POPULATION GENETICS Biology 107/207L Winter 2005 Lab 5. Testing for positive Darwinian selection A growing number of statistical approaches have been developed to detect natural selection at the DNA sequence level (reviews in Kreitman and Akashi 1995; Hughes 1999; Yang and Bielawski 2000; Nielsen 2001). One of the most powerful is the test for positive Darwinian selection in which a protein s rate of nonsynonymous subsitution (d N ) is compared with its rate of synonymous substitution (d S ). Provided that codon bias has not acted to constrain (d S ), a d N /d S ratio (also call the ω ratio) exceeding unity is strong evidence for the continued operation of natural selection favoring amino acid replacement mutations. Recently, maximum-likelihood models of codon substitution have been developed that allow for ω ratios to vary among sites (Nielsen and Yang 1998; Yang et al. 2000) thus enabling the identification of positive selection at individual amino acid sites in a protein-coding gene. These methods appear to offer a number of advantages over earlier pair-wise comparisons of d N and d S among taxa that average ω ratios over all sites and lineages. However, tests for positive selection based on ω ratios > 1 are extremely stringent and will likely fail to identify adaptive evolution when selection is weak and/or episodic or when power is reduced due to limited taxon sampling (see Anisomova et al. 2001). In a wide range of species elevated d N /d S ratios have commonly been reported at two broad classes of genes those involved in host-pathogen interactions (e.g., Hughes and Nei 1988; Smith et al. 1995; Ford 2001) and those functioning in reproduction (e.g., Lee et al. 1995; Metz and Palumbi 1996; Swanson et al. 2001; see recent review by Ford 2002). Despite this emerging generality, the diversity of genes that might experience positive selection in the genome is unclear. Positive selection has been described at proteins as diverse as digestive enzymes, cytochromes, toxins, cytokines, hormones, and antifreeze proteins. The growing list of positively selected genes suggests that diversifying selection may be more common than previously estimated (e.g., Endo et al. 1996) although details of the selective process in many cases remain unknown.
In this lab, we will use maximum-likelihood approaches to test for positive selection at two nuclear genes (pantophysin and S2) and one mitochondrial gene (cytochrome b) among various species of marine fishes belonging to the family Gadidae. The main gene of interest is pantophysin (209 amino acids), an integral membrane protein found in small (< 100 nm) cytoplasmic microvesicles that function in a variety of intracellular shuttling pathways (see Haass et al. 1996; Windoffer et al. 1999). The precise role played by pantophysin in these trafficking pathways is still unknown. However, there is a strong signal of positive selection at this locus in the Atlantic cod, Gadus morhua: Two common alleles are segregating in populations throughout the north Atlantic region that differ by six amino acid substitutions (and no silent changes) and group in one small domain of the protein (Pogson 2001). It is unclear if the elevated rates of replacement changes observed at the PanI locus of G. morhua are due to the unusual polymorphism detected in this species or if similar selection pressures are acting in other related species. The second nuclear gene we will study encodes for the S2 ribosomal protein (183 amino acids). This is a highly conserved gene in most vertebrates and thus is expected to serve as a control (i.e., not exhibit positive selection). We will also include a region of the mitochondrial cytochrome b (cyt b) gene (299 amino acids) that, like S2, is not expected to experience positive selection. We will perform tests for positive selection by running two models implemented by the codeml program of PAML (Phylogenetic Analysis by Maximum Likelihood). The null model we will use is called M7 (beta), which assumes that d N /d S ratios at different amino acid positions in a protein follow a beta distribution. Because the beta distribution is constrained to fall between zero and one, model M7 prevents any amino acid sites from experiencing positive selection (since this necessitates that the d N /d S ratio at a position exceeds unity). The likelihood score of model M7 (l M7 ) is compared to that obtained from model M8 (beta&ω>1), which allows for another group of sites (estimated from the data) to have d N /d S ratios that exceed unity (l M8 ). The likelihood scores of the two models are tested for significance using a standard likelihood ratio test (χ 2 = 2 (l M8 - l M7 ) with 2 d.f.). The codeml program will also perform a Bayes empirical Bayes (BEB) calculation of the posterior probabilities of sites identified as having d N /d S ratios greater than 1. Posterior probabilities above 0.95 for a site provide strong support for the action of positive selection.
Download the PAML program from the following web site: http://abacus.gene.ucl.ac.uk/software/paml.html Directions on downloading the program are given on this home page. There are Windows, Unix, Linux, and Mac OSX versions of PAML available. Download the appropriate archive and Unzip the program into a desired folder. A total of 132 files should be extracted. The data files The codeml program requires a control file codeml.ctl, a data file containing the aligned sequences (e.g., pan.nuc ) and a tree file (e.g., pan.trees ) in the same folder as the executable file in order to run. After unzipping the PAML files delete the default codeml.ctl file. Copy the following control files for each gene from the Bio 107/207 class web site into the PAML folder: pancodeml.ctl, S2codeml.ctl and cytbcodeml.ctl. Then copy the following data files into the PAML folder: pan.nuc, S2.nuc and cytb.nuc. Also copy the following tree files into the same folder: pan.trees, S2.trees, and cytb.trees. To test for positive selection at the pantophysin gene, you must rename the pancodeml.ctl file as codeml.ctl. After the run has completed, the results are printed into the output file pan.out. Re-rename the codeml.ctl back to pancodeml.ctl. Repeat for the S2 and cyt b genes. Running PAML Open up a Command Prompt window from the path Start Programs Accessories Command Prompt. Change the directory to where the PAML program has been installed. For example, if the program is in c:\program Files\PAML then type cd\ Program Files\PAML. Run the codeml program from the command prompt by typing codeml. The codeml program will read the data and then begin iterating through multiple rounds of parameter estimation. Endless hours of fun watching the gibberish on your screen! The time it will take to
perform these runs will depend on the speed of your computer hopefully not more than 3-4 hours per run. Looking at the output files Output files for the three control files are called pan.out, S2.out, and cytb.out. For the file pan.out the results we are interested in are presented at the bottom of page 14 under Model 7: beta (10 categories). This is the output from our null model. On the line beginning lnl (ntime: 35 np: 38) is the likelihood score (with a value, hopefully, close to 2464.479416). Below this are estimates of branch lengths, the transition/transversion ratio (kappa) and parameters for the beta distribution (p and q). Then appear the estimates of dn and ds for all the branches in the tree. Note that the dn/ds ratios are constrained not to exceed 1. The results for model M8 (beta&ω>1) appear below those for M7. Please note the likelihood score for M8 you will need this for the likelihood ratio test. Parameters for the model appear following the branch length estimates. After the line Parameters in beta&ω>1: are listed p 0 (the proportion of selectively constrained sites), p 1 (the proportion of sites experiencing positive selection), parameters for the beta distribution (p and q), and a mean omega ratio (ω) for the positively selected sites. A mean omega ratio significantly above 1.0 is strong evidence for positive selection. The dn and ds values listed for each branch are now maximum-likelihood estimates of the true substitution rates. Below this table is listing of positively selected sites identified by the model and their posterior probabilities. There is a second output file made for each run called rst. (Rename the rst file if you wish to save it after a run, otherwise it will re-written.) This is an extremely detailed file containing reconstructions of ancestral states and substitution patterns. It also lists the actual amino acid substitutions that have occurred along each branch of the phylogeny. This information can provides further insights into the nature of the observed amino acid substitutions (for example between polar and nonpolar, between charged and noncharged residues, etc.). Assignment
1. Run models M7 and M8 on the pantophysin, S2, and cytochrome b data sets as described above. Present likelihood scores, parameter estimates, and listings of positively selected sites (if present) for each locus. Perform likelihood ratio tests for the action of positive selection at each gene. Briefly discuss the similarities and/or differences observed between the patterns of nucleotide substitutions at each gene. 2. If positive selection has been detected at a locus, determine how many branches on the tree exhibit greater numbers of nonsynonymous than synonymous changes. What does this tell us about the history of positive selection at the gene? Also examine the locations of positively selected sites? Do they appear to be clustered or random? What could cause clustering of sites experiencing positive selection? References Anisimova, M., J.P. Bielawski, and Z. Yang. 2001. Accuracy and power of the likelihood ratio tests in detecting adaptive molecular evolution. Mol. Biol. Evol. 18: 1585-1592. Endo, T., K. Ikeo, and T. Gojobori. 1996. Large-scale search for genes on which positive selection may operate. Mol. Biol. Evol. 13: 685-690. Ford, M.J. 2001. Molecular evolution of transferrin: evidence for positive selection in salmonids. Mol. Biol. Evol. 18:639-647. Ford, M.J.. 2002. Applications of selective neutrality tests to molecular ecology. Mol. Ecol. 11:1245-1262. Haass, N.K., J. Kartenbeck, and R.E. Leube. 1996. Pantophysin is a ubiquitously expressed synaptophysin homologue and defines constitutive transport vesicles. J. Cell Biol. 134: 731-746.
Hughes, A.L.. 1999. Adaptive evolution of genes and genomes. Oxford University Press, New York. Hughes, A.L., and M. Nei. 1988. Pattern of nucleotide substitution at major histocompatibility complex class I loci reveals overdominant selection. Nature 335: 167-170. Kreitman, M., and H. Akashi. 1995. Molecular evidence for natural selection. Annu. Rev. Ecol. Syst. 26: 403-422. Lee, Y.-H., T. Ota, and V. D. Vacquier. 1995. Positive selection is a general phenomenon in the evolution of abalone sperm lysine. Mol. Biol. Evol. 12: 231-238. Metz, E.C., and S.R. Palumbi. 1996. Positive selection and sequence rearrangements generate extensive polymorphism in the gamete recognition protein bindin. Mol. Biol. Evol. 13: 397-406. Nielsen, R. 2001. Statistical tests of neutrality in the age of genomics. Heredity 86: 641-647. Nielsen, R., and Z. Yang. 1998. Likelihood methods for detecting positively selected sites and applications to the HIV-1 envelope gene. Genetics 148: 929-936. Pogson, G.H. 2001. Nucleotide polymorphism and natural selection at the pantophysin (PanI) locus in the Atlantic cod, Gadus morhua (L.). Genetics 157: 317-330. Smith, N.H., J. Maynard Smith, and B.G. Spratt. 1995. Sequence evolution of the porb gene of Neisseria gonorrhoeae and Neisseria meningitidis: evidence of positive Darwinian selection. Mol. Biol. Evol. 12: 363-370. Swanson, W.J., and C.F. Aquadro. 2002. Positive Darwinian selection promotes heterogeneity among members of the antifreeze protein multigene family. J. Mol. Evol. 54: 403-410.
Windoffer, R., M. Borchet-Stuhltrager, N.K. Haass, S. Thomas, M. Hergt, C.J. Bulitta, and R.E. Leube. 1999. Tissue expression of the vesicle protein pantophysin. Cell Tissue Res. 296: 499-510. Yang, Z., and J.P. Bielawski. 2000. Statistical methods for detecting molecular adaptation. Trends Ecol. Evol. 15: 496-503. Yang, Z., R. Nielsen, N. Goldman, and A.-M. K. Pedersen. 2000. Codon-substitution models for heterogeneous selection pressures at amino acid sites. Genetics 155: 431-449.