Mapping Modifiers Affecting Muscularity of the Myostatin Mutant (Mstn^Cmpt-dl1Abc) Compact Mouse

^a Institute for Animal Biology, Agricultural Biotechnology Center, H-2101 Gödöll, Hungary,

Corresponding author: László Varga, Agricultural Biotechnology Center, P.O.B. 411, H-2101 Gödöllő, Hungary., varga{at}abc.hu (E-mail)

Communicating editor: N. A. JENKINS

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The hypermuscular Compact phenotype was first noted in a line of mice selected for high body weight and protein content. A new line, based on mice showing the Compact phenotype, was formed and selected for maximum expression of the Compact phenotype. Previously we mapped and identified a 12-bp deletion in the myostatin gene, denoted Mstn^Cmpt-dl1Abc, which can be considered as a major gene responsible for the hypermuscular phenotype. Genetic analysis revealed that full expression of the hypermuscular phenotype requires the action of modifier loci in addition to Mstn^Cmpt-dl1Abc. To map these modifier loci, an interspecific F₂ population was generated between Comp9, an inbred line homozygous for Mstn^Cmpt-dl1Abc, and CAST/Ei, an inbred line generated from Mus musculus castaneus. Selective DNA pooling and genotyping, separately by gender, was carried out within a subpopulation of the F₂ consisting of individuals homozygous for Mstn^Cmpt-dl1Abc. Significant association with hypermuscularity at a false discovery rate (FDR) of 0.05 was found for markers on chromosomes 3, 5, 7, 11, 16, and X. In all cases, the marker allele derived from the Comp9 parent showed a higher frequency in the hypermuscular group and the CAST/Ei allele in the normal group. The modifier loci apparently exerted their effects on muscularity only in the presence of Mstn^Cmpt-dl1Abc.

THE Compact "hypermuscular" mutation arose during a selection experiment in Germany (BUNGER et al. 2001 ). A "Compact" subline was formed and selected for maximum hypermuscularity. Genetic analysis of the Compact line indicated that a major gene, denoted Cmpt, and one or more modifier genes were involved in determining the degree of expression of the Compact phenotype. The putative major gene (Cmpt) was mapped to an 8.2-cM region on mouse chromosome 1 (VARGA et al. 1997 ). After the discovery of the myostatin gene (MCPHERRON et al. 1997 ), we identified the Mstn^Cmpt-dl1Abc deletion in the propeptide region of the gene and mapped the mouse myostatin gene to a point 27.7 cM from the centromere on chromosome 1 (SZABO et al. 1998 ).

Myostatin, belonging to the transforming growth differentiation factor-ß (TGF-ß) superfamily, is a negative regulator of muscle growth. Myostatin was first identified when null-mutant knockout mice exhibited a large and widespread increase in skeletal muscle mass due to an increase of muscle fiber number (hyperplasia) and thickness (hypertrophy; MCPHERRON et al. 1997 ). Subsequently, an 11-bp myostatin deletion in the Belgian Blue and a missense myostatin mutation in the Piedmontese cattle breeds were identified in the C-terminal coding region of Mstn and shown to be responsible for the double-muscled phenotype in these breeds (GROBET et al. 1997 ; KAMBADUR et al. 1997 ; MCPHERRON and LEE 1997 ). Mstn^Cmpt-dl1Abc, however, was found in the propeptide region of the myostatin precursor (SZABO et al. 1998 ). In analogy with TGF-ß1, we hypothesized that the propeptide region might play an important role in the proper folding, efficient secretion, processing, or targeting of the mature growth factor domain of myostatin. Although some of these steps may be seriously impaired by the Mstn^Cmpt-dl1Abc mutation, the activity of mature myostatin is not zero in homozygous Cmpt/Cmpt mutants. Modifier genes may therefore have a significant influence on the phenotype (SZABO et al. 1998 ).

In the current study we report the mapping of modifier loci affecting the expression of the hypermuscularity in the Compact mouse. Our basic assumption is that due to selection for hypermuscularity, the Compact line, in addition to achieving homozygosity for the Mstn^Cmpt-dl1Abc mutation, also accumulated modifier alleles that increased the expression of hypermuscularity in the Compact mouse. To map these modifier genes, a new intersubspecific cross (Cross4) was initiated, which produced a large F₂ population. A genome scan was carried out by selective DNA pooling within the F₂ progeny group homozygous for the Mstn^Cmpt-dl1Abc deletion. Markers on several chromosomes showed linkage with the putative modifiers, and the strongest association was found for markers on chromosomes 16 and X. Mapping of the modifier loci and their identification by positional cloning might provide a route to additional upstream and downstream factors involved in muscle development (VARGA et al. 1997 ).

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mice:
The hypermuscular Compact phenotype was first noted in a line of mice selected for high body weight and high protein content at the Technical University of Berlin. Mice showing the Compact phenotype were separated from the main selection line and a new COMPACT line, based on these individuals, was originated and subjected to selection for maximum expression of the Compact phenotype (F. MAJOR and W. SCHLOTE, personal communication). An inbred line, HCI, was generated within the HCR and used in our previous F₂ crosses (SZABO et al. 1998 ). HCI became extinct at generation 22, and a new inbred strain, Comp9, developed by 16 generations of full-sib mating, was used for the current intersubspecific cross, denoted Cross4. In the present study, five Comp9 males were mated with five CAST/Ei females to produce an F₁ generation. The F₁ progeny were intercrossed to produce 2373 F₂ mice.

Phenotypic analysis:
At 6 weeks of age the F₂ mice were classified for muscularity by visual inspection using a five-grade scale for phenotypes of from 1 (normal) to 5 (most muscular). The classification was always done by G. Müller. Each animal was scored three times on the same day. An animal was characterized by the average of the three independent scores. In this way, the mice distributed across 13 phenotypic categories ranging from 1.00, 1.33, 1.67, ... , to 5.00. The standard deviation among the three scores of the same mouse averaged 1.09 score units. A spread of 0, 1, 2, 3, and 4 within the three scores of the same mouse was found in 78, 9, 11, and 2% and in one individual, respectively. Thus, scoring on the whole was consistent and closely reflected the actual Compact status of the animal. Animals with a wide spread would have ended up with a moderate score and thus would not be included in either the high or the low pools. Thus, they represent lost information, but would not bias the results.

Genotypic analysis:
DNA from tail tips was prepared using standard protocols. A total of 141 microsatellites distinguishing Comp9 and CAST/Ei were typed, using primers purchased from Research Genetics (Huntsville, AL). These microsatellites covered the genome with an average spacing of 10.6 cM between markers (Fig 1). The largest intermarker distance was 17.0 cM. PCR, allele separation, and silver staining were as described (VARGA et al. 1997 ).

View larger version (65K): In this window In a new window Download PPT slide   Figure 1. Chromosomal location of the microsatellite markers used in this study. Formal microsatellite names have been abbreviated. Open squares indicate markers tested only on pools, while open circles represent markers that were individually genotyped. The centimorgan positions of microsatellites were obtained from the Mouse Genome Database, Mouse Genome Informatics Project (October 2002), The Jackson Laboratory, Bar Harbor, Maine (http://www.informatics.jax.org).

Notation:
Genotypes for the Mstn^Cmpt-dl1Abc deletion are denoted as follows: K, Mstn^Cmpt-dl1Abc/Mstn^Cmpt-dl1Abc; H, Mstn^Cmpt-dl1Abc/+; and B, +/+. Marker genotypes are denoted as follows: M_K, the marker genotype homozygous for the allele derived from the Compact line; M_B, the marker genotype homozygous for the allele derived from the CAST/Ei line; and M_H, the heterozygous marker genotype.

DNA pooling and individual genotyping:
We applied the selective DNA pooling strategy (DARVASI and SOLLER 1994 ) for the initial genome scan to map quantitative trait loci (QTL) modifying the expression of the Mstn^Cmpt-dl1Abc mutation. On the hypothesis that the expression of Mstn^Cmpt-dl1Abc is strongly modulated by modifier genes, then in the F₂ population, the allele frequencies of particular modifier genes will differ between pools of homozygous Mstn^Cmpt-dl1Abc animals with low compact score and those with high compact score. Taking into consideration the differences in phenotypic distribution of males and females among F₂ animals of K genotype, four pools were constructed on the basis of gender (male, female) and Compact score (high, low). There were sufficient male K mice of compact score 5.00, and female K mice of compact score 1.00, so that the high muscularity male pool (designated M5K) consisted of animals of score 5.00 only, and the low muscularity female pool (designated F1K) consisted of animals of score 1.00 only. There were only a few male K mice of very low compact score and a few female K mice of very high compact score. Consequently, the low muscularity male pool (designated M1K) had a final composition of 14 animals, score 1.00; 8 animals, score 1.67; 4 animals, score 2.00; and 3 animals, score 2.33. Similarly, the high muscularity female pool (designated F4K) had a final composition of 24 animals, score 4.00; 2 animals, score 4.33; and 1 animal, score 4.67. Total numbers of mice in the various pools were: M1K, 29; M5K, 40; F1K, 39; and F4K, 30. Adding less extreme animals to the M1K and F4K pools reduced the selection applied to the pools, but also reduced the effects of binomial sampling. Thus, net effect on the power of selective genotyping is negligible (LEBOWITZ et al. 1987 ).

At this stage of the cross, 138 mice were selected for the pools, comprising 5.8% of the entire F₂ population. DNA pooling analysis was performed using these mice. Later, as the cross was continued, we were able to choose an additional 51 extreme F₂ animals for the individual selective genotyping, so that a total of 189 animals, 8% of the entire F₂ population (N = 2373), was included in this study.

The pool experiments were carried out with these four pools, using a heterozygous (H) animal from the first generation of the cross as an intensity control. Intensities of the Comp9 and CAST/Ei alleles in these four pools were evaluated "by eye" and independently by two persons. Our mapping strategy was sequential: if we observed allele intensity differences between the pools, we tested an additional flanking marker on the pools. If this result supported the first observation, we individually genotyped the mice from the pools. If there was a significant frequency difference (P < 0.05) between individuals belonging to the high and low pools, subsequent markers along the chromosome were genotyped individually without pooling. This was continued until the significance of the frequency difference fell below the above-mentioned threshold.

False discovery rate and proportion of false positives:
The genome scan performed in the present study is a multiple-test procedure and hence the usual comparison-wise error rate (CWER) significance levels of P = 0.05 and P = 0.01 are not appropriate. LANDER and KRUGLYAK 1995 suggested that in this situation CWER thresholds that control GWER (genome-wise error rate) as though based on a high-density marker map should be used. However, this leads to an exceedingly stringent significance level with much loss of power. As an alternative, WELLER et al. 1998 proposed controlling the comparison-wise false discovery rate (FDR), developed by BENJAMINI and HOCHBERG 1995 in genome scan experiments. An extensive comparison of FDR and GWER methods in the context of interval mapping was carried out by LEE et al. 2002 and indeed showed that power was greater under an FDR criterion than under the LANDER and KRUGLYAK 1995 GWER criterion. The FDR as defined by BENJAMINI and HOCHBERG 1995 is somewhat inappropriate for use in a genome scan (WELLER 2000 ; ZAYKIN et al. 2000 ). A modified version, the adjusted FDR, equivalent to the proportion of false positives (PFP) criterion of SOUTHEY and FERNANDO 1998 , was utilized by MOSIG et al. 2001 .

On the basis of chi-square CWER P-values, comparison-wise FDR and PFP for a given marker, M_i, were calculated as

where P_i is CWER for the ith marker; n is the total number of markers in the scan; n_L is the number of markers in linkage to QTL (since a number of markers will be in linkage to the same QTL, it should be noted that n_L is not an estimate of the number of modifier QTL); and i is the ordinal number of the ith marker, when the markers are ranked according to ascending P-values.

Application of FDR and PFP in the present experiment required some adjustment due to the sequential nature of the scan. As noted above, only markers that showed qualitative indications of a frequency difference between the high-compact-score pools and the low-compact-score pools were individually genotyped. Thus, only 50 of 141 markers employed in the scan were individually genotyped and provided CWER P-values. The FDR and PFP analyses were then carried out on the assumption that all markers that would have given a low P-value had been identified by the pool scan. Thus, the markers were ranked on the basis of their P-value on individual genotyping, but n = 141 was taken as the total number of markers in the scan. This assumption is conservative. If any such markers were not included among the individually genotyped markers, they would (if included) have increased the rank score (i number) of the individually genotyped markers following them in the ranking and hence reduced their FDR or PFP values. Because of the way in which FDR and PFP were calculated, markers with very high P-values obtained calculated FDR and PFP values >1.0. Since this is not possible, these were adjusted to a value of 1.0 and are shown as 1.0* in Table 1.

The number of markers in linkage to QTL, n_L, was estimated as follows. On the null hypothesis, the expected number of P-values in the range 0.00–0.099 is n/10. Let n_D be the number of markers analyzed by individual genotyping and having CWER P-values in this range. On the conservative assumption that all markers having a P-value in this range would have been flagged by pool genotyping and individually genotyped, n_L will equal the excess of n_D over n/10; i.e., n_L = n_D - n/10. This, too, is a conservative estimate, since any markers that were not individually genotyped that would have been included in this class and would have increased the estimate of n_L.

Chi-square test for linkage of markers with QTL modifying the Compact phenotype in homozygous Mstn^Cmpt-dl1Abc animals:
Let NM_K(M1K) be the number of presentations of genotype M_K in the M1K phenotypic group, and let NM_K(M5K), NM_K(F1K), etc., be defined accordingly. Then, under the null hypothesis of no linkage, and assuming equal numbers of progeny in each phenotypic group,

while under linkage we have

Then, letting

and

under the null hypothesis E(U) = E(L), we can test for deviation from expectation by ² = (U - L)²/(U + L) with 1 d.f.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Phenotypic data:
A similar pattern of recessive hypermuscularity was found in both sexes (Fig 2). In spite of their general similarities, however, the two sexes showed some differences in their overall F₂ phenotypic distributions. The female distribution is skewed to the lower categories, while the male distribution is skewed to the higher categories. This tendency is in agreement with the phenotypic difference between males and females observed in the inbred Comp9 strain, where males were uniformly 5.00, while 31% of the females were 3.67 and 69% were 4.00. Similarly, in the F₂, female category F5.00 is completely missing and only a single animal was in F4.67 and two animals in the F4.33 category.

View larger version (27K): In this window In a new window Download PPT slide   Figure 2. Distribution of the three genotypes (K, MstnCmpt-dl1Abc/MstnCmpt-dl1Abc; H, MstnCmpt-dl1Abc/+; B, +/+) for F2 males and females within the phenotypic categories. The 13 phenotypic categories derived from the average of three independent visual classifications using a five-grade scale (1, normal muscularity; 5, extreme hypermuscularity).

In the F₁, only 1 female out of 59 (1.69%), and 9 males out of 67 (13.43%) presented a compact score >1.0. In the F₂, 23.11% of females and 38.6% of males presented a compact score >1.0. Thus, in both F₁ and F₂, the males show a greater proportion of higher value scores than the females. Among the F₂ males the proportion of individuals presenting a compact score >1.0 (38.63%) is significantly greater than the expected proportion of individuals with K genotype, indicating that some individuals from the other two genotypic categories (H and B) must achieve compact scores >1.00 (see below). This is supported by the presence of an appreciable number of F₁ males presenting compact scores >1.00. Among the F₂ females the proportion of animals presenting a compact score >1.0 (23.11%) is equal to the expected proportion of K individuals.

Distribution of the Mstn^Cmpt-dl1Abc genotype among the Compact phenotypic categories:
Mstn^Cmpt-dl1Abc was genotyped across the complete F₂ population of Cross4. The proportion of the three genotypic categories (B = 25.3%, H = 50.7%, and K = 23.9%) was in good agreement with the expected 1:2:1 Mendelian ratio.

Among F₂ animals showing the B genotype, only 25 of 312 males (8.01%) and 2 of 289 females (0.69%) presented phenotypic categories >1.0 (Fig 2). There was a certain oscillation in those categories where a score of 2 occurred more frequently, because this score was used much less frequently than at the classification of the previous crosses. This uneven classification distorted the phenotypic distribution to some extent, but the trend is clear. With the exception of a single B genotype male presenting a compact score of 2.33, all of these were limited to compact scores of 1.33 or 1.67. Among F₂ animals showing the H genotype, the corresponding proportions were 30.02 and 7.29%, respectively. It is noteworthy that there was a greater proportion of higher compact scores among the heterozygous (H) F₂ male and female individuals (30.02 and 7.29%, respectively) than among the F₁ male and female individuals (14.93 and 1.69%, respectively), all of which are heterozygous. The differences between phenotypic distribution of H animals in F₁ and F₂ are highly significant by chi-square contingency test for the F₂ males, but somewhat under the significance threshold for the F₂ females. The greater proportion of higher Compact categories among F₂ individuals with H genotype as compared to F₁ individuals with H genotype is consistent with the notion that recessive modifier genes that come to a homozygous state in the F₂, but not in the F₁, contribute significantly to the increased compact score. Among F₂ animals showing the K genotype, the proportions showing compact scores >1.0 were 91.91 and 78.72%, for males and females, respectively. In both males and females having K genotypes, there was a broad, virtually uniform distribution of compact scores (Fig 2). The variable expressivity of the K genotype in the F₂, not seen in the Comp9 parent line (data not shown), is also consistent with the presence of modifier genes affecting K genotype expression.

Mapping of Mstn^Cmpt-dl1Abc modifier genes:
Since only the Mstn^Cmpt-dl1Abc homozygous animals were involved in the experiment, chromosome 1 where Mstn resides was omitted from the mapping process. This chromosome will be explored by individual genotyping in a future experiment. In all, a total of 141 markers were genotyped on the four phenotypic pools (Fig 1). Thus, there were a total of 564 marker-by-phenotypic-group pool genotypings. On visual inspection of the stained gels, markers on chromosomes 2, 3, 5, 6, 7, 8, 11, 14, 15, 16, 17, and X showed apparent allele-frequency differences between K and B genotypes within one or more marker-by-group combinations. For these markers all four phenotypic groups were individually genotyped. When this was done, the tested combinations for markers on chromosomes 2, 6, 8, 14, 15, and 17 did not show appreciable frequency differences in any of the four groups. Therefore, these chromosomes were not investigated further by individual genotyping. For chromosomes 3, 5, 7, 11, 16, and X, one or more of the individually genotyped marker-by-group combinations were significant at the P < 0.05 level by chi-square (test data not shown). For these chromosomes, all four groups were then individually genotyped for the flagged marker, and additional markers were individually genotyped in both directions along the chromosome and tested for significance by chi-square, as described in MATERIALS AND METHODS, until a clear decrease in significance level was obtained.

Table 1 shows the results of individual genotyping by chromosome and marker. The frequency of M_K and M_B genotypes for the individually genotyped markers on chromosomes 3, 5, 7, 11, 16, and X were examined in the four phenotypic groups (M1K, F1K, M5K, and F4K). With two minor exceptions and two major exceptions, the frequency of M_K genotype was greater than the frequency of M_B genotype in the M5K and F4K phenotypic groups, and the frequency of the M_B genotype was greater than the frequency of the M_K genotype in the M1K and F1K phenotypic groups. For markers on the X chromosome, the frequency of the M_H genotype was greater than that of the M_B genotype in the F4K phenotypic group, while the reverse was true for the F1K phenotypic group. Thus, the modifier genes appear to have equivalent effects in males and females.

The minor exceptions noted above, which we attribute to sampling variation, were markers D5Mit229 in F4K and D7Mit30 and D7Mit299 in M1K and M5K. The major exceptions, which appear to be too large to be due to sampling and are unexplained, were a large excess of M_K genotypes for markers D5Mit229–D5Mit221 in F1K and a marked excess of M_H genotype for markers D7Mit56 and D7Mit22 in M5K. Consequently, the F1K phenotypic group was not included in linkage tests for all chromosome 5 markers, and the M5K phenotypic group was not included in linkage tests for all chromosome 7 markers. With these exceptions, the overall consistency of marker effects in all four phenotypic groups and across all markers in a chromosome argues strongly that the observed frequency differences among marker genotypes within phenotypic groups are due to a common factor exerting its effect in all groups and on all markers within the same chromosome. In all cases, the marker derived from the Compact parent was associated with a greater compact score, and the marker derived from the CAST/Ei parent was associated with a lower compact score. This is consistent with the presence of one or more modifier QTL with positive alleles derived from the Compact parent in each of the tagged chromosomes. The consistency of effect across phenotypic groups justifies the test procedure of pooling genotype data across all four phenotypic groups for the chi-square analysis. CWER P-values resulting from the chi-square analysis for the markers on the tagged chromosomes are also shown in Table 1.

Of the total of 141 markers tested by selective DNA pooling, 50 markers were further analyzed by individual genotyping. On the null hypothesis, the expected number of P-values in the range 0.00–0.099 is 14. Of the markers analyzed by individual genotyping, 39 had CWER P-values in the range 0.00–0.099. On the conservative assumption that all markers having a P-value in this range would have been flagged by pool genotyping and individually genotyped, there is an excess of 25 markers over the expected in this class. The excess is highly significant by chi-square test. This is consistent with the presence of some factor, presumably linked segregating modifier QTL, which caused a greater-than-chance deviation of observed-to-expected marker genotype classes within some of the marker x group combinations. Thus, the excess of 28 markers represents a minimum estimate of n_L, the number of markers in linkage to modifier QTL.

Calculating FDR and PFP gave a value of P = 0.0092 as cut-off CWER significance levels for both FDR and PFP = 0.05. On this basis, significant markers were found on chromosomes 3, 5, 7, 11, 16, and X. Chromosomes 3, 5, 11, 16, and X had 1, 1, 2, 6, and 8 markers, respectively, significant at P < 0.001; chromosomes 3, 5, and 7 each had three markers, significant at P 0.0092. Thus, the results of this analysis support the presence of modifier genes on these six chromosomes (Fig 3).

View larger version (108K): In this window In a new window Download PPT slide   Figure 3. Chi-square test for mapping modifier loci affecting hypermuscularity in F2 animals homozygous for the MstnCmpt-dl1Abc deletion. Significant association at an FDR of 0.05 was found for markers on chromosomes 3, 5, 7, 11, 16, and X. Similar results were obtained using the Map Manager QTX program (http://mapmgr.roswellpark.org/mmQTX.html; our unpublished results). The results of the individual genotypings were evaluated separately for males (), for females (), and also for the gender-merged data set (). P-values are plotted on the y-axis; centimorgan distances from the centromere are plotted on the x-axis, together with the position of the microsatellite markers used for individual genotyping. Note that peak values represent the marker with lowest CWER P-value and are not point estimates of QTL location, as given by interval mapping programs.

Fig 3 shows plots of the chi-square values of the individual markers along the chromosomes. In these plots the peak is at the marker with highest chi-square value. This is not the best estimate of the QTL location, which would ordinarily be located, by interval mapping programs, between the two highest peaks. However, the plots do show the general regions within which the QTL is to be found. The point estimates of QTL location given by interval mapping come with wide confidence intervals, so the actual loss of information is negligible. Considering the chromosomes individually, narrow regions of strong effect were found on chromosome 3 (markers D3Mit206 and D3Mit224 from centimorgan 17 to 25); on chromosome 5 (markers D5Mit239 and D5Mit26 from centimorgan 56 to 64); and on chromosome 7 (at marker D7Mit83 from centimorgan 23 to 26). Male and female peaks on chromosome 3 are separated by only 6 cM and hence almost certainly represent a single modifier gene. Broad regions of effect were found on chromosome 11 (markers D11Mit24–D11Mit199 from centimorgan 26 to 61) with peaks at D11Mit24 (from centimorgan 26 to 32) and D11Mit41 (from centimorgan 46 to 51) and on chromosome 16 (all markers from centimorgan 11 to 58) with peaks at D16Mit143 (centimorgan 10.8) and D16Mit136 (centimorgan 28.2). Chromosome X showed a broad and exceptionally strong region of effect extending from DXMit126 to DXMit99 (from centimorgan 19 to 66), with a single peak at DXMit128 (centimorgan 34.7). In chromosomes 3 and 7, female effects were more powerful than male effects. In the remaining chromosomes, male effects were more powerful than female effects. Significant effects for both males and females were found in chromosomes 3, 11, 16, and X. Effects in males only were found on chromosome 5; effects in females only were found on chromosome 7.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Previous genetic analysis of the Compact trait through crosses of Compact lines to normal phenotype BALB/c mice indicated that a major gene and one or more modifier genes were involved in determining the degree of expression of the Compact phenotype. On this basis we speculated that the Compact phenotype derives primarily from a major mutation, which arose de novo or was already present in the selection line, and that the selection line was also segregating at one or more modifier loci, which came to phenotypic expression only in conjunction with the major Compact mutation (VARGA et al. 1997 ; SZABO et al. 1998 ). In subsequent work, the putative major gene (Cmpt) was mapped to a small region on chromosome 1 very close to the D1Mit237 marker and syntenic to the segment of bovine chromosome 2 containing the double-muscle mh gene.

After the discovery of the myostatin gene (MCPHERRON et al. 1997 ), we identified the Mstn^Cmpt-dl1Abc deletion in the propeptide region of the mouse myostatin (Mstn) gene in homozygous Cmpt animals and mapped the mouse Mstn gene to the same small region on chromosome 1 containing the putative Cmpt gene. On this basis, we identified the Mstn^Cmpt-dl1Abc deletion as the causative mutation responsible for the hypermuscular Compact phenotype (SZABO et al. 1998 ).

To map the putative modifiers, we produced a large F₂ population from a Comp9 x CAST/Ei intersubspecific cross, Cross4. Within the F₂ progeny, mice homozygous for the normal myostatin allele were almost exclusively of normal phenotype. Although in one of the three scorings of individual mice a few male mice and even fewer female mice of this genotype were occasionally given compact scores or 2 or even 3, these were almost always accompanied by two scores of 1.0 and hence can be attributed to phenotyping error. Thus, there is no indication in these data that the Cmpt mutation was acting as a "sensitizing mutation" (MATIN and NADEAU 2001 ) in which a major gene (Mstn^Cmpt-dl1Abc) increases the expression and penetrance of loci affecting the trait, but which normally has only small effects that are difficult to detect. Rather, the modifier loci appear to act only in the presence of Mstn^Cmpt-dl1Abc.

In the Compact crosses reported previously, inheritance of the Compact phenotype showed intermediate dominance in males and was recessive in females (VARGA et al. 1997 ). In the present cross (Cross4) where the new Comp9 strain was employed, a similar pattern of recessive hypermuscularity was found in both sexes. A comparable pattern of similar F₁ and F₂ male and female phenotypic distributions was observed in a previous Comp9 x BALB/c intercross (Cross3; data not shown). This indicates that differences in the pattern of inheritance of the Compact phenotype in Cross4, as compared to the previous crosses, can be attributed to the different genetic character of the new Compact strain, Comp9, rather than to the genetic character of the CAST/Ei parent line in Cross4.

F₂ mice homozygous for the Mstn^Cmpt-dl1Abc deletion distributed over all the phenotypic classes, including normal muscularity. Thus, using the terminology of GREENBERG 1993 , the Mstn^Cmpt-dl1Abc deletion in the myostatin gene is necessary, but not sufficient, for the expression of extreme hypermuscularity. It is plausible that the observed wide variation in mutant expression was due to the influence of the putative modifier genes that are also segregating in this F₂ population. To map these modifier QTL, we carried out a whole genome scan, using only F₂ mice that were homozygous for the Mstn^Cmpt-dl1Abc deletion. Significant association at an FDR of 0.05 was found for markers on chromosomes 3, 5, 7, 11, 16, and X. With a few exceptions, markers examined in these regions showed a higher frequency of the Comp9-derived allele in the high-compact-score group and a higher frequency of the CAST/Ei-derived allele in the low-compact-score group. Thus, our mapping results support the hypothesis that modifier alleles with positive effect on hypermuscularity were present in the Compact lines and were exposed for selection by the presence of the Mstn^Cmpt-dl1Abc deletion, and in this way increased in frequency by the continued selection for hypermuscularity in this line. In the present study, we utilized only part of the potential mapping power of our large F₂ population. By identifying additional recombinant K individuals in the regions containing putative QTL and by genotyping all recombinants for multiple markers across the putative QTL-containing regions, it should be possible to increase map resolution up to fourfold (RONIN et al. 2003 ) to define more precise locations for the modifier loci.

Although it may be premature to speculate on the identity of the modifier genes located on these chromosomes, we note with interest that a number of attractive candidate genes are located near some of the sharper peaks uncovered in the present mapping effort. Recent articles have shown that myostatin is a downstream target of MyoD1 (SPILLER et al. 2002 ), which is a key regulator of skeletal myogenesis on chromosome 7 (23.5 cM), while myostatin can downregulate the activity of MyoD1 through Smad3, thus inhibiting myoblast differentiation, which results in hyperplasia and hypertrophy (LANGLEY et al. 2002 ). Another candidate gene in the same region of chromosome 7 (28.5 cM) is the gene Pcsk6 (for the pro-protein convertase subtilisin/kexin type 6), an enzyme known to be involved in the proteolytic processing of precursors of TGF-ß family members (AKAMATSU et al. 1999 , AKAMATSU et al. 2000 ; CONSTAM and ROBERTSON 1999 , CONSTAM and ROBERTSON 2000 ). Chordin (chromosome 16 at 14 cM) is known to bind the TGF-ß-like bone morphogenetic proteins and sequester them in latent complexes (ABREU et al. 2002 ), raising the possibility that it may also have a role in myostatin regulation. Androgen receptor (Ar) is found 1.3 cM distal to the strongest peak detected on chromosome X at 35.7 cM. The direct androgen regulation mediated by Ar might also influence the expression of myostatin (MA et al. 2001 ), as in the case of TGF-ß (CHIPUK et al. 2002 ).

Innumerable experiments document the importance of modifier genes among the causes of variable expression and penetrance of spontaneous, engineered, and induced mutants in mice. A number of studies have gone on to map the modifier loci. Most of these involved modifiers affecting penetrance of genetically determined abnormal phenotypes (NEUMANN et al. 1994 ; LETTS et al. 1995 ; CRYNS et al. 2002 ). Only two, as in the present study, considered effects of modifiers on expressivity (ALFRED et al. 1997 ; WOO et al. 1997 ). Both were successful in mapping modifiers of this nature. Determining and classifying the effects of modifier genes is of great importance in understanding the range of phenotypes shown by Mendelian disease traits in humans (NADEAU 2001 ) and the effects of engineered or spontaneous mutants in agricultural species. Positional cloning of modifiers can also be expected to expose new elements involved in the regulation of major or engineered genes that may be of physiological significance in their own right.

	ACKNOWLEDGMENTS

We thank Sóvári Krisztina, Katonáné Lénárd Magdolna, and Galli Györgyné for their excellent technical assistance. This research was supported by grant no. T26024 from Hungarian Scientific Research Fund (OTKA).

Manuscript received January 11, 2003; Accepted for publication May 2, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ABREU, G. J., C. COFFINIER, J. LARRAIN, M. OELGESCHLAGER, and E. M. DE ROBERTIS, 2002  Chordin-like CR domains and the regulation of evolutionarily conserved extracellular signaling systems. Gene 287:39-47.[Medline]

AKAMATSU, T., Y. MATSUDA, K. TSUMURA, J. TADA, and M. N. PARVIN et al., 1999  Subtilisin-like proprotein convertase PACE4 (SPC4) is a candidate processing enzyme of bone morphogenetic proteins during tooth formation. Dev. Dyn. 216:481-488.[Medline]

AKAMATSU, T., Y. MATSUDA, K. TSUMURA, J. TADA, and M. N. PARVIN et al., 2000  Highly regulated expression of subtilisin-like proprotein convertase PACE4 (SPC4) during dentinogenesis. Biochem. Biophys. Res. Commun. 272:410-415.[Medline]

ALFRED, J. B., K. RANCE, B. A. TAYLOR, S. J. PHILLIPS, and C. M. ABBOTT et al., 1997  Mapping in the region of Danforth's short tail and the localization of tail length modifiers. Genome Res. 7:108-117.[Abstract/Free Full Text]

BENJAMINI, Y. and Y. HOCHBERG, 1995  Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. B 57:289-300.

BUNGER, L., A. LAIDLAW, G. BULFIELD, E. J. EISEN, and J. F. MEDRANO et al., 2001  Inbred lines of mice derived from long-term growth selected lines: unique resources for mapping growth genes. Mamm. Genome 12:678-686.[Medline]

CHIPUK, J. E., S. C. CORNELIUS, N. J. PULTZ, J. S. JORGENSEN, and M. J. BONHAM et al., 2002  The androgen receptor represses transforming growth factor-ß signaling through interaction with smad3. J. Biol. Chem. 277:1240-1248.[Abstract/Free Full Text]

CONSTAM, D. B. and E. J. ROBERTSON, 1999  Regulation of bone morphogenetic protein activity by pro domains and proprotein convertases. J. Cell Biol. 144:139-149.[Abstract/Free Full Text]

CONSTAM, D. B. and E. J. ROBERTSON, 2000  SPC4/PACE4 regulates a TGFbeta signaling network during axis formation. Genes Dev. 14:1146-1155.[Abstract/Free Full Text]

CRYNS, K., M. P. VAN SPAENDONCK, K. FLOTHMANN, A. M. VAN ALPHEN, and P. H. VAN DE HEYNING et al., 2002  Vestibular dysfunction in the epistatic circler mouse is caused by phenotypic interaction of one recessive gene and three modifier genes. Genome Res. 12:613-617.[Abstract/Free Full Text]

DARVASI, A. and M. SOLLER, 1994  Selective DNA pooling for determination of linkage between a molecular marker and a quantitative trait locus. Genetics 138:1365-1373.[Abstract]

GREENBERG, D. A., 1993  Linkage analysis of "necessary" disease loci versus "susceptibility" loci. Am. J. Hum. Genet. 52:135-143.[Medline]

GROBET, L., L. J. MARTIN, D. PONCELET, D. PIROTTIN, and B. BROUWERS et al., 1997  A deletion in the bovine myostatin gene causes the double-muscled phenotype in cattle. Nat. Genet. 17:71-74.[Medline]

KAMBADUR, R., M. SHARMA, T. P. SMITH, and J. J. BASS, 1997  Mutations in myostatin (GDF8) in double-muscled Belgian Blue and Piedmontese cattle. Genome Res. 7:910-916.[Abstract/Free Full Text]

LANDER, E. and L. KRUGLYAK, 1995  Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat. Genet. 11:241-247.[Medline]

LANGLEY, B., M. THOMAS, A. BISHOP, M. SHARMA, and S. GILMOUR et al., 2002  Myostatin inhibits myoblast differentiation by down-regulating MyoD expression. J. Biol. Chem. 277:49831-49840.[Abstract/Free Full Text]

LEBOWITZ, R. J., M. SOLLER, and J. S. BECKMANN, 1987  Trait based designs for determination of linkage between marker loci and quantitative trait loci. Theor. Appl. Genet. 73:556-562.

LEE, H., J. C. M. DEKKERS, M. SOLLER, M. MALEK, and R. L. FERNANDO et al., 2002  Application of false discovery rate to quantitative trait loci interval mapping with multiple traits. Genetics 161:905-914.[Abstract/Free Full Text]

LETTS, V. A., N. J. SCHORK, A. J. COPP, M. BERNFIELD, and W. N. FRANKEL, 1995  A curly-tail modifier locus, mct1, on mouse chromosome 17. Genomics 29:719-724.[Medline]

MA, K., C. MALLIDIS, J. ARTAZA, W. TAYLOR, and N. GONZALEZ-CADAVID et al., 2001  Characterization of 5'-regulatory region of human myostatin gene: regulation by dexamethasone in vitro. Am. J. Physiol. Endocrinol. Metab. 281:1128-1136.

MATIN, A. and J. H. NADEAU, 2001  Sensitized polygenic trait analysis. Trends Genet. 17:727-731.[Medline]

MCPHERRON, A. C. and S.-J. LEE, 1997  Double muscling in cattle due to mutations in the myostatin gene. Proc. Natl. Acad. Sci. USA 94:12457-12461.[Abstract/Free Full Text]

MCPHERRON, A. C., A. M. LAWLER, and S.-J. LEE, 1997  Regulation of skeletal muscle mass in mice by a new TGF-ß superfamily member. Nature 387:83-90.[Medline]

MOSIG, M. O., E. LIPKIN, G. KHUTORESKAYA, E. TCHOURZYNA, and M. SOLLER et al., 2001  A whole genome scan for quantitative trait loci affecting milk protein percentage in Israeli-Holstein cattle, by means of selective milk DNA pooling in a daughter design, using an adjusted false discovery rate criterion. Genetics 157:1683-1698.[Abstract/Free Full Text]

NADEAU, J. H., 2001  Modifier genes in mice and humans. Nat. Rev. Genet. 2:165-174.[Medline]

NEUMANN, P. E., W. N. FRANKEL, V. A. LETTS, J. M. COFFIN, and A. J. COPP et al., 1994  Multifactorial inheritance of neural tube defects: localization of the major gene and recognition of modifiers in ct mutant mice. Nat. Genet. 6:357-362.[Medline]

RONIN, Y., A. KOROL, M. SHTEMBERG, E. NEVO, and M. SOLLER, 2003  High-resolution mapping of quantitative trait loci by selective recombinant genotyping. Genetics 164:1657-1666.[Abstract/Free Full Text]

SOUTHEY, B., and R. L. FERNANDO, 1998 Controlling the proportion of false positives among significant results in QTL detection. Proceedings of the 6th World Congress on Genetics Applied to Livestock Production, Armidale, Australia, Vol. 26, pp. 221–224.

SPILLER, M. P., R. KAMBADUR, F. JEANPLONG, M. THOMAS, and J. K. MARTYN et al., 2002  The myostatin gene is a downstream target gene of basic helix-loop-helix transcription factor MyoD. Mol. Cell. Biol. 22:7066-7082.[Abstract/Free Full Text]

SZABÓ, G., G. DALLMANN, G. MÜLLER, L. PATTHY, and M. SOLLER et al., 1998  A deletion in the myostatin gene causes the compact (Cmpt) hypermuscular mutation in mice. Mamm. Genome 9:671-672.[Medline]

VARGA, L., G. SZABÓ, A. DARVASI, G. MÜLLER, and M. SASS et al., 1997  Inheritance and mapping of compact (Cmpt), a new mutation causing hypermuscularity in mice. Genetics 147:755-764.[Abstract]

WELLER, J. I., 2000  Using the false discovery rate approach in the genetic dissection of complex traits: a response to Zaykin et al. Genetics 154:1919.[Free Full Text]

WELLER, J. I., J. Z. SONG, D. W. HEYEN, H. A. LEWIN, and M. RON, 1998  A new approach to the problem of multiple comparisons in the genetic dissection of complex traits. Genetics 150:1699-1706.[Abstract/Free Full Text]

WOO, D. D., D. K. NGUYEN, N. KHATIBI, and P. OLSEN, 1997  Genetic identification of two major modifier loci of polycystic kidney disease progression in pcy mice. J. Clin. Invest. 100:1934-1940.[Medline]

ZAYKIN, D. V., S. S. YOUNG, and P. H. WESTFALL, 2000  Using the false discovery rate approach in the genetic dissection of complex traits: a response to Weller et al.. Genetics 154:1917-1918.[Free Full Text]

This article has been cited by other articles:

				B. D. Rodgers and D. K. Garikipati Clinical, Agricultural, and Evolutionary Biology of Myostatin: A Comparative Review Endocr. Rev., August 1, 2008; 29(5): 513 - 534. [Abstract] [Full Text] [PDF]

				P. L. Johnson, J. C. McEwan, K. G. Dodds, R. W. Purchas, and H. T. Blair Meat quality traits were unaffected by a quantitative trait locus affecting leg composition traits in Texel sheep J Anim Sci, December 1, 2005; 83(12): 2729 - 2735. [Abstract] [Full Text] [PDF]

				A. Lionikas, D. A. Blizard, G. S. Gerhard, D. J. Vandenbergh, J. T. Stout, G. P. Vogler, G. E. McClearn, and L. Larsson Genetic determinants of weight of fast- and slow-twitch skeletal muscle in 500-day-old mice of the C57BL/6J and DBA/2J lineage Physiol Genomics, April 14, 2005; 21(2): 184 - 192. [Abstract] [Full Text] [PDF]

				L. Varga, O. Pinke, G. Muller, B. Kovacs, E. Korom, G. Szabo, and M. Soller Mapping a Syntenic Modifier on Mouse Chromosome 1 Influencing the Expressivity of the Compact Phenotype in the Myostatin Mutant (MstnCmpt-dl1Abc) Compact Mouse Genetics, January 1, 2005; 169(1): 489 - 493. [Abstract] [Full Text] [PDF]