not all examined traits gave a signal for linkage despite the fact of biological relatedness. Also, some genes encoding proteins that play an important role in myogenesis (e.g., myogenin, IGF1) did not show signs of linkage to explain variation in muscle mass or strength. A number of reasons could explain the discrepancies in these findings.
First, one must consider the modest effect that any single gene would be expected to have on muscle mass or strength. These phenotypes are complex multifactorial traits that likely cannot be explained by a single gene, and, in addition, environmental factors contribute significantly to the observed variability. Moreover, one might assume that estimated muscle cross-sectional area and strength are also partly determined by gene-gene interactions as well as gene-environment interactions, which make it harder to determine a single gene effect. Second, single-point rather than the more powerful multipoint linkage analysis was performed because this was an explorative study where the primary goal was to screen the myostatin pathway for its potential in explaining interindividual differences in estimated muscle cross-sectional area and strength. Third, the power of this study was limited. With optimal assumptions on our maximum of 204 pairwise comparisons, power to detect a major QTL of 15% is 79.5% (n = 204 pairs; shared residual variance = 70%; recombination fraction = zero). However, for most markers this maximum number of pairs was not attained, and to detect a more realistic QTL of 10% effect size with a smaller sample (n = 120 pairs) and less shared residual variance (60%), power can drop to 30% (24).
Recently, more papers are being published that question the tendency to impose too stringent criteria in linkage studies to avoid false discoveries and thereby possibly losing true signals (4, 35). In addition, since this study was based on a biological hypothesis, specific candidate genes were used rather than equally spaced markers lacking any a priori biological relationship with the phenotype. Therefore, it can be argued that with a limited number of well-chosen candidate genes, conclusions can be drawn when the LOD score does not reach the critical value of 3.
Because of the exploratory nature of this study, patterns of (suggestive) linkage over different phenotypes are more interesting than individual LOD scores. The patterns seen in the region of 3 of 10 candidate genes indicate that this was a successful approach and strongly suggest that these loci (GDF8, CDKN1A, and MYOD1) may explain part of the interindividual variance of knee strength.
In conclusion, this study was the first explorative linkage study in humans to see whether the myostatin pathway might explain interindividual differences in estimated muscle cross-sectional area and knee strength. The present findings suggest that the chromosomal regions 2q32.2, 6p21.2, and 13q14.2 might harbor potential QTLs for skeletal muscle strength with GDF8, CDKN1A, and MYOD1 as good candidate genes. Since 3 of 10 candidate genes revealed suggestive single-point linkage on a limited sample size, it indicates that the myostatin pathway plays a role in human variation of muscle strength. Further studies on a larger sample and using a more dense marker saturation of these candidate regions (allowing multipoint linkage analysis) are required to confirm these results.
![]() |
GRANTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
ACKNOWLEDGMENTS |
---|
The following web sites for electronic databases are relevant to this paper: National Center for Biotechnology Information UniSTS web sitehttp://www.ncbi.nlm.nih.gov/genome/sts; the Genetic Power Calculator web site (http://statgen.iop.kcl.ac.uk/gpc/qtllink.html), and UCSC Genome Bioinformatics web site (http://genome.ucsc.edu/).
![]() |
FOOTNOTES |
---|
Address for reprint requests and other correspondence: M. A. Thomis, Dept. Sport and Movement Sciences, Faculty of Physical Education and Physiotherapy, Katholieke Universiteit Leuven, Tervuursevest 101, B-3001 Leuven (Heverlee), Belgium (E-mail: martine.thomis{at}flok.kuleuven.ac.be).
10.1152/physiolgenomics.00224.2003.
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|