The aim of the study was to explore the genetic architecture influencing weight of fast- and slow-twitch skeletal muscles. The weights of the slow-twitch soleus, the mixed gastrocnemius, the fast-twitch tibialis anterior (TA), and extensor digitorum longus (EDL) muscles were 11–34% greater (P < 0.001) in 200-day-old C57BL/6J (B6) than in DBA/2J (D2) mice. Male muscles were 13–28% larger than female (P < 1 × 10−5, no strain by sex interaction). The sex-related difference in muscle weight, however, varied significantly among the 23 derivative BXD recombinant inbred (RI) strains (strain by sex interaction for soleus, P < 0.01; TA, P < 1 × 10−4; EDL, not significant; and gastrocnemius, P < 0.001). Quantitative trait loci (QTL) affecting muscle weight were mapped in an F2 intercross of B6 and D2 mice (B6D2F2) and BXD RIs. A total of 10 autosomal, muscle-specific, but not muscle-type-specific, QTL, explaining a total of 5.4, 7.7, 22.9, and 8.6% of phenotypic variance for soleus, TA, EDL, and gastrocnemius muscles, respectively, were found across chromosomes 1 (Chr 1), 2, 3 (female-specific), 5 (two), 6, 7, 8, and 9 in B6D2F2 mice. The QTL on Chr 8 for EDL and the female-specific QTL on Chr 3 for gastrocnemius muscles were statistically significant, but the remaining QTL were at the suggestive level of statistical significance. Ten QTL on Chr 1, 2, 4, 5, 7, 8, 14, 17 (two), and 19 were identified in BXD RIs. Half of the QTL in BXD RIs had pleiotropic effects and were at the suggestive level of significance (except for the significant QTL for gastrocnemius muscle on Chr 17). The B6D2F2 nominated QTL on Chr 8 for EDL weight was validated in BXD RIs (P < 0.03). Support intervals for the QTL on Chr 1 and 5 overlapped between B6D2F2 and BXD RIs. An epistatic interaction between markers on Chr 1 and 17 affected gastrocnemius weight in BXD RIs. The interaction was not, however, validated in the B6D2F2 population. Our results indicate that the differences in muscle weight in the B6 and D2 segregating populations were the outcome of a polygenic system, with each factor contributing a small amount to the phenotypic variance and the genetic architecture affecting muscle weight was muscle specific, but not muscle-type specific, and in some instances sex specific.
- muscle size
- quantitative trait locus
as well as performing its primary function of generating power, skeletal muscle also performs other secondary roles. It serves as a source of energy and amino acid during critical illness and starvation, contributes to thermoregulation, and passively protects bone and viscera. The size (weight) of skeletal muscle has a major impact on all of these vital functions, and the relative importance of the different skeletal muscle functions varies through the life span of an individual. There is a large variability in skeletal muscle size among individuals, and the influences of environmental factors like exercise, microgravity, hormonal influences, and critical illness have been well explored (14, 15, 22, 26, 29, 30, 48). On the other hand, our understanding of genetic factors affecting muscle mass is incomplete.
In agricultural science, phenotypic selection in cattle, sheep, and pigs has been used to obtain breeds and strains with increased muscle mass (1, 11, 23). The quest for specific genes causing alterations in muscle mass has identified several candidates. In cattle, myostatin has been recognized as a powerful suppressor of muscle mass growth (20, 21). In mice, selection for high carcass protein content resulted in formation of a new mouse line with a “compact” phenotype, featuring an increase in muscle mass (43) due to remarkable hyperplasia, and to some extent hypertrophy, of muscle fibers (38).
Insulin-like growth factors I and II (IGF-I, IGF-II) are also well known modulators of muscle weight. Overexpression of IGF-I induces hypertrophy and reverses aging-related atrophy in rodents (2, 8, 37). Alternative splicing of IGF-I results in an autocrine, muscle-tissue-specific mechanogrowth factor (MGF) that is expressed in response to strenuous exercise or injury (16). The transfection of MGF cDNA into mature mouse muscle increased muscle mass by up to 20% and fiber size by 25% within 2 wk (17).
Notwithstanding these important advances in the understanding of the contribution of individual genes to muscle mass, the genetic architecture underlying muscle size is likely to include many other genes. A recent study by Brockmann and coauthors (7) identified the existence of seven quantitative trait loci (QTL) and epistatic interactions that accounted for over 32% of quadriceps weight variance in a cross between the high-growth DU6i mouse strain and DBA/2 (7). However, muscle-specific and muscle-type-specific differences were not addressed in this study. Data from several lines of evidence lead to the notion that genetic influences on muscle size differ between different muscles. In particular, differing patterns of muscle degeneration from clinical observations in patients with hereditary myopathies, along with the different contractile, metabolic, and structural properties found in muscles of the fast- and slow-twitch type, suggest that while some genes exert a gross effect in all muscles, others are specific to groups of muscles based on type or function (12, 18). This assumption is supported by the recent discovery of muscle-specific QTL in domestic pigs (Sus scrofa) (34, 39, 44).
Our pilot experiments indicated that four hindlimb muscles, one slow-twitch, one mixed, and two fast-twitch were 11 to 34% heavier in the C57BL/6J (B6) than in the DBA/2J (D2) strain. We hypothesize that the genetic architecture of muscle weight is muscle specific and muscle-type specific. The aim of this study was to investigate the genetic architecture of slow soleus, fast tibialis anterior (TA), and extensor digitorum longus (EDL) and mixed gastrocnemius muscle weights. In an attempt to search for QTL affecting weight of the four different muscles, we used two segregating populations derived from B6 and D2 strains: a B6D2F2 intercross and 23 BXD recombinant inbred (RI) strains. The results from this study have been presented in short form elsewhere (31).
MATERIALS AND METHODS
All animals in this study (C57BL/6J, B6; DBA/2J, D2; 23 BXD recombinant inbred, RI; strains and B6D2F2 animals) were bred in the barrier colony developed at the Center for Development and Health Genetics (CDHG), the Pennsylvania State University. The colony was derived from cesarean section derivation of Jackson Laboratory stock onto Swiss foster mothers. The B6D2F2 intercross was generated in four birth cohorts of ∼100 mice at 1-mo intervals. Each birth cohort of B6D2F2 consisted of an equal number of four possible mating types of F1 males and females, i.e., B6D2 × B6D2, B6D2 × D2B6, D2B6 × B6D2, and D2B6 × D2B6 [D2B6 is the F1 offspring of D2 female (female is always the first member of a mating pair) and B6 male, where B6D2 is the offspring of a B6 female and a D2 male and so on]. This permitted comparison of the effect of the B6 and D2 Y chromosomes in males of the F2 generation. Four hundred B6D2F2 mice, born to 50 primiparous females (range 8–13 pups per litter) began this study. From this group, a total of 188 females and 192 males survived until dissection. Mice of the 23 BXD RI strains were weaned from both primiparous and multiparous females (range, 2–12 pups). The BXD RI study design called for 12 males and 12 females within each strain. Early death reduced the number of animals in some strains or even required additional breeding (BXD13). A total of 249 females and 251 males from BXD RIs survived until dissection. All animals were housed at room temperature (70–72°F) on a 12:12-h light/dark cycle and fed ad libitum with sterilizable Purina 5010 (designed to be equivalent to Purina 5001 after sterilization). At the age of ∼200 days, B6 (215 ± 2 days, range 212–218 days), D2 (214 ± 3, range 211–219 days), B6D2F2 (203 ± 10, range 184–226 days), and BXD (215 ± 4, range 199–221 days) mice were killed by cervical dislocation between 8 AM and 12 PM, i.e., in the first half of the light cycle. Procedures were approved by the Penn State Animal Care and Use Committee.
A variety of phenotypic measures (blood pressure, urine osmolarity, blood chemistry, and behavior) were assessed within the barrier before death. These were carried out in triplicate at 4-wk intervals for the B6D2F2 and once for the BXD RI mice. Animals were weighed on an electronic balance (Ohaus Scout) to 0.1-g accuracy before death, and nose-to-anus distance (with an accuracy of ±1 mm) was recorded immediately after. Multiple tissues were harvested from these animals, and skeletal muscles were dissected from the right hindlimb. The gastrocnemius, soleus, TA, and EDL muscles were gently dissected and weighed on a Mettler AE50 balance. Then the muscles were frozen and stored for further analyses.
The B6D2F2 mice (n = 394) were genotyped at 97 microsatellite markers spaced at ∼15- to 20-cM intervals throughout the genome (Table 1). DNA for genotyping was extracted from tail-tips obtained at weaning. There were 623 microsatellite markers selected for the 23 BXD RIs (47).
The Windows front-end of QTL Cartographer package (version V2.0) and MapManager QTX (version b14) were used for linkage analysis in the combined population including both sexes (3, 35). Analyses are based on likelihood ratio statistics in both QTL Cartographer and MapManager QTX packages. Multiple interval mapping available in QTL Cartographer was used in B6D2F2 population because it combines QTL mapping with the analysis of genetic architecture of a trait, i.e., multiple putative QTL and epistatic interactions are fitted into a model to estimate positions, effects, and interactions of multiple QTL (49). Interval mapping was used in BXD RIs and also in B6D2F2 population (QTL Cartographer). Within-sex linkage analysis in B6D2F2 population was conducted on a trait if genotype at a marker closest to the peak of a QTL interacted with sex (see Statistical Analysis, below) and for all muscles in BXD RIs, because statistically significant sex by strain interactions were found for all muscles. MapManager QTX program was used to run permutations and to search for epistatic interactions between pairs of markers in both B6D2F2 and BXD RI populations.
Suggestive, significant, and highly significant thresholds of statistical significance (values corresponding to 37th, 95th, and 99.9th percentiles, respectively) were determined for each muscle from 10,000 permutations (9, 10). Searching for epistatic interactions, the threshold of statistical significance was set at P = 1 × 10−5 and P = 1 × 10−7 in B6D2F2 and BXD RI populations, respectively, corresponding to P < 0.05 after correction for genome-wide comparisons (9). A default Bayes information criterion for model selection was used in the multiple interval mapping procedure.
There were 372 and 486 soleus, 367 and 491 TA, 367 and 485 EDL, and 357 and 493 gastrocnemius muscles available for analyses of B6D2F2 and BXD RI populations, respectively. The influences of age, sex, litter size, mating type, and birth cohort were estimated via correlation analyses in B6D2F2. The effects of age and sex were investigated in BXD RIs (litter size was not available, whereas birth cohorts did overlap with strains in BXD RIs and therefore could not be used as correction factors). In B6D2F2 mice, weight of all muscles significantly correlated with sex and litter size; in addition, soleus, TA, and EDL muscles correlated with age, and soleus and EDL correlated with birth cohort. In BXD RIs, weight of all muscles significantly correlated with sex, whereas age correlated with soleus, TA, and gastrocnemius muscles. Weights of all muscles were corrected for statistically significant independent variables (except for sex) via multiple regression analyses. The residuals of muscle weight correlated significantly with body weight and body length. Body length was preferred as a factor to adjust for body size, because correcting muscle weight for body weight is to some degree a correction for itself. Because regression slopes for covariates may be different for females and males, standardized residuals were calculated via regression analysis within each sex. Muscle weight of progenitors was also corrected for body length. These residuals were combined in all further analyses and referred to as muscle weight unless otherwise stated. After correction for body length, there was no statistically significant correlation between muscle weight and body weight.
Weight residuals of different muscles were significantly and positively correlated. To identify the putative underlying factor(s) that explain the pattern of correlations, we conducted principal component analysis (SPSS 11.0 statistical software package) on residuals of four muscles in both B6D2F2 and BXD RI populations. The first principal component (PC1) was included in additional analyses.
Test for normality of distribution (Kolmogorov-Smirnov test) was carried out before running analysis of variance and linkage analyses (SPSS 11.0 statistical software package). The residuals of weight of all muscles in B6D2F2 population did not deviate significantly from normal distribution. In BXD RIs, soleus weight was normally distributed both within sex and within each strain. Deviations from normality were observed for three other muscles in males (the most extreme deviation was for EDL; skewness = 1.67) and within some strains. These deviations were not consistent among strains (no strain deviating in more than one muscle) or related to a specific muscle (three strains exhibited significant deviation within each muscle), suggesting that it is a consequence of the small sample within strain rather than a scale effect. Attempts to correct the distribution would result in normality within sex but not within each of the strains; therefore, no transformation was applied.
Heritability (narrow sense) of muscle weight was estimated from a one-way ANOVA by strain in BXD RIs as a SSbetween strains/SStotal (13). The broad sense heritability (13) was estimated based on variances of B6D2F2 intercross, B6, D2, and BXD RIs where H2 is broad sense heritability, VP is phenotypic variance represented by the variance of B6D2F2 intercross, and VG is genetic variance, calculated as a difference between phenotypic and environmental variance. Environmental variance was estimated by averaging variances of B6, D2, and 23 BXD RIs. Both broad and narrow sense heritability were estimated on the residuals of muscle weight.
Linkages to X chromosome in B6D2F2 intercross were analyzed separately from autosomes because males are hemizygous for the X and in females only homozygous individuals give unambiguous information due to random inactivation of the X chromosome (24). The markers genotyped on the X chromosome are located proximal to the pseudo-autosomal region (with the possible exception of the most distal one, http://www.informatics.jax.org/searches/marker_report.cgi) and therefore cannot be treated as pseudo-autosomals. In these analyses, the X chromosome was divided into four intervals, each flanked by two adjacent markers, and only individuals homozygous for the same allele in flanking markers were selected for the two-way analyses of variance (ANOVA) with sex and intervals as the factors. The effect of the X chromosome in BXD RIs was also analyzed separately from autosomes via one-way ANOVA on a marker (14 markers in total).
Strain means of the residuals of muscle weight calculated for BXD RIs passed the normality test (Kolmogorov-Smirnov test) and were used to conduct linkage analyses. The reliability of strain means was tested in the following manner: median scores of the raw muscle weight of odd- and even-numbered mice within each BXD male and female group were computed. Pearson correlation coefficients between odd and even median scores for each strain ranged between 0.85 and 0.90 for all muscles of both genders, except female soleus, r = 0.76. By comparison, body weight correlated 0.87 and 0.94 in females and males, respectively. Thus reliability of the strain means, which to some extent depends on accuracy of the dissection and weighing procedure, appears high and satisfactory for all muscles.
A standard two-way ANOVA was used to estimate sex by strain interaction in progenitors and BXD RIs and sex by genotype interaction for QTL-flanking markers in the B6D2F2 intercross.
A significant (P < 1 × 10−4 for each muscle) strain effect on residuals of weight was observed in all four muscles. The raw weights of the four muscles are shown in Table 2. B6 muscle raw weights were heavier than D2. The raw weight of all four muscles was significantly larger in males than females (P < 1 × 10−5 for each muscle). There was no statistically significant interaction between strain and sex. Proportionately, raw muscle weight was 11–34% heavier in B6 than D2 (across sexes) and 13–28% heavier in males than females (across strains).
Two-way ANOVA of body weight indicated statistically significant differences due to strain (P < 0.005) and sex (P < 1 × 10−7). The D2 mice were significantly heavier than B6, and males were significantly heavier than females (B6, 31.8 ± 2.8 and 25.4 ± 2.0 g; D2, 35.1 ± 4.1 and 28.3 ± 4.2 g, males and females, respectively). There was no interaction between strain and sex. Thus sex differences in raw muscle weight were paralleled by sex differences in body weight, but strain differences in the muscle and body weight were inversely related. The ANOVA on body length indicated a statistically significant effect of sex (P < 1 × 10−7) but no significant strain effect nor strain by sex interaction (Table 2).
As was the case for the progenitor strains, the effect of sex on the raw muscle weight was highly statistically significant in B6D2F2 mice. Although the age of the B6D2F2s was somewhat younger than of the progenitors, the mean weight of B6D2F2 muscles was closer to the weight of the B6 strain than to that of D2 (Table 2). Body weights of B6D2F2 male and female mice were 39.3 ± 5.0 and 28.3 ± 4.4 g, respectively.
Pearson correlation coefficients were calculated among residuals of weight of individual muscles, body weight, and length (Table 3). Residual weight of different muscles intercorrelated positively, and they also correlated positively with body weight and length, although in general, the correlation between muscles and body weight and length tended to be weaker than that between muscles. A positive correlation between muscles in the B6D2F2 population is consistent with a pleiotropic genetic effect, but the correlation also represents the environmental influences in B6D2F2. To identify the common factor affecting all muscle weights, principal component analysis was conducted on the residuals of muscle weight. The first principal component (PC1) was found to account for 56% of total muscle weight variance in B6D2F2 mice. It has to be noted that the PC1 reflects both genetic and environmental, i.e., phenotypic, influences on muscle weight in B6D2F2 population.
The F1 mating design (see materials and methods) permitted comparison of males carrying a Y chromosome of B6 (n = 95) or D2 (n = 97). There were no statistically significant differences in muscle weight between mice carrying the B6 and D2 Y chromosome.
BXD RI strains.
The effect of strain on the residuals of muscle weight in BXD RIs was highly significant for all four muscles (P < 1 × 10−55). The smallest raw muscle weights were typically observed in the BXD 22, 32, and 40 strains, and the largest were observed in the BXD 8, 29, and 33 strains. Considering individual raw muscle weight strain means, the percent differences between largest and smallest for males and females were as follows: soleus 50 and 48%; TA 37% and 33%; EDL 36% and 37%; gastrocnemius 30% and 32%. Genetic correlations between muscles were calculated from strain muscle weight means (5) and ranged from 0.587 (P < 0.01) to 0.813 (P < 0.01, Table 4). PC1 estimated across sexes in BXD RIs accounted for 71% of variance. The muscle-to-body weight relationship was weaker, with correlations ranging from 0.005 to 0.344 (ns, Table 4). The distribution of raw BXD RI strain means suggests a polygenic effect, rather than a major gene, affecting muscle weight (Fig. 1). Strain means of PC1 representing the influence of predominantly genetic factors common to all muscles were entered into the WebQTL program (http://www. webqtl.org) to search for correlations with other published phenotypes. The analysis indicated weak correlations between PC1 and body weight (r ranged between −0.253 and 0.216, not significant; Refs. 4, 46, 50). A strong negative correlation was observed between PC1 and methamphetamine-induced climbing score, r = −0.703 (nominal P = 0.001; Ref. 19).
Strain by sex interaction was statistically significant for all muscles (soleus, P < 0.01; TA, P < 1 × 10−4; EDL, P < 0.05; and gastrocnemius, P < 0.001). Female-to-male raw muscle weight ratio of sex-specific strain means was calculated and ranged from 67 to 94% for soleus, 74 to 100% for TA, 73 to 93% for EDL, and 74 to 92% for gastrocnemius muscle, and the distribution of BXD RI strain means indicated a polygenic effect in all muscles (not shown). In some strains, e.g., BXD 6, 12, and 32, female-to-male weight ratio was higher across all muscles (85–100%), whereas in others, e.g., BXD 24, it was lower (74–80%). Female-to-male ratio was not related to the mean muscle weight of a strain, e.g., the ratio was similar between BXD 6 and BXD 32, whereas all four muscles of BXD 6 were larger than corresponding muscles of BXD 32 strain (for details see Supplemental Table A, available at the Physiological Genomics web site).1 The female-to-male ratio was interrelated in all muscles with correlations from 0.46 (P < 0.05) to 0.841 (P < 1 × 10−6). The strongest correlations were observed for gastrocnemius muscle with TA (r = 0.841, P < 1 × 10−6) and EDL (r = 0.545, P < 0.01) and for TA with EDL (r = 0.523, P < 0.05). The correlation coefficients of female-to-male ratio for soleus were 0.472 (P < 0.05), 0.460 (P < 0.05), and 0.486 (P < 0.05) with TA, EDL, and gastrocnemius muscles, respectively.
Heritability of the residual of muscle weight was calculated within sex because of a statistically significant strain by sex interaction in BXD RIs. The narrow sense heritability of muscle weight was estimated to be 0.61 and 0.62 for soleus, 0.66 and 0.60 for TA, 0.49 and 0.57 for EDL, and 0.66 and 0.55 for gastrocnemius in females and males, respectively. This indicates that heritability of muscle weight is similar for slow-twitch (soleus), fast-twitch (TA, EDL), and mixed muscles (gastrocnemius). In between-sex comparison, the narrow sense heritability was the same in males and females for soleus and similar for TA, EDL, and gastrocnemius muscles. The narrow sense heritability estimated in combined males and females population was 0.57, 0.58, 0.50, and 0.54 for soleus, TA, EDL, and gastrocnemius muscles. The broad sense heritability of muscle weight was estimated to be 0.36 and 0.23 for soleus, 0.17 and 0.23 for TA, and 0.41 and 0.42 for gastrocnemius in females and males, respectively (corresponding estimates in combined male and female populations were 0.24, 0.25, and 0.40). For the EDL muscle, the broad sense heritability resulted in the negative estimates in both sexes.
The residuals of muscle weights and the PC1 were subjected to an interval mapping procedure to search for QTL affecting muscle weight. The interval mapping identified one significant QTL affecting EDL weight and several suggestive QTL influencing different muscles (Fig. 2). Multiple interval mapping enables simultaneous mapping of multiple QTL, identification of possible epistatic interactions, and estimation of r2 for the final model (49). It has been shown that multiple interval mapping is useful in improving estimates of position and in enhancing statistical power (27, 49). Therefore, residuals of soleus, TA, EDL, and gastrocnemius muscle weights and PC1 were subjected to multiple interval mapping analysis. Some of the QTL identified by interval mapping were not found to be statistically significant in multiple interval mapping analysis. We are reporting QTL that were at least suggestive in the interval mapping analyses and statistically significant in multiple interval mapping [chromosomes 1, 2, 3, 5 (two), 6, 8, and 9]. Based on multiple interval mapping analyses, estimated effect of a model accounted for 5.3%, 8.2%, 12.6%, 8.5%, and 4.3% of variance for soleus, TA, EDL, gastrocnemius, and PC1, respectively (Table 5). There were no QTL with pleiotropic effect. If a QTL was observed on the same chromosome for different muscles, e.g., soleus and EDL on chromosome 5, then these were in different locations (Fig. 2). Therefore, all identified QTL appeared to be muscle specific. Multiple interval mapping identified a QTL for PC1 on chromosome 1 (Table 5). An increase in muscle weight was associated with the B6 allele in six and with D2 in four of the QTL. To test the hypothesis that QTL may have sex-specific effects in the B6D2F2 intercross, the ANOVA on the marker closest to the QTL peak was performed. Statistically significant (following Bonferroni correction) sex by marker interaction was found for gastrocnemius muscle on chromosome 3 (P < 0.05). Within-sex multiple interval mapping analysis identified QTL in females but not in males (Fig. 3).
Epistatic interactions were tested between the identified QTL as well as between pairs of other markers. Neither multiple interval mapping analysis between the QTL using QTL Cartographer nor genome-wide search for pairs of interacting markers using MapManager QTX identified any significant interaction.
The X chromosome was analyzed separately from autosomes via two-way ANOVA with sex and genotype as the factors (see materials and methods). There were 52–77 females and 128–148 males included into the analysis across different muscles. These analyses did not identify any interval on the X chromosome with a statistically significant effect on residuals of muscle weight.
To test whether identified QTL are also influencing body weight in B6D2F2 mice, body weight was included as a covariate into ANOVA conducted on the marker closest to the QTL peak. None of the QTL effects were removed by this adjustment.
BXD RI strains.
Search for the QTL affecting strain mean muscle weight was done using the interval mapping procedure since the multiple interval mapping would generate a nonreliable output due to the limited sample size in combination with the dense set of microsatellite markers. The interval mapping identified one significant QTL and several suggestive QTL on chromosomes 1, 2, 4, 5, 7, 8, 14, 17 (two), and 19 (Fig. 4). Most of the QTL affected several muscles; however, QTL on chromosomes 5, 7, 8, 14, and distal QTL on chromosome 17 influenced a single muscle. The QTL on chromosomes 7, 8, 14, and 19 appeared in one sex more than the other; however, ANOVA on strain means with sex and genotype at a marker did not indicate statistically significant interactions between main effects. The B6 allele was associated with a greater muscle weight in 6 of 10 QTL (Table 6). In a situation of limited power due to small sample size, estimate of variance accounted for by QTL becomes highly overestimated (33), and we have therefore omitted those estimates.
To test whether identified QTL also influence body weight of BXD RIs, body weight was included as a covariate into ANOVA conducted on the marker at the QTL peak. None of the QTL effects were removed by this adjustment.
A genome-wide search of interactions was only carried out between markers with nonduplicate genotypes across 23 BXD RIs (432 markers), thereby minimizing the number of comparisons. Interaction between markers D1Mit134 and D17Mit3 affected weight of gastrocnemius muscle (genome wide P < 0.05) in a way that presence of D2 alleles at D1Mit134 was necessary to cause an increase or decrease in weight in combination with B6 or D2 alleles at D17Mit3, respectively (Fig. 5).
Having two segregating populations of the same progenitors offers the possibility to verify significant QTL nominated in B6D2F2 intercross. One of the nominated QTL was verified in BXD RI analyses. Within the 1-LOD drop support interval of the QTL nominated for EDL muscle on chromosome 8 (58.3 cM, 1-LOD interval 53 cM, end of chromosome), a QTL was found affecting the same muscle in BXD RIs (52 cM, 1-LOD 47–56 cM). The D2 allele was associated with an increase in muscle weight in both populations. Three markers, D8Mit211 (46.3 cM), D8Mit113 (49.5 cM), and D8Mit271 (54.1 cM), were within the 1-LOD drop support interval of BXD RIs. The one-way ANOVA on EDL muscle with each of them as a factor indicated the strongest linkage with D8Mit113 (P < 0.009, nominal), fulfilling the criteria for linkage verification (28). The QTL on chromosome 3 affecting gastrocnemius muscle in B6D2F2 females was not observed in the BXD RI strains.
Effects of the markers on the X chromosome were analyzed via the one-way ANOVA. The analysis, however, did not find any statistically significant effects.
A significant epistatic interaction was identified in BXD RIs. For validation of nominated interactions the threshold of significance may be lowered to nominal P < 0.01 because only two markers flanking each nomination would be compared. Because epistatic interaction was observed in both sexes of BXD RIs, a two-way ANOVA on gastrocnemius weight in combined sexes of B6D2F2 population was carried out. Interaction between markers D1Mit87 and D17Mit123 was found to be statistically significant at a nominal P < 0.036 (corrected P = 0.144). Furthermore, the nature of the interaction in B6D2F2 population (B6B6 × B6B6 ≤ B6B6 × D2D2 and D2D2 × B6B6 ≤ D2D2 × D2D2, genotypes at D1Mit87 × D17Mit123) was different from that which was found in BXD RIs (B6B6 × B6B6 ≈ B6B6 × D2D2 ≈ D2D2 × B6B6 > D2D2 × D2D2). Thus the nominated interaction has not been validated.
Muscle weight is a predictor of several important aspects of physiological function including contractile power, thermoregulation, amino acid pool, and passive protection of other organs. Therefore, understanding of the genetic architecture of muscle size is of significant interest to basic, applied, and clinical science.
Body weight is strongly influenced by muscle mass, and highly significant correlations between these measures were reported in juvenile (42 days) mice (7). In the 200-day-old mice in this study, on the other hand, both phenotypic and also genetic correlations between strain means (reflecting genetic commonality of traits) of muscle and body weight were smaller (Table 3 and 4) than those previously observed. Weak correlation between muscle and body weight was supported by the observation that the inclusion of body weight as a covariate did not eliminate any QTL in either B6D2F2 or BXD RI populations. Taken together it suggests that the genetic architecture underlying body weight and muscle mass diverges with age, perhaps reflecting differences in body composition, e.g., accumulation of fat, between 42- and 200-day-old mice. In this study, two 200-day-old populations were derived from B6 and D2 mice, B6D2F2 intercross, and BXD RI strains and were used to search for chromosomal intervals containing genes affecting residuals of weight of both fast- and slow-twitch skeletal muscles. The muscle weight differences between the progenitor strains and among BXD RIs as well as estimated heritability suggested that the B6/D2 lineage was suitable to search for QTL affecting muscle weight. The availability of genome sequence of both B6 and D2 strains makes segregating populations of these strains even more attractive for linkage studies, because it facilitates the search for polymorphisms. Genetic architectures of muscles involved in different tasks and possessing different functional and histochemical properties are likely to differ. In recent studies, a single muscle or a whole group of muscles were measured in mice, thus the possibility of muscle specificity was not addressed (7, 36). In agricultural animals, muscle-specific QTL were reported in several crosses of domestic pigs (34, 39, 44). The hypothesis of muscle specificity, but not muscle-type specificity (fast- vs. slow-twitch), was supported in the B6D2F2 intercross. Specific anatomical structure, histological and functional differences among muscles in combination with a complex control of expression of Myf5, a myogenic regulatory factor, may be a substantial modifier of individual muscle weight (41). Principal component analysis in the B6D2F2 intercross indicated a substantial degree of commonality among the four muscles, with the PC1 accounting for 56% of muscle variance. This commonality, however, contains both genetic and environmental components in B6D2F2 population. Presence of significant linkage in PC1 would be consistent with pleiotropic influence on muscle weight. The single suggestive QTL for PC1, mapped on the chromosome 1, was within the region of subthreshold linkages in soleus, EDL, gastrocnemius muscles and suggestive QTL affecting TA muscle with the same allelic model (Fig. 2).
Muscle specificity of QTL in the B6D2F2 intercross was somewhat contradicted by the significant genetic correlation of the weight among muscles and apparent pleiotropic QTL identified in BXD RIs (Table 5 and 6). The lack of pleiotropic effects in the B6D2F2 sample might be explained in two different ways: 1) one can hypothesize that the common causation is not genetically based, or 2) common causation might be attributed to many loci of small effect that could not be detected because of confounding influences of local factors and insufficient power. The significant correlation between strain means of muscle weights and apparent pleiotropic QTL in BXD RI strains speak in favor of the second alternative. A lack of power might have also precluded detection of epistatic effects in the B6D2F2 intercross, whereas epistatic interactions were also reported to affect muscle size in other studies (7, 36).
Significant strain by sex interactions in BXD RIs and sex-specific QTL in B6D2F2 population suggested that the genetic architecture of muscle weight varied between males and females. In addition, distribution of female-to-male weight ratios among strains indicated a polygenic effect. The mechanisms underlying sex-specific effects are unknown but may arise from sex hormone regulation of the polymorphic genes or interactions between mitochondrially linked or Y-chromosome-linked genes.
Having two segregating populations of the same progenitors offers the possibility of QTL verification. BXD RIs are also advantageous because denser recombination may narrow down the chromosomal region harboring a gene. As shown by Lander and Kruglyak (28), an attempt to replicate linkages that are only significant at a “suggestive” level of statistical significance may result in confirmation of spurious QTL. Therefore only QTL that are above the highly significant and significant thresholds of statistical significance are suitable subjects for verification. In the present analysis of the B6D2F2 population, two QTL, on chromosome 3 affecting gastrocnemius muscle in females and on chromosome 8 affecting EDL weight in combined male and female population, reached the appropriate levels of statistical significance (Table 2, Fig. 2). The first QTL, however, was not found in BXD RIs, whereas the QTL on chromosome 8 was validated according to the criterion for QTL replication (28).
Aside from those QTL which reached the formal levels of significance, it is important to mention several other findings which may be useful to future investigators. Among QTL which reached the suggestive threshold of significance, those on chromosomes 1 and 5 paralleled QTL found in BXD RIs for allelic model and support interval (Table 5 and 6). A QTL on chromosome 2 that reached a suggestive level of statistical significance (B6 allele associated with increased weight) affected EDL muscle in B6D2F2 population and had a pleiotropic influence on all muscles in BXD RIs with the same allelic model (Table 5 and 6). The support intervals of 1-LOD drop were 1–6 cM apart between the populations. As it has been discussed by Blizard and Darvazi (6), a 1-LOD drop support interval is not a reliable representation of 95% confidence interval in F2 generation, and in many cases at least a 2-LOD drop support interval is required. Therefore, the effects may represent the influence of the same gene(s) on chromosome 2 in both populations. Finally, a QTL affecting gastrocnemius weight on chromosome 4 that reached the suggestive level of statistical significance in B6D2F2 interval mapping analyses (but was not significant in multiple interval mapping analyses) was supported in BXD RIs analyses (Fig. 2 and 4).
Verification of B6D2F2 nominated QTL in BXD RIs may be hindered by random effects of sampling or multiple genes underlying a specific QTL effect. For example, B6D2F2 nominated QTL and/or epistatic interactions may reflect the influence of multiple loci in a chromosomal region, and analysis with RI strains that have multiple opportunities for recombination during their generation may separate the effects of loci clustered within a specific chromosomal region. The shortcomings of the present sample of BXD RIs, such as 1) low power because of a limited sample size (n = 23 RIs), meaning that a QTL has to be of a substantial effect size to be detected, and 2) nonsyntenic associations, may have reduced chances to identify or verify QTL (47). The nonsyntenic association found between the genotypes on chromosome 1 (∼16–26 cM) and chromosome 9 (∼28–35 cM) may have precluded detection of QTL through those regions in BXD RIs. Pearson correlation coefficients between genotypes of the microsatellite markers of the two regions ranged from r = 0.455 (P < 0.05) to r = 0.760 (P < 0.01) between different markers. These regions matched the position of the QTL on chromosomes that affected TA muscle weight in B6D2F2 mice (Table 5). Furthermore, in B6D2F2 analyses there was an opposite allelic model for the increase in muscle weight of the QTL on chromosomes 1, D2 associated with greater weight, and 9, B6 associated with greater weight. The positive correlation between the regions would therefore be likely to diminish the effect of the individual loci considered separately. Thus, due to the combination of small effect size with limited power of a group of 23 BXD RIs, verification of the QTL, as well as absence of it, should be considered cautiously.
Finally, three QTL identified in the present study matched the positions of muscle weight QTL reported by other groups: Cmpt phenotype locus found in a cross of BALB/c and compact mice on chromosome 1 (42) and a forearm muscle size QTL in a cross of MRL/MPL and SJL/J on chromosome 14 and 17 (36).
As noted, muscle weight is determined by the amount of its constituent proteins and retained fluid. Therefore, genes involved in signaling, transcription, translation and degradation of proteins, as well as ion transport, may contribute to the difference in muscle weight. Studies in mice and domestic animals indicated that myostatin (20) and IGF-I and IGF-II (32) are powerful modifiers of muscle mass. Deletion of the proto-oncogene FOS induced growth retardation in mice (25, 45). Polymorphism of the FOS gene was also associated with skeletal muscle fiber traits in pigs (40). We searched the web site at http://www.informatics.jax.org/searches/marker_form.shtml for the candidate genes within a 1-LOD support interval of the significant and suggestive QTL in B6D2F2 and BXD RI populations. Products of the following genes have a potential influence on muscle weight in B6 and D2 segregating populations: myostatin, β/slow myosin heavy chain isoform, signal transducer and activator of transcription 1 and 4 on chromosome 1; titin on chromosome 2; IGF-II binding protein 3 on chromosome 6. It has to be noted that the genetic effect of myostatin is dominant/recessive in mice (38), whereas the QTL on chromosome 1 (the region containing the myostatin gene) was additive in the B6D2F2 population of the present study. Therefore, the observed QTL effect may be influenced by another gene(s) in close proximity rather than by myostatin itself. No strong muscle-specific candidate gene is known for QTL on chromosomes 3, 4, 5, 7, 8, 9, 14, 17, and 19.
The results from this study show that the difference in muscle weight and sex-related variation of muscle weight in B6 and D2 segregating populations is due to a polygenic system and that the genetic architecture may be sex specific and muscle specific, but not muscle-type specific. The availability of BXD RIs and the fact that the sequence of both B6 and D2 strains recently became available will facilitate the search for polymorphic loci and eventually for genes, extending our knowledge on genetic architecture affecting muscle weight.
We are grateful to Dr. F. Yu, Dr. N. Sharkey, C. Vandenberg, and J. Heckman for excellent help with the tissue harvesting.
This study was supported by National Institutes of Health Grants AG-14731, AR-45627, and AR-47318 and by European Commission Grants BMH4-CT96-0174 and QLK6-CT-2000-0530.
↵1 The Supplementary Material for this article (Table A) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00103.2003/DC1.
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: A. Lionikas, Penn State Univ., 101 Amy Gardner House, University Park, PA 16802 (E-mail:).
- Copyright © 2003 the American Physiological Society