The Genom of Homo sapiens.pdf

484 CLARK ET AL.be under adaptive evolution fall into four main categories:skeletal development, neurogenesis, reproduction,and homeotic transcription factor genes (Table 4).For example, the homeotic transcript factor genes CDX4,HOXA5, HOXD4, MEOX2, POU2F3, MIXL1, andPHTF play key roles in early development and haveModel 2 p-values less than 0.05. TRAF5 plays a key rolein osteoclast proliferation and may be implicated in acceleratedgrowth of the long bones in the leg (Kanazawaet al. 2003). TRAF5 shows adaptive evolution along withsix other skeletal developmental genes. At least 10 genesinvolved in neurogenesis processes, including axonalguidance and synapse remodeling, have low Model 2 p-values. For example, the SIM2 transcription factor hasbeen implicated in human Down syndrome and memorydefects in mice (Chrast et al. 2000). FOXN1, or wingedhelix nude, encodes a transcription factor involved in keratingene expression. Mutations in this gene causeathymia, resulting in a severely compromised immunesystem. Developmental defects in Drosophila andCaenorhabditis elegans are also observed when this geneis mutated. A plausible hypothesis is that the relative hairlessnessof humans compared to chimps is in part determinedby FOXNI. Hypotheses like this are generated inabundance by studies such as this, and an exciting aspectof the work is that such hypotheses are amenable to futuretesting.The anatomy and physiology of reproduction are strikinglydifferent between humans and chimpanzees. Severalgenes involved in pregnancy appear to exhibit nonneutralevolution (Table 4). For example, the progesterone receptor(PGR) is involved with maintenance of the uterus andmay be involved in the acrosome reaction (Gadkar et al.2003). The reproductive hormone receptors GNRHR andMTNR1A also have significant Model 2 p-values.Several genes associated with the development of hearingappear to have undergone adaptive evolution (Table4). α-Tectorin, which shows the most significant Model 2p-value, plays an important role in the tectorial membraneof the inner ear. When it is mutated, humans showhigh-frequency hearing loss (Mustapha et al. 1999) andmouse knockout mice are deaf. Other genes under human-specificselection, DIAPH1, FOXI1, and EYA4,cause hearing loss in humans when mutated.CONCLUSIONSThere has been considerable interest in obtaining thegenome sequence of the chimpanzee, our closest relative,because of the notion that, by comparing our twogenomes, it might be possible to infer which genetic differencesare responsible for the morphological, physiological,and behavioral factors that differentiate us. At 1%sequence divergence, however, we expect there to beroughly 3 million base pairs of sequence difference, andthe discrimination between substitutions that are totallyunimportant and substitutions that are causal to our biologicaldifferences appears to be a steep challenge. Fortunately,the phylogenetic approach offers a promise tomake progress on this problem. With multiple relatedspecies arranged on a phylogenetic tree, models ofmolecular evolution can place the mutations on particularlineages of the tree. This information can be used to inferwhat DNA sequence changes have occurred specificallyalong the lineage subsequent to the node representing ourcommon ancestry with chimpanzee, and reflectingchanges that occurred in our line of descent since thattime. The challenge that remains is that many of thesechanges will have arisen purely because our populationsize is finite, and because random mutations, providedthey are not too deleterious, may go to fixation by randomdrift. It is humbling to consider that potentially a largeportion of the genomic differences between humans andchimps have arisen by such a purely neutral process. Ifone asks, “What are the genes that make us human?”,these random changes may surely be an important class ofgenes that carry this label.In this paper, we apply methods that have been used bymany others to infer which genes have been undergoingpositive or adaptive evolution. The idea is based on therelative rates of substitution at silent (synonymous) and atreplacement (nonsynonymous) nucleotide positions inthe gene. Strictly neutral genes are expected to have equalrates of substitution for these two classes of sites, whilemost genes have some selective constraint and show considerabledeficit of replacement changes. Formal statisticalapproaches allow us to test the null hypothesis that thechanges are compatible with neutrality, and to make quiteincisive tests of alternative hypotheses about the way thatselection has acted (Yang and Nielsen 2002).The application of this inferential approach identified along list of genes for which it is all too easy to tell an evolutionarystory about how these genes are important forhuman–chimp differences. However, it should be emphasizedthat these approaches are strictly exploratory andthat they really only highlight hypotheses to be tested byadditional data collection at several levels. The finding ofpositive selection in genes such as α-tectorin suggeststhat there may be differences in the hearing acuity of humansand chimpanzees, and the data available on the subjectare too sparse to properly address the issue. This motivatesspecific hearing tests of chimpanzees, with theidea that aspects of vocal speech may place additional requirementson hearing not faced by speechless chimpanzees.For every gene cited in this paper, there are additionalexperiments that must be done to solidify theevidence that these genes may be involved inhuman–chimpanzee differences. In the case of significantdifferences in biological processes, the case based onlyon DNA sequences becomes more compelling, mostlybecause each test involves many genes showing aberrant(nonneutral) behavior. That amino acid catabolismshould be a biological process showing rapid adaptiveevolution suggests further research into the physiology ofdigestion of low- versus high-protein diets, and considerationof the differences among primates in diet. Dietarychanges are not the only thing that might be driving thisdifference. Demands on protein synthesis during braindevelopment might be the driver of this signal of past naturalselection. Despite all these uncertainties, and the