Lizards in the genus Anolis have radiated extensively within and among islands in the Caribbean. Here, I provide a prospectus for identifying genes underlying adaptive phenotypic traits in anoles. First I review patterns of diversification in Anolis and the important morphological axes along which divergence occurs. Then I discuss two features of anole diversification, the repeated, convergent evolution of ecomorphs, and phenotypic divergence among populations within species, that provide opportunities to identify genes underlying adaptive phenotypic variation. While small clutch size and difficulty with captive rearing currently limit the utility of quantitative trait locus analyses, comparative analyses of gene expression, and population genomic approaches are promising.
The diversity of organisms for which whole genomic sequences are known offers an exciting opportunity for evolutionary biologists and ecologists to exploit genomic data, and associated methods, to address long-standing questions in comparative biology, ecology, and evolution. Not only can comparative analyses of genomes from diverse taxa generate novel insight into the evolution of whole genomes, but also extending genomic information and approaches from sequenced taxa to other related taxa offers the opportunity to address a number of fundamental questions in ecology and evolution.
Identifying genes underlying ecologically important phenotypic traits is a central goal of ecological and evolutionary genetics (Feder and Mitchell-Olds 2003; Storz 2005; Lee and Mitchell-Olds 2006; Ellegren and Sheldon 2008; Stinchcombe and Hoekstra 2008). Advances in technology, theory, and applications now allow analysis of genes underlying important phenotypic variation in wild populations of nonmodel organisms. Recent examples include analyses of coat color in beach mice (Hoekstra et al. 2006), size and shape of the bill in Galapagos finches (Abzhanov et al. 2004, 2006), adaptive variation in wild sunflowers (Kane and Rieseberg 2007; Sapir et al. 2007; Baack et al. 2008), and size, shape, body armor, and pigmentation in three-spine sticklebacks (Colosimo et al. 2004; Coyle et al. 2007; Miller et al. 2007; Albert et al. 2008; Makinen et al. 2008).
My goal here is to assess the opportunity to exploit the complete genomic sequence of the lizard, Anolis carolinensis, to address ecological and evolutionary questions in the adaptive radiation of the genus Anolis. Anoles have become a model for the study of adaptive radiation, and genomic approaches promise novel insight into the genetic changes underlying adaptive radiation. Difficulty with captive breeding of anoles currently limits their utility for genetic analyses requiring a known pedigree, but comparative analyses of gene expression across species, and population genomic analyses of closely related but phenotypically varying populations within species, provide opportunities to identify loci involved in adaptation. Here I briefly review Anolis as a model system in ecology and evolution, and discuss approaches to identify the underlying genetic basis for ecologically and evolutionarily important phenotypic variation. I also discuss the problem of speciation in Anolis, which is a neglected component of their adaptive radiation, present a model for speciation in Anolis, and discuss how genomic information and approaches may assist in understanding the entire breadth of speciation and adaptive diversification of anoles.
Background on lizards in the genus Anolis
Lizards in the genus Anolis have become a model system for studies of adaptive radiation, community ecology, morphology and performance, behavior, communication, local adaptation, and biotic diversification (Losos 1994; Losos et al. 1998; Losos 2001; Losos et al. 2003; Losos et al. 2004; Losos et al. 2006a, 2006b). The genus Anolis (Hass et al. 1993; Poe 1998; Frost et al. 2001) contains upwards of 400 species distributed throughout the neotropics, Caribbean, and south-eastern United States. The large adaptive radiation of more than 150 species in the Caribbean has received the most attention.
Recently, a draft genome assembly of the Green Anole (A. carolinensis) was completed (http://Broad.mit.edu/ftp/pub/assemblies/reptiles/lizard/AnoCar1.0). The genome of A. carolinensis is the first non-avian reptilian genome and its phylogenetic placement makes it valuable for comparative analyses of genomic structure and evolution in both reptiles and mammals. Importantly, the complete genome of A. carolinensis makes Anolis one of the few “genome enabled” adaptive radiations and thus provides the opportunity to study the genetics of adaptation. The question before us is: “How can we use the genome of A. carolinensis to further understand adaptation and diversification in the genus Anolis?”
Convergent evolution of anoles
Anoles have undergone independent adaptive radiations on each of the islands in the Greater Antilles (Cuba, Hispaniola, Puerto Rico, and Jamaica) (Losos et al. 1998; Jackman et al. 1999). A striking pattern of convergent evolution of morphology and behavior is demonstrated by the independent evolution of similar “ecomorphs” on each island (Losos et al. 1998). Ecomorphs are categorizations of behavior, ecology, and morphology that reflect microhabitat and substrate specialization. Ecomorphological divergence facilitates coexistence of different species and has resulted in complex, multispecies assemblages on each island in the Greater Antilles. Cuba and Hispaniola have a complicated geological history and their anole faunas may be comprised of species assemblages that evolved on isolated landmasses that subsequently coalesced to form the current islands (Hass et al. 1993; Glor et al. 2004). Thus, Cuba, with 55 species, and Hispaniola, with 40 species, contain the most diverse anole faunas. In contrast, Puerto Rico and Jamaica, where anole species assemblages have largely evolved in situ, host smaller faunas (11 and 7 species, respectively). In each case, similar ecomorphs have evolved.
Ecomorphs differ primarily in limb length, girdle dimensions, number of subdigital lamellae and dimensions of the skull (Losos 1990a, 1990b; Irschick et al. 1997). While ecomorphological divergence has occurred convergently on each of the Greater Antilles (Losos et al. 1998), it is not known if convergence resulted from the same or different changes in genes and development in each instance. Nor is it known whether changes in structural genes and/or regulatory regions contributed to adaptive morphological change (Hoekstra and Coyne 2007, Wray 2007). However, the fact that most phenotypic variation among ecomorphs occurs on a limited number of phenotypic axes highlights the genetic and developmental systems that we may target for analyses.
Anoles in both the Greater and Lesser Antilles occupy a wide range of habitats and many species show substantial variation in morphology associated with habitat. The most detailed studies of geographic variation in anoles have been on species from the Lesser Antilles. On volcanic, topographically complex, single-species islands in the Lesser Antilles (Guadeloupe, Dominica, Martinique), species are geographically variable, with population differentiation apparently reflecting local adaptation. Most of the species vary primarily in adult male color and pattern, but also in size, shape and scalation, and several subspecies have been described in each instance (Lazell 1964, 1972). Evidence for local adaptation derives from analyses of selection (Malhotra and Thorpe 1991, 2000; Thorpe et al. 2005) and reduced gene flow among divergent populations (Schneider 1996; Ogden and Thorpe 2002). Despite substantial variation in morphology among populations within islands in the Lesser Antilles, most populations have not diverged to the extent that different ecomorphs have (Knox et al. 2001).
Populations of some species on the Greater Antilles also show substantial geographic variation, although these have been less studied. For example, the Anolis distichus complex from Hispaniola comprises eight subspecies (Case 1990; Schwartz and Henderson 1991) that differ again in size, scalation, and, in adult males, color and pattern, especially of the dewlap, an extensible throat fan that is important in social communication and species recognition.
Because ecomorphological divergence is an important component of adaptive radiation of Anolis, populations within species that reflect the early stages of ecomorphological divergence are of particular interest. Anolis sagrei populations, originally derived from Cuba, but now occupying many islands in the Caribbean, vary in size, scalation, relative limb length, and lamellae number (Lister 1976; Kolbe et al. 2007). Recent studies revealed that populations of A. sagrei occupying different habitats in the Bahamas vary in size and relative limb length in a direction that mirrors those among ecomorphs (Losos et al. 1994; Calsbeek and Smith 2007; Calsbeek et al. 2007). Also, experimental manipulation of populations of A. sagrei in the Bahamas demonstrated that evolution of size and relative limb length can occur very rapidly and is associated with available height and diameter of perches, as well as with presence of predators (Losos et al. 1997, 2001, 2004, 2006b). Because of their shared genetic background, these populations may provide an exceptional opportunity to discover the genetic basis of adaptive change during the early stages of ecomorphological divergence.
Speciation in Anolis
Anoles are highly visual animals and there is strong sexual dimorphism in color and pattern in most Caribbean species. The males are territorial and use visual displays, including extension of a throat fan, or dewlap, along with stereotypical movement patterns, as social signals. The color of the dewlap varies among species and is likely important in species recognition (Losos 1985; Case 1990), as are markings and colors on the head and body. It is likely that divergence in color and pattern, especially of the dewlap, is an important step in the evolution of reproductive isolation among species.
Nearly all described subspecies of Anolis in the Caribbean differ primarily in color and pattern of adult males (Lazell 1964, 1972). Local adaptation in color and pattern among populations living in different habitats (e.g., lowland rainforest versus montane forest versus dry forest) has led to reduced gene flow among populations (Ogden and Thorpe 2002), suggesting that local adaptation may be an important step in speciation. Divergence in mate recognition systems, without ecomorphological divergence, suggests that divergence in mate-recognition signals, and perhaps reproductive isolation, precedes ecomorphological divergence. Ecomorphological divergence may then subsequently result from ecological interactions among reproductively isolated species in sympatry. This model contrasts with a proposed general model of adaptive radiation (Streelman and Danley 2003) in which habitat divergence is followed by (eco) morphological divergence, with divergence in communication (i.e., mate-recognition signals) occurring last.
Both sexual and ecological selections are important in producing differences in communication signals in anoles. Local adaptation of signal colors to light environment may very quickly lead to reduced gene flow among populations. Ecological selection via predation could also influence body and dewlap colors, with animals in areas of high predation being less conspicuous. This prediction is supported by a general pattern of anoles on dry islands, where vegetation is sparse and open, tending to be grey-brown and cryptic, whereas anoles in rainforests and wet habitats tend to be bright green and often contrasting with the background. Therefore, an important step in the speciation of anoles likely involves divergence in genes and gene expression that control color and pattern of the dewlap, head, and body. There are several cases in both the Lesser and Greater Antilles where closely related populations within species differ dramatically in color and pattern of adult males and these populations are good targets for population genomic analyses to identify genes affecting variation in color and pattern.
Identifying genes underlying ecologically and evolutionarily important phenotypic variation
Comparative analyses of gene expression among ecomorphs
Convergent evolution is pervasive among anole species. But what is the genetic basis of convergence? A promising approach to identify genes involved in convergent evolution of morphology is through comparative analysis of gene expression during developmental stages when morphological differences among species are established. Recent studies of development in anoles suggest that differences among ecomorphs in the length of limbs are established early during limb-bud development (T. J. Sanger, unpublished data). Similarly, differences in the size of the toe pad and number of lamellae appear to be established during a narrow window of development when the toe pad is formed. It is expected that differences in cranial morphology and other characters likewise will be established during specific stages of morphogenesis. Comparative analysis of gene expression from specific tissues at developmental stages when interspecific differences are established provides a good opportunity to identify genes responsible for adaptive phenotypic evolution.
But, how best to compare gene expression among species that may have diverged many millions of years ago? Until recently, gene expression was analyzed primarily using DNA microarrays. While the technical and analytical aspects of microarray gene expression analyses have advanced significantly in recent years, it has become clear that there are limitations to DNA microarrays in cross-species comparisons (Gilad et al. 2005). Given nearly 160,000 expressed sequence tags (ESTs) that have been produced from diverse tissues of A. carolinensis, DNA microarrays could be designed that contain nearly all expressed genes from that species. However, comparative analysis of gene expression using microarrays designed for A. carolinensis may be limited to comparisons of closely related species. Given the age of divergence within Anolis (∼50 million years) an approach that allows comparative analyses of gene expression among more distantly related species is desirable.
A recently developed approach, known as RNA sequencing or cDNA censusing, uses massively parallel sequencing technology to directly sequence mRNA or cDNA from target tissues (Torres et al. 2008; Wold and Myers 2008). The sequence reads are then compared against a reference genome, transcripts identified to gene, and their frequency tallied. This approach does not require sequencing of entire cDNAs, as short sequence reads of 25–30 bases are sufficient to place most reads on a reference genome. However, longer reads are desirable since they provide additional information on splice variants and nucleotide polymorphisms within genes. The results of RNA sequencing or cDNA censusing are spectra of EST frequencies which can be compared directly among species to discern differences in gene expression. This approach is promising as it accesses the entire transcriptome, but it is not without its own set of problems. Presently, short reads of 25–40 bases that are produced by some sequencing technologies may result in 15–20% of reads being difficult to assign to a single region in the reference genome due to multiple copies of short sequences in the genome or poor matches of sequence to reference (Wilhelm et al. 2008; Wold and Myers 2008). The latter becomes more of a problem as species examined become more evolutionarily distant from the species from which the reference sequence was derived. Again, longer sequence reads may provide an advantage in placing sequences on a reference genome. In a best case scenario, Torres and others (2008) were able to map 97% of EST’s from Drosophila melanogaster, with read lengths of ∼200 base pairs, to the D. melanogaster genome, and 90% of the sequences were unambiguously mapped to annotated genes. RNA sequencing is the most promising approach for comparative analyses of gene expression among species of Anolis.
Population genomics of intraspecific phenotypic variation
Because population differentiation is the first step in adaptive radiation, it is important to understand the genetic basis of adaptation at this level. Several recent reviews discuss approaches to identify loci underlying phenotypic variation within and among populations (Feder and Mitchell-Olds 2003; Beaumont and Balding 2004; Slate 2005; Storz 2005; Ellegren and Sheldon 2008; Stinchcombe and Hoekstra 2008). There are three general approaches. All rely on genome-wide surveys of variation using genetic markers (e.g., microsatellites, SNPs, or AFLPs) to identify regions under selection. The first, and most general, approach has come to be known as “Population Genomics” (Black et al. 2001; Goldstein and Weale 2001; Jorde et al. 2001). The term “population genomics” refers to analyses of variation at a large number of marker loci to identify those linked to regions under selection, while controlling for demographic events that affect levels of genetic variation across the entire genome. The simplest application of population genomic scans is to use the variance in allele frequency at marker loci to calculate a measure of divergence or variation between two populations and identify marker loci that have higher or lower values than would be expected by chance (Luikart et al. 2003; Schlotterer 2003; Beaumont and Balding 2004; Storz 2005; Stinchcombe and Hoekstra 2008). These outlier marker loci may be linked to regions under selection and thus have a pattern of population differentiation that differs from neutral genetic variation across the genome. It is unlikely that the marker itself is the target of selection, but if these markers are mapped, or can be localized in a published genome, then candidate loci in regions to which they are linked can be identified. Variation upstream, and downstream of the marker(s) can then be examined directly to identify genes and regulatory regions that may be, or have recently been, targets of selection (see e.g., Sutter et al. 2007). If markers are developed from sequence reads of the target species, those reads may be mapped to the A. carolinensis genome and, assuming shared synteny, genes, or other functional features in the region would be candidate loci.
The population genomics approach does not require knowledge of the location of markers, or even of the existence of phenotypic variation among populations (e.g., Bonin et al. 2006), but if such information exists, then statistical associations between phenotypes and genetic markers can be detected using linkage disequilibrium (LD) mapping or quantitative trait locus (QTL) analyses (Stinchcombe and Hoekstra 2008). QTL analysis requires the measurement of phenotype and marker genotype in a controlled breeding design, but can be applied to natural populations in cases in which pedigrees can be determined unambiguously (Slate 2005). QTL analyses are most appropriate to organisms that can be readily reared in captivity and that have large numbers of offspring and, therefore, anoles may not be good subjects for QTL analyses. LD mapping, on the other hand, may be applicable to natural populations. LD mapping in natural populations will require many more marker loci than QTL analysis (hundreds or thousands instead of dozens) because the breeding design of QTL analysis results in large linkage groups. Natural populations are likely to have smaller linkage groups due to a long history of recombination but, with sufficient marker density, small linkage groups may provide higher resolution for identifying regions under selection since any marker consistently associated with phenotype is likely to be closely linked to candidate loci. As with any association study, LD mapping requires controlling for population structure and careful attention must be paid to the impact of population structure on statistics for association (Freedman et al. 2004; Kimmel et al. 2007).
Which of these approaches is applicable to populations of anoles? Because of difficulty in husbandry and small clutch size, QTL analyses may not be feasible for Anolis in the near term. However, we may exploit population genomic and LD studies in closely related, but phenotypically varying, populations to identify regions of the genome that underlie important variation in body size, limb length, color and pattern, and other characteristics. Therefore, the most fruitful approaches are likely to involve population genome surveys using microsatellite or SNP markers to identify markers associated with phenotypes of interest. Marker loci can then be mapped by comparing the sequences from which they were derived to the genome sequence of A. carolinensis. Closely related populations of Anolis, in which most neutral genetic variation is still segregating in both populations (i.e., those in which genetic drift has not fixed populations for many loci), but which are fixed or nearly so for phenotypic characters potentially under selection, offer the best opportunity for identifying marker loci associated with phenotypic variants. Populations of A. sagrei in the Bahamas, and several polytypic species in the Lesser Antilles, are good targets.
Summary and conclusions
Despite difficulties with captive breeding of Anolis, comparative analyses of gene expression across species and population genomic surveys and LD mapping in natural and experimental populations, offers great opportunity to identify and study genes underlying ecologically and evolutionarily important variation. The completion, and future annotation, of the A. carolinensis genome will facilitate the study of genes involved in adaptation, speciation, and adaptive radiation by providing a framework for identifying candidate genes and gene regions underlying adaptive phenotypic evolution. The radiation of Anolis in the Caribbean is a textbook example of adaptive radiation and the integrated historical and experimental approaches applied to date have resulted in a deep and rich understanding of the causes of diversification in anoles (Losos 2007). Information on the genetic basis of adaptation and diversification in this group will complete the picture and provide insight into the genetic basis and architecture of adaptation and adaptive radiation.
Thanks to Daniel Janes, Christopher Organ, and Nicole Valenzuela for organizing the Reptile genomics symposium. Thanks to Travis Glenn, Nicholas Crawford, Jonathan Losos, and Hopi Hoekstra for insightful discussions, and thanks to three anonymous reviewers for constructive criticism on the manuscript. Special thanks to Thomas Sanger for sharing his work in progress on embryological development in Anolis.