A premature stop codon in ACTN3 resulting in α-actinin-3 deficiency (the ACTN3 577XX genotype) is common in humans and reduces strength, muscle mass, and fast-twitch fiber diameter, but increases the metabolic efficiency of skeletal muscle. Linkage disequilibrium data suggest that the ACTN3 R577X allele has undergone positive selection during human evolution. The allele has been hypothesized to be adaptive in environments with scarce resources where efficient muscle metabolism would be selected. Here we test this hypothesis by using recently developed comparative methods that account for evolutionary relatedness and gene flow among populations. We find evidence that the ACTN3 577XX genotype evolved in association with the global latitudinal gradient. Our results suggest that environmental variables related to latitudinal variation, such as species richness and mean annual temperature, may have influenced the adaptive evolution of ACTN3 577XX during recent human history.
Citation: Friedlander SM, Herrmann AL, Lowry DP, Mepham ER, Lek M, North KN, et al. (2013) ACTN3 Allele Frequency in Humans Covaries with Global Latitudinal Gradient. PLoS ONE 8(1): e52282. doi:10.1371/journal.pone.0052282
Editor: Toomas Kivisild, University of Cambridge, United Kingdom
Received: July 25, 2012; Accepted: November 16, 2012; Published: January 24, 2013
Copyright: © 2013 Friedlander et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing interests: The authors have declared that no competing interests exist.
Genes important for metabolism include the α-actinin family that code for contractile proteins in muscle . In humans, two different genes code for skeletal muscle α-actinins: ACTN2 is expressed in all skeletal muscle fibers, while ACTN3 is expressed in fast-twitch muscle fibers. Mammalian ACTN2 and ACTN3 genes appear to have arisen from a gene duplication that occurred over 310 million years ago, before the divergence of birds and mammals . Following the duplication of the ancestral ACTN 2/3 gene, ACTN3 evolved specific expression patterns in fast twitch muscle fibers, which are responsible for generating force at high velocity. Yet, approximately 18% of the human population is homozygous for an allele (R577X) that contains a premature stop codon in ACTN3 , . In the homozygous condition (577XX genotype) this leads to a complete deficiency of the α-actinin-3 protein, although α-actinin-2 partially compensates for the deficiency . The absence of α-actinin-3 in humans has been shown to reduce strength, muscle mass, and fast-twitch fiber diameter, but increase skeletal muscle metabolic efficiency and resistance to fatigue . Consequently, the ACTN3 577XX genotype has been hypothesized to be adaptive where efficient muscle metabolism is concerned, such as in environments with scarce resources or where endurance running has an impact on survival . Together, these Darwinian hypotheses predict a significant evolutionary relationship of ACTN3 577XX genotype frequency and global latitudinal patterns of biodiversity. This relationship is predicted because factors related to food acquisition and hunting, such as species richness or mean annual temperature, show well-established global latitudinal patterns .
The ACTN3 577X substitution likely preceded the arrival of anatomically modern humans in Europe and Asia 40,000–60,000 years ago . Since the evolution of modern humans, approximately 100,000–200,000 years ago, populations have occupied a wide range of habitats and adapted to use a variety of resources. Differences in environmental resources can drive even small genetic changes, such as single nucleotide polymorphisms, to high frequency in human populations. For example, using demographic HapMap data, allele frequencies related to subsistence have been found to change across different ecoregions . It is therefore plausible that differences in the frequency of ACTN3 577XX may have arisen during the recent past, specifically in relation to the global latitudinal gradient. Here we test the ACTN3 577XX adaptation hypothesis using global genotype and biodiversity data, with Bayesian comparative methods that account for phylogeny and migration.
ACTN3 XX frequency data were obtained from the literature (Table 1). These data were derived from studies that compared the frequencies of the ACTN3 polymorphism in elite athletes to the frequencies in control groups of the general population. We only used frequency data from the control groups (n = 3,351). ACTN3 XX frequency data were Arcsin transformed before statistical analysis, a transform suitable for nominal and percentage data .
Using the World Wildlife Fund Ecoregion database , we determined the average latitude, mean annual temperature (Co), net primary productivity (grams of carbon), and total species richness (tetrapods) data for each country/region. Although some of these metrics have changed over the time period in which ACTN3 polymorphisms arose in human populations, they nonetheless show a consistent relationship with the global latitudinal gradient . For example, the latitudinal biodiversity gradient (increasing numbers of species and higher taxa from the poles to the tropics) has been a striking large-scale pattern that extends back through the Mesozoic into the Paleozoic . However, it should be noted that large climatic changes have occurred (e.g. glaciation and aridification of North Africa) over the time in which differences in ACTN3 allele frequencies arose among human populations. Even so, we expect differences in our gene frequency data to have arisen relatively recently in line with modern climatic variation (as opposed to a novel mutation that may have had specific adaptive value in a past environment).
We curated our data by compiling a list of every ecoregion within each country. The list was constructed first using maps to form overestimated rectangles of minimum and maximum latitude and longitude around each country to account for irregular borders. Every ecoregion within these rectangles was gathered and narrowed down further using Google Earth (we limited Russia to west of Mongolia). By using the average latitudinal and longitudinal coordinates of each ecoregion, we were able to see if the ecoregion fell within the borders of the country in question. Ecoregions within the country were retained, while ecoregions outside of the country were removed from the list. One exception was Israel, which had no ecoregions within its borders, so the two closest ecoregions were used.
We produced a Bayesian posterior distribution of phylogenetic trees from whole mitochondrial sequences for the following human lineages: Aboriginal Australian (NCBI 134303155), Bantu (NCBI 160426755), Israeli (NCBI 145968101), Japanese (NCBI 61287226), Chinese (NCBI 292597164), Kenyan (NCBI 13272962), Papua New Guinean (NCBI 256946671), Lithuanian (NCBI 301017707) Russian (NCBI 290555741), Swedish (NCBI 215789486), Nigerian (EMBL AF346985), Italian (EMBL AF346988), Greek (EMBL GQ129165), Spanish (NCBI 74475826), and Ethiopian (NCBI 156459510). The whole mitochondrial genome of Homo Neanderthalensis (NCBI 196123578) was included as an outgroup.
The program MUSCLE  was used to align the genome sequences and a posterior distribution of trees was inferred with the program BayesPhylogenies . BayesPhylogenies implements a reversible jump mixture model that fits more than one model of sequence evolution to the data without partitioning. We used a GTR model with gamma-distributed rate variation across sites (4 rates) on one chain for 20,000,000 iterations, sampling every 2,000 iterations. We also implemented the reversible-jump mixture model for pattern heterogeneity. We checked for convergence by examining a time-series plot for the log-likelihoods in Tracer .
Statistical tests that accounted for phylogeny (regression and ancestral character state reconstruction) were performed in the program BayesTraits  (http://www.evolution.rdg.ac.uk). BayesTraits accounts for confounding phylogenetic relatedness in continuous data using a generalized least square (PGLS) approach. It can also account for phylogenetic uncertainty by integrating analyses over more than one tree (or a distribution of trees). Before we ran the regression analysis, we tested whether a random walk model of character evolution or a directional model of evolution best fit our data. We found no evidence for directional evolution (highest Bayes factor = 1.8, see below). Our analysis also estimated phylogenetic signal (λ), a parameter that assesses the degree to which covariation among trait values follows the phylogeny during the random walk and regression analysis –. The Bayesian Markov Chain Monte Carlo (MCMC) settings in BayesTraits were as follows: 2,050,000 iterations with a sampling frequency of 100. The rate deviation setting (which determines acceptance of new proposals during the MCMC process) was set at 0.5 for the random walk test and 0.05 for the regression.
Significance was determined for the random walk test using Bayes Factors, calculated as two times the difference between the harmonic means of log-likelihoods for the two models. Bayes factors of three or more indicate positive evidence, five of more strong evidence, and 10 or more very strong evidence , . Significance testing for the regression was determined by calculating the percent of the posterior distribution for β (the slope parameter) that fell outside of the null expectation (a slope equal to zero).
Statistical tests that account for both phylogeny and migration were performed in the statistical programming language R  using the package MCMCglmm , . MCMCglmm can be used to fit generalized linear mixed models using Markov chain Monte Carlo techniques. Such models can account for non-independence of the data from population level dynamics (migration) and phylogenetic relatedness by including their matrices as random affect components. We followed the general procedure in Stone et al.  to construct the mixed model, using the program Migrate-n v3.2.16  to generate a migration matrix. The migration matrix was generated using default settings with 150 full mitochondrial genome sequences (see Supplemental Information). For the migration matrix, the nearest positive definite matrix was determined with the R function nearPD (the script is available in the Supplemental Information).
Results and Discussion
Recent work  has shown that genome-wide SNP data can be used to reliably reconstruct human phylogenetic/demographic history. Comparative statistical methods can now be used to analyze trait data within a species by accounting for both phylogenetic relatedness and for gene flow , . In this manuscript, we build regression models that account for phylogenetic relatedness and gene flow to show that the ACTN3 577XX genotype may have evolved in association with global latitudinal patterns of biodiversity.
We began by inferring a Bayesian posterior distribution of phylogenetic trees (Figure 1A), which matches the generally accepted demographic relationships among world-wide human populations . Next we regressed ACTN3 577XX population frequency onto latitude, while estimating phylogenetic signal over the distribution of trees (and thereby accounting for phylogenetic uncertainty). We found strong phylogenetic signal in the regression model relating ACTN3 577XX frequency and latitude (maximum likelihood estimate of λ = 1). The model relating ACTN3 577XX frequency and latitude (absolute value) was supported, with a slope parameter for the generalized linear mixed model of 0.003 (lower 95% CI = 0.000, upper 95% CI = 0.005, p-value = 0.06). It should be noted here the importance of properly controlling for relatedness among the data – a standard linear regression model returns the same slope, but with an associated p-value of 0.006 and an R2 of 0.45. Note that goodness of fit is not reported in MCMCglmm, but an R2 of 0.32 is reported in BayesTraits, which controls for phylogeny, but not migration.
A, The Bayesian posterior consensus tree for the human populations used in this study (inferred from whole mitochondrial sequences). The numbers on the branches indicate the proportion of trees in the posterior distribution that contain a given clade. Branch lengths are in the number of changes per site. B, The relationship between ACTN3 577XX genotype frequency and latitude. The generalized linear mixed model that accounts for both phylogenetic relatedness and migration is arcsin(ACTN3XX) = 0.035–0.003 *Latitude (p-value for slope = 0.06).
As discussed earlier, many factors covary with latitude that could explain ACTN3 577XX frequency variation. We tested only the most plausible factors to avoid overt testing of multiple hypotheses: species richness, mean annual temperature, and net primary productivity (natural log). We found no evidence for phylogenetic signal in the regression model relating ACTN3 577XX frequency and total vertebrate species richness (maximum likelihood estimate of λ = 0), but high levels of phylogenetic signal (maximum likelihood estimate of λ = 1) for both mean annual temperature and primary productivity. Accounting for phylogeny and migration, the model relating ACTN3 577XX frequency and total vertebrate species richness was supported, with a slope parameter of −0.23 (lower 95% CI = −0.36, upper 95% CI = −0.08, p-value = 0.016), as was the relationship between ACTN3 577XX frequency and mean annual temperature (slope of −0.006, lower 95% CI = −0.0126, upper 95% CI = 0.001, p-value = 0.09). However, we found no support for a model relating ACTN3 577XX frequency and net primary productivity (slope of −0.067, lower 95% CI = −0.31, upper 95% CI = 0.161, p-value = 0.56).
For present day latitude and associated data (such as species richness) to be meaningful in an evolutionary analysis, it must fairly represent latitudinal data of the past where genetic associations may have evolved. The relationship between latitudinal gradients and species richness, for example, is consistent through the recent past, with taxa preferentially originating in the tropics and spreading toward the poles , . Moreover, the changes we analyze (differences in allele frequency) could have shifted relatively recently with respect to environmental change. However, it should be noted that we did not include marine species richness data in our analysis for coastal areas – and this could affect our species richness model. Moreover, although this study brings together disparate forms of data (which we view as a strength), the sample size for human populations was relatively small. Another limitation is the use of countries as demographic units, since countries vary in size, ecological diversity, as well as population size and heterogeneity (though our migration matrix should help account for this). For example, the 577X substitution is thought to have arisen in populations migrating out of Africa (Ethiopia) roughly 40,000–60,000 years ago . Our analysis could, therefore, be confounded by including African population data, but 40,000 years is potentially enough time to allow for gene flow to increase the 577X allele frequency within African populations. Given the mechanical and physiological roles of ACTN3 variation in the 577XX genotype, long-range linkage disequilibrium associated with ACTN3 577X (signature of positive selection) , and our results presented here, a neutral (demographically-driven) model of evolution is not supported. Available evidence instead suggests that changes in ACTN3 577XX frequency were driven by changing environmental conditions and resources associated with latitudinal migration.
We thank S.V. Edwards, C. Balakrishnan, D. Janes, N. Hobbs, F. Hobbs, and D. Rand for comments and support of this work. We are indebted to two anonymous reviewers whose comments substantially improved this paper. We also thank Jarrod Hadfield for help with creating the generalized linear mixed model in R.
Conceived and designed the experiments: SMF ALH DPL ERM. Performed the experiments: SMF ALH DPL ERM. Analyzed the data: SMF ALH DPL ERM CO. Contributed reagents/materials/analysis tools: SMF ALH DPL ERM CO. Wrote the paper: SMF ALH DPL ERM ML KNN CO.
- 1. MacArthur DG, North KN (2004) A gene for speed? The evolution and function of alpha-actinin-3. Bioessays 26: 786–795. doi: 10.1002/bies.20061
- 2. Vincent B, De Bock K, Ramaekers M, Van den Eede E, Van Leemputte M, et al. (2007) ACTN3 (R577X) genotype is associated with fiber type distribution. Physiological Genomics 32: 58–63. doi: 10.1152/physiolgenomics.00173.2007
- 3. Mills MA, Yang N, Weinberger RP, Vander Woude DL, Beggs AH, et al. (2001) Differential expression of the actin-binding proteins, α-actinin-2 and-3, in different species: implications for the evolution of functional redundancy. Human Molecular Genetics 10: 1335–1346. doi: 10.1093/hmg/10.13.1335
- 4. North KN, Yang N, Wattanasirichaigoon D, Mills M, Easteal S, et al. (1999) A common nonsense mutation results in a-actinin-3 deficiency in the general population. Nature Genetics 21: 353–354. doi: 10.1038/7675
- 5. Lek M, Quinlan KGR, North KN (2010) The evolution of skeletal muscle performance: gene duplication and divergence of human sarcomeric alpha-actinins. Bioessays 32: 17–25. doi: 10.1002/bies.200900110
- 6. MacArthur DG, Seto JT, Raftery JM, Quinlan KG, Huttley GA, et al. (2007) Loss of ACTN3 gene function alters mouse muscle metabolism and shows evidence of positive selection in humans. Nature Genetics 39: 1261–1265. doi: 10.1038/ng2122
- 7. Yang N, MacArthur DG, Gulbin JP, Hahn AG, Beggs AH, et al. (2003) ACTN3 genotype is associated with human elite athletic performance. American Journal of Human Genetics 73: 627–631. doi: 10.1086/377590
- 8. Gaston KJ (2000) Global patterns in biodiversity. Nature 405: 220–227. doi: 10.1038/35012228
- 9. Mills M (2001) Differential expression of the actin-binding proteins, [alpha]-actinin-2 and -3, in different species: implications for the evolution of functional redundancy. Hum Mol Genet 10: 1335–1346. doi: 10.1093/hmg/10.13.1335
- 10. Hancock AM, Witonsky DB, Ehler E, Alkorta-Aranburu G, Cynthia Beall C, et al. (2010) Human adaptations to diet, subsistence, and ecoregion are due to subtle shifts in allele frequency. Proceedings of the National Academy of Sciences, USA 107: 8924–8930. doi: 10.1073/pnas.0914625107
- 11. Sokal RR, Rohlf FJ (1995) Biometry: The Principles and Practice of Statistics in Biological Research. New York: W. H. Freeman and Co. 887 p.
- 12. Olson DM, Dinerstein E, Wikramanayake ED, Burgess ND, Powell GVN, et al. (2001) Terrestrial ecoregions of the world: a new map of life on earth. BioScience 51: 933–938. doi: 10.1641/0006-3568(2001)051[0933:teotwa]2.0.co;2
- 13. Gaston KJ (2007) Latitudinal gradient in species richness. Current Biology 17: R574–R574. doi: 10.1016/j.cub.2007.05.013
- 14. Jablonski D, Roy K, Valentine JW (2006) Out of the tropics: evolutionary dynamics of the latitudinal diversity gradient. Science 314: 102–106. doi: 10.1126/science.1130880
- 15. Edgar RC (2004) MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Research 32: 1792–1797. doi: 10.1093/nar/gkh340
- 16. Pagel M, Meade A (2004) A phylogenetic mixture model for detecting pattern-heterogeneity in gene sequence or character-state data. Systematic Biology 53: 571–581.
- 17. Rambaut A, Drummond AJ (2007) Tracer v1.4, Available from http://beast.bio.ed.ac.uk/Tracer. Accessed 2011 Dec 5.
- 18. Pagel M (1999) Inferring the historical patterns of biological evolution. Nature 401: 877–884. doi: 10.1038/44766
- 19. Pagel MD (1997) Inferring evolutionary processes from phylogenies. Zoologica Scripta 26: 331–348. doi: 10.1111/j.1463-6409.1997.tb00423.x
- 20. Revell LJ, Harmon LJ, David C, Collar DC (2008) Phylogenetic signal, evolutionary process, and rate. Systematic Biology 57: 591–601.
- 21. Freckleton RP, Harvey HP, Pagel M (2002) Phylogenetic analysis and comparative data: A test and review of evidence. American Naturalist 160: 712–726. doi: 10.1086/343873
- 22. Gilks WR, Richardson S, Spiegelhalter DJ (1996) Introducing Markov chain Monte Carlo. In: Gilks WR, Richardson S, Spiegelhalter DJ, editors. Markov Chain Monte Carlo in Practice. London: Chapman and Hall. pp. 1–19.
- 23. Raftery AE (1996) Hypothesis testing and model selection. In: Gilks WR, Richardson S, Spiegelhalter DJ, editors. Markov Chain Monte Carlo in Practice. London: Chapman and Hall. pp. 163–188.
- 24. R Development Core Team (2011) R: A Language and Environment for Statistical Computing. http://www.R-project.org. Vienna, Austria: R Foundation for Statistical Computing.
- 25. Hadfield JD (2010) MCMC Methods for Multi-Response Generalized Linear Mixed Models: The MCMCglmm R Package. Journal of Statistical Software 33: 1–22.
- 26. Hadfield JD, Nakagawa S (2010) General quantitative genetic methods for comparative biology: phylogenies, taxonomies and multi-trait models for continuous and categorical characters. Journal of Evolutionary Biology 23: 494–508. doi: 10.1111/j.1420-9101.2009.01915.x
- 27. Stone GN, Nee S, Felsenstein J (2011) Controlling for non-independence in comparative analysis of patterns across populations within species. Philosophical Transactions of the Royal Society B: Biological Sciences 366: 1410–1424. doi: 10.1098/rstb.2010.0311
- 28. Beerli P (2009) How to use migrate or why are markov chain monte carlo programs difficult to use? In: Bertorelle G, Bruford MW, Hau HCR, A., Vernesi C, editors. Population Genetics for Animal Conservation, volume 17 of Conservation Biology. Cambridge UK: Cambridge University Press. pp. 42–79.
- 29. Li JZ, Absher DM, Tang H, Southwick AM, Casto AM, et al. (2008) Worldwide human relationships inferred from genome-wide patterns of variation. Science 319: 1100–1104. doi: 10.1126/science.1153717
- 30. Yang N, Macarthur DG, Wolde B, Onywera VO, Boit MK, et al. (2007) The ACTN3 R577X polymorphism in East and West African athletes. Medicine and Science in Sports and Exercise 39: 1985–1988. doi: 10.1249/mss.0b013e31814844c9
- 31. Shang X, Huang C, Chang Q, Zhang L, Huang T (2010) Association between the ACTN3 R577X polymorphism and female endurance athletes in China. International Journal of Sports Medicine doi: 10.1055/s-0030-1265176
- 32. Moran CN, Yang N, Bailey MES, Tsiokanos A, Jamurtas A, et al. (2007) Association analysis of the ACTN3 R577X polymorphism and complex quantitative body composition and performance phenotypes in adolescent Greeks. European Journal of Human Genetics 15: 88–93. doi: 10.1038/sj.ejhg.5201724
- 33. Eynon N, Duarte JA, Oliveira J, Sagiv M, Yamin C, et al. (2009) ACTN3 R577X polymorphism and Israeli top-level athletes. International Journal of Sports Medicine 30: 695–698. doi: 10.1055/s-0029-1220731
- 34. Paparini A, Ripani M, Giordano GD, Santoni D, Pigozzi F, et al. (2007) ACTN3 Ggenotyping by real-time PCR in the Italian population and athletes. Medicine and Science in Sports and Exercise 39: 810–815. doi: 10.1097/mss.0b013e3180317491
- 35. Ginevičien? V, Pranckevičien? E, Milašius K, Kučinskas V (2010) Relating fi tness phenotypes to genotypes in Lithuanian elite athletes. Acta Medica Lituanica 17: 1–10. doi: 10.2478/v10140-010-0001-0
- 36. Druzhevskaya AM, Ahmetov II, Astratenkova IV, Rogozkin VA (2008) Association of the ACTN3 R577X polymorphism with power athlete status in Russians. European Journal of Applied Physiology 103: 631–634. doi: 10.1007/s00421-008-0763-1
- 37. Norman B, Esbjörnsson M, Rundqvist H, Österlund T, Walden Fv, et al. (2009) Strength, power, fiber types, and mRNA expression in trained men and women with different ACTN3 R577X genotypes. Journal of Applied Physiology 106: 959–965. doi: 10.1152/japplphysiol.91435.2008