Human adaptive evolution at Myostatin (GDF8), a regulator of muscle growth

Abstract

Myostatin (GDF8) is a negative regulator of muscle growth in mammals, and loss-of-function mutations are associated with increased skeletal-muscle mass in mice, cattle, and humans. Here, we show that positive natural selection has acted on human nucleotide variation at GDF8, since the observed ratio of nonsynonymous:synonymous changes among humans is significantly greater than expected under the neutral model and is strikingly different from patterns observed across mammalian orders. Furthermore, extended haplotypes around GDF8 suggest that two amino acid variants have been subject to recent positive selection. Both mutations are rare among non-Africans yet are at frequencies of up to 31% in sub-Saharan Africans. These signatures of selection at the molecular level suggest that human variation at GDF8 is associated with functional differences.

Figure A1.

Table of polymorphism for diploid data for GDF8. Data are shown for an African American (AFRAM) panel, a European (EURO) panel, and a single chimpanzee (only sites polymorphic in humans are shown for the chimpanzee sequence). Surveyed regions are P, E1, E2, and E3 (see fig. 1A). Alignment positions are indicated at the top, relative to downstream or upstream distance (in bp) from the ATG-coding start position. Polymorphisms in coding regions are shaded. Consensus sequence is denoted along the top, with sites that cause replacement polymorphisms shown in bold and silent polymorphisms shown in italics. For all samples, the demarcations are as follows: identity to consensus sequence is indicated by a dot (·), deletion is indicated by a dash (—), and heterozygote A/- is indicated by an X. All other notations are standard nucleotide International Union of Pure and Applied Chemistry (IUPAC) codes. Inferred phased haplotypes are listed on the right, with haplotype labels (A–K). Sequences of the inferred phased haplotypes are shown at the bottom. Haplotypes bearing the intermediate frequency replacement changes at amino acid sites 55 and 153 (alignment positions 163 and 2246, respectively) are marked in bold.

Figure 1.

A, Sequences and number of occurrences of the inferred haplotypes (A–K) from GDF8 in an African American panel (AFRAM) and a European panel (EURO). Surveyed regions are P (encompassing the putative cis-promoter region), E1, E2, and E3 (spanning exons 1, 2, and 3, respectively). Alignment positions are indicated at the top, relative (in base pairs) to the ATG coding start position. Polymorphisms in coding regions are shaded. The consensus sequence is denoted at the top, with sites that cause replacement polymorphisms shown in bold and silent polymorphisms shown in italics. A dot (·) indicates identity to consensus sequence, and a dash (—) indicates a deletion. Haplotypes bearing the intermediate frequency replacement changes Ala55Thr and Lys153Arg (nucleotide alignment positions 163 and 2246, respectively) are shown in bold. Haplotypes of haplogroup 55 and haplogroup 153 are indicated by black and gray ovals, respectively. B, Haplotype tree for GDF8. Each circle represents an inferred haplotype (A–K) in a size proportional to the haplotype frequency in the combined sample (AFRAM and EURO). Hatch marks represent single mutations labeled according to the alignment positions in panel A. Positions of replacement changes are indicated in boldface numbers. Replacement changes at amino acid sites 55 and 153 (alignment positions 163 and 2246, respectively) are boxed. Positions of silent changes are italicized. Haplotypes of haplogroup 55 and haplogroup 153 are indicated by black and gray circles, respectively. The haplotype inference method is known to be inaccurate for rare (q<0.01) frequency polymorphisms. Mutations 307 and 5273 are both at low frequency in the sample (q=0.0068) and therefore may have been incorrectly assigned to a haplotype with mutation 2246. Alternative haplotype inferences for individual AFRAM_B11 would be (C,Y) (HapY: AG|GCGCA|—|TA|T) and, for individual EURO_B08, would be (C,Z) (HapZ: AG|GCGCG|A|TA|C) (see panel A and fig. A1).

Figure 2.

A, EHH at varying genetic distances from the core regions that include A2246G and G163A at GDF8. The EHH for each haplotype at a frequency >7% is displayed as follows. The core haplotypes of haplogroups 153 and 55 are indicated by a solid gray line and a solid black line, respectively. These extended haplotypes exhibit high levels of EHH at a distance of up to 0.2 cM from the core. All other core haplotypes are indicated by dashed lines. B, EHH relative to population frequency. EHH values were calculated for each core haplotype at a genetic distance of 0.2 cM. An empirical distribution of EHH values was produced for core haplotypes from 10 anonymous genomic regions across chromosome 2 in the HapMap YRI panel. These data were compared with core haplotypes from the genomic region spanning ∼300 kb centered on GDF8. Values of EHH for the GDF8 core haplotypes representing haplogroups 153 (gray arrow) and 55 (black arrow) are significantly higher than other core haplotypes within the same population frequency bin (P=.01 and P=.03, respectively).

Figure 3.

Population frequencies for GDF8 replacement nucleotide polymorphisms G163A and A2246G (corresponding to haplogroups 55 and 153, respectively) in a worldwide panel (HGDP-CEPH Diversity Panel). For each population, the respective pie chart denotes the frequencies of haplogroups 55 (dark shading) and 153 (light shading). Population and haplogroup details are shown in table 3.