escaping the mouse trap: the selection of new evo-devo model species

10
Escaping the Mouse Trap: The Selection of New Evo-Devo Model Species MICHEL C. MILINKOVITCH AND ATHANASIA TZIKA y Laboratory of Evolutionary Genetics, Institute for Molecular Biology & Medicine, Universite´Libre de Bruxelles (ULB), 6041 Gosselies, Belgium ABSTRACT Among the many, sometimes contradictory, criteria that have been used for promoting model species, the most prominent has probably been their relevance for understanding human biology. Recently however, the debate has partly shifted from the search for evolutionary conservation (medicine-driven models) to a better understanding of the generative mechanisms underlying biological diversity (Evo-Devo-driven models). Integration of multiple disciplines, beyond developmental genetics and evolutionary molecular genetics, as well as of innovative technologies will help biologists to open the massive realm of living species to genome manipulation and phenotypic investigation. However, a consensual list of model species must still be reached for optimizing the interplay between in silico analyses and in vivo experiments, and we claim that the Evo-Devo community should play a more energetic role in this endeavor. We discuss here a few criteria and limitations of major relevance to the choice of model species for Evo-Devo studies, and promote the use of a pragmatic approach. Finally, given the difficulties related to manipulating and breeding model species, we suggest the development of Evo-Devo virtual zoos maintaining breeding colonies of a selected set of species and from which eggs or staged embryos are available on order. J. Exp. Zool. (Mol. Dev. Evol.) 308B:337– 346, 2007. r 2007 Wiley-Liss, Inc. How to cite this article: Milinkovitch MC, Tzika A. 2007. Escaping the mouse trap: the selection of new evo-devo model species. J. Exp. Zool. (Mol. Dev. Evol.) 308B: 337–346. THE EARLIEST MODEL SPECIES Most of what is known on fundamental cellular processes comes from a handful of model organ- isms, with the bacterium Escherichia coli (and its bacteriophages) and the bakers/brewers yeast Saccharomyces cerevisiae leading the way because of the ease and speed with which they can be grown and manipulated in the laboratory. How- ever, investigation of metazoan biology requires additional models because understanding of physiological systems, and their interactions, cannot be tackled efficiently through cell culture or in vitro approaches. One major assumption justifying the use of model species has been that biological phenomena uncovered in these species could be extrapolated to other metazoans in general and to humans in particular, i.e., for potentially improving human health and well- being. However, the a priori chances that this extrapolation is valid depend on the sum of branch lengths between human and the model species because, on average, closely related species have more similar genomes and share more metabolic/ physiological and developmental pathways than distantly related species do. Hence, when it came to choosing metazoan model species, many researchers somewhat sacrificed the advantages of small size, short generation time, and ease of manipulation that characterize models such as the nematode worm Caenorhabditis elegans, or the fruit fly Drosophila melanogaster, for species humans can more readily relate to such as the laboratory mouse (Mus musculus). The justifica- tion of the mouse as a model for improving human Published online 22 May 2007 in Wiley InterScience (www. interscience.wiley.com). DOI: 10.1002/jez.b.21180 Received 18 February 2007; Revised 28 March 2007; Accepted 30 April 2007 Grant sponsor: Communaute ´ Franc - ais de Belgique; Grant number: ARC 1164/20022770; Grant sponsor: National Fund for Scientific Research Belgium (FNRS). Correspondence to: Michel C. Milinkovitch, Laboratory of Evolu- tionary Genetics, Institute for Molecular Biology & Medicine, Universite´ Libre de Bruxelles (ULB), 12 rue Jeener & Brachet, 6041 Gosselies, Belgium. E-mail: [email protected] y PhD candidate at the ‘‘Fonds pour la formation a ` la Recherche dans l’Industrie et dans l’Agriculture (FRIA)’’, Belgium. r 2007 WILEY-LISS, INC. JOURNAL OF EXPERIMENTAL ZOOLOGY (MOL DEV EVOL) 308B:337–346 (2007)

Upload: michel-c-milinkovitch

Post on 11-Jun-2016

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Escaping the mouse trap: the selection of new Evo-Devo model species

Escaping the Mouse Trap: The Selection of NewEvo-Devo Model Species

MICHEL C. MILINKOVITCH� AND ATHANASIA TZIKAy

Laboratory of Evolutionary Genetics, Institute for Molecular Biology & Medicine,Universite Libre de Bruxelles (ULB), 6041 Gosselies, Belgium

ABSTRACT Among the many, sometimes contradictory, criteria that have been used forpromoting model species, the most prominent has probably been their relevance for understandinghuman biology. Recently however, the debate has partly shifted from the search for evolutionaryconservation (medicine-driven models) to a better understanding of the generative mechanismsunderlying biological diversity (Evo-Devo-driven models). Integration of multiple disciplines, beyonddevelopmental genetics and evolutionary molecular genetics, as well as of innovative technologieswill help biologists to open the massive realm of living species to genome manipulation andphenotypic investigation. However, a consensual list of model species must still be reached foroptimizing the interplay between in silico analyses and in vivo experiments, and we claim that theEvo-Devo community should play a more energetic role in this endeavor. We discuss here a fewcriteria and limitations of major relevance to the choice of model species for Evo-Devo studies, andpromote the use of a pragmatic approach. Finally, given the difficulties related to manipulating andbreeding model species, we suggest the development of Evo-Devo virtual zoos maintaining breedingcolonies of a selected set of species and from which eggs or staged embryos are available on order.J. Exp. Zool. (Mol. Dev. Evol.) 308B:337– 346, 2007. r 2007 Wiley-Liss, Inc.

How to cite this article: Milinkovitch MC, Tzika A. 2007. Escaping the mouse trap:the selection of new evo-devo model species. J. Exp. Zool. (Mol. Dev. Evol.) 308B:337–346.

THE EARLIEST MODEL SPECIES

Most of what is known on fundamental cellularprocesses comes from a handful of model organ-isms, with the bacterium Escherichia coli (andits bacteriophages) and the bakers/brewers yeastSaccharomyces cerevisiae leading the way becauseof the ease and speed with which they can begrown and manipulated in the laboratory. How-ever, investigation of metazoan biology requiresadditional models because understanding ofphysiological systems, and their interactions,cannot be tackled efficiently through cell cultureor in vitro approaches. One major assumptionjustifying the use of model species has been thatbiological phenomena uncovered in these speciescould be extrapolated to other metazoans ingeneral and to humans in particular, i.e., forpotentially improving human health and well-being. However, the a priori chances that thisextrapolation is valid depend on the sum of branchlengths between human and the model speciesbecause, on average, closely related species have

more similar genomes and share more metabolic/physiological and developmental pathways thandistantly related species do. Hence, when it cameto choosing metazoan model species, manyresearchers somewhat sacrificed the advantagesof small size, short generation time, and ease ofmanipulation that characterize models such as thenematode worm Caenorhabditis elegans, or thefruit fly Drosophila melanogaster, for specieshumans can more readily relate to such as thelaboratory mouse (Mus musculus). The justifica-tion of the mouse as a model for improving human

Published online 22 May 2007 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jez.b.21180

Received 18 February 2007; Revised 28 March 2007; Accepted 30April 2007

Grant sponsor: Communaute Franc-ais de Belgique; Grant number:ARC 1164/20022770; Grant sponsor: National Fund for ScientificResearch Belgium (FNRS).�Correspondence to: Michel C. Milinkovitch, Laboratory of Evolu-

tionary Genetics, Institute for Molecular Biology & Medicine,Universite Libre de Bruxelles (ULB), 12 rue Jeener & Brachet, 6041Gosselies, Belgium. E-mail: [email protected] candidate at the ‘‘Fonds pour la formation a la Recherche dansl’Industrie et dans l’Agriculture (FRIA)’’, Belgium.

r 2007 WILEY-LISS, INC.

JOURNAL OF EXPERIMENTAL ZOOLOGY (MOL DEV EVOL) 308B:337–346 (2007)

Page 2: Escaping the mouse trap: the selection of new Evo-Devo model species

medicine has always been quite explicit. Forexample, the first 50 years of mouse genetics(Paigen, 2003), which correspond to the first halfof the twentieth century, were dominated by thestudy of the genetic bases of both spontaneoustumor incidence and susceptibility to transplantedtumors, which eventually led to the discovery ofretroviruses/oncogenes and the major histocom-patibility complex, respectively. In the next 5decades, the genetic mouse system was extendedto many additional aspects of mammalian biologysuch as sex determinism, biochemical genetics,physiology, behavior, etc.

Specific practical features (beyond ease andspeed of growth) facilitating experimental manip-ulation have been used for selecting model species(e.g., Krebs, ’75). For example, D. melanogaster ischaracterized by, among others, the existence ofeasy-to-score morphological variation (affectingbristles, wing veins, compound eyes, etc.), thepresence of only three pairs of autosomal, and onepair of sex, chromosomes that can be directlyvisualized as giant (polytene) chromosomes of thelarval salivary glands, and the absence of meioticrecombination in males (facilitating genetic stu-dies). Each of these features made possible thepioneering work of T.H. Morgan, A.H. Sturtevant,C.B. Bridges, and H.J. Muller, i.e., the chromo-some theory of heredity, as well as the discoveryof sex and genetic linkage and of ionizing-radia-tion-induced mutations (Sturtevant, ’65). A largeproportion of the Drosophila community of thattime exploited these new concepts and techniquesto generate a highly accurate whole-genomemutational scan (Lindsley et al., ’72). Less thana decade later, a genome-wide mutational screenaimed at systematically finding embryonic lethalmutants led to the identification of the first 15 lociinvolved in segmental patterning of the Drosophi-la larva and initiated the concept of multilevelspatial organization during development(Nusslein-Volhard and Wieschaus, ’80). Theabove-mentioned breakthroughs led to not lessthan three Nobel Prizes (Morgan in 1933, Mullerin 1946, and Nusslein-Volhard and Wieschaus in1995). Not bad for a tiny fly with fat chromosomes.

Similarly, the roundworm C. elegans has beenselected as an important model organism becauseits largely invariant complete cell lineage, fromthe fertilized egg to the adult, as well as its fullneural connectivity, have been determined withgreat precision. These features, added to thesignificant advantage of being tolerant to freezingfor long-term storage, have made the roundworm

a model of choice to study cell differentiation,programmed cell death, and neural mechanisms.

FROM MAPS TO FULL-GENOMESEQUENCES AND GENETIC

ENGINEERING

As the molecular genetic revolution kicked in,a constant concern has been the possibility tomanipulate the genome of model species suchthat, nowadays, one can hardly claim model statusto a species if it fails to meet that criterion.For decades, researchers have used thermaland electrical stimuli to transform E. coli andS. cerevisiae with foreign DNA, as well as homo-logous recombination for precise insertion/dele-tion/replacement of genetic elements. In Metazoa,besides traditional screens using mutagenesisinduced by X-ray radiations or ethyl-methane-sulphonate (that can only identify the firstessential function of a gene), Drosophila scientistshave developed and applied ingenious approachesthat allow, in principle, screening for any pheno-type in any cell at any stage of development[reviewed in St Johnston (2002)]. Even for themouse model, ever since the first transgenic micehave been generated in the early 1980s (Gordonet al., ’80; Wagner et al., ’81; Palmiter et al., ’82),multiple techniques have been developed forperforming genotype- or phenotype-driven experi-ments, i.e., identify the function(s) of any givenknown sequence or identify the sequence(s)associated with any given phenotype of interest,respectively. These approaches have been madepossible by (i) the use of homologous recombina-tion with positive/negative selection of recombi-nants in stable (immortal) and totipotentembryonic stem (ES) cells, and (ii) tissue-specificactivation/inactivation techniques (Doetschmanet al., ’87; Thomas and Capecchi, ’87; Mansouret al., ’88; Koller et al., ’89; Thompson et al., ’89;Thomas and Capecchi, ’90; Lakso et al., ’92; Orbanet al., ’92; Furth et al., ’94; Wigley et al., ’94).Methods have even been developed to generategenomic macro-rearrangements through exploita-tion of loxP sites inserted at different positionsof the genome in different mouse strains (Heraultet al., ’98). The importance of these technicaladvances in making the mouse a crucibleof genetic manipulation should not be under-estimated. For example, despite that it exhibitsmajor advantages in behavioral and learningstudies as well as for investigating physiologicalparameters associated with susceptibility to

M.C. MILINKOVITCH AND A. TZIKA338

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 3: Escaping the mouse trap: the selection of new Evo-Devo model species

cancer, hypertension, obesity, diabetes, and auto-immune diseases, the rat remains today a modelspecies less attractive than the mouse, mostlybecause of the unavailability of suitable totipotentES cells. Similarly, ongoing efforts seek to develophomologous recombination and suitable ES cellsin additional model species, including the zebra-fish (Fan et al., 2006), as well as in humans fortransplantation and gene therapy (Zwaka andThomson, 2003).

Fortunately, two approaches have great promisesfor genome manipulation (including gene target-ing) in basically any species. The first, and byfar the most costly and complex, is cloning, i.e.,in vitro transformation, selection, and screeningof primary cells followed by nuclear transfer intooocytes. Since the birth of Dolly, the first mammalderived from an adult somatic cell (Wilmut et al.,’97), genetically modified (in culture) cell lineshave been used for generating knock-in and knock-out sheep and pigs (McCreath et al., 2000; Denninget al., 2001; Phelps et al., 2003). Successful applica-tions of this technique to the rat (Zhou et al., 2003),goat (Baguisi et al., ’99), horse (Galli et al., 2003),mule (Woods et al., 2003), cat (Shin et al., 2002),gaur (Lanza et al., 2000), African Wildcat (Gomezet al., 2004; Lee et al., 2005), and dog (Lee et al.,2005) suggest that it could be generalized to manyspecies. These methods have even been put forwardas possible efficient tools for the preservation ofendangered species (Blomquist, ’98; Corley-Smithand Brandhorst, ’99; Begley, 2000; Lee, 2001).Genetic engineering followed by nuclear transferremains, however, limited by high costs andconfounding issues such as low efficiency of genetargeting and developmental abnormalities. At leastin mammals, many problems are related to geneticreprogramming of the transplanted nucleus, aphenomenon whose effectiveness seems highlydependent on the levels of epigenetic modification[especially at imprinted loci; Rideout et al. (2001)]exhibited by the donor cell.

The second approach for targeted controlledgene expression is based on RNA interference(RNAi; Fire et al., ’98), a phenomenon, for whichAndrew Z. Fire and Craig C. Mello (2006) wererecently awarded the Nobel Prize in Medicine, andby which small stretches of double-stranded RNAcan reduce or eliminate the expression of genescontaining the same sequence. RNAi and relatedsilencing pathways have not only revolutionizedour understanding of gene regulation, but therecently developed technologies that take advan-tage of RNAi are also anticipated to become

efficient tools for knocking down any particulargene in possibly any species (Hannon and Rossi,2004; Mello and Conte, 2004). For example, small-hairpin RNA (shRNA)-expressing constructs canbe introduced as transgenes in model animalsthrough pronuclear injection or lentiviral infec-tion and generate specific and stable silencingof gene expression (e.g., Rubinson et al., 2003;Tiscornia et al., 2004; Peng et al., 2006). Althoughmost studies of specific gene knock-down by theuse of RNAi are performed in mice, similarapproaches are being successfully employed in,among others, the fruit fly, the zebrafish(Liu et al., 2005), Xenopus, and the rat.

Obviously, genetic manipulation is facilitated bythe accessibility of the full genome sequence ofthe species of interest. The availability of genomesequences annotated for gene identification andstructure/function prediction of, e.g., encoded pro-teins, has become of such large interest that specificgenome idiosyncrasies have even been put forwardas an argument by itself for choosing model species.For example, the compact nature of the yeast,Arabidopsis, or pufferfish genomes has been usedfor promoting these organisms as model species,whereas, e.g., the low G1C content of the Hydralarge genome makes it difficult to generate andscreen genomic libraries.

IDENTIFYING AND UNDERSTANDINGPHENOTYPES

Maybe unexpectedly, one of the most importantlimiting factors to understand the function of genesis not genetic engineering or sequence data avail-ability but the difficulties with which phenotypescan be identified. Indeed, many genes exert func-tions that cannot be investigated by simple exam-ination of a few general morphological and/or physiological parameters. A genetically modifiedmouse could have a phenotype of primary impor-tance but this phenotype might be very difficult toevidence. This is one of the reasons why the rat hasremained a valued model (despite its inferiority incomparison with the mouse, as far as geneticengineering is concerned): it is much easier tophenotype a rat than a mouse, simply because theformer is significantly bigger than the latter.

Transparent mice occupy a significant percen-tage of mouse geneticists’ dreams, but this dreamis made true in another species, the zebrafish(Danio rerio), whose development and organogen-esis can be observed in great details (i.e., it istraceable at the level of individual cells), thanks to

SELECTING NEW MODEL SPECIES 339

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 4: Escaping the mouse trap: the selection of new Evo-Devo model species

its transparent embryos. That asset brings allother advantages of zebrafish, i.e., powerfulgenetics tools, robustness of the embryos, andpermeability to small drug molecules (Nusslein-Volhard and Dahm, 2002; Liu et al., 2005), to fullappreciation and greatly contributed to its successas a vertebrate model species. Fortunately forresearchers using other model species such asmice and rats, imaging technologies and physiolo-gical measuring techniques have recently beenminiaturized and adapted for their use with smallspecies. In recognition of the major importanceof reliable phenotypic data for achieving the fullpotential of genomic information and geneticengineering, initiatives have been undertaken forestablishing collections of baseline phenotypicdata on commonly used and genetically diverseinbred strains of model species. Particularlynoteworthy is the coordinated international effortnamed the Mouse Phenome Project (http://www.jax.org/phenome) aimed at identifying appropriatestrains for, among others, physiological testing,drug discovery, toxicology studies, mutagenesis,disease onset and susceptibility, QTL analyses,identification of new genes, and unraveling theinfluence of environment on genotype/phenotype.

Another source of difficulty for phenotyping isthe epigenetic effects among genes. Indeed, multi-ple reports indicate that a full knock-out mouse(i.e., homozygous for the null mutation) at a locussupposedly essential for an identified specificprocess can yield no visible phenotype becauseanother gene has been co-opted during develop-ment to fulfill the function of the invalidated gene.This problem boils down to differences of geneticbackground among species, an issue that bringsadditional consequences in the choice of optimalmodel species (see below).

THE SEARCH FOR EVOLUTIONARYCONSERVATION (MEDICINE-DRIVEN

MODELS)

As indicated above, most of the researchperformed so far with model species has beenjustified by the potential power of these species forunderstanding human biology. For example, thejustification put forward for the use of Drosophilaas a model species has rarely been that it exhibitsa body plan characterizing the most diverse andsuccessful group of metazoan (the insects) butrather by its ‘‘reasonable similarity with humans’’[sic]. Classically, it is emphasized, e.g., in grantapplications, that 60–70% of the known human

disease genes have a recognizable match in thegenome of fruit flies and that the species is alreadybeing used as a genetic model for development,fertility, longevity, learning, degenerative diseases(such as Parkinson’s, Huntington’s, and Alzhei-mer’s), and drug susceptibility. Similar justifica-tions are used for other model species, such as thezebrafish (Ernest et al., 2000; Fishman, 2001).However, the significant phylogenetic distancesthat separate humans from fruit flies, zebrafish, oreven mice, can restrict the suitability of thesespecies for medical research. For example, ahuman gene of interest may have no ortholog inthe mouse [as it is the case for the Kallmann’ssyndrome human neurogenetic disorder; but seeCobb et al. (2006) for the interesting approach, inthe case of human short-stature syndromes, ofusing a close mouse paralog of the target humangene absent from the mouse genome] or a mutantmouse might not exhibit the same phenotype asthe corresponding human mutant (as it is the casefor Lesch-Nyhan’s disease and Ataxia-Telangiec-tasia). Such situations have been used as justifica-tions for ongoing projects that aim at gene-targeting and cloning of non-human primates(Norgren, 2004). In specific cases, non-murinemodels exhibit physiologies, metabolisms, ororgans of major interest in medicine [mechanismsof artheriosclerosis development in rabbit; organtransplantation in pig; nutritional infertility inthe musk shrew, Suncus murinus (Temple, 2004);etc.] promoting the development of genetic en-gineering and nuclear transfer techniques in thesespecies (Dinnyes and Szmolenszky, 2005).

Finally, how the overall gene conservatismamong species is perceived depends on the waythe statistics are presented. It is, e.g., indeedstaggering that human and chimp differ only byabout 1% in sequence of the protein-codingportion of their genomes, but this observationtranslates into the fact that the two species exhibitat least one difference in amino acid sequence in55% of their proteins, a figure that rises to 95% forhuman vs. mouse (Coyne, 2005). So, are humansand chimps 99% identical or do they exhibit half oftheir genes with potentially adaptive differences?

UNDERSTANDING VARIATION(EVO-DEVO-DRIVEN MODELS)

Obviously, in the context of Evo-Devo, it is themassive realm of living species that should,ideally, be opened to genome manipulation andphenotypic investigation. Indeed, the interests of

M.C. MILINKOVITCH AND A. TZIKA340

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 5: Escaping the mouse trap: the selection of new Evo-Devo model species

evolutionary developmental biologists go wellbeyond widely conserved physiologies and devel-opmental processes/patterns as they seek tounderstand the generative mechanisms under-lying biological diversity. Uncovering thesemechanisms will require the merging of twosuccessful disciplines, if not cultures or metaphy-sics (Minelli, 2003), that remained largely sepa-rated despite the obvious conceptual links betweenthe two: molecular developmental biology andevolutionary molecular genetics. Indeed, on onehand, molecular developmental biologists havefocused on the use of a handful of model organismsfor deciphering the fascinating processes by whichcells differentiate, as well as tissues, organs, andorganisms grow and develop. On the other,evolutionary molecular geneticists have investi-gated the modes and tempos of DNA and proteinevolution in a multitude of organisms (fromviruses to vertebrates) and developed the labora-tory techniques and analytical methods allowingto infer phylogenies, reconstruct population his-tories, uncover hidden biodiversity, and detectselection and stochastic patterns in laboratory andnatural populations. But Evo-Devo, if it claims thestatus of becoming a successful discipline, willrequire more than developmental genetics andevolutionary molecular genetics. It will needintegrating evolutionary biology in the broadestmeaning of the term. For example, it is likely thatcomparative anatomy and paleontology should‘‘determine the agenda of a collaboration betweenmolecular evolution and mechanistic molecularbiology’’ (Wagner and Larsson, 2003) because,after all, the ultimate questions tackled by mostEvo-Devo researchers is how the diversity ofphenotypes has been generated. It is the samerequirement of meeting the real world of organ-isms that has recently revived the appreciation ofintegrating ecological aspects into Evo-Devo(Dusheck, 2002; Gilbert and Bolker, 2003). Indeed,the use of inbred lines of model species [selected,among others, for their robustness to environ-mental perturbation; Bolker (’95)], raised andbred in very controlled and stable laboratoryenvironments, has somehow eluded the impor-tance of developmental plasticity (reaction norms),i.e., the fact that environmental factors cansignificantly influence developmental pathways.A significant portion of this plasticity is adaptiveas demonstrated by spectacular examples oforganisms modifying their development followingenvironmental cues indicating the presence ofsymbionts (Nyholm et al., 2000; Stappenbeck

et al., 2002) or of predators (Van Buskirk, 2002).The comparison of current model species (with flatreaction norms) and new model species (withlarger developmental plasticity) might providehints at the mechanisms underlying (in)sensitivityto environmental conditions.

Hence, one major challenge in Evo-Devo will beto adapt the tremendous knowledge and sophisti-cated technologies accumulated on ‘‘classical’’model species (allowing to perform genetic crosses,genome engineering, gene expression analyses,and efficient phenotyping) to model organismsfrom a wider assortment of lineages on the treeof life in a wider set of environmental conditions.On one hand, developmental geneticists haveto accept the invasion of new species on theirbenches and in their animal rooms, and shouldbetter value the importance of working sometimeswith natural populations. On the other, evolu-tionary biologists and ecologists have to accept theconstraints of working with model species andincorporate in their research programs the tech-nological advances of genome engineering andphenotyping.

One could argue that the freedom experiencedby evolutionary molecular geneticists in theirchoice of species should be extended to the newdiscipline of Evo-Devo. This will, undoubtedly,happen partly: there is no reason we shouldrestrain anyone to use Evo-Devo approaches forpossibly investigating any character of interest inany possible (group of) species. It is certain thatinteresting phenotypes can be found in manylineages such that some level of subjectivity will(should) be maintained. However, as demon-strated in the past, promoting the use of the sameset of model species increases the efficiency withwhich techniques and analytical approaches aredeveloped simply through collaboration, emula-tion, and establishment of public databases.For example, access to the sequences of fullgenomes greatly facilitates genetic manipulationsand investigation of the evolution of new genefunctions. However, despite the large increase insequencing throughput [thanks to the develop-ment of capillary technologies and fluorescence-based enzymatic-driven chemistries as well as,more recently, high-density picolitre reactors(Margulies et al., 2005)], the genomes of eukaryotesare large enough to constrain, at least in the next5–10 years, the sequencing of full genomes to aselected small set of species. Given the necessity tooptimize the interplay between in silico analysesand in vivo experiments, it is crucial to establish

SELECTING NEW MODEL SPECIES 341

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 6: Escaping the mouse trap: the selection of new Evo-Devo model species

the list of preferred species for which the genomewill be sequenced in priority, and this shouldbe done with more input from the Evo-Devocommunity (Tzika and Milinkovitch, 2007). Forexample, the Reptilian Genomics Working Group(www.reptilegenome.com) has recently suggestedthe sequencing of the full genome from Thamno-phis sirtalis as the representative species forsnakes. However, the species is ovoviviparous, akey reproductive feature that seriously limits theaccessibility of embryos unless the mother iskilled. The corn snake (Elaphe guttata) mightconstitute a better alternative model species in theEvo-Devo context (Tzika and Milinkovitch, 2007)because it is oviparous, such that embryos can beincubated artificially for investigation at any stageof the second half of development (i.e., after egglaying).

CRITERIA FOR CHOOSINGNEW MODEL SPECIES

Beside the practical criteria of, among others,small size, short generation, abundant progeny,ease of manipulation and housing/breeding, acces-sibility of phenotyping, and genome manipulationtechniques, there are other parameters thatshould be considered when listing preferred modelspecies. We will discuss below a few criteria thatwe think will greatly influence the intrinsicexplanatory power of model species in Evo-Devostudies. We claim that Evo-Devo considerationsshould be taken into account for the choice ofspecies that will be proposed as models for thewhole community of biologists. Evo-Devo biolo-gists should play a proactive and energetic role inthis endeavor because these choices are likely tohave a profound impact on years of research tocome.

Although intuitive (but see, Jenner and Wills,2007), the criterion of evolutionary divergenceamong species has been poorly used for guidingthe choice of model species. For example, despitethat several orders within Reptilia (includingbirds) are more genetically diverse than mammals,it is striking that the latest version of the database(v44, April 2007) produced by the EnsemblProject, which generates and maintains automaticannotation on selected eukaryotic genomes(www.ensembl.org/), includes 20 mammalian andfive teleost fish genomes, but only one bird and noreptile. Current proposals for full genome sequen-cing (www.genome.gov/10002154) correct onlyvery partially such phylogenetic biases. However,

the phylogenetic-distance criterion is limited by atleast two parameters: (i) the rate of phenotypictransformation is highly variable among lineages,and (ii) variation worth investigation exists atmultiple phylogenetic levels (one should not focusonly on major transformations). Intermingledwith the phylogenetic-distance criterion, the‘‘ancestrality’’ of a model species is a decisivefactor. For example, the zebrafish is often con-sidered a ‘‘canonical vertebrate’’ (Fishman, 2001)because the common ancestor of all vertebrateswas a fish. This statement is, however, of limitedvalue because, on average, no extant species isintrinsically more ancestral than any other. So,the real, non-trivial, question is: what is theconsidered species a model of (i.e., at whathierarchical level(s) of the phylogeny)? For exam-ple, is the zebrafish a model species for chordates,for vertebrates, for bony fishes (Osteichtyes), forray-finned fishes (Actinopterygii), for teleost fishes(Teleostei), or for the family Cyprinidae? Theanswer heavily depends on what characters oneis interested into: anatomical, physiological, devel-opmental, genomic, ecological, behavioral, etc. Forexample, as far as anatomy is concerned, thezebrafish might be a better model for teleost fishesthan for vertebrates because the species exhibitsmultiple characters that seem ancestral for theformer group (contrary to other model teleostssuch as the fugu or stickleback), whereas it is veryderived (as all teleost fishes are) in respect to thevertebrate ancestor (Metscher and Ahlberg, ’99).The intuitive reluctance of considering the skele-ton of an amniote as possibly more ancestral thatthose of the zebrafish or Xenopus, demonstratesthat the view of a ‘‘transformist’’ pyramid of life(with humans at the top) is still pervasive.Furthermore, although structural and genomicsimplicity are valid criteria for choosing modelspecies [e.g., the cephalochordate amphioxus;Holland et al. (2004)], it should be carefullyinvestigated whether the observed simplicity isancestral for the group and not a secondary(derived) simplification as it is likely the case for,e.g., flatworms or myzostomes.

These considerations indicate that the choice ofa species as a model for a given taxonomic groupwill necessarily depend, at least partially, on thecharacters of interest. Hence, defining a list of trueinnovations will greatly help determine the list offocal model species in Evo-Devo studies. This isthe very reason why robust and extensive (mole-cular) phylogenetic hypotheses will be so central:evolutionary trees constitute the basic framework

M.C. MILINKOVITCH AND A. TZIKA342

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 7: Escaping the mouse trap: the selection of new Evo-Devo model species

on which character changes are mapped.Obviously, morphologists in general and paleon-tologists in particular will also play a key role(Hall, 2002) in that endeavor for accurateinference of ancestral states and character trans-formations. Indeed, the choice of a model speciesfor the study of a set of morphological innovationsshould be partly guided by the quality of the fossilrecord pertaining to the anatomical features ofinterest. For example, the rich fossil record ofXenarthra could undoubtedly help in Evo-Devostudies aimed at understanding the genetic basisof their adaptations to burrowing and eatinginsects, whereas the poor fossil record of Chir-optera will be of little use to decipher thegenerative mechanisms of the early anatomicaltransformations that led to active flight in thatclade. Another example is the exceptional pre-servation of teeth in the Cenozoic fossil record, aclear incentive for using the Evo-Devo approachto investigate the specialization and differentia-tion of dentition in eutherian mammals.

Understanding the genetic and developmen-tal basis of convergences in morphological, ecolo-gical, and physiological characteristics is probablyone of the most important challenges in Evo-Devo.Indeed, the mapping of characters on robustmolecular phylogenies has demonstrated extensiveand multiple convergences of ecologically specia-lized species (ecomorphs), e.g., of cichlid fishes(Kocher et al., ’93; Ruber et al., ’99), ranid frogs(Bossuyt and Milinkovitch, 2000), Anolis lizards(Losos et al., ’98), and mammals (e.g., betweensome Afrotherian and Eulipotyphla insectivores).Similarly, the snake-like body form has evolvedmultiple times independently in squamate reptiles(Wiens et al., 2006). Such analyses will require astrategy of choosing model organisms based ontheir traits rather than phylogenetic position (oreven sum of branch lengths) per se.

In relation to both the problems of (i) theancestral/derived nature of character states, and(ii) the non-independence of characters, Wagner(2001) has put forward an important cautionarynote: there might be intrinsic limits to what canbe experimentally proven in terms of causalexplanation for specific evolutionary innovations.More specifically, when using the genotype–phe-notype map model (Schuster et al., ’94; Stadleret al., 2001), i.e., the mapping of the underlyinggenotype space to the corresponding morphospace(phenotypes), the likelihood of a given phenotypictransformation depends on the neighborhoodrelationships of the relevant sets of genotypes.

In other words, a given phenotypic transi-tion might require a specific mutation to occur ina genome ‘‘that is poised for this transition’’(Wagner, 2001). Hence, in the context ofusing model species to decipher the causal ex-planation for specific evolutionary innovations, itis not sufficient that the model species exhibitsthe ancestral state of the relevant phenotype, itmight also require to exhibit the pertinentancestral genetic background. Obviously, the like-lihood that a given extant species exhibits anoverall ancestral genetic background is exceed-ingly low. However, we think that there is roomfor some optimism. First, the relevance andintensity of the genetic background problem willdepend on (i) the nature of the character analyzed(two sets of neutral genotypes, corresponding tothe ancestral and derived phenotypes, respec-tively, might share a large border, such thatthe accessibility of the derived phenotype fromthe ancestral phenotype might be reasonablyhigh); and (ii) the age of the innovation investi-gated (the more recent the novelty, the lessproblematic is its investigation). Hence, thegenetic background problem will not necessarilyheavily affect experimental investigation of allevolutionary novelties. Second, different lineagesdisplaying the same ancestral state of a givencharacter of interest will hold different geneticbackgrounds such that one of them might exhibitthe ancestral appropriate background. For exam-ple, some marsupial species might prove, justby chance, to constitute better models than othermarsupial species for understanding the repro-ductive physiology novelties associated with therise of eutherians. Third, the genotype–phenotypemap model predicts that transition probabilitiesbetween phenotypic states can be asymme-tric, i.e., the A-D (Ancestral-Derived) and D-A(Derived-Ancestral) transitions can exhibit verydifferent probabilities (Wagner, 2001). We thinkthis is good news for Evo-Devo, because it meansit might be easier, in some cases, to produce theD-A transition by manipulating the genome of aspecies exhibiting the derived state than toproduce the A-D transition by manipulating thegenome of a species exhibiting the ancestralstate. In other words, species characterized byderived (rather than ancestral) states of investi-gated characters should also be considered aspossible models for Evo-Devo studies. All thesethoughts discussed above constitute additionalmotivations for lengthening the list of modelspecies.

SELECTING NEW MODEL SPECIES 343

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 8: Escaping the mouse trap: the selection of new Evo-Devo model species

PRAISE FOR A PRAGMATIC APPROACHAND FOR EVO-DEVO VIRTUAL ZOOS

As demonstrated by the non-exhaustive list ofissues raised above, the criteria that are relevantto the choice of a set of model species are multipleand can even be contradictory. This is intrinsicto the highly multidisciplinary nature of Evo-Devo. For example, a lineage can be characterizedby a unique and dramatic set of derived characterstates that makes it particularly appealing forEvo-Devo studies, but the group might lackrepresentatives that offer reasonable hope forconstituting a widespread model species (e.g., takethe extreme case of cetaceans). Furthermore,there are more subtle incompatibilities amongcriteria. For example, genetic variability of themost commonly used laboratory mouse strains hasbeen purposefully and dramatically reduced bothby inbreeding (to facilitate identification andanalysis of mutations, through reduction of thephenotypic variance caused by multiple geneticbackgrounds) and by selection (to increase, e.g.,fecundity). The laboratory mouse has been soefficiently selected for breeding all year long inartificial conditions that photoperiodism has con-comitantly dwindled. Hence, investigation of howthe hypothalamic-pituitary-gonadal-pineal path-way is involved in some major aspects of circadianrhythms (and their alterations as models ofcircadian disorders in humans) requires the useof other species such as Octodon degus, a moder-ate-sized hystricomorph rodent from central Chile(Lee, 2004). It is somewhat ironic that themechanisms causing phenotypic variability arepartly investigated by generating X-ray or ethyl-methane-sulphonate-induced mutants in modelanimal strains with uniform genetic background,whereas some of the phenotypic variation presentin the ancestral populations (i.e., before controlledinbreeding) was probably particularly relevant,given the likely adaptive nature of this variation.

We think there is no other possibility thanpromoting the use of a pragmatic optimizationapproach (not void of a significant amount ofsubjectivity) incorporating criteria such as, amongothers, phylogenetic position as well as numberand nature of the ancestral/derived characterstates of the model species, evolutionary diver-gence with other model species and/or selectedancestral nodes, level of variability characterizingthe taxon each chosen species belongs to, ease withwhich the species representatives can be handled,housed, and bred, and the protection status of the

species. Compromises will have to be made, as itis simply impossible to find species that combineall the possible advantageous features discussedabove. We applied elsewhere (Tzika and Milinko-vitch, 2007) such an approach for a set of speciesthat could serve as the workhorses for Evo-Devoresearch within amniotes hoping it will be used asa starting point for in-depth analysis with theinput from morphologists, paleontologists, animalbreeders, physiologists, developmental biologists,and molecular phylogeneticists.

Finally, given the difficulties and costs related todeveloping new model species (Burian, ’93), itwould be particularly beneficial for the Evo-Devocommunity if samples from different breedingcolonies of different model species could be madeavailable to everyone for a reasonable share ofcredit or funding. The ultimate solution to thisproblem would be the set-up of Evo-Devo ‘‘zoos’’where species mentioned above are kept, and wild-type/mutant eggs or staged embryos are availableon order. For example, the diversity of phenotypesamong domestic chicken breeds is similar to thatof dogs, and eggs can be bought and shipped fromjust about anywhere for incubation. This makescollecting diverse embryos for study very cheapand much easier than maintaining them all.Although such an endeavor for multiple Evo-Devomodel species would require significant organiza-tion and funding, similar projects with thenematode worm C. elegans, or the fruit flyD. melanogaster (see, e.g., the links to multiplestock collections listed on www.flybase.org),demonstrate its feasibility.

ACKNOWLEDGMENTS

Our manuscript has benefited from discussionswith Giuseppe Fusco, Denis Headon, OlivierLambert, Alesandro Minelli, Christophe Remy,and Gunter Wagner, as well as from comments bytwo anonymous reviewers.

LITERATURE CITED

Baguisi A, Behboodi E, Melican DT, Pollock JS, DestrempesMM, Cammuso C, Williams JL, Nims SD, Porter CA, MiduraP, Palacios MJ, Ayres SL, Denniston RS, Hayes ML, ZiomekCA, Meade HM, Godke RA, Gavin WG, Overstrom EW,Echelard Y. 1999. Production of goats by somatic cellnuclear transfer. Nat Biotechnol 17:456–461.

Begley S. 2000. Cloning the endangered. Newsweek 136:56–57.

Blomquist RF. 1998. Cloning endangered animal species?Valparaiso University law review Valparaiso University 32:383–418.

M.C. MILINKOVITCH AND A. TZIKA344

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 9: Escaping the mouse trap: the selection of new Evo-Devo model species

Bolker JA. 1995. Model systems in developmental biology.Bioessays 17:451–455.

Bossuyt F, Milinkovitch MC. 2000. Convergent adaptiveradiations in Madagascan and Asian ranid frogs revealcovariation between larval and adult traits. Proc Nat AcadSci USA 97:6585–6590.

Burian R. 1993. How the Choice of Experimental OrganismMatters. J Hist Biol 26:351–367.

Cobb J, Dierich A, Huss-Garcia Y, Duboule D. 2006. A mousemodel for human short-stature syndromes identifies Shox2as an upstream regulator of Runx2 during long-bonedevelopment. Proc Nat Acad Sci USA 103:4511–4515.

Corley-Smith GE, Brandhorst BP. 1999. Preservationof endangered species and populations: a role for genomebanking, somatic cell cloning, and androgenesis? MolReprod Dev 53:363–367.

Coyne JA. 2005. Switching on evolution. Nature 435:1029–1030.

Denning C, Burl S, Ainslie A, Bracken J, Dinnyes A, Fletcher J,King T, Ritchie M, Ritchie WA, Rollo M, de Sousa P, TraversA, Wilmut I, Clark AJ. 2001. Deletion of the alpha(1,3)galactosyl transferase (GGTA1) gene and the prion protein(PrP) gene in sheep. Nat Biotechnol 19:559–562.

Dinnyes A, Szmolenszky A. 2005. Animal cloning by nucleartransfer: state-of-the-art and future perspectives. ActaBiochimica Polonica 52:585–588.

Doetschman T, Gregg RG, Maeda N, Hooper ML, Melton DW,Thompson S, Smithies O. 1987. Targeted correction ofa mutant HPRT gene in mouse embryonic stem cells.Nature 330:576–578.

Dusheck J. 2002. It’s the ecology, stupid! Nature 418:578–579.

Ernest S, Rauch G-J, Haffter P, Geisler R, Petit C, Nicolson T.2000. Mariner is defective in myosin VIIA: a zebrafishmodel for human hereditary deafness. Hum Mol Genet 9:2189–2196.

Fan L, Moon J, Crodian J, Collodi P. 2006. Homologousrecombination in zebrafish ES cells. Transgenic Res 15:21–30.

Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE, MelloCC. 1998. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391:806–811.

Fishman MC. 2001. Genomics. Zebrafish—the canonicalvertebrate. Science 294:1290–1291.

Furth PA, St Onge L, Boger H, Gruss P, Gossen M, Kistner A,Bujard H, Hennighausen L. 1994. Temporal control of geneexpression in transgenic mice by a tetracycline-responsivepromoter. Proc Nat Acad Sci USA 91:9302–9306.

Galli C, Lagutina I, Crotti G, Colleoni S, Turini P, PonderatoN, Duchi R, Lazzari G. 2003. Pregnancy: a cloned horse bornto its dam twin. Nature 424:635.

Gilbert SF, Bolker JA. 2003. Ecological developmental biology:preface to the symposium. Evol Dev 5:3–8.

Gomez MC, Pope CE, Giraldo A, Lyons LA, Harris RF, KingAL, Cole A, Godke RA, Dresser BL. 2004. Birth of AfricanWildcat cloned kittens born from domestic cats. Cloningstem cells 6:247–258.

Gordon JW, Scangos GA, Plotkin DJ, Barbosa JA, Ruddle FH.1980. Genetic transformation of mouse embryos by micro-injection of purified DNA. Proc Nat Acad Sci USA 77:7380–7384.

Hall BK. 2002. Palaeontology and evolutionary developmentalbiology: a science of the nineteenth and twenty-firstcenturies. Palaeontology 45:647–669.

Hannon GJ, Rossi JJ. 2004. Unlocking the potential of thehuman genome with RNA interference. Nature 431:371–378.

Herault Y, Rassoulzadegan M, Cuzin F, Duboule D. 1998.Engineering chromosomes in mice through targeted meioticrecombination (TAMERE). Nat Genet 20:381–384.

Holland LZ, Laudet V, Schubert M. 2004. The chordateamphioxus: an emerging model organism for developmentalbiology. Cell Mol Life Sci 61:2290–2308.

Jenner RA, Wills MA. 2007. The choice of model organisms inevo-devo. Nat Rev 8:311–314.

Kocher TD, Conroy JA, McKaye KR, Stauffer JR. 1993.Similar morphologies of cichlid fish in Lakes Tanganyikaand Malawi are due to convergence. Mol Phylogenet Evol 2:158–165.

Koller BH, Hagemann LJ, Doetschman T, Hagaman JR,Huang S, Williams PJ, First NL, Maeda N, Smithies O.1989. Germ-line transmission of a planned alteration madein a hypoxanthine phosphoribosyltransferase gene byhomologous recombination in embryonic stem cells. ProcNat Acad Sci USA 86:8927–8931.

Krebs HA. 1975. The August Krogh principle. J Exp Zool 194:221–226.

Lakso M, Sauer B, Mosinger B Jr, Lee EJ, Manning RW, YuSH, Mulder KL, Westphal H. 1992. Targeted oncogeneactivation by site-specific recombination in transgenic mice.Proc Nat Acad Sci USA 89:6232–6236.

Lanza RP, Cibelli JB, Diaz F, Moraes CT, Farin PW, Farin CE,Hammer CJ, West MD, Damiani P. 2000. Cloning of anendangered species (Bos gaurus) using interspecies nucleartransfer. Cloning 2:79–90.

Lee BC, Kim MK, Jang G, Oh HJ, Yuda F, Kim HJ, ShamimMH, Kim JJ, Kang SK, Schatten G, Hwang WS. 2005. Dogscloned from adult somatic cells. Nature 436:641.

Lee K. 2001. Can cloning save endangered species? CurrBiol 11:R245–R246.

Lee TM. 2004. Octodon degus: a diurnal, social, and long-livedrodent. ILAR J/Nat Res Council, Institute of LaboratoryAnimal Resources 45:14–24.

Lindsley DL, Sandler L, Baker BS, Carpenter AT, Denell RE,Hall JC, Jacobs PA, Miklos GL, Davis BK, Gethmann RC,Hardy RW, Steven AH, Miller M, Nozawa H, Parry DM,Gould-Somero M. 1972. Segmental aneuploidy and thegenetic gross structure of the Drosophila genome. Genetics71:157–184.

Liu WY, Wang Y, Sun YH, Wang Y, Wang YP, Chen SP, ZhuZY. 2005. Efficient RNA interference in zebrafish embryosusing siRNA synthesized with SP6 RNA polymerase. DevGrowth Differ 47:323–331.

Losos JB, Jackman TR, Larson A, Queiroz K, Rodriguez-Schettino L. 1998. Contingency and determinism in repli-cated adaptive radiations of island lizards. Science 279:2115–2118.

Mansour SL, Thomas KR, Capecchi MR. 1988. Disruption ofthe proto-oncogene int-2 in mouse embryo-derived stemcells: a general strategy for targeting mutations to non-selectable genes. Nature 336:348–352.

Margulies M, Egholm M, Altman WE, Attiya S, Bader JS,Bemben LA, Berka J, Braverman MS, Chen YJ, Chen Z,Dewell SB, Du L, Fierro JM, Gomes XV, Godwin BC, He W,Helgesen S, Ho CH, Irzyk GP, Jando SC, Alenquer ML,Jarvie TP, Jirage KB, Kim JB, Knight JR, Lanza JR,Leamon JH, Lefkowitz SM, Lei M, Li J, Lohman KL, Lu H,Makhijani VB, McDade KE, McKenna MP, Myers EW,Nickerson E, Nobile JR, Plant R, Puc BP, Ronan MT, Roth

SELECTING NEW MODEL SPECIES 345

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b

Page 10: Escaping the mouse trap: the selection of new Evo-Devo model species

GT, Sarkis GJ, Simons JF, Simpson JW, Srinivasan M,Tartaro KR, Tomasz A, Vogt KA, Volkmer GA, Wang SH,Wang Y, Weiner MP, Yu P, Begley RF, Rothberg JM. 2005.Genome sequencing in microfabricated high-density pico-litre reactors. Nature 437:376–380.

McCreath KJ, Howcroft J, Campbell KH, Colman A, SchniekeAE, Kind AJ. 2000. Production of gene-targeted sheep bynuclear transfer from cultured somatic cells. Nature405:1066–1069.

Mello CC, Conte D Jr. 2004. Revealing the world of RNAinterference. Nature 431:338–342.

Metscher BD, Ahlberg PE. 1999. Zebrafish in context: uses of alaboratory model in comparative studies. Dev Biol 210:1–14.

Minelli A. 2003. The development of animal form. Onto-geny, morphology, and evolution. Cambridge: CambridgeUniversity Press.

Norgren RB Jr. 2004. Creation of non-human primateneurogenetic disease models by gene targeting and nucleartransfer. Reprod Biol Endocrinol 2:40.

Nusslein-Volhard C, Dahm R. 2002. Zebrafish. Oxford: OxfordUniversity Press. 303p.

Nusslein-Volhard C, Wieschaus E. 1980. Mutations affecting seg-ment number and polarity in Drosophila. Nature 287:795–801.

Nyholm SV, Stabb EV, Ruby EG, McFall-Ngai MJ. 2000.Establishment of an animal-bacterial association: recruitingsymbiotic vibrios from the environment. Proc Nat Acad SciUSA 97:10231–10235.

Orban PC, Chui D, Marth JD. 1992. Tissue- and site-specificDNA recombination in transgenic mice. Proc Nat Acad SciUSA 89:6861–6865.

Paigen K. 2003. One hundred years of mouse genetics: anintellectual history. I. The classical period (1902–1980).Genetics 163:1–7.

Palmiter RD, Brinster RL, Hammer RE, Trumbauer ME,Rosenfeld MG, Birnberg NC, Evans RM. 1982. Dramatic growthof mice that develop from eggs microinjected with metallothio-nein-growth hormone fusion genes. Nature 300:611–615.

Peng S, York JP, Zhang P. 2006. A transgenic approach forRNA interference-based genetic screening in mice. Proc NatAcad Sci USA 103:2252–2256.

Phelps CJ, Koike C, Vaught TD, Boone J, Wells KD, Chen SH,Ball S, Specht SM, Polejaeva IA, Monahan JA, Jobst PM,Sharma SB, Lamborn AE, Garst AS, Moore M, Demetris AJ,Rudert WA, Bottino R, Bertera S, Trucco M, Starzl TE, DaiY, Ayares DL. 2003. Production of alpha 1,3-galactosyl-transferase-deficient pigs. Science 299:411–414.

Rideout WM 3rd, Eggan K, Jaenisch R. 2001. Nuclear cloningand epigenetic reprogramming of the genome. Science 293:1093–1098.

Ruber L, Verheyen E, Meyer A. 1999. Replicated evolution oftrophic specializations in an endemic cichlid fish lineage fromLake Tanganyika. Proc Nat Acad Sci USA 96:10230–10235.

Rubinson DA, Dillon CP, Kwiatkowski AV, Sievers C, Yang L,Kopinja J, Rooney DL, Ihrig MM, McManus MT, Gertler FB,Scott ML, Van Parijs L. 2003. A lentivirus-based system tofunctionally silence genes in primary mammalian cells, stemcells and transgenic mice by RNA interference. Nat Genet33:401–406.

Schuster P, Fontana W, Stadler PF, Hofacker IL. 1994. Fromsequences to shapes and back: a case study in RNAsecondary structures. Proc R Soc Lond B 255:279–284.

Shin T, Kraemer D, Pryor J, Liu L, Rugila J, Howe L, Buck S,Murphy K, Lyons L, Westhusin M. 2002. A cat cloned bynuclear transplantation. Nature 415:859.

St Johnston D. 2002. The art and design of genetic screens:Drosophila melanogaster. Nat Rev 3:176–188.

Stadler BM, Stadler PF, Wagner GP, Fontana W. 2001. Thetopology of the possible: formal spaces underlying patternsof evolutionary change. J Theor Biol 213:241–274.

Stappenbeck TS, Hooper LV, Gordon JI. 2002. Developmentalregulation of intestinal angiogenesis by indigenous microbesvia Paneth cells. Proc Nat Acad Sci USA 99:15451–15455.

Sturtevant AH. 1965. A history of genetics. New York: Harperand Row. 167p.

Temple JL. 2004. The musk shrew (Suncus murinus): a modelspecies for studies of nutritional regulation of reproduction.ILAR J/Nat Res Council, Institute Laboratory AnimResources 45:25–34.

Thomas KR, Capecchi MR. 1987. Site-directed mutagenesis bygene targeting in mouse embryo-derived stem cells. Cell 51:503–512.

Thomas KR, Capecchi MR. 1990. Targeted disruption of themurine int-1 proto-oncogene resulting in severe abnormal-ities in midbrain and cerebellar development. Nature 346:847–850.

Thompson S, Clarke AR, Pow AM, Hooper ML, Melton DW.1989. Germ line transmission and expression of a correctedHPRT gene produced by gene targeting in embryonic stemcells. Cell 56:313–321.

Tiscornia G, Tergaonkar V, Galimi F, Verma IM. 2004. CRErecombinase-inducible RNA interference mediated by lenti-viral vectors. Proc Nat Acad Sci USA 101:7347–7351.

Tzika A, Milinkovitch MC. 2007. A pragmatic approach forselecting Evo-Devo model species in amniotes. In: Minelli A,Fusco G, editors. Evolving pathways; key themes inevolutionary developmental biology. Cambridge: CambridgeUniversity Press. in press.

Van Buskirk J. 2002. A comparative test of the adaptiveplasticity hypothesis: relationships between habitat andphenotype in Anuran larvae. Am Nat 160:87–102.

Wagner GP. 2001. What is the promise of developmentalevolution? Part II: a causal explanation of evolutionaryinnovations may be impossible. J Exp Zool 291:305–309.

Wagner GP, Larsson HCE. 2003. What is the promiseof developmental evolution? III. The crucible of develop-mental evolution. J Exp Zool B: Mol Dev Evol 300B:1–4.

Wagner TE, Hoppe PC, Jollick JD, Scholl DR, Hodinka RL,Gault JB. 1981. Microinjection of a rabbit beta-globin geneinto zygotes and its subsequent expression in adult mice andtheir offspring. Proc NatAcad Sci USA 78:6376–6380.

Wiens JJ, Brandley MC, Reeder TW. 2006. Why does a trait evolvemultiple times within a clade? Repeated evolution of snake-likebody form in squamate reptiles. Evolution 60:123–141.

Wigley P, Becker C, Beltrame J, Blake T, Crocker L, HarrisonS, Lyons I, McKenzie Z, Tearle R, Crawford R, Robins A.1994. Site-specific transgene insertion: an approach. ReprodFertil Dev 6:585–588.

Wilmut I, Schnieke AE, McWhir J, Kind AJ, Campbell KH.1997. Viable offspring derived from fetal and adult mam-malian cells. Nature 385:810–813.

Woods GL, White KL, Vanderwall DK, Li GP, Aston KI, BunchTD, Meerdo LN, Pate BJ. 2003. A mule cloned from fetalcells by nuclear transfer. Science 301:1063.

Zhou Q, Renard JP, Le Friec G, Brochard V, Beaujean N,Cherifi Y, Fraichard A, Cozzi J. 2003. Generation of fertilecloned rats by regulating oocyte activation. Science 302:1179.

Zwaka TP, Thomson JA. 2003. Homologous recombination inhuman embryonic stem cells. Nat Biotechnol 21:319–321.

M.C. MILINKOVITCH AND A. TZIKA346

J. Exp. Zool. (Mol. Dev. Evol.) DOI 10.1002/jez.b