use with caution: developmental systems divergence and

53

YALE JOURNAL OF BIOLOGYAND MEDICINE 82 (2009), pp.53-66.Copyright © 2009.

REVIEW

Use with caution: Developmental systemsdivergence and potential pitfalls of animalmodels

Vincent J. LynchDepartment of Ecology and Evolutionary Biology, Yale University, New Haven,Connecticut

Transgenic animal models have played an important role in elucidating gene functions andthe molecular basis development, physiology, behavior, and pathogenesis. Transgenic mod-els have been so successful that they have become a standard tool in molecular geneticsand biomedical studies and are being used to fulfill one of the main goals of the post-ge-nomic era: to assign functions to each gene in the genome. However, the assumption thatgene functions and genetic systems are conserved between models and humans is takenfor granted, often in spite of evidence that gene functions and networks diverge during evo-lution. In this review, I discuss some mechanisms that generate functional divergence andhighlight recent examples demonstrating that gene functions and regulatory networks di-verge through time. These examples suggest that annotation of gene functions based solelyon mutant phenotypes in animal models, as well as assumptions of conserved functionsbetween species, can be wrong. Therefore, animal models of gene function and human dis-ease may not provide appropriate information, particularly for rapidly evolving genes andsystems.

Biomedical studies of animal modelshave played an important role in elucidatingthe molecular basis and pathogenesis ofhuman disease, as well as in developing andtesting novel therapies [1,2].As early as the1930s, animal models such as the fruit fly(Drosophila melanogaster), zebrafish(Danio rerio), and mouse (Mus musculus)were being used to study the embryology

and development of tissue and organ sys-tems [1,2]. These early studies significantlyadvanced our understanding of embryologyand development by taking advantage offorward-genetic approaches to describe andcharacterize mutants, but it was not until thedevelopment of target transgenesis that itbecame possible to generate specific animalmodels of human disease.

To whom all correspondence should be addressed: Vincent J. Lynch, Department of Ecologyand Evolutionary Biology, Yale University, PO Box 208106, New Haven, CT 06520-8106;E-mail: [email protected].

†Abbreviations: Dfd, Drosophila homolog Deformed; PRL, prolactin; PLCZ1, phospholipaseC zeta 1; NLS, nuclear localization signal; MHC, major histocompatibility; PGR, proges-terone receptor; tin, tinman; DSD, developmental system drift; R-fng, Radical-fringe; TSPY,testis-specific protein, Y-encoded; DAF, Derived Allele Frequency; ACR, Acrosin; DAZ,Deleted in Azoospermia; DAZL, DAZ-like; PLCZ1, phospholipase C zeta 1; NLS, nuclear lo-calization signal; KOMP, Knockout Mouse Project.

The development of transgenic animalsin the early 1980s was a major breakthroughin molecular genetics and marked the begin-ning of modern biomedical studies. Genetargeting in model organisms enabled thedeletion (“knockout”) or replacement(“knockin”) of any gene in the genome, al-lowing investigation of that gene’s functionin development, morphology, physiology,and even behavior [1-3]. Gene targeting hasbeen so successful in elucidating the molec-ular function of genes that it has become astandard tool in molecular genetics and bio-medical studies and now is being used to re-solve one of the major goals of thepost-genomic era: assigning a function toeach gene in the genome [4]. Despite thegreat success of animal models in studies ofgene function and disease, those studies arebased on the critical underlying assumptionthat gene functions and developmental sys-tems are conserved between models and hu-mans — which may not always be the case.

In this review, I outline some of themechanisms that lead to evolutionary diver-gence in protein functions and gene regulatorynetworks and discuss how divergence be-tween species can complicate studies ofmodelorganisms. I begin with a brief summary ofthe biological justifications for using animalmodels to study development and disease anddiscuss how principles emerging from evolu-tionary developmental biology challenge theassumption of conserved genetic architecture.Finally, I provide some examples of diver-gence in protein function and gene regulationbetween mouse — a common model organ-ism for studying gene function and disease—and humans.While model organisms are a vi-tally important part of biomedical research,the assumption that functions are conservedbetweenmodels and other organisms needs tobe critically examined on a case-by-case basis.

EVOLUTIONARY ANDDEVELOPMENTAL JUSTIFICATIONSFOR ANIMAL MODELS

Comparative anatomists have long rec-ognized the similarity of tissue and organsystems between diverse animals, particu-

larly within mammals. Indeed, these sharedsimilarities led early comparative embryol-ogists to formalize the concept of “homol-ogy,” or similarity between characteristicsbecause of shared ancestry, as a distinct bi-ological phenomenon [5]. The numeroussimilarities in nervous, cardiovascular, en-docrine, immune, musculoskeletal, and re-productive system development andfunction between model organisms such asthe mouse, fish, fly, and humans provide astrong evolutionary foundation that the de-velopmental processes giving rise to homol-ogous structures is likely to be similarbetween different species.

Perhaps the strongest evidence that ho-mologous genes control the development ofhomologous structures is the finding thatproteins between remarkably divergent or-ganisms can be functionally equivalent,often despite great divergence in both pro-tein similarity and time. For example, thefirst use of transgenics to demonstrate thathomologous genes functioned in homolo-gous ways was between the human Hoxtranscription factor HoxB-4 and itsDrosophila homolog Deformed (Dfd†).McGinnis and colleagues (1990) testedwhether the humanHoxB-4 gene could sub-stitute for the regulatory functions of Dfd inDrosophila embryos by inserting the HoxB-4 gene into the Drosophila genome. Amaz-ingly, the human HoxB-4 gene can properlyregulate endogenous Dfd expression in de-veloping embryonic and larval cells throughits autoregulatory element [6]. Thus, the au-toregulatory function of the human andDrosophila homologs has been conservedafter more than 800 million years of diver-gence.

Probably the best-known example of ho-mologous proteins initiating homologous de-velopmental programs is the transcriptionfactor Pax6. Loss of function mutations inPax6 result in similar eye malformations inmice, humans, andDrosophila [7-9], indicat-ing that Pax6 has a conserved function in eyeformation, despite the vast differences be-tween vertebrate and invertebrate eyes. Ec-topic expression of eyeless, the Drosophilahomolog of the vertebrate Pax6 gene, in the

54 Lynch: Developmental systems divergence and pitfalls of animal models

antennae, legs, and wings of flies induces eyedevelopment on these body parts, indicatingeyeless functions a “master regulator” thatswitches on the developmental program foreye formation [10]. Remarkably, ectopic ex-pression of the mouse Pax6 gene inDrosophila embryos induces the formation ofwell developed compound eyes [10], whileexpression of the Drosophila eyeless gene inXenopus embryos leads to the induction ofvertebrate eyes [11]. These data strongly indi-cate the function of Pax genes as high-levelregulatory switches that activate the gene reg-ulatory network leading to eye developmentis conserved between animals.

Numerous other examples of functionalequivalence have been identified betweenorthologous genes from very different or-ganisms [12-19] and between paralogousgenes from ancient gene and genome dupli-cations [20-27]. The broad consensus result-ing from these studies is that developmentalsystems do not change much through evolu-tion, and, therefore, studies of gene functionin one species are directly applicable to an-other. But is this conclusion justified?

CHALLENGES TO THEASSUMPTION OF CONSERVEDGENETIC ARCHITECTURE

I have briefly reviewed some of the ev-idence suggesting that protein functions andthe developmental systems to which theycontribute are so conserved during evolutionthat biomedical studies in model organismsare directly applicable to human develop-ment, physiology, and disease. However, nu-merous studies have found that homologousproteins do not necessarily remain function-ally equivalent during evolution [28-55]. Forexample, the transcription factor HoxA-11has evolved a novel ability to activate pro-lactin (PRL) expression in endometrial stro-mal cells in placental mammals, whileHoxA-11 from non-placental mammals likeopossum, platypus, and chicken lack activa-tion functions and can only repress PRL ex-pression. Similarly, mouse phospholipase Czeta 1 (PLCZ1) is actively transported intothe nucleus shortly after fertilization, but

PLCZ1 from other mammals lack the nu-clear localization signal (NLS) found inmouse PLCZ1 and do not enter the nucleus.Furthermore, evolutionary modification ofgene regulatory networks and developmen-tal systems are mediated through changes inhow genes are regulated and function [56-60], thus the assumption that homologousgenes function in homologous ways be-tween models and humans needs to be criti-cally examined. In the following section, Ireview some of the ways that developmentalsystems can diverge and discuss how thesefactors can bias biomedical studies that as-sume conservation of regulatory systems be-tween models and humans.

Positive selection promotes functionaldivergence

Many powerful methods for detectingadaptive molecular evolution have been de-veloped over the last decade. These methodscompare the rate of synonymous (non-amino acid changing) and non-synonymous(amino acid changing) nucleotide substitu-tions to infer the strength and direction of se-lection acting on protein coding genes[61-63]. Since synonymous nucleotide sub-stitutions do not lead to amino acid changes,they are not exposed to natural selection act-ing on the protein structure/function and ac-cumulate in the genome at a more or lessclock-like, or neutral, rate. On the otherhand, non-synonymous nucleotide changesgenerate amino acid changes and are, there-fore, exposed to the action of selection. If anamino acid change is deleterious for thestructure/function of a protein, the new alle-les’ frequency in the population will remainlow and eventually disappear. If, however,the amino acid change is beneficial becauseit generates a novel function or contributesother beneficial effects, the frequency of thatallele in the population will increase until itcompletely replaces the ancestral allele.Thus, when amino acid changes are adap-tive, the rate of fixation in the populationwill increase above the background rate ofsynonymous substitutions [61-63].

Numerous cases of molecular adapta-tion have been identified in various systems

55Lynch: Developmental systems divergence and pitfalls of animal models

from viruses to humans with the commontheme that positive selection is very oftenassociated with an escalating arms race,such as between immune genes and para-sites, or the emergence of novel functions.For example, one of the first demonstrationsof molecular adaption was in the antigenrecognition site of major histocompatibility(MHC) genes. It was shown that amino acidsubstitutions occurred more rapidly at thislocus than either synonymous substitutionsor amino acid changes in other parts of theprotein [64,65]. Positive selection can alsoact to modify ancestral functions or generatenovel ones. For instance, the progesteronereceptor (PGR) evolved extremely rapidlyin humans and chimpanzees with the major-ity of amino acid substitutions occurring inparts of the protein important for transcrip-tional activity [66]. Remarkably, theepisodes of rapid PGR evolution are coinci-dent with changes in the mechanism of par-turition in higher apes, suggesting that usingrodents as a model organism to study PGRactions in humans and other primates maynot produce meaningful results.

One of the major findings from recentgenome-wide scans for adaptively evolvinggenes is that positive selection may be morecommon than previously thought [67-69].These analyses suggest that positive selec-tion in humans is strongest for X-linkedgenes and genes related to the immune sys-tem, reproduction, and sensory perception[69]. Unlike early studies of positive selec-tion in humans, which focused on a fewgenes of interest (PRM1, PRM2, SRY,FOXP2, G6PD, and MC1R), many of thegenes identified from genome-wide scansare novel [69]. They include several geneswith testis-specific expression (USP26,C15orf2, and HYAL3), immune regulation(CD58, APOBEC3F, and CD72), tumorantigens (SAGE1 andMAGEC2), and otherswith unknown functions (FLJ46156I,ABHD1, and LOC389458) [67-69]. Many ofthe genes that were positively selected in hu-mans are involved in resistance to malaria(HBB, CD40L, FY, and CD36), HIV (CCR5-∆32) or other infectious diseases (MHC).Other genes under selection are involved in

skin pigmentation (OCA2, MYO5A,DTNBP1, TYRP1, and SLC24A5), diet (LCTand ADH), brain development (MCPH1,ASPM, CDK5RAP2, CENPJ, GABRA4,PSEN1, SYT1, SLC6A4, and SNTG1) andbone morphogenesis (BMP3, BMPR2, andBMP5), among many others [67-69].Analy-sis of Affymetrix array data indicated thatpositively selected genes are more tissue-specific than other genes and more fre-quently expressed in spleen, testes, liver, andbreast than other tissues [67]. Remarkably,the number of positively selected genes issubstantially smaller in humans than inchimpanzees [70]. These results suggest thatknockouts of mouse genes to identify theirfunctions in humans may not identifyhuman-specific functions, particularly forquickly evolving genes under positive selec-tion. Indeed, a recent study (discussed inmore detail below) found that genes withdifferential effects when mutated in humansvs. mice are commonly associated with in-creased rates of evolution and positive se-lection [96].

Orthologous proteins, divergentnetworks

One of the best studied gene regulatorynetworks specifies the development of theheart [71,72]. The fundamental functionalcomponents of hearts are the cardiac musclecells that express an array of contractile pro-teins, such as actin, myosin, troponin, andtropomyosin. The first heart-like organsevolved more than 500 million years ago inthe ancestor of extant bilaterians and proba-bly resembled the simple tube-like hearts oftunicates, amphioxus, and Drosophila [71].During evolution, this simple, single-layeredtube, which pumped via peristaltic contrac-tions, evolved into a more efficient pump,beating via synchronous contractions withthick muscular chambers functionally spe-cialized for receiving (atria) and sending(ventricles) blood [71]. Interestingly, musclecells are developmentally and evolutionarilyderived from mesoderm, a phylogeneticallyolder tissue type [71].

While the structure of invertebrate andvertebrate hearts are dramatically different,



Figure 1. Gene regulatory networks for heart development in Drosophila (top) and verte-brates (middle); shared linkages are shown at the bottom. A core set of regulatory genesare used in common between insect and vertebrates and are linked in a similar way inconserved subcircuits of the gene network. Gray boxes highlight different ways that thesame two nodes of the network are linked in Drosophila and vertebrates. Orthologs ofmany regulatory genes in vertebrate heart formation are not known in Drosophila. Imageused with permission from [72].

the core gene regulatory network for heartformation is highly conserved and governedby homologous transcription factors and sig-naling molecules in all animals (Figure 1)[71,72]. This conserved core of regulatorscontrols cardiac cell fate, the expression ofgenes for contractile proteins, and the mor-phogenesis of the heart [71]. For example,initiation of both Drosophila and mousehearts during development is dependent onthe homologous transcription factors tinman(tin) and Nkx2.5, respectively. Detailed net-work analysis showed that both tin andNkx2.5 function by regulating many of thesame target genes in their respective species.However, gene swap experiments that at-tempted to rescue tin-null flies with its mam-malian homolog Nkx2.5 uncovered that theywere only partially equivalent. Specifically,the mouse Nkx2.5 gene could completelyrescue the expression of some target genes,like FascIII, and partially rescue others likeMEF2, but the Nkx2.5 gene could not rescuethe expression of eve and zfb-1 in the devel-oping Drosophila heart [47,50].

Similar to tin and Nkx2.5, humanOtx1/2 genes can rescue theirDrosophila or-tholog, otd, in nervous system development[46,73]. The reverse, however, is only par-tially true. Otd can replace most Otx1/2functions in mouse nervous system develop-ment, but it is unable to rescue the develop-ment of the mesencephalon, cerebellarfoliation, or the lateral semicircular canalsof the inner ear [74]. These differences infunction between otd and Otx1/2 likely re-sult from divergence in their ability to regu-late downstream target genes. Montalta-Heet al. (2002) tested this possibility usingwhole genome microarrays in Drosophila,overexpressing otd andOtx2. Amazingly, ofthe 287 and 682 genes that responded to otdand Otx2 overexpression, respectively, only90 were shared target genes. Thus, eventranscription factors that are functionallyequivalent with respect to the developmentof a particular structure can be nonequiva-lent with respect to regulation of gene net-works at the systems level.

The functional equivalence oftin/Nxk2.5with respect to FascIII andMEF2

expression, but not eve and zfb-1, in the de-veloping heart and the divergence ofotd/Otx2 target genes and functions in spe-cific parts of nervous system development isparticularly revealing. FascIII is a cell sur-face antigen expressed in derivatives of vis-ceral mesoderm, andMEF2 is a transcriptionfactor that is a central component of generegulatory networks in all muscle types[47,50]. Thus, they are likely ancient targetsof tin/Nxk2.5, whose expression in musclecells predates the origin of the heart. How-ever, the ability of tin, but notNkx2.5, to reg-ulate eve and zfb-1 suggests these genes areeither derived, clade-specific targets of tin inDrosophila or Nkx2.5 has lost the ability toregulate them during the evolution of mam-mals. Similarly, the function of otd/Otx2 innervous system development is ancient, andtheir ability to regulate partially overlappingsets of target genes reflects the underlyinghomologies in nervous system developmentin bilaterians. Nevertheless, clade-specificfunctions have evolved for both genes, sug-gesting that while the basic developmentalplan of nervous system development is sim-ilar between bilaterians, otd/Otx2 have func-tionally diverged with respect to their targetgene repertoires and the ability to direct thedevelopment of species-specific neural struc-tures.

Like the divergence in protein func-tions, gene regulatory networks can divergeover time. Thus, homologous proteins maynot have equivalent functions in regulatorynetworks, or may not even be part of thesame network between different species.This suggests that functional genomic andbiomedical studies should be careful whendetermining gene function and network po-sition in models and applying it to otherspecies, particularly when the phylogeneticdistance between the species grows.

Developmental and physiologicalsystems drift

It is generally a reasonable assumption— and fundamental to biomedical studiesusing animal models — that homologousdevelopmental processes govern the devel-opment and function of homologous cells,


tissues, and organs. Indeed, animal modelshave provided invaluable information on de-velopment, disease, and gene function.However, a growing number of studies haveuncovered surprising divergence in develop-mental pathways, often without overtchanges in phenotype [75]. This phenome-non, the development of homologous sys-tems via divergent processes, has beentermed developmental system drift (DSD).

DSD is emerging as a general processthat arises after species diverge. Indeed, thedivergence in target genes between insectand mammalian tin/Nxk2.5 and otd/Otx2likely are special cases of DSD. While thedivergence between insect and mammaliandevelopmental systems may have occurredover 100 million years, DSD can arise overmuch shorter time frames. For example, tho-racic bristles are missing in hybrids betweenDrosophila melanogaster and Drosophilasimulans, even though the pattern of tho-racic bristles is identical between D.melanogaster and D. simulans [76]. Moredramatically, the second (T2) and third (T3)thoracic segments of hybrids between D.subobsura and D. madierensis are partiallytransformed into the first segment (T1)[77,78].

Hints of DSD also have been discov-ered in limb development and spermatogen-esis. Radical-fringe (R-fng), an extracellularprotein that binds Notch during cell signal-ing, is expressed at the dorsal-ventral bound-ary of the limb bud in mouse and chicken,where it promotes the formation of the api-cal epidermal ridge (AER; Johnston et al.1997). Remarkably, while R-fng is sufficientto induce AER formation in chickens [79],mouse knockouts of R-fng have no defectsin limb development [80]. A particularly in-teresting example of DSD is the testis-spe-cific protein, Y-encoded (TSPY) gene. Oneor more copies of the TSPY gene are foundon the Y chromosome of all placental mam-mals, and, as its name implies, its expressionis restricted to the testis [81]. Remarkably,although TSPY is functional in rat,Mus mus-culus has a single-copy TSPY pseudogenethat does not generate a functional tran-script; knockin of multiple human TSPY

does not alter spermatogenesis [81]. Obvi-ously, this natural knockout does not affectfertility in mice and restoration of functionalcopies of TSPY does not alter fertility, sug-gesting that the spermatogenesis has driftedaway from TSPY-dependence in mice.

Recent studies suggest that the molecu-lar basis of DSD likely is transcriptional reg-ulation. Various elements of transcriptionalregulation play central roles in development,homeostasis, and evolution. Initial studiescomparing genome-wide binding sites fortwo stem cell-specific transcription factorsfound large differences in binding site loca-tions between human and mouse [82]. In amore comprehensive analysis, Odom et al.(2007) systematically compared the bindingof four tissue-specific transcription factors(FOXA2, HNF1A, HNF4A, and HNF6)using ChIP-chip on hepatocytes purifiedfrom human and mouse livers. Despite theconserved function of these transcriptionfactors, between 41 percent and 89 percentof their binding sites were species-specific[83]. Moreover, genome-wide analysis ofpolymorphisms in human populations usingthe Derived Allele Frequency (DAF) test ofnatural selection found that many human-and primate-specific transcription factorbinding sites in cis-regulatory elements areevolving under positive selection [68].These results suggest that divergence ingene regulation is adaptively driven and canaccumulate particularly fast without leadingto obvious differences in phenotype [54,84].Thus, even morphologically similar organscan develop from divergent gene regulatorynetworks.

Pseudoorthology

The differential gain and loss of genesis an important source of functional varia-tion between species. Differences in genecontent between species arise from lineage-specific gains via duplication and losses viadeletion or inactivation. A consequence ofgene duplication is often functional differ-entiation because one copy is no longer con-strained to perform the ancestral functionand is free to acquire novel functions. Thisprocess of neofunctionalization has been


proposed to be the main source of new pro-tein functions during evolution [85].

While gene duplication followed byneofunctionalization is a major source ofmolecular novelty, it can also mask true or-thology— particularly if there is differentialloss of duplicate genes in two species. Forexample, if gene A duplicates before a spe-ciation event to produce paralogs A and A’,differential loss of duplicates after a subse-quent speciation can lead to one speciesmaintaining Awhile the other maintains A’.If the initial duplication event was associ-ated with neofunctionalization, then even

though both species apparently have a copyof the A gene, the gene’s functions will notbe the same and can be very different.

Unfortunately, there are few studies ofhow common differential gene loss is, thus,it is not clear how often it may bias func-tional genomic annotation of gene functionor biomedical studies of gene functions [86].However, a recent study of the mammalianbeta-globin gene family suggests it can becommon among some genes. Opazo et al.(2008) used a comparative genomic ap-proach to investigate β-globin gene turnoverin placental mammals. Interestingly, they


Figure 2. β-globin gene family evolution in eutherian mammals. Two successive duplicationsof a proto-ε gene gave rise to the γ and η genes in the stem-lineage of eutherian mammalsafter their divergence from marsupials. Consequently, the full complement of embryonic glo-bin genes (hemoglobin ε, γ and η) was present in the common ancestor of the eutherianmammals. The η gene was lost in the common ancestor of Xenarthrans and Afrotherians,and the γ gene was lost in Xenarthrans after divergence from the Afrotherians. The η and γgenes were independently lost in the Euarchontoglires and Laurasiatherians, respectively.Crosses indicate lineage-specific gene losses. Image used with permission from [87].

found that there was large-scale differentialloss and retention of genes after an initial ex-pansion of the non-adult portion of the β-globin gene cluster in the common ancestorof placental mammals (Figure 2). The dif-ferential sorting and species-specific dupli-cation/neofunctionalization of ε-, γ-, andη-globin gene lineages among groups of pla-cental mammals has produced species dif-ferences in the functions of hemoglobinisoforms, thereby generating variation in thecomplement of globin genes among mam-mals [87].

FUNCTIONAL DIVERGENCEBETWEEN MOUSE AND HUMANGENES

The examples discussed above indicatethat functional divergence in genes and reg-ulatory networks accumulate during evolu-tion. This suggests that a fundamentalassumption of comparative genomics andbiomedical studies, i.e., that gene functionsare conserved between model organisms andhumans, may be too simplistic. Indeed, be-yond case-by-case studies and anecdotal ev-idence, there have been few systematicexaminations to verify the assumption offunctional conservation between models andhumans. Examples of functional divergence,however, have received little attentionwithin the biomedical community, despite agrowing number of studies finding thatchanges in protein function may be commonduring mammalian evolution.

Perhaps one of the most dramatic exam-ples of an essential human gene that is non-essential in mouse is Acrosin (ACR). ACRenables sperm to penetrate the extracellularmatrix and fertilize oocytes by proteolyzingthe zona pellucida. Given the role of ACR infertilization, it is not surprising that numer-ous studies have implicated ACR in humanmale infertility [88,89]. Unexpectedly, how-ever, male ACR-null mice are fertile in spiteof the complete absence of acrosin proteaseactivity in the sperm [90]. Thus, ACR is notessential for sperm penetration of the zonapellucida or fertilization in mice, despite itsnecessity for human fertility. Interestingly,

mouse sperm have much lower levels ofACR than other rodents, such as rat or ham-ster, suggesting that other proteases mayfunction in place of ACR during mouse fer-tilization [91]. In fact, other serine proteasesin addition to ACR are present in mousesperm, but not in rat or hamster, which maybe an indication of developmental systemdrift in mammalian fertilization systems[91].

Like ACR, Deleted in Azoospermia(DAZ) andDAZ-like (DAZL) are involved inmale germ cell differentiation and function.DAZL is highly conserved during evolutionand has been isolated from species as di-verse as mammals, frogs, fish, and worms;DAZ is only found in apes and other OldWorld monkeys [92]. Thus, the DAZ geneduplicated sometime after Old World andNew World monkeys diverged but beforethe radiation of the Old World monkeys. Inmost mammals, the ancestralDAZL is suffi-cient to complete gametogenesis; however,in humans, deletions removing the Y-chro-mosomal DAZ gene are often associatedwith azoospermia or oligospermia [93], indi-cating DAZ is necessary for spermatogene-sis. Interestingly, most amino acids in DAZare under strong selective constraint, sug-gesting they are important for protein struc-ture/function. Nevertheless, a few sitesevolving under diversifying selection sug-gest they are involved in a molecular armsrace or that these sites have differentspecies-specific functions [94]. Lineage-specific analysis indicated that human mem-bers of this gene family were evolving bypositive Darwinian selection, suggestingDAZ may have evolved human-specificfunctions [94].

Vogel et al. (2002) studied the functionsof the human DAZL and DAZ genes inmouse DAZL-null mice and found that bothDAZL and DAZ enabled production ofprophase spermatocytes. However, bothhuman genes failed to promote differentia-tion into mid- to late pachytenes. The humanDAZL transgene led to more early germ cellscompared with that observed in DAZ.Human DAZL and DAZ can only substitutefor early functions of the mouseDAZL, sug-


gesting gene- and species-specific functionhave evolved inDAZL/DAZ after human andmouse diverged and/or afterDAZL/DAZ du-plicated, particularly in their ability to drivemid- to late spermatogenesis. Thus, eventhough both human DAZL/DAZ and mouseDAZ are essential genes, the essentiality hasdiverged.

During fertilization, sperm trigger a se-ries of oscillations in the concentration ofintracellular calcium, which initiates eggactivation. The major protein delivered bythe sperm that induces the calcium flux isan isoform of phospholipase C termedphospholipase C zeta 1 (PLCZ1). Remark-ably, mouse PLCZ1 is transported into thenucleus after fertilization, but PLCZ1 from

other species, such as human, rat, and fish,are not. Ito et al. (2008) traced this differ-ence in nuclear localization ability to anovel nuclear localization signal (NLS) thatevolved in mouse within a preexisting ly-sine/arginine rich region of the protein(Figure 3). All PLCZ1 genes tested by Ito etal. (2008) could induce Ca2+ oscillationswhen expressed in mouse eggs, but the ac-tivity was highly variable. For example,Ca2+ oscillations induced by human PLCZ1continued far beyond the time of pronu-clear formation, whereas oscillations in-duced by rat PLCZ1 stopped far beforepronuclear formation. Ito and colleaguesfound that PLCZ1 sequestration into thepronucleus participates in termination of


Figure 3. Evolution of a novel nuclear localization signal (NLS) in PLCZ1. (a) Logo of the X-Y linker region from 31 species of amniotes. In this representation of a multiple sequencealignment the conservation of amino acid residues in the protein are indicated by the heightof the column (Bit Score) while the conservation of specific amino acids in at that site areshown by the height of letters within columns. Note that conservation is generally poor. (b)Superimposed structural models of human and mouse PLCZ1 proteins. Note that while thestructure of human and mouse PLCZ1 proteins are generally the same, the X-Y linker regionis much longer in humans leading to a longer disordered loop than in mouse.

Ca2+ oscillations of mouse embryos, but notother mammals, indicating the mechanismthat terminates Ca2+ oscillations after fer-tilization has diverged significantly, evenbetween the closely related mouse and rat[95].

The examples of ACR, DAZ/DAZL, andPLCZ1 show that homologous proteins arenot necessarily functionally equivalent, butthese case studies only hint that a larger pat-tern of functional divergence may exist. Re-cently, Liao and Zhang (2008) systematicallycompared the phenotype of null mutationsfor 120 essential human genes, i.e., genes as-sociated with death before puberty or infertil-ity, which have been knocked out in mice.Remarkably, ~22 percent of human essentialgenes were non-essential in mouse. Changesin essentiality between species were associ-ated with adaptive evolution of the proteins,consistent with previous findings thatchanges in protein function are often associ-ated with rapid evolution. By examining theevolution of genes that have changed essen-tiality between humans and mice in a broadercontext, they found that positive selectionwas much more common in the human line-age than in the rodent, suggesting that thedifferences in essentiality arose duringhuman evolution [96].

CONCLUSIONS AND OUTLOOKBiomedical studies of animal models

have played incredibly important roles inelucidating the molecular basis develop-ment, physiology, behavior, and pathogene-sis of human disease. Transgenic models areso successful in elucidating the molecularfunction of genes that they are now a stan-dard tool in molecular genetics and biomed-ical studies and are being used to assignfunctions to each gene in the genomethrough initiatives such as the KnockoutMouse Project (KOMP). Yet the assumptionthat gene functions and developmental sys-tems are conserved between models and hu-mans is taken for granted, in spite of theevidence that gene functions and gene net-works can diverge through evolution. Themechanisms and examples of functional di-

vergence discussed above indicate that genestudies in model organisms need to be ex-amined carefully for several potentialbiases. Annotation of gene functions basedsolely on mutant phenotypes in animal mod-els, as well as assumptions of conservedfunctions between species, may be mislead-ing. Perhaps more importantly, animal mod-els of human disease may not provideappropriate information on gene function,particularly for rapidly evolving genes andsystems. It is vitally important to demon-strate equivalence between the model andhuman gene with respect to the particularfunction under study to avoid spurious con-clusions.

REFERENCES1. Lieschke GJ, Currie PD. Animal models ofhuman disease: zebrafish swims into view.Nature Reviews Genetics. 2007;8:250-67.

2. Rosenthal N, Brown S. The mouse ascend-ing: perspectives for human-disease models.Nat Cell Biol. 2007;9(9):993-9.

3. Hau J. Animal models for human disease.Berlin, Germany: Humana Press Inc.; 2008.

4. Koonin EV. Orthologs, paralogs and evolu-tionary genomics.Annu Rev Genet. 2005;39:309-38.

5. Amundson R. The changing role of the em-bryo in evolutionary thought: roots of Evo-Devo. Cambridge: Cambridge UniversityPress; 2005.

6. McGinnis N, Kuziora MA, McGinnis W.Human Hox-4.2 and Drosophila Deformedencode similar regulatory specificities inDrosophila embryos and larvae. Cell.1990;63(5):969-76.

7. Niimi T, Clements J, GehringWJ, et al. Dom-inant-negative form of the Pax6 homologeyeless for tissue-specific loss-of-functionstudies in the developing eye and brain indrosophila. Genesis. 2002;34(1-2):74-5.

8. Hill RE, Favor J, Hogan BLM, et al. MouseSmall eye results from mutations in a paired-like homeobox-containing gene. Nature.1991;354:522-5.

9. Jordan T, Hanson I, Zaletayev D, et al. Thehuman PAX6 gene is mutated in two patientswith aniridia. Nat Genet. 1992;1:238-332.

10. Halder G, Callaerts P, GehringWJ. Inductionof ectopic eyes by targeted expression of theeyeless gene in Drosophila. Science.1995;267(5205):1788-92.

11. Onuma Y, Takahashi S, Asashima M, et al.Conservation of Pax 6 function and upstreamactivation by Notch signaling in eye devel-opment of frogs and flies. Proc Acad Nat SciUSA. 2002;99(4):2020-5.


12. Shimeld SM. An Amphioxus Gli Gene Re-veals Conservation of Midline Patterning andthe Evolution of Hedgehog Signalling Diver-sity in Chordates. PLoSONE. 2007;2(9):e864.

13. Yu H, Schurr MJ, Deretic V. Functionalequivalence of Escherichia coli sigma E andPseudomonas aeruginosaAlgU: E. coli rpoErestores mucoidy and reduces sensitivity toreactive oxygen intermediates in algU mu-tants of P. aeruginosa. J Bacteriol.1995;177(11):3259-68.

14. Liu M, Su M, Lyons G, et al. Functional con-servation of zinc-finger homeodomain genezfh1/SIP1 in Drosophila heart development.Dev Genes Evol. 2006;216(11):683-93.

15. Wang W, Grimmer JF, Van De Water TR, etal. Hmx2 and Hmx3 Homeobox Genes Di-rect Development of the Murine Inner Earand Hypothalamus and Can Be FunctionallyReplaced by Drosophila Hmx. Dev Cell.2004;7(3):439-53.

16. Bickar D, Lehninger A, Brunori M, et al.Functional equivalence of monomeric (shark)and dimeric (bovine) cytochrome c oxidase.J Inorg Biochem. 1985;23(3-4):365-72.

17. Zhang P, Tan HT, Pwee KH, et al. Conserva-tion of class C function of floral organ devel-opment during 300 million years of evolutionfrom gymnosperms to angiosperms. Plant J.2004;37(4):566-77.

18. McIntosh KB, Bonham-Smith PC. Establish-ment of Arabidopsis thaliana ribosomal pro-tein RPL23A-1 as a functional homologue ofSaccharomyces cerevisiae ribosomal proteinL25. Plant Mol Biol. 2001;46(6):673-82.

19. Wang VY, Hassan BA, Bellen HJ, et al.Drosophila atonal Fully Rescues the Pheno-type of Math1 Null Mice: New FunctionsEvolve in New Cellular Contexts. Curr Biol.2002;12(18):1611-6.

20. Hirth F, Loop T, Egger B, et al. Functionalequivalence of Hox gene products in thespecification of the tritocerebrum during em-bryonic brain development of Drosophila.Development. 2001;128(23):4781-8.

21. Coronado R, Ahern CA, Sheridan DC, et al.Functional equivalence of dihydropyridine re-ceptor alpha1S and beta1a subunits in trigger-ing excitation-contraction coupling in skeletalmuscle. Biol Res. 2004;37(4):565-75.

22. Pan L, Yang Z, Feng L, et al. Functionalequivalence of Brn3 POU-domain transcrip-tion factors in mouse retinal neurogenesis.Development. 2005;132(4):703-12.

23. Roman BB, Wieringa B, KoretskyAP. Func-tional Equivalence of Creatine Kinase Iso-forms in Mouse Skeletal Muscle. J BiolChem. 1997;272(28):17790-4.

24. Chia IV, Costantini F. Mouse Axin andAxin2/Conductin Proteins Are FunctionallyEquivalent In Vivo. Mol Cell Biol.2005;25(11):4371-6.

25. Bouchard M, Pfeffer P, Busslinger M. Func-tional equivalence of the transcription factorsPax2 and Pax5 in mouse development. De-velopment. 2000;127(17):3703-13.

26. FleischmannA, Hafezi F, Elliott C, et al. Fra-1 replaces c-Fos-dependent functions inmice. Genes Dev. 2000;14(21):2695-700.

27. Acampora D, Annino A, Puelles E, et al.OTX1 compensates for OTX2 requirement inregionalisation of anterior neuroectoderm.Gene Expr Patterns. 2003;3(4):497-501.

28. Adamska M, Matus DQ, Adamski M, et al.The evolutionary origin of hedgehog pro-teins. Curr Biol. 2007;17(19):R836-7.

29. Anan K, Yoshida N, Kataoka Y, et al. Mor-phological Change Caused by Loss of theTaxon-Specific Polyalanine Tract in Hoxd-13. Mol Biol Evol. 2007;24(1):281-7.

30. Averof M, Akam M. Hox genes and the di-versification of insect and crustacean bodyplans. Nature. 1995;376(6539):420-3.

31. BranfordWW, Benson GV, Ma L, et al. Char-acterization of Hoxa-10/Hoxa-11 Tran-sheterozygotes Reveals FunctionalRedundancy and Regulatory Interactions.Dev Biol. 2000;224(2):373-87.

32. Chen M,Abele R, Tampe R. Functional Non-equivalence of ATP-binding Cassette Signa-ture Motifs in the Transporter AssociatedwithAntigen Processing (TAP). J Biol Chem.2004;279(44):46073-81.

33. EmiliA, Greenblatt J, Ingles CJ. Species-spe-cific interaction of the glutamine-rich activa-tion domains of Sp1 with the TATAbox-binding protein. Mol Cell Biol.1994;14(3):1582-93.

34. Falciani F, Hausdorf B, Schroder R, et al. Class3 Hox genes in insects and the origin of Zen.Proc Natl Acad Sci USA. 1996;93:8479-84.

35. Fondon JW III, Garner HR. Molecular ori-gins of rapid and continuous morphologicalevolution. Proc Natl Acad Sci USA.2004;101(52):18058-63.

36. Galant R, Carroll SB. Evolution of a tran-scriptional repression domain in an insectHox protein. Nature. 2002;415(6874):910-3.

37. Gerber HP, Seipel K, Georgiev O, et al. Tran-scriptional activation modulated by ho-mopolymeric glutamine and proline stretches.Science. 1994;263(5148):808-11.

38. Grenier JK, Carroll SB. Functional evolutionof the Ultrabithorax protein. Proc Natl AcadSci USA. 2000;97(2):704-9.

39. Hanks MC, Loomis CA, Harris E, et al.Drosophila engrailed can substitute formouse Engrailed1 function in mid-hindbrain,but not limb development. Development.1998;125(22):4521-30.

40. Hsia CC,McGinnisW. Evolution of transcrip-tion factor function. Current Opinion in Ge-netics & Development. 2003;13(2):199-206.

41. Hyman CA, Bartholin L, Newfeld SJ, et al.Drosophila TGIF Proteins Are TranscriptionalActivators.MolCellBiol. 2003;23(24):9262-74.

42. Kellerer S, Schreiner S, Stolt CC, et al. Re-placement of the Sox10 transcription factorby Sox8 reveals incomplete functional equiv-alence. Development. 2006;133(15):2875-86.


43. Li X, Liang J, Yu H, et al. Functional conse-quences of new exon acquisition in mam-malian chromodomainY-like (CDYL) genes.Trends Genet. 2007;23(9):427-31.

44. Lohr U, Yussa M, Pick L. Drosophila fushitarazu: a gene on the border of homeoticfunction. Curr Biol. 2001;11(18):1403-12.

45. Montalta-He H, Leemans R, Loop T, et al.Evolutionary conservation of otd/Otx2 tran-scription factor action: a genome-wide mi-croarray analysis in Drosophila. Genome Biol.2002;3(4):research0015.1-0015.15.

46. Nagao T, Leuzinger S, Acampora D, et al.Developmental rescue of Drosophila cephalicdefects by the human Otx genes. Proc AcadNatl Sci USA. 1998;95(7):3737-42.

47. Park M, Lewis C, Turbay D, et al. Differen-tial rescue of visceral and cardiac defects inDrosophila by vertebrate tinman-relatedgenes. Proc Acad Natl Sci USA.1998;95(16):9366-71.

48. Prager EM, Wilson AC. Slow evolutionaryloss of potential for interspecific hybridiza-tion in birds. Proc Acad Natl Sci USA.1975;72:200-4.

49. Punzo C, Plaza S, Seimiya M, et al. Func-tional divergence between eyeless and twinof eyeless in Drosophila melanogaster. De-velopment. 2004;131(16):3943-53.

50. Ranganayakulu G, Elliott DA, Harvey RP, etal. Divergent roles for NK-2 class homeoboxgenes in cardiogenesis in flies and mice. De-velopment. 1998;125(16):3037-48.

51. RonshaugenM,McGinnis N,McGinnisW.Hoxprotein mutation and macroevolution of the in-sect body plan. Nature. 2002;415(6874):914-7.

52. Stolt CC, Lommes P, Friedrich RP, et al. Tran-scription factors Sox8 and Sox10 performnon-equivalent roles during oligodendrocytedevelopment despite functional redundancy.Development. 2004;131(10):2349-58.

53. TanayA, RegevA, Shamir R. Conservation andevolvability in regulatory networks: The evolu-tion of ribosomal regulation in yeast. ProcAcadNatl Sci USA. 2005;102(20):7203-8.

54. Tuch BB, Galgoczy DJ, Hernday AD, et al.The Evolution of Combinatorial Gene Regu-lation in Fungi. PLoS Biology. 2008;6(2):e38.

55. Lynch VJ, Tanzer A, Wang Y, et al. Adaptivechanges in the transcription factor HoxA-11are essential for the evolution of pregnancyin mammals. Proc Acad Natl Sci USA.2008;105(39):14928-33.

56. Carroll SB. Evolution at TwoLevels: OnGenesand Form. PLoS Biology. 2005;3(7):e245.

57. Carroll SB. Endless FormsMostBeautiful. NewYork:W.W. Norton & Company, Inc.; 2005.

58. Wagner GP, Lynch VJ. Molecular evolutionof evolutionary novelties: the vagina anduterus of therian mammals. J Exp ZoologPart B Mol Dev Evol. 2005;304B(6):580-92.

59. Lynch VJ, Wagner GP. Resurrecting the roleof transcription factor change in developmen-tal evolution. Evolution. 2008;62(9):2131-54.

60. Wagner GP, Lynch VJ. The gene regulatorylogic of transcription factor evolution. Trendsin Ecology and Evolution. 2008;23(7):377-85.

61. Kimura M. Evolutionary rate at the molecularlevel. Science. 1968;217:624-6.

62. Kimura M. The Neutral Theory of molecularevolution. Cambridge: Cambridge UniversityPress; 1983.

63. King JL, Jukes TH. Non-Darwinian Evolu-tion. Science. 1969;164(3881):788-98.

64. Hughes AL, Nei M. Pattern of nucleotidesubstitution at major histocompatibility com-plex class I loci reveals overdominant selec-tion. Nature. 1988;335:167-70.

65. Chen FC, Chen CJ, Ho JY, et al. Identifica-tion and evolutionary analysis of novel exonsand alternative splicing events using cross-species EST-to-genome comparisons inhuman, mouse and rat. BMC Bioinformatics.2006;7(1):136.

66. Chen C, Opazo JC, Erez O, et al. The humanprogesterone receptor shows evidence ofadaptive evolution associated with its abilityto act as a transcription factor. Molecular Phy-logenetics and Evolution. 2008;47(2):637-49.

67. Kosiol C, Vinar T, da Fonseca RR, et al. Pat-terns of Positive Selection in Six MammalianGenomes. PLoS Genet. 2008;4(8):e1000144.

68. Voight BF, Kudaravalli S,Wen X, et al.AMapof Recent Positive Selection in the HumanGenome. PLoS Biology. 2006;4(3):e72.

69. Sabeti PC, Schaffner SF, Fry B, et al. PositiveNatural Selection in the Human Lineage. Sci-ence. 2006;312(5780):1614-20.

70. Bakewell MA, Shi P, Zhang J. More genesunderwent positive selection in chimpanzeeevolution than in human evolution. Proc NatlAcad Sci USA. 2007;104(18):7489-94.

71. Olsen ER. Gene regulatory networks in theevolution and development of the heart. Sci-ence. 2006;313:1922-7.

72. Davidson EH, Erwin DH. Gene regulatorynetworks and the evolution and animal bodyplans. Science. 2006;331:796-800.

73. Leuzinger S, Hirth F, Gerlich D, et al. Equiv-alence of the fly orthodenticle gene and thehuman OTX genes in embryonic brain devel-opment of Drosophila. Development.1998;125(9):1703-10.

74. Acampora D,Avantaggiato V, Tuorto F, et al.Murine Otx1 and Drosophila otd genes shareconserved genetic functions required in in-vertebrate and vertebrate brain development.Development. 1998;125(9):1691-702.

75. True JR, Haag ES. Developmental system driftand flexibility in evolutionary trajectories. Evo-lution and Development. 2001;3(2):109-19.

76. Takano T. Loss of notum macrochaetae as aninterspecific hybrid anomaly betweenDrosophila melanogaster and D. simulans.Genetics. 1998;129:1435-50.

77. Khadem M, Krimbas CB. Studies of thespecies barrier between Drosophila madeiren-


sis and Drosophila subobscura. Hereditas.1991;114:189-95.

78. Johnston SH, Rauskolb C, Wilson R, et al. Afamily of mammalian Fringe genes implicatedin boundary determination and the Notch path-way. Development. 1997;124(11):2245-54.

79. Laufer E, Dahn R, Orozco OE, et al. Expres-sion of Radical fringe in limb-bud ectodermregulates apical ectodermal ridge formation.Nature. 1997;386:366-73.

80. Moran JL, Levorse JM, Vogt TF. Limbs movebeyond the radical fringe. Nature.1999;399:742-3.

81. Schubert S, Skawran B, Dechend F, et al.Generation and Characterization of a Trans-genic Mouse with a Functional HumanTSPY. Biol Reprod. 2003;69(3):968-75.

82. LohYH,Wu Q, Chew JL, et al. The Oct4 andNanog transcription network regulatespluripotency in mouse embryonic stem cells.Nat Genet. 2006;38(4):431-40.

83. Odom DT, Dowell RD, Jacobsen ES, et al.Tissue-specific transcriptional regulation hasdiverged significantly between human andmouse. Nat Genet. 2007;39(6):730-2.

84. Tuch BB, Li H, Johnson AD. Evolution ofEukaryotic Transcription Circuits. Science.2008;319(5871):1797-9.

85. Ohno S. Evolution by gene duplication.Berlin, Germany: Springer-Verlag; 1970.

86. Studer RA, Robinson-Rechavi M. How con-fident can we be that orthologs are similar,but paralogs differ? Trends Genet. In Press,Corrected Proof.

87. Opazo JC, Hoffmann FG, Storz JF. Differen-tial loss of embryonic globin genes during theradiation of placental mammals. Proc NatlAcad Sci USA. 2008;105(35):12950-5.

88. Koel C, Tuhin D, Baidyanath C, et al.Acrosin activity as a potential marker for

sperm membrane characteristics in unex-plained male infertility. Fertility and Sterility.2005;83(1):104-9.

89. Shimizu Y, Kodama H, Fukuda J, et al. Evi-dence of proacrosin molecule abnormality asa possible cause of low acrosin activity andunexplained failure of fertilization in vitro. JAndrol. 1997;18(3):281-8.

90. Baba T, Azuma S, Kashiwabara S, et al.Sperm from mice carrying a target mutationof the acrosin gene can penetrate the oocytezona pellucia and effect fertilization. J BiolChem. 1994;269(50):31845-9.

91. KazuoY, HondaA, Kashiwabara S, et al. Dif-ference of acrosomal serine protease systembetween mouse and other rodent sperm. De-velopmental Genetics. 1999;25(2):115-22.

92. Vogel T, Speed RM, RossA, et al. Partial res-cue of the Dazl knockout mouse by thehuman DAZL gene. Mol Hum Reprod.2002;8(9):797-804.

93. Vogt P, Edelmann A, Kirsch S, et al. HumanY chromosome azoospermia factors (AZF)mapped to different subregions inYq11. HumMol Genet. 1996;5(7):933-43.

94. Bielawski JP, Yang Z. Positive and NegativeSelection in the DAZ Gene Family. Mol BiolEvol. 2001;18(4):523-9.

95. Ito M, Shikano T, Oda S, et al. Difference inCa2+ Oscillation-Inducing Activity and Nu-clear Translocation Ability of PLCZ1, anEgg-Activating Sperm Factor Candidate, Be-tweenMouse, Rat, Human, andMedaka Fish.Biol Reprod. 2008;78(6):1081-90.

96. Liao BY, Zhang J. Null mutations in humanand mouse orthologs frequently result in dif-ferent phenotypes. Proc Natl Acad Sci USA.2008;105(19):6987-92.


use with caution: developmental systems divergence and

Documents