* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Tiie Need for Bioinformatics in Evo-Devo
Survey
Document related concepts
Public health genomics wikipedia , lookup
History of genetic engineering wikipedia , lookup
Quantitative trait locus wikipedia , lookup
Biology and consumer behaviour wikipedia , lookup
Artificial gene synthesis wikipedia , lookup
Gene expression profiling wikipedia , lookup
Dual inheritance theory wikipedia , lookup
Adaptive evolution in the human genome wikipedia , lookup
Designer baby wikipedia , lookup
Nutriepigenomics wikipedia , lookup
Genome evolution wikipedia , lookup
Genome (book) wikipedia , lookup
Gene expression programming wikipedia , lookup
Population genetics wikipedia , lookup
Transcript
Articles .. .., Integrating Evolution and Deveiopment: Tiie Need for Bioinformatics in Evo-Devo PAULA M. MABEE This article is an overview of concepts relating to the integration ofthe genotype and phenotype. One ofthe major goals of evolutionary developmental biology, or evo-devo, is to understand the transformation of morphology in evolution. This goal can be accomplished by synthesizing the data pertaining to gene regulatory networks and making use ofthe increasingly comprehensive knowledge of phylogenetic relationships and associated phenotypes. I give several examples of recent success in connecting these different biological levels. These examples help illuminate the "black box" between genotype and phenotype and illustrate a few ofthe technical and bioinformatic challenges ahead. The key concept of modularity unites genetic, developmental, and evolutionary approaches, because modules are the units of evolution. Primitive and derived network modules interact in development to create the phenotype. To obtain a systems-level understanding of evolutionary phenotypic change, bioinformatics approaches involving ontologies need to be applied, and new methods of visualization need to be developed. Keywords: bioinformatics, genetics, modularity, evolution, development T WO of the most exciting and difficuit chaiienges in biology are to understand how morphology, or anatomical form, is created in development and to determine how such developmental morphology is transformed in evolution. These questions unite the emergent field of "evo-devo," the interdisciplinary study of evolution and development, which has become one of the most rapidly moving fields in biology. A number of excellent reviews and books document the history, knowledge base, and syntheses in this field (Gilbert 2003, Minelli 2003, Holland 2004, Carroll 2005). As evo-devo scientists sharpen their focus on comparing developmental morphology, it becomes more obvious that the comparisons must be "read" within a phylogenetic framework (Baum et al. 2005). In fact, tracing the pattern of evolutionary transformation of the developing phenotype against the evolutionary tree of life defines a significant problem set for evo-devo (Wagner and Larsson 2003). In addition, the impact of various developmental mechanisms (Arthur 2004) can be assessed against the pattern defined by the tree of life. Explaining the pattern of evolutionary change in morphology in an integrative fashion requires a systems approach—synthesizing knowledge from various biological levels encompassed in development (e.g., DNA sequence, regulatory networks, cell-cell and tissue interactions), evolution (e.g., phylogenetic relationships, population level), and ecology. www.biosciencemag.org The critical need to relate genotypic to phenotypic data, and the difficulty of that endeavor, is echoed in many recent papers (Hall 2003, Smith 2003, Irish and Benfey 2004, Kuratani 2004). Despite the battery of tools being used to address the question of how an organism is built (i.e., how the developing phenotype, the observable expression ofthe genotype, is created in development), the connection between genotype and phenotype remains poorly understood. Large-scale genome sequencing has resulted in a "parts catalog" of complete information about what is in the genome, and yet the "user's manual," the genetic and molecular basis and rules by which morphology is assembled, is poorly understood (Kitano 2002). It is a gargantuan leap to go fiom genotype to phenotype using tools from developmental genetics. Progress is being made by chipping away at genotype-phenotype relationships in the context of developmental genetics of model species, and another jump, aided by tools of phylogenetics, has been taken to connect development to the directionality and timing of evolutionary change. But it is the tools of computation, of bioinformatics and ontologies (i.e., structured Paula M. Mabee (e-mail: [email protected]) is an evo-devo biologist at the University of South Dakota, Vermillion, SD 57069. She studies skeletal development and evolution in fishes. © 2006 American Institute of Biological Sciences. April 2006 / Vol. 56 No. 4 • BioScience 301 Articles vocabularies that make databases interoperable), that are now required to merge the biological levels of knowledge. With improved methods ofdata mining and analysis, evolutionary questions can be answered with the desired mechanistic depth. The purpose of this article is to explain a key concept— namely, modularity—that spans the gap from genotype to phenotype; to describe a few of the recent successes in connecting genotype and phenotype; and to champion the need for bioinformatics tools to facilitate more rapid and higherlevel progress. Modularity Modules are a piece of common conceptual ground for evolutionary and developmental biologists because they span and unite biological levels. Biologists working at different levels, however, have different conceptions of modules: Geneticists conceive of modularity in terms of gene organization and function, developmental biologists consider the modularity of gene regulatory networks, and many evolutionary biologists think in terms of phenotypic "characters." Modularity at the morphological level is somewhat intuitive: An organism is composed of discrete parts, and these parts are semiindependent in development (figure 1) and evolution. Morphological modules as entities are assumed to have a developmental genetic integrity, and as such they are individuated (Wagner 1989, Geeta 2003). orphology Developmental genes Figure 1. The black box between genotype and phenotype is filled here with the modules (represented by colored cylinders) that function in development to create morphology (more generally, the phenotype). Blue and red indicate the evolutionary status ofthe modules, blue for primitive and red for derived. Primitive and derived modules interact here to form the typical mosaic of primitive and derived features (characters or modules) that comprise an organism. The limb bud is an excellent example of modularity at both morphological and genetic levels. The paired limbs have a clear morphological identity and a limited degree of connectivity to the rest of the body at the morphological level. At a genetic level, the regulatory cascades that control limb development (Tabin et al. 1999) constitute a patterning module that is conserved between forelimbs and hindlimbs. Other examples of morphological modules include 302 BioScience • April 2006 / Vol. 56 No. 4 actinopterygian fish fins (Mabee et al. 2002), the mouse mandible (morphometrically partitioned into submodules; Klingenberg et al. 2003), and bird feathers (hierarchically nested into metamodules in development; Prum and Dyck 2003). Morphological modules, inasmuch as they represent units of homology, may be considered equivalent to the morphological characters of phylogenetic systematics. Evolutionary characters are the homologous structures that vary across species, and multiple characters are analyzed together to generate phylogenies. Although systematists do not have "modularity" as a working concept and evo-devo biologists do not think in terms of "characters," it is appropriate to consider these terms equivalent, because both refer to an individuated entity that is genetically determined, homologous, and maintained across taxa. In working terms, because so little is understood about the modules underlying the phenotype, systematists cannot use them as characters to infer phylogeny at this time. As details about how modules are integrated are better understood, these data can be added to phylogenetic analyses. Underlying modularity at the morphological level—and responsible for it—are integrated developmental networks (see Schlosser and Wagner 2004). Modules have been defined, in fact, as "networks of interacting elements behaving as relatively independent units of development or function" (Schlosser and Thieffry 2000). At the level of gene interactions, modules may range from interactions among a few genes to large-scale genetic networks, such as the complex regulatory gene network for endomesoderm specification in the sea urchin (figure 2; Davidson et al. 2002). This complex regulatory network underlies the specification of a single cell type (endomesoderm). Glearly, the development of even a single aspect of morphology, which is controlled by many connected networks of regulatory genes, must be staggeringly complex. Interestingly, though, the complexity of body plan development is said to be matched by that of some human-engineered systems (Gsete and Doyle 2002). A Boeing 777 generates a volume of information similar to that in the human genome every minute: 150,000 modules, organized through protocols into networks (Csete and Doyle 2002). Genetic modules, or "cassettes" of genes used to effect a common Sanction (Rudel and Sommer 2003), interact with one another to ultimately specify phenotypes. Thus, modules that themselves interact, that vary semi-independently, and that together specify a modular morphology (figure 1) are an important link between genotype and phenotype. One might say that such molecular modules, the gene interactive networks and their connecting protocols or rules (Ravasz et al. 2002), are the building blocks of development and evolution. The network properties of development are increasingly discussed in the literature of development and of bioinformatics (Ravasz et al. 2002, Qin et al. 2003, Anholt 2004, Cork and Purugganan 2004). Development of the body plan is controlled by large networks of regulatory genes, and it essentially flows from the correct temporal and spatial transcription of www.biosciencemag.org Articles Endomesoderm Specification to 30 Hours UCl-HGSK-3 ,, , Nucl. Jan 18, 2006 frizzled I Mat Acadherin -HMat SoxBll iMat Otx nP-TCF ECNS 9- o — unkn mes/end rep SU(H) < •—j-HO?-GSK-3i-v/ nP-TCF O unkn vegetal activ SoxBl Krl Bml/2/4 Eve Sm50 Mspl30 Msp-L L j i r SuTx CAPK Dpt _^ Sm30 Acadherin Ficolin "CyP FvMol,2,3 Decorin Ubiq=ubiquitous: Mat = maternal; activ = activator; rep = repressor; unkn = unknown; Nucl. = nuclearization; x - P-catenin source; np-TCF = nuclearized b-3-catenin-Tcfl; ES = early signal; ECNS = early cytoplasmic nuclearization system; Zyg. N. = zygotic Notch Pks Hoxll/13b OrCT Apobec GelsoUn Apobec Kakapo Gelsolin Endol6 Copyright ® 2001-2006 Hamid Bolouri and Eric Davidson Figure 2. Regulatory gene network for endomesoderm specification (reproduced with permission from Eric Davidson; http://sugp.caltech.edu/endomes/j. Forty genes are currently known to be involved in specifying this single cell type (Davidson et al. 2002). Each short horizontal Une from which a bent arrow extends to indicate transcription represents the as-regulatory element that is responsible for expression ofthe gene named in the domain shown. The architecture ofthe network is based on perturbation and expression data, on data from ds-regulatory analyses for several genes, and on other experiments referenced at the Web site cited above. such genes, which require the correct use of the instructions encoded in ds-regulatory DNA (Stern 2000, Revilla-iDomingo and Davidson 2003). Evolution of phenotype occurs through changes in these networks. Because gene regulatory networks underlie the processes of both development and evolution, unraveling their architecture in appropriately chosen species will be the key to understanding how genomes control development and how they evolve (Revillai-Domingo and Davidson 2003). The discovery of gene transcription modules from the rapidly accumulating gene expression data has been aided recently by new analytical methods in bioinformatics (Lee TI et al. 2002, Lee I et al. 2004, www. biosciencemag. org Kloster et al. 2005). Uncovering the modular architecture of the transcriptional networks that comprise developmental programs (Rast 2003) and understanding the developmental and evolutionary ramifications of their topology are critical new areas in evo-devo. Underlying the phenotype and the genetic networks is another level of modularity at the genetic regulatory level. The modularity of these genetic regulators or "switches" (Carroll 2005) is perhaps underappreciated by those working at higher levels. The ds-regulatory DNA of many genes is organized into independent modules that direct or repress transcription in specific tissues at particular times in development. The modApril 20061 Vol. 56 No. 4 • BioScience 303 Articles ules themselves are constructed from multiple binding sites for individual transcription factors (Arnone and Davidson 1997). Modular regulatory switches are used for building modular animals (Carroll 2005), and switches are thus the critical connection to evolution: Such regulation permits evolutionary change to occur in one part ofa structure, independent of other parts. Ultimately, modularity at the genetic level of regulation is the secret to modularity at the morphological level, and modularity is the key to building complexity (Carroll 2005), as evidenced by the overwhelming number of evolutionary modifications that consist of specializing, tinkering with, and modifying the number of existing modular parts. How do modules evolve? Modularity has been argued to be the most critical aspect of order in living organisms and their ontogenies, and the attribute that most strongly facilitates evolution (Raff 1996). Modularity brings together development and evolution in several respects (Schlosser and Wagner 2004). First, we know that in development, the modularity of gene regulation is translated into semi-independent genetic regulatory networks, and these specify the discrete features ofthe phenotype. Until the advent of molecular systematics, such features, or characters of the phenotype, were almost exclusively used by systematists to infer phylogenetic relationships. The inferential methods of phylogenetic systematics assume character independence, and the likelihood of inferring false relationships is increased by correlated features. Formulation ofa character thus might be seen as equivalent to delineating a phenotypic module that has been maintained and modified by evolution. Essentially, developmental modularity underlies the evolutionary modularity of systematic characters. Because at some level the entire phenotype is correlated, the empirical problem becomes that of determining levels of correlation. With enough developmental data on the underlying networks and appropriate informatics, character dependencies can be evaluated, and these can be built into the methods of analysis. This is a nontrivial undertaking that relies on heavy use of phylogenetic and developmental model databases, interoperability, tools, and other programmatic infrastructure. Phylogenetic information about how individual species are related is necessary for determining which modules are ancestral and which are derived (Maddison and Maddison 1992), for learning whether a new morphology is due to the gain or loss ofa module, and for finding out whether a module has been derived once or many times. Development can be represented as a mosaic of primitive and derived developmental modules interacting to form primitive or derived morphological features. Phylogenetic hypotheses are thus critical in interpreting the patterns of evolution of modules—at any hierarchical level. Without a phylogenetic fi-amework, even the most basic questions cannot be framed in a meaningful way. Applying phylogenetic methods to gene networks and other modules will yield answers to vital questions in thefieldof evo-devo (box 1). Fortunately, a great deal 304 BioScience • April 2006 / Vol. 56 No. 4 of progress has been made toward understanding the phylogenetic relationships of major animal and plant groups, as well as the relationships within many smaller clades. Ontologies: Connecting evolutionaiy and developmental anatomy The phenotype figures prominently in the work of developmental biologists, who place major emphasis on connecting gene expression and function to phenotypic effects. Mutagenesis (i.e., isolating single gene mutants that have discrete effects on body pattern) has been a primary approach in understanding normal development and in identifying the body-patterning and "toolkit" genes of evolution (Greenspan 2001, Carroll 2005). The control gene eyeless, for example, was identifiedfi-omthe mutants v^thout eyes in Drosophila. Other approaches include the removal or misexpression of a gene product during development, followed by examination of the effects on the resultant phenotype (Stern 2000). The data resulting from these approaches are voluminous, and discovering the complex gene networks that connect sequence to phenotype will require long-term bioinformatics efforts and new tools. Ontologies, or controlled vocabularies, are one approach that is used to connect databases. Ontologies formally represent hierarchical relationships between defined biological concepts, such that the vocabularies can be used by both humans and computers to exchange and explore information (HoUoway 2002, Blake 2004). Controlled vocabularies in the form of words or phrases are used in ontologies to bring different meanings or synonyms together under a single clear term; free text is avoided, and the data can be computed. Bio-ontologies clarify scientific discussions by providing a shared vocabulary to communicate results and to explore data. They also make data exploration, inference, and data mining computationally possible. The development of ontologies was instigated by molecular biologists to promote collaborations between the medical informatics and bioinformatics ontology communities (Holloway 2002). The Gene Ontology (GO) project (Harris et al. 2004) is perhaps the most widely known in the molecular world, consisting of structured, controlled vocabularies (> 17,500 terms) and classifications that cover several domains of molecular and cellular biology {www.geneontology.org). It consists of three ontologies at different hierarchical levels (molecular function, biological process, and cellular component attributes of gene products). These ontologies are orthologous (relatable) to one another and thus interoperable (i.e., they can work together). Use ofthe GO is widespread, and many biological resources, such as UniProt (Apweiler et al. 2004) and Protein Data Bank, annotate their data in GO terms (Wolstencroft et al. 2005). The result is that by searching for all proteins associated with a particular GO term, data fi'om many different sources can be retrieved automatically and efficiently. A single GO term has associated with it all known synonyms, and thus all related terms are searched simultaneously. www. biosciencemag. org Articles Box 1 . Evolution and development: Developmental network-, phenotype-, and systems-level questions. The pattern of phylogenetic history is critical for tests, predictions, and investigations in evolutionary developmental biology, or evo-devo. Patterns of character evolution must be synthesized across the full spectrum of biological levels to answer the major questions below and those subsumed within them. Because of the complexity and volume of these data, new methods of visualization and analysis will be required to fuUy grasp the patterns of such evolutionary changes. Developmental network-level questions: • What is the pattern of evolution of gene regulatory network modules? • What are the specific changes that have occurred in a particular gene network as it is transformed in evolution, and exactly where have these have occurred? • Where exactly does the remodeling of developmental pathways occur? (cis-acting elements? Protein function?) • What is the frequency and nature of parallel co-option of genetic networks? (Co-option appears to occur frequently in evolution, as evidenced by the parallel independent co-option of Pax-6, DU, and tinman to pattern eyes, limbs, and hearts, respectively, in insects and vertebrates; Jockusch and Ober 2004.) • Developmental pathways may be redeployed in other tissues (heterotopy) or at other developmental times (heterochrony), or both. How often and under what circumstances does this happen? • What are the specific bases for constraint in gene networks? • What are the network properties that promote resilience or enhance evolvahility? (The interactivity among genes indicates that there might be considerable flexibility in the capacity ofthe genome to respond to diverse conditions; Greenspan 2001.) • How is constraint at the morphological level related to that at the network level? • Which sites in a gene network are most conserved or constrained, and which are most labile? • What types of changes are most common? (Cork and Purugganan [2004] predict that genes functioning early in a genetic pathway are subject to stronger stabilizing selection than downstream loci, since mutations in these genes are likely to have greater pleiotropic effects and affect all downstream phenotypes.) Phenotype-level questions: • What is the developmental basis for the phenotypic characters (modules) of evolution? • Is there a "signature" modular composition to the morphological characters (modules) of systematics? • What are the probabilities of different types of changes within modules? • Are new linkages between submodules made to produce modular characters? • Are sets of modules (at different or similar biological levels) correlated evolutionarily? (By mapping multiple modules simultaneously, the patterns of character association can tested for correlation. Levels of correlation can be quantified.) Systems-level questions (requiring information from ecology and environment as well): • What accounts for major novelties? • Generally, what accounts for homology? • Can we generalize about homoplasy? Why are some characters susceptible to parallel evolution? • Why are there trends or "metapatterns" in evolution? More recently, researchers have developed phenotypic ontologies (Bard 2005) within model organism communities (zebrafish, mouse, fly), motivated by the desire to answer the developmental questions of how phenotype is generated from genes. Phenotypic effects of different treatments for a model species (e.g., mouse) can be compared and analyzed together with genetic sequence data to piece together the www. biosciencemag.org underlying genetic architecture. There are now about 15 anatomical ontologies, many of which are linked to organism databases (Bard 2005) (e.g., Zebrafish Anatomical Ontology, http://zftn.org; Drosophila database [Flybase], http:// flybase.bio.indiana.edu; and Edinburgh Mouse Atlas Project, http://genex.hgu.mrc.ac.uk). These bio-ontologies ofthe phenotype (Gkoutos et al. 2004), or "anatomies" (Bard 2005), have April 2006 / Vol. 56 No. 4 • BioScience 305 Articles sprung up without significant input from comparative evolutionary morphologists or systematists. To address many evolutionary questions (box 1), species' phenotypes must be compared. Unlike the phenotypes contained in developmental databases, the phenotypes of evolutionary biology are not the result of short-term laboratory perturbation or mutation, but rather the result of millions of years of natural mutation. Phenotypic variation across species is much broader than that observed within a mutated model species. To understand the evolutionary changes in the developmental genetic and epigenetic inputs that are involved in building and individuating species, phenotypic ontologies for individual species must be connected to each other, as well as to ontologies at lower biological levels (e.g., GO). Evolutionary comparisons operate at a high tier in systems biology, namely, at the level of continuity (and modification) of the phenotype across the tree of life. Similarity among phenotypes is due to the continuity of inherited information (i.e., homology; Van Valen 1982, Roth 1984). Homology is an evolutionary concept; homologous features are similar because they are inherited from a common ancestor—that is, they are similar because of their common genealogy. Homology, a common ground for developmental and evolutionary biologists, is central to all comparative biology (Bock and Cardew 1999). It is central in phylogeny reconstruction, because characters have to be homologues to achieve a correct hypothesis of evolution using phylogenetic methods. The invisible evolutionary threads connecting organisms are those of continuity—of homology—and thus ultimately ontologies must be connected through homologous features or modules. The developmental genetic basis of convergent morphological change There is some evidence—and it is becoming almost a central tenet of evo-devo—that the evolution of regulatory elements is the primary source of phenotypic change during the course of evolution. Regulatory elements, including cis elements, enhancers, promoters, and trans regulatory factors, in combination with a host of other proteins that are part ofthe transcriptional machinery (Brivanlou and Darnell 2002), activate or repress gene transcription. Recent molecular advances in several model systems have demonstrated the correlation of morphological variation with variation in regulatory regions. These regulatory elements appear to be highlyflexible,and evolutionary changes in their roles and expression domains seem fairly common. Regulatory mutations in key developmental control genes appear to provide a general mechanism for selectively altering expression in specific structures while preserving expression at other sites required for viability (Carroll 2000, Stern 2000, Tautz 2000). Several recent studies of evolutionary changes in the regulatory elements underlying phenotypic evolution (Gompel and Carroll 2003, Sucena et al. 2003, Shapiro et al. 2004) provide examples of making the much-needed connection from genot)^e to phenotype. Interestingly, these studies come from examining the development of parallel or convergent similarity within 306 BioScience • April 2006 / Vol. 56 No. 4 small clades, as opposed to the focus on novel morphologies (e.g., the pentameral echinoderm body plan from a bilateral ancestor; Arenas-Mena et al. 2000, Peterson et al. 2000, Popodi and Raff 2001) that has characterized much ofthe research in evo-devo. Convergence describes cases in which similar derived morphologies are produced from different ancestral morphologies, though the terms convergence and parallelism are used interchangeably in several subdisciplines of biology (Wiens et al. 2003), including evo-devo. Regulatory changes are involved in the convergent loss of features (Sucena et al. 2003, Shapiro et al. 2004). The evolutionary modification of "just a few developmental hotspots" (Richardson and Brakefield 2003) may underlie parallel evolutionary changes at the phenotypic level (Richardson and Brakefield 2003, Sucena et al. 2003). The examples below demonstrate such modifications, and they provide an opportunity to describe how advances in bioinformatics can significantly enhance understanding of these processes. Morphological convergent reduction in pelvic fins of three- spine sticklebacks. Marine threespine sticklebacks, Gasterosteus aculeatus, have given rise to freshwater stickleback populations that have evolved complete or partial loss ofthe pelvic skeleton in less than 10,000 generations. Shapiro and colleagues (2004) identified a major chromosome region controlling loss of pelvic structures in a natural population of sticklebacks. They found an altered pattern of expression for the gene Pitxl in some tissues but not in others, consistent with a regulatory mutation that disrupts expression in both the prospective pelvic region and the caudal fin. Their data suggest that crs-acting regulatory mutations in Pitxl are a major cause of pelvic reduction in this rapidly evolving system; the coding regions of Pitxl are unchanged between populations with and without pelvic fins. In this and many other cases in which evolutionary changes have been traced to regulatory alterations in major developmental control genes, the actual DNA sequences responsible for tissuespecific expression differences are unknown. Regulatory mutations are much more difficult to identify at the molecular level, because the regulatory regions can be located far from the gene of interest. The upshot is that it appears that similar regulatory mutations in key developmental control genes are responsible for the parallel morphological changes in fins during stickleback evolution. With the appropriate bioinformatics resources in place, researchers could quickly find all of the examples of evolutionary loss of the pelvic fins in fish evolution. They could quickly access and examine images of pelvic fins in the most recent ancestors of thefishesthat lost them, and they could see stages (when they exist) in the gradual loss of such structures. The genetic underpinnings of pelvic fin loss in sticklebacks, which is known, could be compared with that in examples of natural mutation, and in fact the stickleback genes and regulators could be used as candidates to examine in otherfishes.By connecting the breadth of parallel evoluwww. biosciencemag. org Articles tionary phenotypes with the developmental genetic depth from a well-understood system, important, overarching evodevo questions (box 1) can be answered. Morphological convergent loss of larval hairs in Drosophila. The pattern of trichomes (larval "hairs") on the dorsal and lateral cuticle varies among the larvae of various Drosophila species; "thin" trichomes have been lost to varying degrees in four species in the D. virilis clade. When the presence or absence of thin trichomes is mapped (Sucena et al. 2003) on a phylogeny for the group (Spicer and Bell 2002), at least three evolutionarily independent losses of trichomes are required. It had been previously determined that the transcription factor Svb (shavenbaby/ovo, svb/ovo) acts to switch cells between naked cuticle and the production of trichomes (Payre et al. 1999). Using in situ hybridization, Sucena and colleagues (2003) demonstrated that the expression of svb is strictly correlated with the pattern of trichomes on the cuticle along the anterior-posterior axis. In other words, in naked species (those lacking thin trichomes), svb transcription is absent and the corresponding cells differentiate naked cuticle. Combining this finding with data from intraspecific crosses, Sucena and colleagues (2003) concluded that the transcriptional enhancer that promotes svb expression in the domain that produces thin trichomes is turned off in the lineages that lack the trichomes. Regulatory changes in svb expression are involved in all cases of parallel loss in the D. virilis clade. Although determining exactly how the cis-regulatory region of svb has been altered will require identification and characterization ofthis genomic area in these species—which, as in the case ofthe stickleback, is technically difficult—it is clear that parallel changes in the regulation ofthe svb/ovo gene underlie all independent cases of morphological convergence. Like the stickleback fish example, this demonstrates that apparently identical developmental genetic mechanisms may underlie parallel evolutionary changes. Although evolutionary biologists have long recognized that identical developmental genetic mechanisms may not underlie homology (Roth 1988), the finding that identical developmental regulatory changes may, in fact, underlie parallelism underscores the need for phylogenetic information in determining homology (Sucena et al. 2003). If only the developmental genetic information, and not the phylogenetic information, were available for the D. virilis species group, the absence of trichomes at the first larval stage would be considered homologous, because changes in regulation ofthe same gene (svb) produce this pattern. Phylogeny, however, indicates that this loss must have happened independently, at least three times. Sucena and colleagues (2003) point out that the modular nature of svb regulatory regions permits alterations in part of the cuticular pattern without pleiotropic consequences (Stern 2000, Sucena and Stern 2000). Caution is advisable even when phylogenetic history and morphology indicate a common (homologous) developmental basis. There are now several examples of morphology, such as vulva position in nematodes (Sommer 2000, Rudel and Sommer 2003) and www. biosciencemag. org wing polyphenism in ants (Abouheif and Wray 2002), in which differently modified gene networks underlie apparent homology. As in the first example, with the appropriate bioinformatics resources in place, researchers could quickly query for data across insect taxa to find other examples of such phenotypic changes in evolution. They could promptly examine images of morphological features, and again, the well-known set of genes and regulators could be searched for parallel use at a broader level. Pigment and trichome patterns in flies. On a phylogeny for 13 Drosophila species, Gompel and Carroll (2003) mapped the pattern of melanie pigment and trichomes on the pupal abdominal segments against aspects oibab2 gene regulation. They inferred from phylogeny and character distribution that, ancestrally, Bab2 repressed pigment formation and also regulated trichome development in the abdominal epidermis. In the evolution ofthe group, males of several species evolved repression of bab in the posterior abdomen (allowing pigment to form). Convergence in pigment pattern was achieved through similar, independent regulatory changes in bab. Additionally, from the phylogeny they could see a spectrum of speciesspecific modulations of Bab2 expression. This showed them that bab controls pigment and trichomes independently, and that their evolution can be uncoupled. They suggested that changes in cis to bab2, in regulatory elements responding directly or indirectly fo one or more body plan regulators, were responsible for the diversification of Bab2 expression. Like the first two examples, this detailed study of fruit fiies adds potential developmental genetic depth to fhe multitude of parallel changes across insects. Ontologies would provide a link across these databases and enable comparative morphologists and developmental geneticists to exchange information. Many of the examples of convergent evolution in evodevo have involved the loss or reduction of a particular morphological feature within a small clade, and they show that parallel changes in genetic regulatory networks underlie the loss of morphological features such as eyes in cavefish, wings in ants, and pelvicfinsin sticklebackfishes(Carroll et al. 2001, Abouheif and Wray 2002, Jenner 2004, Yamamoto et al. 2004, Tian and Price 2005). Convergence in regulatory gene expression domains, which is probably more common than is generally acknowledged, can arise for several different reasons (Wray 2002). The ease with which a morphological feature may be lost is consistent with the evolutionary idea that the loss ofa morphological feature is more probable than gain (reviewed in Jenner 2004), which in systematics has been formalized as Dollo's law (Farris 1977). If these empirical developmental data are representative, they support the use of this assumption in phylogenetic inference. Conclusions Understanding the interplay between developmental and phylogenetic constraints in the evolution ofthe phenotype is April 20061 Vol 56 No. 4 • BioScience 307 Articles an overarching goal ofthe field of evo-devo. The examples of convergent evolution related above have shown that parallel changes in genetic regulatory networks underlie the loss of morphological features. Whether this will prove true across disparate phylogenetic levels will require additional empirical research—research at the genomic and phenotypic levels of multiple species—and informatics to facilitate data mining and analysis. Practically, shared ontologies for homologous parts, characters, or modules ofthe phenotype must be referenced, so thaf growing databases in genomics and evolution can connect and help address evolutionary questions (box 1). Currently, developmental biologists and evolutionary morphologists individuate or parse the phenotype differently (Geefa 2003), but bioinformatics efforts are under way to facilitate communication across disciplines. Better analytical tools and more powerfiil methods of analysis wiU make it possible to understand the staggering level of complexity underlying phenotypes and the constraints of the developmental genetic architecture thaf underlie phenotypic modules. Bioinformatics methods that facilitate visualization of multidimensional genotype and phenotype data simultaneously (Tao et al. 2005) are crucial to biological research in this postgenomic era. Acknowledgments I thank Matthew Greenstone for his encouragement in developing this overview, and Patricia Crotwell and anonymous reviewers for their suggestions. References cited Abouheif E, Wray GA. 2002. Evolution of the gene network underlying wing polyphenism in ants. Science 297:249-252. Anholt RR. 2004. Genetic modules and networks for behavior: Lessons from Drosophila. Bioessays 26: 1299-1306. Apweiler R, et al. 2004. UniProt: The Universal Protein knowledgebase. Nucleic Acids Research 32: D115-D119. Arenas-Mena C, Gameron AR, Davidson EH. 2000. Spatial expression of Hox cluster genes in the ontogeny of a sea urchin. Development 127: 4631^643. Arnone MI, Davidson EH. 1997. The hardwiring of development: Organization and fiinction of genomic regulatory systems. Development 124: 1851-1864. Arthur W. 2004. Biased Embryos and Evolution. Gambridge (United Kingdom): Gambridge University Press. Bard JBL. 2005. Anatomies: The intersection of anatomy and bioinformatics. Journal of Anatomy 206: t-16. Baum DA, Smith SD, Donovan SSS. 2005. The Tree-Thinking Ghallenge. Science 310: 979-980. Blake ). 2004. Bio-ontologies—fast and furious. Nature Biotechnology 22: 773-774. Bock GR, Gardew G, eds. 1999. Homology. New York: Wiley. Brivanlou AH, Darnell JE. 2002. Signal transduction and the control of gene expression. Science 295:813-818. Garroll SB. 2000. Endless forms: The evolution of gene regulation and morphological diversity. Gell 101: 577-580. . 2005. Endless Forms Most Beautiful: The New Science of Evo Devo and the Making of the Animal Kingdom. New York: Norton. Garroll SB, Grenier JK, Weatherbee SD. 2001. From DNA to Diversity: Molecular Genetics and the Evolution of Animal Design. London: Blackwell Science. Gork JM, Purugganan MD. 2004. The evolution of molecular genetic pathways and networks. Bioessays 26:479-484. 308 BioScience • April 2006 / Vol. 56 No. 4 Gsete ME, Doyle JG. 2002. Reverse engineering of biological complexity. Science 295:1664-1669. Davidson EH, et al. 2002. A genomic regulatory network for development. Science 295: 1669-1678. Farris JS. 1977. Phylogenetic analysis under Dollo's Law. Systematic Zoology 26: 77-88. Geeta R. 2003. Structure trees and species trees: What they say about morphological development and evolution. Evolution and Development 5:609-^21. Gilbert SE 2003. The morphogenesis of evolutionary developmental biology. International Journal of Developmental Biology 47:467-477. Gkoutos GV, Green ECJ, Mallon A-M, Blake A, Greenaway S, Hancock JM, Davidson D. 2004. Ontologies for the description of mouse phenotypes. Gomparative and Functional Genomics 5: 545-551. Gompel N, Garroll SB. 2003. Genetic mechanisms and constraints governing the evolution of correlated traits in drosophilid flies. Nature 424: 931-935. Greenspan RJ. 2001. The flexible genome. Nature Reviews Genetics 2: 383-387. Hall BK. 2003. Evo-Devo: Evolutionary developmental mechanisms. International Journal of Developmental Biology 47: 491-495. Harris MA, et al. 2004. The Gene Ontology (GO) database and informatics resource. Nucleic Acids Research 32: D258-D261. Holland PWH. 2004. The fall and rise of evolutionary developmental biology. Pages 261-275 in Williams DM, Forey PL, eds. Milestones in Systematics. Boca Raton (FL): GRG Press. Holloway E. 2002. Meeting review. From genotype to phenotype: Linking bioinformatics and medical informatics ontologies. Gomparative Functional Genomics 3: 447-450. Irish VF, Benfey PN. 2004. Beyond Arabidopsis. Translational biology meets evolutionary developmental biology. Plant Physiology 135:611-614. Jenner RA. 2004. When molecules and morphology clash: Reconciling conflicting phylogenies of the Metazoa by considering secondary character loss. Evolution and Development 6: 372-378. Jockusch EL, Ober KA. 2004. Hypothesis testing in evolutionary developmental biology: A case study from insect wings. Journal of Heredity 95: 382-396. Kitano H. 2002. Systems biology: A brief overview. Science 295:1662-1664. Klingenberg GP, Mebus K, Auffray JG. 2003. Developmental integration in a complex morphological structure: How distinct are the modules in the mouse mandible? Evolution and Development 5: 522-531. Kloster M, Tang G, Wingreen NS. 2005. Finding regulatory modules through large-scale gene-expression data analysis. Bioinformatics 21:1172-1179. Kuratani S. 2004. Evolution of the vertebrate jaw: Gomparative embryology and molecular developmental biology reveal the factors behind evolutionary novelty. Journal of Anatomy 205: 335-347. Lee I, Date SV, Adai AT, Marcotte EM. 2004. A probabilistic functional network of yeast genes. Science 306: 1555-1558. Lee TI, et al. 2002. Transcriptional regulatory networks in Saccharomyces cerevisiae. Science 298: 799-804. Mabee PM, Grotwell PL, Burke AG, Bird NG. 2002. Evolution of median fin modules in the axial skeleton offishes.Journal of Experimental Zoology: Molecular and Developmental Evolution 294: 77-90. Maddison WP, Maddison DR. 1992. MacGlade: Analysis of Phylogeny and Gharacter Evolution, version 3.0. Sunderland (MA): Sinauer. Minelli A. 2003. The Development of Animal Form: Ontogeny, Morphology, and Evolution. London: Gambridge University Press. Payre F, Vincent A, Garreno S. 1999. ovo/svb integrates Wingless and DER pathways to control epidermis differentiation. Nature 400:271-275. Peterson KJ, Arenas-Mena G, Davidson EH. 2000. The A/P axis in echinoderm ontogeny and evolution: Evidence from fossils and molecules. Evolution and Development 2:93-101. Popodi E, Raff RA. 2001. Hox genes in a pentameral animal. Bioessays 23: 211-214. Prum RO, Dyck J. 2003. A hierarchical model of plumage: Morphology, development, and evolution. Journal of Experimental Zoology: Molecular and Developmental Evolution 298: 73-90. www. biosciencemag. org Articles Qin H, Lu HHS, Wu WB, Li W-H. 2003. Evolution of the yeast protein interaction network. Proceedings of the National Academy of Sciences 100: 12820-12824. Raff RA. 1996. The Shape of Life. Chicago: University of Chicago Press. Rast JP. 2003. Development gene networks and evolution. Journal of Structural and Functional Genomics 3: 225-234. Ravasz E, Somera AL, Mongru D, Oltvai ZN, Barabasi A-L. 2002. Hierarchical organization of modularity in metabolic networks. Science 297: 1551-1555. Revilla-i-Domingo R, Davidson EH. 2003. Developmental gene network analysis. International Journal of Developmental Biology 47: 695-703. Richardson MK, Brakefield PM. 2003. Developmental biology: Hotspots for evolution. Nature 424: 894-895. Roth VL. 1984. On homology. Biological Journal ofthe Linnean Society 22: 13-29. . 1988. The biological basis of homology. Pages 1-26 in Humphries CJ, ed. Ontogeny and Systematics. New York: Columbia University Press. Rudel D, Sommer RJ. 2003. The evolution of developmental mechanisms. Developmental Biology 264: 15-37. Schlosser G, Thieffry D. 2000. Modularity in development and evolution. Bioessays 22:1043-1045. Schlosser G, Wagner GP. 2004. Modularity in Development and Evolution. Chicago: University of Chicago Press. Shapiro MD, Marks ME, Peichel CL, Blackman BK, Nereng KS, Jonsson B, Schluter D, Kingsley DM. 2004. Genetic and developmental basis of evolutionary pelvic reduction in threespine sticklebacks. Nature 428: 717-723. Smith KK. 2003. Time's arrow: Heterochrony and the evolution of development. International Journal of Developmental Biology 47:613-621. Sommer RJ. 2000. Comparative genetics: A third model nematode species. Current Biology 10: R879-R881. Spicer GS, Bell CD. 2002. Molecular phylogeny of the Drosophila virilis species group (Diptera: Drosophilidae) inferred from mitochondrial 12S and 16S ribosomal RNA genes. Annals of the Entomological Society of America 95: 156-161. www.biosciencemag.org Stern DL. 2000. Evolutionary developmental biology and the problem of variation. Evolution: International Journal of Organic Evolution 54: 1079-1091. Sucena E, Stern DL. 2000. Divergence of larval morphology between Drosophila sechellia and its sibling species caused by cis-regulatory evolution of ovo/shaven-baby. Proceedings of the National Academy of Sciences 97: 4530-4534. Sucena E, Delon I, Jones I, Payre F, Stern DL. 2003. Regulatory evolution of shavenbaby/ovo underlies multiple cases of morphological parallelism. Nature 424: 935-938. Tabin CJ, Carroll SB, Panganiban G. 1999. Out on a limb: Parallels in vertebrate and invertebrate limb patterning and the origin of appendages. American Zoologist 39: 650-663. Tao Y, Friedman C, Lussier YA. 2005. Visualizing information across multidimensional post-genomic structured and textual databases. Bioinformatics 21:1659-1667. Tautz D. 2000. Evolution of transcriptional regulation. Current Opinion in Genetics and Development 10: 575-579. Tian NM, Price DJ. 2005. Why cavefish are blind. Bioessays 27: 235-238. Van Valen LM. 1982. Homology and causes. Journal of Morphology 173: 305-312. Wagner GP. 1989. The biological homology concept. Annual Review of Ecology and Systematics 20: 51-69. Wagner GP, Larsson HC. 2003. What is the promise of developmental evolution? Ill: The crucible of developmental evolution. Journal of Experimental Zoology: Molecular and Developmental Evolution 300:1-4. Wiens JJ, Chippindale PT, Hillis DM. 2003. When are phylogenetic analyses misled by convergence? A case study in Texas cave salamanders. Systematic Biology 52: 501-514. Wolstencroft K, McEntire R, Stevens R, Tabernero L, Brass A. 2005. Constructing ontology-driven protein family databases. Bioinformatics 21: 1685-1692. Wray GA. 2002. Do convergent developmental mechanisms underlie convergent phenotypes? Brain, Behavior and Evolution 59: 327-336. Yamamoto Y, Stock DW, Jeffery WR. 2004. Hedgehog signaling controls eye degeneration in blind cavefish. Nature 431: 844-847. April 2006 / Vol. 56 No. 4 • BioScience 309