Evolutionary Anthropology 23:188–200 (2014)

ARTICLE

Genetic and Developmental Basis for Parallel Evolution and its Significance for Hominoid Evolution PHILIP L. RENO

Greater understanding of ape comparative anatomy and evolutionary history has brought a general appreciation that the hominoid radiation is characterized by substantial homoplasy.1–4 However, little consensus has been reached regarding which features result from repeated evolution. This has important implications for reconstructing ancestral states throughout hominoid evolution, including the nature of the Pan-Homo last common ancestor (LCA). Advances from evolutionary developmental biology (evo-devo) have expanded the diversity of model organisms available for uncovering the morphogenetic mechanisms underlying instances of repeated phenotypic change. Of particular relevance to hominoids are data from adaptive radiations of birds, fish, and even flies demonstrating that parallel phenotypic changes often use similar genetic and developmental mechanisms. The frequent reuse of a limited set of genes and pathways underlying phenotypic homoplasy suggests that the conserved nature of the genetic and developmental architecture of animals can influence evolutionary outcomes. Such biases are particularly likely to be shared by closely related taxa that reside in similar ecological niches and face common selective pressures. Consideration of these developmental and ecological factors provides a strong theoretical justification for the substantial homoplasy observed in the evolution of complex characters and the remarkable parallel similarities that can occur in closely related taxa. Thus, as in other branches of the hominoid radiation, repeated phenotypic evolution within African apes is also a distinct possibility. If so, the availability of complete genomes for each of the hominoid genera makes them another model to explore the genetic basis of repeated evolution.

Philip L. Reno studies primate and vertebrate developmental evolution. He is an Assistant Professor in the Department of Anthropology at The Pennsylvania State University, where he uses mouse models to determine the genetic basis for differential skeletal growth and the loss of penile spines. E-mail: [email protected]

Key words: repeated evolution; convergence; stickleback; Ardipithecus; Pierolapithecus; adaptive radiation

C 2014 Wiley Periodicals, Inc. V

DOI: 10.1002/evan.21417 Published online in Wiley Online Library (wileyonlinelibrary.com).

The frequent occurrence of homoplasy, or what Darwin5 called “adaptive resemblances,” has long been known to be fundamental to the evolutionary process. Primate evolution has proven to be no exception. While most shared phenotypes and genotypes are inherited from a common ancestor, evolutionary phylogenies must typically accommodate a substantial degree of homoplasy.2 Phylogenetic analyses based on diverse and robust datasets are appropriately evaluated based on their parsimony, with preferred scenarios involving fewer independent evolutionary transitions.4 However, when the focus of analysis is the reconstruction of specific ancestral conditions, particularly those that repre-

sent adaptive suites, strict adherence to parsimony in character state evolution will likely result in a substantial number of erroneous hypotheses. Accordingly, it can be beneficial to expand the notion of parsimony to include both ecological and developmental considerations during ancestral reconstruction. One of the fundamental principles revealed by evo-devo is that animals are constructed through the redeployment of a relatively small set of homologous “tool kit” genes that are reused in different developmental and evolutionary contexts.6–8 This has resulted in the startling discovery that even in greatly diverged groups the repeated evolution of certain structures occurred through the reuse of homologous genes.7 Classic examples include the eyes (eyeless/Pax6), heart (tinman/Nk2), and limbs (distal-less/ Dlx5/6) in animals as diverse as flies and vertebrates.6,9 This paradox is explicable by the realization that deep conservation of the gene regulatory networks involved in animal development actually facilitates the repeated evolution of similar phenotypes. Such examples are not limited to repeated evolution in distantly related species in which homoplasy is readily apparent. Study of the mechanisms underlying finer scale evolutionary change has revealed similar examples of the reuse of shared genetic and developmental mechanisms to produce parallel adaptive changes in closely related species. Such phenomena appear to blur the distinction between homology and homoplasy.10,11 Developmentally, characters that result from similar genetic mechanisms are essentially the “same” and, in one sense, homologous (for example, the

ARTICLE

Genetic and Developmental Basis for Parallel Evolution and its Significance for Hominoid Evolution 189

Glossary cis-regulation — modification of gene expression by enhancers (activators or repressors) lying on the same chromosome as the gene. Cautious clambering — deliberate use of multiple supports during climbing and successive bridging in both above- and below-branch positions. Convergent evolution — the independent occurrence of similar phenotypes and/or genetic features in divergent lineages. Enhancer — a cis-regulatory DNA sequence that, when bound to transcription factor(s), modifies expression of a gene. Enhancers act as activators or repressors, depending on which transcription

first lumbar vertebrae of chimpanzees and humans). However, a discussion of ancestral character states implies historical continuity so that, for example, the first lumbar that follows the human 12th or the chimpanzee 13th thoracic vertebra should no longer be considered homologs. Here, repeatedly evolved characters will be considered homoplasies, regardless of whether they occur through parallelism or convergence. After a brief discussion of the potential for repeated evolution of locomotion in hominoids, I will review a few cases of dramatic parallel evolution in animals more conducive to genetic and developmental analyses. We will see that identifying the proximate mechanisms underlying these radiations can be directly applied to hominoid evolution. Application of this broader ecological and developmental context demonstrates that the potential for extensive homoplasy in hominoids, and even within African apes, is not unprecedented.

REPEATED EVOLUTION IN HOMINOIDS All extant apes exhibit numerous postcranial specializations that commonly are attributed to adaptation for

factors are bound at a given point and can be located a megabase or more from the genes they regulate. Evolvability — the capacity of a population to produce adaptive genetic variation that is available to selection. Evolvability can be influenced by various factors, including the limitation of deleterious effects of mutation, preexisting regulatory complexity, and standing genetic variation. Parallel evolution — the independent occurrence of similar phenotypes via common gene pathways or mechanisms. Commonly, but not universally, parallel evolution occurs in closely related species.

below-branch locomotion or suspension; these include longer forelimbs than hind limbs, elongated manual digits, laterally oriented shoulder joints with scapulae placed on the dorsum of a broad torso, invagination of the spine into the thorax and abdomen, and the absence of a tail (Fig. 1).3 Gibbons and siamangs use a distinctive form of richochetal brachiation involving extreme forelimb elongation.12 The larger-bodied great apes have evolved features that are useful for supporting their large mass in the arboreal canopy. These features include stiff backs, which are due to a reduced lumbar column containing three to four vertebrae that are entrapped between cranially elongated iliac blades, and a short compliant mid-foot.13 Orangutans rarely come to the ground and use their highly mobile limbs to climb cautiously in the upper canopy. African apes are specialized for knuckle-walking during terrestrial travel, yet to various extents forage and sleep in trees, necessitating frequent vertical climbing.14 Hominoid evolution has confirmed Darwin’s observation that evolution is not required to follow the most parsimonious path.2,4 The fossil record provides numerous examples of experimentation with forelimbdominated locomotion (Fig. 1).1 Early

Quantitative trait locus (QTL) — a site containing one or more genes that underlie variation in a measurable trait. Repeated evolution, or homoplasy — the independent attainment of a similar phenotype through either convergent or parallel evolution. Standing genetic variation — the presence of more than one allele at a locus in a population. Transcription factor — a protein (for example, Pitx1 or Hox) that, by binding to a DNA recognition sequence, regulates gene expression.

Miocene Proconsul demonstrates that tail loss and some aspects of hominoid elbow morphology evolved in the context of above-branch arboreal quadrupedalism.15 Conserving a narrow torso typical of pronograde primates, Nacholapithecus of the African Middle Miocene has unusually large forelimbs,16 suggesting climbing behaviors distinct from those of living apes.3 Pierolapithecus shows some derived features such as spinal invagination, dorsally placed scapulae, and ulnar styloid reduction combined with relatively short digits indicating palmigrade stance.17,18 This suggests that the reorganization of the spine, thorax, and scapula, also observed in Early Miocene Morotopithecus based on its derived lumbar vertebrae,19 may be decoupled from the highly specialized suspensory adaptations of modern apes, which typically include forelimb elongation and a short, stiff lumbar column (Fig. 1).13,17–20 Even clearly suspensory Late Miocene Hispanopithecus has relatively short and robust metacarpals and ulnar morphology suggesting a diverse locomotor repertoire that included at least occasional above-branch palmigrady.21–23 Sivapithecus shares evident craniofacial affinities with orangutans, yet is thought to have the postcrania of a pronograde

190 Reno

ARTICLE

Figure 1. Phylogeny illustrating the locomotor diversity and evolution of thoracic reorganization in hominoids. Hypothesized relationships of key fossils are indicated by dashed lines.3,4 Explicitly suspensory/below-branch species are shown in gray. Proconsul and Sivapithecus indicate that ancestral hominoids were above-branch quadrupeds. However, multiple locomotor experiments occurred within hominoids, indicated by the enlarged forelimbs of Nacholapithecus, the specialized skeleton of Oreopithecus, and long fingers of Hispanopithecus. Some taxa (bold text) also undergo spinal invagination, broadening of the thorax, and dorsal positioning of the scapula, indicated by a lumbar transverse process (LTP) on the pedicle (Morotopithecus and Pierolapithecus) or a reduced retroauricular region of the pelvis (Ardipithecus). Pierolapithecus and Ardipithecus lack many of the other derived below-branch specializations of extant hominoids. Inset: Illustration of the functional importance of migration of the LTP from the vertebral body to the pedicle of the neural arch. Old World monkeys reflect the ancestral condition similar to Proconsul, with large extensor musculature crucial for supporting the long back of acrobatic above-branch quadrupeds. This shift accompanied invagination of the spine into the broadened thorax and dorsally placed scapula, providing a greater range of glenohumeral stability. Dorsal migration of the LTP reduced the size of the supportive extensor musculature. Such derived vertebral morphology implies a cautious clamberer in a more generalized long-backed hominoid or suspension when combined with reduction in lumbar number in great apes. Inset image used with permission from Owen Lovejoy and Linda Spurlock.

quadruped.24,25 Although Oreopithecus has orangutan-like limb proportions and five lumbar vertebrae similar to gibbons,26 its specialized dentition demonstrates that it is not a close relative of living great apes.27 This suggests that at least four separate forays into suspensory locomotion occurred within hominoids (gibbons, orangutans, African apes, and Oreopithecus). Numerous models have been proposed over the past century to describe the ancestral condition from which humans evolved.28,29 However, this diversity in extant and fossil ape locomotion makes phylogenetic reconstruction and the inference of ancestral character states difficult.2,4 With genetic data firmly rooting humans within the African ape clade

(Fig. 1),30,31 any scenario of human origins is essentially a model of gorilla and chimpanzee origins as well. For example, one reasonable hypothesis is that the locomotor similarities of African apes are homologs mandating that our LCA with chimpanzees was also a knuckle-walking suspensory and vertical climbing ape.4,29,32–34 Richmond, Begun, and Strait29 found support for this hypothesis in potential knuckle-walking features that arguably limit extension of the radiocarpal and midcarpal joints, fusion of the os centrale, and irregularly shaped (“keeled”) second carpometacarpal joints that are also observed in Australopithecus and Homo.29 Pilbeam33 and Williams34 note the lower variation in human vertebral formulae relative to apes and argue for recent

selection of lumbar lengthening for bipedalism. On the other hand, the patterns of ontogeny and variation of the carpals suggest that chimpanzee and gorilla methods of knucklewalking are biomechanically distinct.35–37 Kivell and Schmitt point out that gorillas lack the claimed extension-limiting features observed in the chimpanzee wrist, reflecting the more columnar stance of the larger ape.37 In addition, McCollum et al.38 observe that bonobos differ from chimpanzees and gorillas in the total number of precoccygeal vertebrae, suggesting partial independent lumbar reduction in African apes. These data suggest that terrestrial knuckle-walking and reliance on vertical climbing and suspension evolved independently in African apes. While consensus has yet to be reached on the specifics, recently multiple scenarios have posited that African apes and humans evolved from a more generalized hominoid. Ward suggests a partially derived ancestor reminiscent of Oreopithecus or Pierolapithecus.3 The former would indicate an LCA with an essentially suspensory bauplan requiring only the simple refinements observed in most great apes.4 Crompton, Vereecke, and Thorpe28 have proposed that the PanHomo LCA was an arboreal orthograde clamberer that engaged in frequent arm-assisted bipedalism analogous, in many respects, to orangutans, with longer forelimbs than hind limbs and a short, stiff lumbar column. Similarly, Wolpoff39 argued that the likely smaller mass of the LCA may have favored arboreal bipedalism and limited selection for suspensory features such as a short back that are observed in larger apes.39 In each of these models, terrestrial knuckle-walking adaptations evolved in parallel in chimpanzees and gorillas.28,39 A more generalized Pierolapithecuslike ancestor would suggest retention of above-branch clambering as a significant component of the ancestral locomotor repertoire18 and might indicate that the fundamental modern hominoid adaptation is not for suspension per se, but for the ability to maintain shoulder stability under a range of forelimb loading positions.13

ARTICLE

Genetic and Developmental Basis for Parallel Evolution and its Significance for Hominoid Evolution 191

Box 1. Ardipithecus ramidus: Expanding the Window on the LCA

Box Figure 1. Retention of nonsuspensory features in Ardipithecus and the differences in African ape knuckle-walking and belowbranch adaptations may indicate that the clade evolved in parallel from a clamberer with a broad thorax with spinal invagination, lateral shoulder stability, long lumbar column, and palmigrade stance. The dashed circle indicates the potential duration of a more generalized clambering morphotype in each lineage.

While it is rare to find true ancestors in the fossil record, the discoveries over the past 20 years of Ar. ramidus, Sahelanthropus tchadensis, and Orrorin tugenesis from the Late Miocene and Early Pliocene can be highly informative regarding the nature of the Pan-Homo LCA.40,41 The best represented is the 4.4 Ma Ar. ramidus assemblage, including ARA-VP-6/500 (“Ardi”), a partial skeleton that included elements of the teeth, skull, limbs, and pelvis and excellent preservation of the hands and feet, along with more than 100 additional individuals.40 Derived cranial-dental characters, such as a shortened cranial base and reduction of the canines with loss of the functional honing complex, are shared with Sahelanthropus and Australopithecus.90,91 The pelvis is broad and short, freeing the lumbar column from articulation with the flared iliac blades, facilitating lordosis. The pelvis also displays evidence

of the distinctive human and australopithecine separate ossification center for the anterior inferior iliac spine and a muscular conformation favoring hip stabilization during the single support phase of bipedal gait.92 Despite indications of bipedality on the ground, the abducted great toe and long ischium (providing robust leverage for the hamstring muscles) shows Ardipithecus still relied on arboreal climbing.93 More informative are ancestral characters that are more reminiscent of primitive hominoids than of living African apes, such as similar fore- and hind limb lengths and short metacarpals relative to phalanges (Table 1). Furthermore, some apparently ancestral characters of Ardipithecus link modern humans and primitive hominoids to the exclusion of extant apes. These characters include a long fulcrumating tarsus used for propulsion in both bipedal and abovebranch quadrupedal gaits. A recent analysis by Almecija et al.94 shows that the Orrorin femur also displays greater morphometric affinity to early Miocene hominoids such as Proconsul than it does to African apes. Ardipithecus exhibits a reduced retroauricular region of the ilium. This suggests that, like modern humans, Australopithecus, extant apes, and some extinct hominoids (Pierolapithecus, Oreopithecus, and Morotopithecus) Ardi had already undergone spinal invagination and broadening of the thorax for dorsal placement of the scapula. This indicates that thoracic reorganization is a homologous character in African apes and humans. Ardi lacks crucial suspensory indicators, such as a more rigid and energy-dissipating second and third carpal-metacarpal joints (central joint complex), elongated palm, loss of the deltopectoral crest, and lumbar entrapment by the ilia observed in African apes (Table 1).95 This combination of

features, along with Ardi’s relatively large size (50 kg), suggests that while in the trees it was not a primitive arboreal quadruped in the manner of most Old World monkeys and Proconsul, nor did it rely on the various below-branch suspensory locomotor modes of the extant apes. Instead, it likely combined a mixture of orthograde and pronograde postures as a palmigrade cautious climber using simultaneous grasping and careful bridging between multiple supports.41 Multiple hypotheses have been proposed to account for this morphology and behavior. Potentially, the cardinal features of the human lineage, such as terrestrial bipedalism and feminization of the male canine, may have evolved in parallel with the ancestors of Australopithecus, suggesting that species like Sahelanthropus, Orrorin, and Ardipithecus may lie off of the human branch.96 However, the characters that link these fossils to Australopithecus and Homo are extensive.40,90 Such a hypothesis requires an explanation of why evolutionary experimentation with bipedalism and canine feminization was so prominent at the Miocene-Pliocene transition compared to the 20 Ma history of diversity in quadrupedal locomotor modes across hominoids. Alternatively, the affinities between Ardipithecus and archaic hominoids could represent reversals from an African ape-like LCA.97 While reversals associated with the evolution of bipedalism are plausible (such as a longer lumbar spine), the loss of previously evolved suspensory and vertical climbing features in a large-bodied arboreal hominoid such as Ardi is difficult to reconcile. The scenario presented during the original description of the Ar. ramidus assemblage is that the LCA of African apes and humans may have been a more generalized arboreal palmigrade cautious clamberer

192 Reno

ARTICLE

TABLE 1. Character States for Ar. ramidus and Their Implications for Polarity and Locomotion in the Pan-Homo LCA.a

Character

Earliest Condition on Human Branch

Extant Ape Condition

Limb proportions (Intermembral index)

Equal forelimb and hind Longer forelimbs than limb length in Ardipihind limbs (>100) thecus (89–91) (parallelism)

Metacarpal proportions

Short metacarpals relative to phalanges in Ardipithecus

Elongated metacarpals relative to phalanges in extant apes (parallelism) for suspension/ vertical climbing Deltopectoral crest Prominent deltopectoral Reduced deltopectoral crest in Ardipithecus, crest in extant apes Australopithecus, & (parallelism) Homo Lumbar number 6 lumbar vertebrae in 3–4 lumbar in extant (based on zygopo- Australopithecus33,38,98 great apes physeal (parallelism) orientation) Sacral morphology Broad sacrum with 4–5 Great apes with narrow segments in Australosacrum with 5–7 segpithecus & early ments (lumbar sacralHomo38; breadth ization)33; caudal implied for Ardipithelumbar entrapment cus by preserved between ilia innominate (parallelism) Spinal invagination Lumbar transverse proc- LTPs on pedicle and broad rib cage in esses (LTPs) on pedicle great apes (homoloand broad rib cage in gous in African apes; Australopithecus; potential parallelism in reduction of retroaurorangutan) icular region in Ardipithecus. Shortened midfoot Midfoot complex Moderately rigid midfoot enhances fulcruenhances grasping mation at toe off in during vertical climbArdipithecus, Australoing (parallelism) pithecus, & Homo

a

Examples in Fossil Hominoids

Implication for PanHomo LCA

Equal forelimb and hind Equal forelimb and hind limb in Proconsul (87) limbs; nonsuspensory and Old World monkeys Short metacarpals with Grasping palmigrade phalanges lacking hand; nonsuspensory suspensory features clamberer Pierolapithecus17,18 Prominent deltopectoral Prominent deltopectoral crest; nonsuspensory crest in Proconsul & Sivapithecus24 >7 lumbar in Proconsul and Nacholapithecus99

6 lumbar; nonsuspensory

Intermediate breadth with 4–5 segments in basal hominoids (3 in monkeys 11 with tail loss); 6 segments in Oreopithecus38,100

Intermediately broad sacrum providing mobility of lumbar column; facilitates incipient bipedalism

LTPs on pedicle/body juncture and broad rib cage in Pierolapithecus & Morotopithecus17,19

Broad torso, dorsally placed scapula, lateral shoulder stability13; cautious clamberer

Semi-rigid midfoot in Fulcrumating midfoot; Proconsul & Old World above-branch monkeys propulsion

For further details see Lovejoy and coworkers41 and references therein unless otherwise noted.

with a broad thorax, spinal invagination, dorsal placement of the scapulae, and free lumbar column (Table 1). Such a generalized bauplan, combined with a high degree of ecological plasticity, genetic

Such an ancestor would be compatible with a recent model from Lovejoy and colleagues41 informed by Ardipithecus ramidus, which contains a unique suite of derived bipedal features yet retains characters reminiscent of more primitive hominoids than living African apes (Box 1). This proposed Pan-Homo LCA combines a

diversity, and substantial genomic and developmental regulatory complexity, would permit relatively easy evolution of the unique bipedal features observed in Ardipithecus. This would also set the

reorganized thorax with a longer lumbar spine, a shorter metacarpus, Proconsul-like limb proportions, and a more rigid fulcrumating midfoot.40,41 The absence of general suspensory, vertical-climbing, and/or knuckle-walking specializations implies that all of these features are homoplasies in African apes.13,40

stage for parallel evolution of the vertical climbing (a stiffened spine, elongated forelimb/metacarpus, and compliant midfoot) and knuckle-walking features of African apes (Box Figure 1).

Classical genetic theory posits that novel genotypes arise from stochastic processes, so that similar genotypes are unlikely to evolve repeatedly. Phenotypic homoplasy should thus be rare without intense selection and similar ecological conditions.42,43 Repeated evolution of suspension across hominoids and knuckle-

ARTICLE

Genetic and Developmental Basis for Parallel Evolution and its Significance for Hominoid Evolution 193

walking within African apes may attest to the remarkable adaptive power of selection. However, recent developmental analyses in organisms undergoing adaptive radiation have shown that repeated phenotypic evolution often uses similar genetic and developmental mechanisms.43–45 Genomic properties, such as differences in mutation rate and regulatory architecture, can influence the generation and sorting of alleles, making them less random than previously assumed.42,43,46 This potential to use similar genetic mechanisms independently increases in closely related taxa that share similar genomes. Such genetic, developmental, and ecological factors can potentially facilitate parallel adaptation in a variety of organisms, including hominoids.

REPEATED EVOLUTION AND THE ORGANIZATION OF GENOMIC AND DEVELOPMENT ARCHITECTURE Evo-devo has shown that morphological evolution is influenced by the fact that, like the complex phenotypes they encode, genomes and developmental processes manifest similar organizational properties, such as recognizable homology, modularity, and functional specificity.6 These properties underlie the generation of organismic complexity and evolvability.47 Analogous to the convergent wings of bats, birds, and pterosaurs, which rely on different arrangements of homologous skeletal elements, both flies and vertebrates use modified homologous regulatory networks to construct limbs, despite their independent origin.7,9,48 The deep conservation of regulatory networks using a limited set of “tool kit” genes is illustrated by the success of candidate gene approaches in finding genetic and developmental bases for phenotypic evolution. The radiation of Galapagos finches involves modular changes between beak shapes along their three cardinal axes (depth, width, and length). Following previously established roles of bone morphogenetic protein (BMP) in craniofacial development, Abzhanov

et al.49 determined that the deeper, wider beaks of Galapagos ground finches correlated with increased expression of Bmp4. Also, overexpression of Bmp4 in chicken embryos was sufficient to form a deeper, wider beak.49 Modification of Bmp4 expression also correlates with differences in the width of beaks in chickens and duck bills, illustrating the reuse of a common mechanism underlying modular evolution of beak depth and width, independent of length.50,51 Organismal modularity is pervasive and emerges from the relationships of hierarchical and nested morphogenetic cellular fields organized by the interaction of gene regulatory networks.52 For example, multiple Hox genes are expressed in highly conserved overlapping domains during development of the amniote limb.53 As transcription factors, Hox proteins regulate the expression of batteries of downstream genes that specify cellular behaviors and, ultimately, skeletal growth.54 Such modular relationships appear to influence evolution of the hominoid forelimb, as demonstrated by the co-evolution of distal forearm and finger proportions that develop within a common Hoxd11 expression territory.53 Similarly, conserved patterns of genetic correlations in both mice and Old World monkeys distinguish incisor and molar size variation, and correspond to known patterns of dental homeobox gene expression.55,56 These conserved modular relationships reflect a developmental process that simultaneously generates highly integrated organisms, while retaining the capacity to shape phenotypes with remarkable specificity.57–59 Recently, the literature regarding the genetic basis for evolutionary change has blossomed as a result of the increasing application of comparative genomic analyses to natural populations. Numerous cases of repeated evolution using similar genetic and developmental mechanisms have been uncovered.60 Such examples allow circumvention of evolutionary contingency to recognize generalizations regarding the adaptive, genetic, and developmental foundations of morphological traits.48,61

Recent surveys of published literature designed to account for ascertainment bias have identified more than 100 genes that underlie repeated evolution of similar phenotypes60; they also have revealed that 32% of genetic mapping studies and 55% of candidate gene studies found gene reuse in repeated evolution.43 One particularly revealing model concerning regulatory phenotypic evolution is the modification of skeletal traits in three-spine stickleback fish (Gasterosteus aculeatus). This species has undergone recent radiations from the northern oceans into postglacial coastal freshwater lakes and streams. These isolated populations are remarkable for their parallel evolution of numerous morphological, physiological, and behavioral adaptations to novel environmental conditions (Fig. 2).62 Despite these differences, marine and freshwater sticklebacks can still interbreed, making them excellent models to uncover genetic bases for morphological evolution.63–65 Crosses between the marine and the freshwater fish produce hybrids that, after undergoing genetic recombination in the F2 generation, can be used to link particular genomic regions or quantitative trait loci (QTL) with corresponding phenotypic variation. Given the high degree of conservation in gene regulatory networks across large phylogenetic distances, investigations into the genetic and developmental basis for phenotypic change in various species can be directly applicable to vertebrates, including primates. For example, the stickleback studies described reveal that regulation of the Kit ligand (Kitlg) gene plays a similar role in pigmentation evolution in both the ventral flank and gills of sticklebacks and human skin.65 Such results run counter to the traditional models that phenotypic distance necessitates strong genetic divergence. Instead, shared properties of the underlying genetic and developmental architecture can influence the prevalent mechanisms available to evolution.42 These include the topology of gene regulatory networks, reduction of pleiotropic constraints, co-option of previously existing regulatory motifs,

194 Reno

ARTICLE

Figure 2. Adaptive radiation of threespine sticklebacks. A) Marine sticklebacks (center) have prominent dorsal and pelvic spines (short arrow) and lateral plates (long arrow). Independent colonization of separate freshwater habitats resulted in parallel reduction of the spines, plates, and pigmentation, as well as other physiological and behavioral traits. Modified from Bell and Foster.62 B) Reduction and/or loss of pigmentation, pelvic spine, and lateral plates results from modification of surrounding tissue-specific enhancers (cisregulation) of the targeted gene. Line drawings used by permission from Michael A. Bell.

amounts of standing genetic variation, and differences in mutation rates across the genome.43

NODAL POINTS ON REGULATORY NETWORKS: PITX1 AND STICKLEBACK PELVIC FIN REDUCTION The basic structural unit of regulatory networks is the modulation of gene expression through the binding of transcription factors to cis-regulatory enhancers.6 Due to the nested and hierarchical interactions between networks, nodes are formed that concentrate the reception and propagation of genetic signals.42 The development of all limbs shares a number of key properties, including

proximal to distal outgrowth and the establishment of polarity along the three cardinal axes. Proper limb development is thus dependent on multiple highly conserved gene networks shared between the fore- and hind limbs or the limbs of distantly related species. However, the anatomical differences must also result from variation in the underlying regulatory networks. Pituitary homeobox gene 1 (Pitx1) is one of a few genes that confers hind limb identity in jawed vertebrates,66 making it a prime candidate for mediating hind limb-specific evolutionary change. Marine sticklebacks have a prominent pelvic spine for defense against ingestion by larger fish. However, in freshwater environments with low calcium ion concentration and lack-

ing predatory fish, this spine is often independently lost (Fig. 2B).62 Crossing marine and freshwater fish identified a large effect QTL mapping to the Pitx1 locus. While the freshwater coding sequence of Pitx1 remained unaltered, expression was specifically lost in the presumptive pelvic region, despite continued expression at other anatomical locations.63 The key locus was isolated to a 2.5 kilobase region that, while highly conserved in other fish species, was deleted in pelvicreduced sticklebacks. Transgenes with these deleted sequences rescue Pitx1 expression and spine formation in transgenic freshwater sticklebacks, identifying this loss as the causative mutation for pelvic reduction.46 Limb reduction is one of the more prominent examples of convergence in vertebrate evolution (in snakes, lizards, whales, and manatees). Given its repeated use in freshwater threespine sticklebacks, how broadly does modification of Pitx1 serve as a means for hind limb loss? Nine-spine sticklebacks (Pungitius pungitius) diverged from their three-spine relatives more than 10 Ma and also evolved pelvic-complete and pelvicreduced populations. Hybrids between pelvic-reduced three-spine and nine-spine sticklebacks also fail to form pelvic spines, indicating that Pitx1 is inactivated during pelvic development in both genera.67 In mice, Pitx1 inactivation results in bilateral asymmetry of the reduced hind limbs due to partial rescue of the closely related Pitx2, which is preferentially expressed on the left side of vertebrate embryos.66 As expected, many pelvic-reduced populations of three-spine and nine-spine sticklebacks show a similar pattern of leftbiased pelvic asymmetry. Intriguingly, the rudimentary pelves of Florida manatees (Trichechus manatus latirostris) are also left biased, suggesting a similar Pitx1-mediated mechanism for hind limb reduction in mammals.67 These examples illustrate that across all taxonomic levels, from populations to phyla, similar points of a regulatory pathway can be independently, and potentially preferentially, targeted to produce convergent phenotypic outcomes.

ARTICLE

Genetic and Developmental Basis for Parallel Evolution and its Significance for Hominoid Evolution 195

REGULATORY SPECIFICITY AND THE LIMITATION OF DELETERIOUS EFFECTS While the integration of gene regulatory networks around nodal genes can create paths of least resistance that favor reuse of particular genetic loci, regulatory specificity can minimize pleiotropic effects to make these genes common evolutionary targets. The high conservation and promiscuous interactions across multiple regulatory networks of Pitx1 with roles in hind limb, craniofacial, and pituitary development would intuitively seem to make it a poor target for mediating evolutionary changes.60 However, this ability to be reused in multiple developmental contexts requires highly specific regulation, typically through intricate cis-regulatory complexes that can be co-opted by evolution.6,42 In addition to losing the pelvic spine, freshwater sticklebacks commonly reduce the size and number of armor plates lining their flanks (Fig. 2A). Using similar crosses, Colosimo et al.64 found that regulation of the Ectodysplasin (Eda) gene underlies a substantial majority of armor plate variation. Like Pitx1, inactivation of Eda has pleiotropic effects on multiple skin appendages, such as teeth, hair, and sweat glands in humans and mice, and operates within a larger regulatory network. However, investigation of six of the other genes in the network showed that only the Eda receptor (Edar) gene had an additional minor effect on plate reduction. When compared to these other genes, Eda itself was unique in having a large surrounding cis-regulatory region.68 Similar regulatory specificity also appears to be important in the targeted Kitlg-mediated changes in gill and ventral flank pigmentation, as other anatomical loci fail to show similar reductions in pigmentation and Kitlg expression.65 Thus, the shared capacity of Pitx1, Eda, and Kitlg for regulatory specificity is a key factor that facilitates their ability to produce evolutionarily important phenotypic variation and enables such genes to be frequent targets for evolutionary change.

Figure 3. Evolution of Drosophila wing spot by enhancer co-option. Ancestral wing pigmentation occured through separate enhancers of the yellow gene, which drives expression in the wing (pale gray) and veins (dark gray). Parallel evolution of wing spots occurred through modifications of either enhancer (medium gray). Based on figure from Prud’homme and colleagues.70

CO-OPTION OF EXISTING CISREGULATORY MODULES Many of the previous regulatory examples involve enhancer loss or down-regulation. Such mechanisms are likely to be important for human evolution, as many derived characters (for example, body hair, canine size, and muscle mass) involve phenotypic reduction. Similarly, the loss of penile spines and whiskers has been linked to the deletion of an enhancer of the androgen receptor gene.69 Growth can also be produced by regulatory deletions of gene repressors or loss of activators promoting cell death.69 However, phenotypic evolution also requires the generation of novel regulatory motifs. Theoretically, regulatory novelty should be less likely to result in repeated evolution of similar genetic mechanisms, since it would be unlikely that a multicomponent enhancer complex could evolve multiple times independently. The complex coloration patterns of fruit flies (Drosophila) provide insight regarding the evolution of regulatory novelty. The gain of male-specific wing spots has independently occurred Drosophila at least twice (Fig. 3). The yellow gene codes for an enzyme crucial for making black pigment which, in unspotted flies such as D. melanogaster, is uniformly expressed across the faintly shaded wing as a result of the action of a general wing enhancer. In spotted flies such as D. biarmipes, yellow is expressed at high levels at the anterior distal mar-

gin where the spot forms. This new expression pattern results from a novel regulatory domain immediately adjacent to the wing enhancer. Thus, D. biarmipes co-opted the wing enhancer by adding novel activator and repressor domains to focus yellow expression in a new pattern.57 Interestingly, the other instance of wing spot evolution also relies on a regulatory change of yellow, but in D. tristis the wing enhancer is unmodified. Instead, novel regulatory domains have been introduced near an enhancer within the first intron of yellow that drives expression within the wing veins. Both instances involve co-opting a previously existing enhancer to generate regulatory novelty. As anatomical novelty often occurs through the modification of preexisting structures (fin to leg to wing). So too does genetic and developmental innovation occur by modifying preexisting regulatory architecture. Such co-option is dramatically simpler than assembling the multitude of binding sites de novo.70 It also illustrates another example of mechanistic evolution taking the “path of least resistance” in generating phenotypic novelty.45,71

PARALLEL ADAPTATION VIA STANDING GENETIC VARIATION AND VARIATION IN MUTATION RATES While modification of existing functional regulatory motifs can facilitate repeated evolution, an even simpler

196 Reno

ARTICLE

Figure 4. The role of standing variation in parallel evolution of sticklebacks. Left: Lateral plate reduction occurred through use of a common low-frequency regulatory allele (gray triangle) of the Eda locus found in the ancestral marine population. Selection in separate freshwater populations drove the allele to fixation in parallel. Right: Pelvic spine reduction occurred through loss of a Pitx1 enhancer (triangle) due to novel deletions facilitated by the intrinsic fragility of the ancestral locus (gray).

mechanism is to use variation already present within the population. Haldane72 hypothesized that commonly inherited genetic variants can be as important as exposure to shared selective pressures in producing parallel evolution. Sticklebacks provide an excellent test case because specific genetic variants can be traced through their evolutionary radiation to determine if parallel adaptation relied on the selection of previously existing alleles or novel mutations. In their analysis of stickleback plate reduction, Colosimo et al.64 found that multiple populations from both Pacific and Atlantic Ocean tributaries share the same Eda allele. The degree of divergence of this allele suggested that it separated from the ancestral full-plated allele at least 2 Ma. This might suggest an ancient common divergence of the plate-reduced populations from the marine population. However, other gene sequences indicate that low-plated freshwater morphs tend to be more closely related to nearby marine populations than they are to other freshwater sticklebacks. This contradiction is explained by the presence of the low-plated Eda allele at low frequency (