REVIEWS

A recent search of the \-louse Knockout Database l

revealed alx)ut 15 genes, to date, whose hom()zygou~ null mutant phenotypes have proved to he too subtle for detection hy the laboratories that engineered them. This can occur even for genes whose products have pronounced effects when experimentally over-expressed in an embryo in vim or applied to tissue culture cells. There could be up to 10" bona fide genes (open reading frames) in the mouse genome, and many of these might tum out to earn their keep hy optimizing the phenotype in \vays too ~uhtle for laboratory detection. Yet, genes with reported absence of null mutant phenotype include those with ~uch cnlcial expected functions as glial fIbril­lary acidic protein, tenascin, vimentin, one of the colla­gens, proto-oncogene~ believed to function 10 mtracel­lular signal transduction, and component:o. of the retinoid receptor and hinding protein array2-1.,. The natural interpretation of these ob:o.ervations is that two or more different gene products perform the same essential role: if one gene i~ knocked out, the role i!'l performed hy another, and the phenotype appear:o. to he normal.

A rdated phenomenon is that observl'd, for instance, in mouse follistatin, goo~ecoid and activin null mu­tants l ;-1". SpeCIfic features of the early expressIon pat­terns of these genes in vertebrate emhryo~, and the effel1s of ectopic expression of their products in Xenopus embryo~, had led to the helief that particular early steps in body panern formation would be compromised 10 the null phenotypes. In the event, while there is a definlle phenotype in each case (which relates to other later ex­pression sites of each gene and is quite adequate to account for its existence), the aspects of development corresponding to any role at the earlier expression sitc!'l appear to he untouched It seems that these particular genes are redundant as far as their early expressions arc concerned, but each has a latl'r role 10 whll'h It is uniquely effective. We are aware of additional striking example~ that have not yet been puhli:o.hed, and 11 is with mixed relIef and dishelief that we fmally read of the !'Ionic hedgehog mouse mutant, whose complex pheno­type corresponds almost exactly to what had heen predicrcd on the baSIS of the ~pecific early gene ex­pressions, and the results of experimental embryologi­cal and ectopic gene expression experiment~ 1M.

The aim of this review is to explain the evolutionary origin of such apparently redundant genes and redundant gene expression sites. Our own thought~ are centred on vertebrates, specifically the mouse, because one of us works 10 a medical research institute, but also because, as we shall ~ee, systematic gene duplication might be particularly important in the vertehrJte lmeage. Neverthe­less, it is clear that redundanlY, in all of the forms that we di~cuss, is relatively widespread in other typt..'S of com­plex metazoan development (see examples in Ret's 19, 20l. Severdl other contrihutions to thi~ topic have appeared. in this journal and et.,ewhere (e.g. Ref~ 19--21), whme help in shaping our own thoughts we gratefully acknowledge.

How can redundancy be maintained? Tn explain the maintenancc of redundant genes

over evolutionary time. it is not sufficient to draw the analogy with engineering de.,ign and say that important functlOn~ must he hacked up. "Iothmg could he more

Evolutionary origins and maintenance of redundant gene expression during metazoan development

JONATIIAN COOKE (j<ookcinimr.mrc.ac.uk)

MARTIN A. NOWAK (martin,nowakizoo,ox.ac.uk)

MAARTEN 80ERLUST

JOHN MAYNARD-SMrnI

Various iet'eis of redundancy in det'elopmental gene function appear common in complex meta%oans. There might be no apparent phenot)'pe at many, or et'en any, of a gene's specific expression sites in homozygous nuU mlltant embry'os, Here we ask u'bat underlies tbe origin of sucb arrangements. The generation of families oj genes by duplication has clearly been Important. AdditionaUy, howet'er, selection mlgbt bave drit'en molecularly unrelated genes, wbicb encode proteins of similar physiologicalfunction, to become expressed during tbe same sets oj developmental events (times and places), even though each such gene might InltlaUy have evoll'ed In connection u'ith just one oj these et'ents,

important for ~ur\'ival than the Krebs cycle, yl'l thcre is relati\'ely little redundancy among the genes specifying the necessary enzyme~ As we ~hall ~'t', an Important due to the ongm., of redundancy mIght come from contra~tmg 'developmental' genl'~ wllh thme for univer~al cellular 'hou~ekl'l'plTlg' functions. It is u~dul to di~tinguish thrl'e type~ of pf(X:l'~S that might explain redundancy

Ret:ll/t/dmlt Rem's inc;n'llse/itrless ill S/lhtle 1I'C/ys Ll'wi.~ Wolpert rl'l'l'nlly rl'markl'd Ihal putatively

phenotypelc~s kncx:kout mice ~1)(Juld hl' takl'n from theIr laboratory cage:o. and away from theIr invl'stigator~' microsc()pe~ to a tl'st eVl'ning at till' Opl'r~l. The implI­cation is that, in naturl' a~ ()ppo~cd to thl' lahoratory, an apparently redundant gl'nl' lTll'fl'a~l':o. fitnl'ss, not only as backup when another gene Lllls, but also in more suhtle ways when the othl'r gl'nl' is functIoning per­fectly. Thoma:o. I9 sugge~t~ that this l'Ould Ix: tnIe Ix:­cause two gene~ produl'l' morl' of ~oml' product than one (e.g. rRNA gl'nl's), or mcrl'a~l' fidelity, or Ix:cause together they have ~ome l'ml'rgl'nt function. [t might abo lx:, as Wolpert\ remark iJl1plil'~, that thl' fitnl'~~ advan­tage arises only in ~OlllC l'nvinJTlml'nt~. [n othl'r words, it is unclear whethl'r. to date, it has hccn tlL'lllon~tratl'd that any gene~ are truly fl't\undanl.

The idea that redundancy i~ only apparl'nt rai~l':o. no theorl'tlt'al dlfflcllltle~. EVl'n a Vl'ry ~mall ~l'lccti\'l' advan­tage. of thl' ..,ame order a ... till' gl'rm 11Tll' lllut:.I!ion rate, would he sutllcil'nt to maintain a ~'t'CJnd gl'ne inddimtdy. HO\\'l'\'l'r, it can lx' shcJ\\ n that tlll'rl' are situations in which pcrfl't1 redund~mlY may Ix: mamtamt'd hy selectIon.

TIG SEPTEMBER 1997 VOL 13 No.9

(.{lp\n)(llI" ll)-r t-1't'\ll'r""'-ll'ml' Ltd o\1i n~ht .. n",.l'r\t:>d OlhH-LJ'il"9"" ~1:I)(J PII 'CllhH-'.r;l.c;,l)"'.1012i~-X

360

REVIEWS

A recent search of the \-louse Knockout Database l

revealed alx)ut 15 genes, to date, whose hom()zygou~ null mutant phenotypes have proved to he too subtle for detection hy the laboratories that engineered them. This can occur even for genes whose products have pronounced effects when experimentally over-expressed in an embryo in vim or applied to tissue culture cells. There could be up to 10" bona fide genes (open reading frames) in the mouse genome, and many of these might tum out to earn their keep hy optimizing the phenotype in \vays too ~uhtle for laboratory detection. Yet, genes with reported absence of null mutant phenotype include those with ~uch cnlcial expected functions as glial fIbril­lary acidic protein, tenascin, vimentin, one of the colla­gens, proto-oncogene~ believed to function 10 mtracel­lular signal transduction, and component:o. of the retinoid receptor and hinding protein array2-1.,. The natural interpretation of these ob:o.ervations is that two or more different gene products perform the same essential role: if one gene i~ knocked out, the role i!'l performed hy another, and the phenotype appear:o. to he normal.

A rdated phenomenon is that observl'd, for instance, in mouse follistatin, goo~ecoid and activin null mu­tants l ;-1". SpeCIfic features of the early expressIon pat­terns of these genes in vertebrate emhryo~, and the effel1s of ectopic expression of their products in Xenopus embryo~, had led to the helief that particular early steps in body panern formation would be compromised 10 the null phenotypes. In the event, while there is a definlle phenotype in each case (which relates to other later ex­pression sites of each gene and is quite adequate to account for its existence), the aspects of development corresponding to any role at the earlier expression sitc!'l appear to he untouched It seems that these particular genes are redundant as far as their early expressions arc concerned, but each has a latl'r role 10 whll'h It is uniquely effective. We are aware of additional striking example~ that have not yet been puhli:o.hed, and 11 is with mixed relIef and dishelief that we fmally read of the !'Ionic hedgehog mouse mutant, whose complex pheno­type corresponds almost exactly to what had heen predicrcd on the baSIS of the ~pecific early gene ex­pressions, and the results of experimental embryologi­cal and ectopic gene expression experiment~ 1M.

The aim of this review is to explain the evolutionary origin of such apparently redundant genes and redundant gene expression sites. Our own thought~ are centred on vertebrates, specifically the mouse, because one of us works 10 a medical research institute, but also because, as we shall ~ee, systematic gene duplication might be particularly important in the vertehrJte lmeage. Neverthe­less, it is clear that redundanlY, in all of the forms that we di~cuss, is relatively widespread in other typt..'S of com­plex metazoan development (see examples in Ret's 19, 20l. Severdl other contrihutions to thi~ topic have appeared. in this journal and et.,ewhere (e.g. Ref~ 19--21), whme help in shaping our own thoughts we gratefully acknowledge.

How can redundancy be maintained? Tn explain the maintenancc of redundant genes

over evolutionary time. it is not sufficient to draw the analogy with engineering de.,ign and say that important functlOn~ must he hacked up. "Iothmg could he more

Evolutionary origins and maintenance of redundant gene expression during metazoan development

JONATIIAN COOKE (j<ookcinimr.mrc.ac.uk)

MARTIN A. NOWAK (martin,nowakizoo,ox.ac.uk)

MAARTEN 80ERLUST

JOHN MAYNARD-SMrnI

Various iet'eis of redundancy in det'elopmental gene function appear common in complex meta%oans. There might be no apparent phenot)'pe at many, or et'en any, of a gene's specific expression sites in homozygous nuU mlltant embry'os, Here we ask u'bat underlies tbe origin of sucb arrangements. The generation of families oj genes by duplication has clearly been Important. AdditionaUy, howet'er, selection mlgbt bave drit'en molecularly unrelated genes, wbicb encode proteins of similar physiologicalfunction, to become expressed during tbe same sets oj developmental events (times and places), even though each such gene might InltlaUy have evoll'ed In connection u'ith just one oj these et'ents,

important for ~ur\'ival than the Krebs cycle, yl'l thcre is relati\'ely little redundancy among the genes specifying the necessary enzyme~ As we ~hall ~'t', an Important due to the ongm., of redundancy mIght come from contra~tmg 'developmental' genl'~ wllh thme for univer~al cellular 'hou~ekl'l'plTlg' functions. It is u~dul to di~tinguish thrl'e type~ of pf(X:l'~S that might explain redundancy

Ret:ll/t/dmlt Rem's inc;n'llse/itrless ill S/lhtle 1I'C/ys Ll'wi.~ Wolpert rl'l'l'nlly rl'markl'd Ihal putatively

phenotypelc~s kncx:kout mice ~1)(Juld hl' takl'n from theIr laboratory cage:o. and away from theIr invl'stigator~' microsc()pe~ to a tl'st eVl'ning at till' Opl'r~l. The implI­cation is that, in naturl' a~ ()ppo~cd to thl' lahoratory, an apparently redundant gl'nl' lTll'fl'a~l':o. fitnl'ss, not only as backup when another gene Lllls, but also in more suhtle ways when the othl'r gl'nl' is functIoning per­fectly. Thoma:o. I9 sugge~t~ that this l'Ould Ix: tnIe Ix:­cause two gene~ produl'l' morl' of ~oml' product than one (e.g. rRNA gl'nl's), or mcrl'a~l' fidelity, or Ix:cause together they have ~ome l'ml'rgl'nt function. [t might abo lx:, as Wolpert\ remark iJl1plil'~, that thl' fitnl'~~ advan­tage arises only in ~OlllC l'nvinJTlml'nt~. [n othl'r words, it is unclear whethl'r. to date, it has hccn tlL'lllon~tratl'd that any gene~ are truly fl't\undanl.

The idea that redundancy i~ only apparl'nt rai~l':o. no theorl'tlt'al dlfflcllltle~. EVl'n a Vl'ry ~mall ~l'lccti\'l' advan­tage. of thl' ..,ame order a ... till' gl'rm 11Tll' lllut:.I!ion rate, would he sutllcil'nt to maintain a ~'t'CJnd gl'ne inddimtdy. HO\\'l'\'l'r, it can lx' shcJ\\ n that tlll'rl' are situations in which pcrfl't1 redund~mlY may Ix: mamtamt'd hy selectIon.

TIG SEPTEMBER 1997 VOL 13 No.9

(.{lp\n)(llI" ll)-r t-1't'\ll'r""'-ll'ml' Ltd o\1i n~ht .. n",.l'r\t:>d OlhH-LJ'il"9"" ~1:I)(J PII 'CllhH-'.r;l.c;,l)"'.1012i~-X

360

REVIEWS

Redundant genes and expression sites are selected as backup against mutation

Suppose that two genes have identical developmental roles. Could each be maintained by selection in individ­uals mutant for the other? As Fisher22 showed a long time ago, it is not easy. As explained in Fig. 1(a), if the two genes A and B each perform the same essential role per­fectly, then selection will maintain only the gene with the lower mutation rate: the other will accumulate nonfunc­tional mutations. If mutation rates are exactly equal, then A and B will both survive for a long time although, ulti­mately, one or other will be eliminated by genetic drift. We return to this possibility in the following subsection.

There are, however, situations in which both genes can be maintained indefinitely. The two simplest cases are explained in Fig. l(b). In case <0, two genes perfonn the same role Rl, but with different efficiencies. The less efficient gene B, has a lower mutation rate, and will be maintained by selection in individuals mutant at the A locus. Case (ij) is more complex, but potentially inter­esting. Gene A performo; one role only Rl, and is main­tained because it performs this function more efficiently than gene B. Gene B performs two roles Rl and R2, its product being expressed at the same time and site in development as that of A as well as at one or more other times and sites distinctive to itself. It is maintained as a bona fide gene because of its performance of R2. Why does it continue to perform Rl, which it does less effi­ciently than A? The answer is that it is maintained by selection in individuals mutant for gene A. For this to work, it is necessary that mutations in gene B that cause the loss of its secondary role only, while preserving its primary role, should he . less frequent than mutations

·causing loss of functional gene product and thus loss of both roles. This laner condition seems probable in most cases, because DNA encoding the specific biochemical function of a protein represents a larger mutational tar­get than the regulatory motifs determining its selective synthesis at specific sites.

These mechanisms are interesting, because they sug­gest how more complex systems, involving genes per­forming more than one role, might arise and be evolu­tionarily stable. Fig. 1(e) shows an example with three genes and three roles. In fact, simulated evolution of regulatory systems shows that a system in which each gene has only one role, and each role can be performed by one gene only (j.e. one gene per expression site), can rather easily evolve into one in which there is extensive pleiotropy and redundancy (see also Ref. 23).

(a) e ~.I \~b

A B

(b) (i) ® (ii) ® e t ' t ';>J~2 'A.

~. ~;', A B A If

(c) ® e @) t '~'~~1~~:~ t --, , ~ , ,

A B C

FIGUaE 1. (a) Two genes, A and B perfonn a single role with equal dfkicmcy. Either gene is as efficient on it~ own as art: tx)th acting together. #J..d are Poh are the respective mutation rJtes to (recessive) nonfunctional alleles. a and b. The system is stahle only if #J..d - Ponexactly. The fe'Json is that at equilihrium. the mtt' at which nonfunctional alleles arise hy mutation must exactly equal the mte at which they are eliminated hy sele(tion. BeC-Juse sele<.tive elimination occurs only through double-mutant homozygotes. equal numhers of nonfunctional alleles must he lost for e"dch gene. Therefore. for equilibrium to be possible. such alleles must arise at the same rate for each gene. a situation only Iikdy to occur for genes of closely similar size and sequence charJlteristit's. that is, recent duplicates. (h) Two examples of evolutionarily stable redundancy. In case (j). two genes perfonn the same role (RI). and A is more efficient than B. With complete recessivity. the relevant fitnesses are in the order AlA BIB - AlA bib> ala BIB » ala blh. 'Ole system is ~1able only if #J..d > Pol>. In Glse (ij). B is maintained as a gene because of a second role. R2. but it also perfoons RI. although less efficiently than gene A. This redundant fun(tion is maintained by selection prOVided that Pol > Pol. that is. mutations of B causing total loss of it'i function are more frequent than those <:ausing loss of it~ perfonnance of RI only. (c) A more complex sr.-tern that can he evolutionarily stahle.

redundancies resembling that in Fig. Hbii) will evolve. So generally, following duplications, there will he a race between loss of redundancy by mutation, and its stabilization through acquisition of new roles.

Gene duplication and the origin of redundancy Increasing sophistication in tracing common descent

between protein structures, together with progressive Redundant genes arise by duplication and last a long time btorage of complete small genome sequences, is revealing Suppose that a developmental gene is duplicated. reduplication and redundancy in the elaboration of the There will then be two genes performing the same role basic eukaryotic 'kit' of genes26. But in the vertebrate Iin-or roles, and no loss of fitness if one gene is lost. At eage in particular, duplication events involving whole least initially, their mutation rates are likely to be equal, genomes, or large parts of them, appear to have been a or nearly so. If so, both are likely to !Jurvive for millions of major mode of increase in complexity27-29. This has generations before one is lost by the random aC(.umu- allowed maintenance of overlapping role sel .. (redun-lation of mutations24.25 . But the situation is not stable on dan<."y) because the dupli<."ates retain shared, ancestral the evolutionary time scales involved in, say, the adaptive expression sites. Such conservatism is particularly likely radiation of vertebrates. However, it is also possible that because of the small targets for change by genetic drift, one or both genes will acquire one or more additional offered by the expression-control regions of genes, roles following the duplication, and this seems espe<:ially relative to those offered by the regions encoding the likely for genes whose spatially restricted expressions gene products. Crucial control elements (transcription-underlie metazoan development. In such cases, stable factor-binding sites) are short sequence motifs. alting TIG SEPTEMBER 1997 VOL. 13 No.9

361

REVIEWS

Redundant genes and expression sites are selected as backup against mutation

Suppose that two genes have identical developmental roles. Could each be maintained by selection in individ­uals mutant for the other? As Fisher22 showed a long time ago. it is not easy. As explained in Fig. 1(a), if the two genes A and B each perform the same essential role per­fectly. then selection will maintain only the gene with the lower mutation rate: the other will accumulate nonfunc­tional mutations. If mutation rates are exactly equal. then A and B will both survive for a long time although. ulti­mately, one or other will he eliminated by genetic drift. We return to this possibility in the following subsection.

There are, however, situations in which both genes can be maintained indefinitely. The two simplest cases are explained in Fig. l(b). In case (i), two genes perfonn the same role RI. but with different efficiencies. The less efficient gene B, has a lower mutation rate , and will be maintained by selection in indiViduals mutant at the A locus. Case (ji) is more complex, but potentially inter­esting. Gene A perfonn .. one role only Rl, and is main­tained because it perfonns this function more efficiently than gene B. Gene B performs two roles Rl and R2, its product being expressed at the same time and site in development as that of A as well as at one or more other times and sites distinctive to itself. It is maintained as a bona fide gene because of its performance of R2. Why does it continue to perform Rl, which it does less effi­ciently than A? The answer is that it is maintained by selection in individuals mutant for gene A. For this to work, it is necessary that mutations in gene B that cause the loss of its secondary role only, while preserving its primary role, should he · less frequent than mutations ·causing loss of functional gene product and thus loss of both roles. This laner condition seems probable in most cases, because DNA encoding the specific biochemical function of a protein represents a larger mutational tar­get than the regulatory motifs determining its selectivt: synthesis at specific sites.

These mechanisms are interesting, because they sug­gest how more complex systems, involving genes per­forming more than one role, might arise and be evolu­tionarily stable. Fig. Hc) shows an example with three genes and three roles. In fact, simulated evolution of regulatory systems shows that a system in which each gene has only one role, and each role can be performed by one gene only (i.e. one gene per expression site), can rather easily evolve into one in which there is extensive pleiotropy and redundancy (see also Ref. 23),

Redundant genes arise ~ duplication and last a long time Suppose that a developmental gene is duplicated.

There will then be two genes performing the same role or roles, and no loss of fitness if one gene is lost. At least initially, their mutation rates are likely to be equal, or nearly so. If so, both are likely to !o"Urvive for millions of generations before one is lost by the random ac<."Umu­lation of mutations24•25 . But the situation is not stahle on the evolutionary time scales involved in, say, the adaptive radiation of vertebrates. However, it is also possible that one or both genes will acquire one or more additional roles following the duplication, and this seems espedaUy likely for genes whose spatially restricted expressions underlie metazoan development. In such cases, stable

A B

(ii) e A B

A B c

FIGUU 1_ (a) Two genes, A and B pt:rfonn a single role with equal efficiency. Either gene is as efficient on il~ own as an: tx)th acting together. I'a are I'h are the respective mutation rJtes to (recessive) nonfunctional alleles. a and b. The system i~ stahle only if I'a - 1'1> exactly. The reason is that at equilihrium. the mIt' at which nonfunctional alldes arise hy mutation must exactly equal the rate at which they are eliminated hy selet1ion. Bee-JuSt! selt:<.tive elimination occurs only through douhle-mutant homozygotes. equal numbers of nonfunctional alleles must he lost for each gene. Therefore. for equilibrium to be pUlsihle. such alldes must arise at the same Me for each gene. a situation only likely to occur for geJkS of closely similar size and seqUetK~ char..ll1eristil"s. that is, recent duplicates. (h) Two examples of evolutionarily stable redundancy. )n case (i). two genes perfonn the same role (Rl). and A is more efficient than B. With complete reces.o;ivity. the relevant fitnes.c;es are in the order A/A BIR- AlA bib> ala BIB » ala bib. The system is stahle only if IL. > 1'1>. In t .... se (ij). B is maintained as a gene because of a second role. R2. hut it also perfonns RI . although k'ss effidently than gene A. This redundant funt1ion is maintained by selection prOVided that 1'2 > 1'1. that is. mutations of B causing total Ios.o; of it'i function are more freqUl!nt than those <:ausing loss of il~ performance of Rl only. (c) A more <:omplex system that can he evolutionarily stahle.

redundancies resembling that in Fig. l(bii) will evolve. So generally, follOWing duplications, there will be a race between loss of redundancy by mutation. and its ~1abilization through acquisition of new roles.

Gene duplication and the origin of redundanq Increasing sophistication in tracing common descent

between protein structures, together with progres.'iive ~torage of complete small genome sequences, is revealing reduplication and redundancy in the elaboration of the basic eukaryotic 'kit' of genes26. But in the vertebrate lin­eage in particular, duplication event., involving whole genomes, or large parts of them, appear to have been a major mode of increase in complexity27-29. This has allowed maintenance of overlapping role set .. (redun­dant)') because the duplicates retain shared, ancestral expression sites. Such conservatism is particularly likely because of the small targets for change by genetic drift, offered by the expression-control regions of genes, relative to those offered by the regions encoding the gene products. Crucial control elements (transcription­factor-binding sites) are short sequence motifs, at1ing

TIG SEPTEMBER 1997 VOL. 13 No.9

361

REVIEWS

combinatorially or additively. Relative positions or dis­tances along the ))]\A, and even polarity. of such se­quences relative to the coding region are often not cru­cial to fum.tion. Thus, even if the now redundant role. hence expre)'sion site, of a gene is not sele<.tively main­tained. it might decay only very slowly by mutation and so be a sort of hi),torical bagg'age.

Meanwhile, the acquisition of novel roles by mem­bers of such gene families can occur, by evolution of new expressIon l>ites. This b beCCll11>C control regions of genes can also evolve rapidly under positive selection (e.g. Ref. 30>. Short sequence motifs can move readily between the neighbourhoods of different genes where. agam, their precIse posItions of integratIon mIght not be crucial to their control function. Thul>, for the reasons analY1>Cd in the previoul> sc<.lion and Fig l(b), control inputs to genes and. hence, gene expression sites might he kept in place or might POSItIvely evolve. by seJe<.llon, even if they are functionJe~s in most individuals. We di~cuss pos~ible example), of such coevolved. redun­dant gene fun<.tion in the final se<.tion.

Without subSCribing to naive recapitulationism - the idea that present-day emhryo:, particularly re~emhle ancestral adult forms·~l - we can ahrrec that the earliest­developing l>pecific ~ites for gene expression in embryo. .. will tend to be those associated with the evolutionarily oldest gene roles. For example, axial orgamzation at the ga~trular d()~1 lip. or equivalent. cX'curs hefore subre­gionalization within a brain rudiment. or formation of limb buds; it also occurred in organisms who1>C devel­opment did not include the latter events. In view of the above S('enario of gene duplications followed hy partial diversification of expression, we would expc<.t rela­tively more of the specific hut redundant expression sitel> and. therefore. backed-up developmental roles among genes, to occur in earlier developmental stages. A ... yet, the datahase offers no clear test of this idea. The long­awaited screen of developmental zebra fish mutants~.2, however. has produced le~s than the harvest of single mutant phenotypes affL~ting the earliest steps in verte­hrclte ar<:hite<.ture. than might have been expc(:ted by direct analogy with its Drosophila predeces..,or. There is some eVlden<.'e that rather than representing any tmn­sitional stage towards amniote vertehrates in its hi1>(ory of duplication events. the fbh genome might have under­gone all the events c:harclcterizing craniate vertehratel>. plul> some partial additional ones.

the acquisition of new mil'S by recruitment to new spe­cific sites of expression, coupled with partial redundant), hy retention of the ancestral site. Fewer comparable opportunites have been open to metabolic genes, which are expressed throughout the body and function essen­tially identically everywhere. given the bystem properties wherehy metabolism adaptl> to cdb' varying micro­environments within the organism. Vertebrate exceptions to this generalization. that is. partly tissue-spedfic isotypes of enzymes of Widespread fun<.tion. tend to he results of more recent duplications in restricted parts of the evolu­tionary tree. where a version of an enzyme with special chara<.teristics is adaptive for a p'Jrticular lifestyle and/or tissue. Thus, overall. within complex metazoans. the major mechanism for retention of ancient gene dupli­cat<.'S would appear to have been the acqUisition of novel expression sites for developmental genes. with its accompanymg opportunity for new gene mil'S under­lying the progressive extension of development itself. In much. although not all. of this duplication. the dupli­cate gene products retain functional interchangeability. in appropriate experiment" <e.g. Ref. 33). In engineering terms therefore. the developmental complexity could a:, well be achieved by diversifying the set of spl'Cific expression sites in the one original gene. This proce~ has surely been important aiM). but we record gene duplicationl> and partial redundancies because they have been the means of survival of mndom gene dupli­cates in the hlind evolutionary prcX'ess.

Protection against 'developmental error' There is one further reason why quaSi-redundant

genel>. or gene expression sites. might survive or he positively selected in evolution. Suppose. as before. that two genes A and B perform a partkular role with equal effidenly. If only one of the genes were pres­ent. the developmental l>tep might (X'casionally fail as a result of a developmental accident - an effective failure in the normal proces.o,cs of gene expre!'>sion - mther than an arn.cnce of any gene fun<.tlon through mutation. With two :,cpamte genel>. the likelihood of such a faIl­ure is greatly reduced. It turns out that a low frequency of 1>uch developmental failure:, b l>uffkient to maintain rcdundanlY; in effe<.t. the frequency of developmental expression failure for each gene must merely be higher than the germ line mutation nile to los. .. of fum:tion. or to los,... of the relevant cxprcs!'>ion site in thc other. which is a highly plausible a:'l>umption. But this hypoth-

New roles keep genes alive el>is for redundan<.y phenomena must also recognize Any explanation of redundancy must explain the that. as mentioned above, they are frequent among

follOWing observation. Although redundanty appears to genes with developmental hut more exceptional among be wlde~pread among vertebrJte and pos.'lihly other com- those with 'hou!'>ekl.>t!ping' roles. There are at least two plex metazoan developmental genes, and among genes pOl>l>ible reasons. Firstly. mechanisms of spatiotempo­involved in the control of cell proliferJtion and. pcr- mlly precise and limited gene expression in the emhryo haps, fidelity of DNA replication. 'housekeeping' genes might be much more prone to non-genetic perturb­encoding the central metabolic machinery <the Krebs ation. cau!'>ing failure within individuals. than arc the cycle. respirJtion. and so on) tend not to show survival mechanisms of universal gene exprcs. ... lon. Secondly. cor­of phylogenetically ancient. redundant duplicates. Why r<.~1 fun<.tion of most housekeepmg genes (e.g. enzyme­should this be? Duplicate copies of the genes encoding cataly!>Cd rea<.tionl» might be more stringently specified metabolic enzyme~ must have arisen. given the whole- than are some developmental function!'> (e.g. sequester­sale genome duplication!'> already mentioned. but instead ing TGF-~ Iigandl>. see below. or adju:,tment of cell of redundancy. the result ha~ been the decay of all hut adhesivlty/motility). so that certain mo\(.><"ularly unrelated one duplicate. The naturJI explanation b that. for devel- geneb might. nevertheles. ... have backup value for each opmental genes. duphcation otTered an opportunity for other's rob in development.

TIG SEPTEMBER 1997 VOL. 13 No.9

362

REVIEWS

combinatorially or additively. Relative positions or dis­tances along the ))]\A, and even polarity. of such se­quences relative to the coding region are often not cru­cial to fum.tion. Thus, even if the now redundant role. hence expre)'sion site, of a gene is not sele<.tively main­tained. it might decay only very slowly by mutation and so be a sort of hi),torical bagg'age.

Meanwhile, the acquisition of novel roles by mem­bers of such gene families can occur, by evolution of new expressIon l>ites. This b beCCll11>C control regions of genes can also evolve rapidly under positive selection (e.g. Ref. 30>. Short sequence motifs can move readily between the neighbourhoods of different genes where. agam, their precIse posItions of integratIon mIght not be crucial to their control function. Thul>, for the reasons analY1>Cd in the previoul> sc<.lion and Fig l(b), control inputs to genes and. hence, gene expression sites might he kept in place or might POSItIvely evolve. by seJe<.llon, even if they are functionJe~s in most individuals. We di~cuss pos~ible example), of such coevolved. redun­dant gene fun<.tion in the final se<.tion.

Without subSCribing to naive recapitulationism - the idea that present-day emhryo:, particularly re~emhle ancestral adult forms·~l - we can ahrrec that the earliest­developing l>pecific ~ites for gene expression in embryo. .. will tend to be those associated with the evolutionarily oldest gene roles. For example, axial orgamzation at the ga~trular d()~1 lip. or equivalent. cX'curs hefore subre­gionalization within a brain rudiment. or formation of limb buds; it also occurred in organisms who1>C devel­opment did not include the latter events. In view of the above S('enario of gene duplications followed hy partial diversification of expression, we would expc<.t rela­tively more of the specific hut redundant expression sitel> and. therefore. backed-up developmental roles among genes, to occur in earlier developmental stages. A ... yet, the datahase offers no clear test of this idea. The long­awaited screen of developmental zebra fish mutants~.2, however. has produced le~s than the harvest of single mutant phenotypes affL~ting the earliest steps in verte­hrclte ar<:hite<.ture. than might have been expc(:ted by direct analogy with its Drosophila predeces..,or. There is some eVlden<.'e that rather than representing any tmn­sitional stage towards amniote vertehrates in its hi1>(ory of duplication events. the fbh genome might have under­gone all the events c:harclcterizing craniate vertehratel>. plul> some partial additional ones.

the acquisition of new mil'S by recruitment to new spe­cific sites of expression, coupled with partial redundant), hy retention of the ancestral site. Fewer comparable opportunites have been open to metabolic genes, which are expressed throughout the body and function essen­tially identically everywhere. given the bystem properties wherehy metabolism adaptl> to cdb' varying micro­environments within the organism. Vertebrate exceptions to this generalization. that is. partly tissue-spedfic isotypes of enzymes of Widespread fun<.tion. tend to he results of more recent duplications in restricted parts of the evolu­tionary tree. where a version of an enzyme with special chara<.teristics is adaptive for a p'Jrticular lifestyle and/or tissue. Thus, overall. within complex metazoans. the major mechanism for retention of ancient gene dupli­cat<.'S would appear to have been the acqUisition of novel expression sites for developmental genes. with its accompanymg opportunity for new gene mil'S under­lying the progressive extension of development itself. In much. although not all. of this duplication. the dupli­cate gene products retain functional interchangeability. in appropriate experiment" <e.g. Ref. 33). In engineering terms therefore. the developmental complexity could a:, well be achieved by diversifying the set of spl'Cific expression sites in the one original gene. This proce~ has surely been important aiM). but we record gene duplicationl> and partial redundancies because they have been the means of survival of mndom gene dupli­cates in the hlind evolutionary prcX'ess.

Protection against 'developmental error' There is one further reason why quaSi-redundant

genel>. or gene expression sites. might survive or he positively selected in evolution. Suppose. as before. that two genes A and B perform a partkular role with equal effidenly. If only one of the genes were pres­ent. the developmental l>tep might (X'casionally fail as a result of a developmental accident - an effective failure in the normal proces.o,cs of gene expre!'>sion - mther than an arn.cnce of any gene fun<.tlon through mutation. With two :,cpamte genel>. the likelihood of such a faIl­ure is greatly reduced. It turns out that a low frequency of 1>uch developmental failure:, b l>uffkient to maintain rcdundanlY; in effe<.t. the frequency of developmental expression failure for each gene must merely be higher than the germ line mutation nile to los. .. of fum:tion. or to los,... of the relevant cxprcs!'>ion site in thc other. which is a highly plausible a:'l>umption. But this hypoth-

New roles keep genes alive el>is for redundan<.y phenomena must also recognize Any explanation of redundancy must explain the that. as mentioned above, they are frequent among

follOWing observation. Although redundanty appears to genes with developmental hut more exceptional among be wlde~pread among vertebrJte and pos.'lihly other com- those with 'hou!'>ekl.>t!ping' roles. There are at least two plex metazoan developmental genes, and among genes pOl>l>ible reasons. Firstly. mechanisms of spatiotempo­involved in the control of cell proliferJtion and. pcr- mlly precise and limited gene expression in the emhryo haps, fidelity of DNA replication. 'housekeeping' genes might be much more prone to non-genetic perturb­encoding the central metabolic machinery <the Krebs ation. cau!'>ing failure within individuals. than arc the cycle. respirJtion. and so on) tend not to show survival mechanisms of universal gene exprcs. ... lon. Secondly. cor­of phylogenetically ancient. redundant duplicates. Why r<.~1 fun<.tion of most housekeepmg genes (e.g. enzyme­should this be? Duplicate copies of the genes encoding cataly!>Cd rea<.tionl» might be more stringently specified metabolic enzyme~ must have arisen. given the whole- than are some developmental function!'> (e.g. sequester­sale genome duplication!'> already mentioned. but instead ing TGF-~ Iigandl>. see below. or adju:,tment of cell of redundancy. the result ha~ been the decay of all hut adhesivlty/motility). so that certain mo\(.><"ularly unrelated one duplicate. The naturJI explanation b that. for devel- geneb might. nevertheles. ... have backup value for each opmental genes. duphcation otTered an opportunity for other's rob in development.

TIG SEPTEMBER 1997 VOL. 13 No.9

362

REVIEWS

Conclusions Many processes that might maintain genetic redun­

dancy have been discussed in this review. We can best summarize them in relation to particular cases.

A major source of gene redundancy and redundant expression has clearly been the incomplete role diver­gence between members of gene families arising by duplication. Many such families, such as the TGF-(3-related and WNT (Drosophila wingless)-related genes in vertebrates34•35 are substantial in size with complex pat­terns of partial role overlap. But a particularly clear example is afforded by the two Drosopbila e71g1'YA.iled­related gene duplicates surviving in amniote vertebrates, En1 and En2 (Ref.') 33, 36) (the zebrafish appears to have retained, or added, a third member). These pro­teins share several distinctive structural domains, aside from the engrailed homeobox (DNA-binding domain), and are probably completely functionally interchange­able33• Both are expressed throughout a particular region of the brain rudiment that will foml parts of the mid­and hind-brain, but En1 is expressed earlier. TIle mouse null mutant phenotype, within this expression region, is more severe for En1 alone. TIle En2-null mutant homo­zygote does, however, have a phenotype in this region, thought to be related to the fact that its expression extends into a later period. In this period there appears to be a graded requirement for En activity for somewhat different aspecl"s of- development. Eacll homozygous mutant shows a more severe phenotype if only one functional allele exist') at the other locus, even though the double heterozygote is without apparent pheno­type. Finally, Bn1 alone exhibits several additional, quite different expression sites, some of which (in somite­derived structures) appear to be redundant, but two of which, in limb and sternum rudiment'), show pheno­type in the mutant36.

A second potentially important type of redundancy is the apparent convergence of expressions of sequence­unrelated, but functionally similar genes, at particular sites. A topical example concerns vertebrate neural induc­tion. The secreted proteins noggin, chordin, follistatin and flik (follistatin-Iike) are appropriately expressed in the early embryo for involvement in this intercellular signalling process, in which embryonic ectodeml is directed to form neural tissue·:n-4o. At least tJle first three proteins can also perfonn this inductive role experimen­tally when ed:opicallyexpressed in the fTOg embry037- 39• 111ree unrelated sequence structures are involved, but the proteins have the common property of binding various members of the TGF-f3 family of small secreted proteins41-43 to anlagonize tlleir action, while there is genetic evidence that the Drosophila homologue of Chordin (tJle short-gastrulation gene product) has this propeny as its sole function-i4 . Why are up to four appro­priate genes expressed at the site of induction when apparently one would do?

Because the process is very ancient: and the proteins unrelated, it seems unlil<ely lhal we are looking at an ancestral fundion that has been retained redundantly and has not yet had time to decay. In fact, at least follistatin and noggin are revealed by tJleir mutant mouse pheno­types to have necessary roles at one or more sites later in development CHef. J5; A.P. McMahon, unpublished). 111eir existence as genes is, therefore, explained even if

they are perfectly redundant within neural induction. But one must remember [Fig. l(bij)] that this explanation for expressional or 'role' redundancy requires that mu­tations leading to total loss of function by a gene are more common tllan mutations leading to the loss of one (redundant) role only. A third explanation would be that the genes all have roles in initial neural induction, but subtly different ones, so that individuals with all genes functional have subtly superior neural rudiments than those with any of them inactive. To est'ablish whether tllis is the case will require deeper understand­ing of what is going on at tile molecular level, as well as more subtle anatomical and functional knowledge tl1an is currently available. Finally, there is the possibility that two or more of the genes have essentially the same func­tion in neural induction, and that redundancy has been selected as backup against somatic developmental error whereby one of these fails to express adequately at'the appropriate site. As yet, there is insufficient evidence to enable us to evaluate the relative contributions of these evolutionary mechanisms to tJlis example; what we have tried to do is to distinguish between them as care­fuJly as we can, so that the relevant evidence can be recognized.

Acknowledgements J.C. is grateful for stimulating and informative

discussions with A. Purley, J. Sharpe and K. Hastings during tlle preparation of this article. We also t1lank P. Towers, J. 1110mas and otl1er, anonymous reviewers for constructive comments on drafts.

References 1 http://biomednet.com/cgi-binlmko/mkohome.pl 2 Gomi, H. el al. (995) NeulU'1. ]4, 29-4] 3 Pe)my, M. et al. (995) Elt1BO]. 14, ]590-1598 4 Saga, Y. et al. (992) Genes,De,). 6, ]821-183] 5 Sleindler, D.A. et al. (995)]. Neu,.osci. 15, ]971-1983 6 Colucci, G.E. el al. (994) Cell79, 679-694 7 Rosati, R. et al. (994) Nat. Ge11et. S, 129-135 8 l.owell. C.A .• SOriano, P. and Vamlus. H.P. (1994) Genes

Den.8,387-398 5) Umanoff, H. et al (995) Proc. Nat!. Acad. Sct. u. S. A. 92,

1709-1713 10 Stein, P.L. el al. (994) Genes Dev. 8, 1999-2007 11 Gorry, P. el al. (994) Pmc. Natl. Acad. Sci. U. s. A. 91,

9032-9036 12 Lufkin, T. et al. (993) Proc. NatT. Acad. Sci. U. s. A. 90,

7225-7229 13 Li. E. et al. (993) Proc. Nail. Acad. Sci. u. S. A. 90,

1590-]594 14 Mendelsohn, C. et aT. (994) Dev. Bioi. 166,246-258 15 Matzuk, M.M. et aI, (1995) Nature 374, 360-363 16 Rivera-Perez, J.A. et aT. (995) JJetJeloJ~me1f.t 12](9),

3005-3012 17 VassalJi, A. elol. (1994) Genes De/). 8,414-427 18 Chiang, C.C. et aT. (996) Nature 383, 4Q7-413 19 TIlomas, J.H. (993) 'Jirmds Genet. 9, 395-399 20 Taut:7., D. (992) Bioessa)'s 14,263-266 21 Brookfield,]. (992) Cui,·. BioI. 2,553-554 22 Fisher. ItA. (935) Am. Nat. 69. 446-455 23 NO\\Iak, M.A.. BoerJijst, M., Cooke, J. and

Maynard-Smith. J. (1997) Nat1l.1"e 338, 167-171 24 Kimura, M. and King, J.I •. (1979) Proc. Nat!. Acad. Sci.

U.S.A.76.2858-2R61 25 Takahata, N. and Maruyama, T. (1979) Pmc. Nail. Acad.

Sct. U. S. A. 76, 452]-4525

TIG SEPTEMBER 1997 VOL. J 3 No.9

363

REVIEWS

Conclusions Many processes that might maintain genetic redun­

dancy have been discussed in this review. We can best summarize them in relation to particular cases.

A major source of gene redundancy and redundant expression has clearly been the incomplete role diver­gence between members of gene families arising by duplication. Many such families, such as the TGF-(3-related and WNT (Drosophila wingless)-related genes in vertebrates34•35 are substantial in size with complex pat­terns of partial role overlap. But a particularly clear example is afforded by the two Drosopbila e71g1'YA.iled­related gene duplicates surviving in amniote vertebrates, En1 and En2 (Ref.') 33, 36) (the zebrafish appears to have retained, or added, a third member). These pro­teins share several distinctive structural domains, aside from the engrailed homeobox (DNA-binding domain), and are probably completely functionally interchange­able33• Both are expressed throughout a particular region of the brain rudiment that will foml parts of the mid­and hind-brain, but En1 is expressed earlier. TIle mouse null mutant phenotype, within this expression region, is more severe for En1 alone. TIle En2-null mutant homo­zygote does, however, have a phenotype in this region, thought to be related to the fact that its expression extends into a later period. In this period there appears to be a graded requirement for En activity for somewhat different aspecl"s of- development. Eacll homozygous mutant shows a more severe phenotype if only one functional allele exist') at the other locus, even though the double heterozygote is without apparent pheno­type. Finally, Bn1 alone exhibits several additional, quite different expression sites, some of which (in somite­derived structures) appear to be redundant, but two of which, in limb and sternum rudiment'), show pheno­type in the mutant36.

A second potentially important type of redundancy is the apparent convergence of expressions of sequence­unrelated, but functionally similar genes, at particular sites. A topical example concerns vertebrate neural induc­tion. The secreted proteins noggin, chordin, follistatin and flik (follistatin-Iike) are appropriately expressed in the early embryo for involvement in this intercellular signalling process, in which embryonic ectodeml is directed to form neural tissue·:n-4o. At least tJle first three proteins can also perfonn this inductive role experimen­tally when ed:opicallyexpressed in the fTOg embry037- 39• 111ree unrelated sequence structures are involved, but the proteins have the common property of binding various members of the TGF-f3 family of small secreted proteins41-43 to anlagonize tlleir action, while there is genetic evidence that the Drosophila homologue of Chordin (tJle short-gastrulation gene product) has this propeny as its sole function-i4 . Why are up to four appro­priate genes expressed at the site of induction when apparently one would do?

Because the process is very ancient: and the proteins unrelated, it seems unlil<ely lhal we are looking at an ancestral fundion that has been retained redundantly and has not yet had time to decay. In fact, at least follistatin and noggin are revealed by tJleir mutant mouse pheno­types to have necessary roles at one or more sites later in development CHef. J5; A.P. McMahon, unpublished). 111eir existence as genes is, therefore, explained even if

they are perfectly redundant within neural induction. But one must remember [Fig. l(bij)] that this explanation for expressional or 'role' redundancy requires that mu­tations leading to total loss of function by a gene are more common tllan mutations leading to the loss of one (redundant) role only. A third explanation would be that the genes all have roles in initial neural induction, but subtly different ones, so that individuals with all genes functional have subtly superior neural rudiments than those with any of them inactive. To est'ablish whether tllis is the case will require deeper understand­ing of what is going on at tile molecular level, as well as more subtle anatomical and functional knowledge tl1an is currently available. Finally, there is the possibility that two or more of the genes have essentially the same func­tion in neural induction, and that redundancy has been selected as backup against somatic developmental error whereby one of these fails to express adequately at'the appropriate site. As yet, there is insufficient evidence to enable us to evaluate the relative contributions of these evolutionary mechanisms to tJlis example; what we have tried to do is to distinguish between them as care­fuJly as we can, so that the relevant evidence can be recognized.

Acknowledgements J.C. is grateful for stimulating and informative

discussions with A. Purley, J. Sharpe and K. Hastings during tlle preparation of this article. We also t1lank P. Towers, J. 1110mas and otl1er, anonymous reviewers for constructive comments on drafts.

References 1 http://biomednet.com/cgi-binlmko/mkohome.pl 2 Gomi, H. el al. (995) NeulU'1. ]4, 29-4] 3 Pe)my, M. et al. (995) Elt1BO]. 14, ]590-1598 4 Saga, Y. et al. (992) Genes,De,). 6, ]821-183] 5 Sleindler, D.A. et al. (995)]. Neu,.osci. 15, ]971-1983 6 Colucci, G.E. el al. (994) Cell79, 679-694 7 Rosati, R. et al. (994) Nat. Ge11et. S, 129-135 8 l.owell. C.A .• SOriano, P. and Vamlus. H.P. (1994) Genes

Den.8,387-398 5) Umanoff, H. et al (995) Proc. Nat!. Acad. Sct. u. S. A. 92,

1709-1713 10 Stein, P.L. el al. (994) Genes Dev. 8, 1999-2007 11 Gorry, P. el al. (994) Pmc. Natl. Acad. Sci. U. s. A. 91,

9032-9036 12 Lufkin, T. et al. (993) Proc. NatT. Acad. Sci. U. s. A. 90,

7225-7229 13 Li. E. et al. (993) Proc. Nail. Acad. Sci. u. S. A. 90,

1590-]594 14 Mendelsohn, C. et aT. (994) Dev. Bioi. 166,246-258 15 Matzuk, M.M. et aI, (1995) Nature 374, 360-363 16 Rivera-Perez, J.A. et aT. (995) JJetJeloJ~me1f.t 12](9),

3005-3012 17 VassalJi, A. elol. (1994) Genes De/). 8,414-427 18 Chiang, C.C. et aT. (996) Nature 383, 4Q7-413 19 TIlomas, J.H. (993) 'Jirmds Genet. 9, 395-399 20 Taut:7., D. (992) Bioessa)'s 14,263-266 21 Brookfield,]. (992) Cui,·. BioI. 2,553-554 22 Fisher. ItA. (935) Am. Nat. 69. 446-455 23 NO\\Iak, M.A.. BoerJijst, M., Cooke, J. and

Maynard-Smith. J. (1997) Nat1l.1"e 338, 167-171 24 Kimura, M. and King, J.I •. (1979) Proc. Nat!. Acad. Sci.

U.S.A.76.2858-2R61 25 Takahata, N. and Maruyama, T. (1979) Pmc. Nail. Acad.

Sct. U. S. A. 76, 452]-4525

TIG SEPTEMBER 1997 VOL. J 3 No.9

363

REVIEWS

26 Wolfe. K H. and Shields. D.C (}9<)7) .Yat/lre 3~r'. 70H-7U 27 Lundin. LO (1lJ'-)j) GellomicsI6.1-19 28 Garda-Femandl'z. J. and Holland. I' W. ( 1(94) Naill re

:\70. '56~'56() 29 Storm. E.E. el al (1994) ,Yallln! ~,(~. 639-64.3 30 LI. X. and :-'011.1\1. (1994) ,\·(lllIr(':\67. H,3--8"" 31 Dc Beer. G. (1910) Emln)'os alld Allceslors,

Cl.lrl'ndon Pre~~ 32 Haffter. P. el til. (19%) ["JeI>e/opmell/1.B, 1-:16 33 Joyner, A (1996) 1h'I/(ls (;ell£'t 12, 1-;""'20 34 I.\'on~, K 1\1 ,J()nl'~, C\I and Hogan. IH.M (1991)

'f'rl'lItis Gellel 7. 40H-i 12 35 Sld()\\. A. (1992) Pmc. ,\'all Acad. So. (' S A. H9, '509B-'i102 36 \X'ur~t. W. Allerharh. A B and Joyner. A.L. (1994)

["JeI'e/opmeIll120.206-;.....2075 37 L:.unb. T M. et til (199:1) SCience 262. 71j-71R 38 Sa~;II. Y. 1'1 (/1 (199'1) .\'alllr£'376 . . ~:\j-.:t~6 39 Hellllll,ltl-i3manloli. A . Kelly. (Hi and Melton. 0 A

( 19<)1) Cell ....... 2R'}-29S

T ran~criptional regulation of tis1>ue-spedfic genes is controlled by cell-type-specific promoters or enhancers that interpret unique combinations of tr.lm.cription factors in different cell types. Studies of gene regulation in cul­tured cells have led to the identification of numerou~ tissue-spedfic promoters and enhancers. However, it i~ becoming Increasingly clear that the regulatory DNA sequences that are Important for tmnscription in tissue culture are often different from those that direct correct ti~sue ~pecificity and temporospatial re!,'lliation in l'i11(). Recent studies of gene regulation in the skeletal, cardiac and ~mooth muscle cell lineages of transgenic mice have revealed that the complete developmental expression pat­terns of individual muscle genes arc frequently depend­ent on composites of independent cis-acting regulatory region!'>, or mexlules, each of which directs a portion of the expression pattern of a gene. 111US, a regulatory mod­ule that directs trJn~cription in it ~uhset of mu!'>cle cells at a specific time and place in the embryo might be com­pletely silent in other muscle cells of the same type. The strict temporospatial specificity of myogenic regulatory modules reveals surprising molecular heterogeneity among muscle cells within the ~ame lineage, and sug­gests the existence of myogenic subprogmms of gene ex­presS)(ln that arc established through comhinatorial inter­acti()n~ between muscle-~pecific, positionally restricted and widely expressed tmnscription factor~.

Here, we review the results of several recent analyses of muscle transgene expression and consider the impli­cations of these studies for understanding the molecular mechafllsms that generate muscle cell diversity. Thb type of modularity of Cis-acting regulatory ~cqllences provides a means of generating a multitude of patterns of gene expression from a finite number of regulatory elements and is emerging as a common theme in the developmental control of gene expre~sion in other cell types in organisms ranging from Drosophila to humans l .

Diversity of muscle cell types

40 Patel. K et at 09<)(}) [Jel'. BIoi 178. 327-.-\42 41 Nakamllf'J, T .. Sugino. K .. Titani. K. and Sligino. H. (1991>

] Bioi Chem 266. 194.32-19437 42 Zimmennan. L 13 . De Jesll~-EM'Uhar. J.M. and Harland, R.M.

<19%) eel/B(,. '5l)<H)06 43 Piccolo, s .. Sa~ai, B.L and De R()hertl~. E.M. <ll)96) Cell

H6, ;8<)-;9H 44 Bieh~, B. Francob. Y and Bier, E (1996) Gelles Del' 10.

2922-2934

J. Cooke is 111 the Dit'isioll of Del'elopmelltat Nellrobiologv, Sa/lOlllll IlIslillllefor Jfedical Research. The RitiRell'ay. Mrlllfill. LOlldol/. eK i\lF71M M.A. Nowak alld M. Boerlljst are m Ihe lJepartment of ZOO/OlD'. SO/llh Parks Road. Oxford. UK OX I 3P

J. Maynard-Smith IS ill /he Sch(Jol oj BIOI{)J,1iclIl SCIences. r:llil'ersIZI' o/SUSS(!x. Falmer. Bnp,h/oll. {'K B: .. /1 9QG

Modular regulation of muscle gene transcription: a mechanism for muscle cell diversity

AN11IONY B. FlRUUJ (firuUi@utsw.swmed.edu)

ERIC N. OlSON (eolson@ha.mon.swmed.edu)

Skeleta~ cardiac and smooth muscle ceUs express overlapping sets 0/ muscle-specific genes, such that some muscle genes are expressed in only a single muscle ceQ type, whereas others are expressed in mulliple muscle ceQ Uneages. Recent studies in transgenic mice bave ret'ealed that, in many cases, mulliple, independent ds-regulatory regions, or modules, are required to direct 'he complete developmental pattern 0/ expression o/indlvidual muscle-specific genes, even within a single muscle ceQ type. The temporospattal speciflcity o/these myogenic regulatory modules is established by unique combinations o/transcription/actors and bas ret'ealed unanticipated diversity In tbe regulatory programs tbat control muscle gene expression. This type of composite regUlation 0/ muscle gene expression appears to reflect a general strategy for 'he control 0/ ceU-specific gene expression.

Skeletal, cardiac and smooth muscle cells originate during embryogenesIs from different mesodermal precur­sor cell populations. All skeletal muscle in vertebrates is

deriv<..>d from the S()mites, except for S()me mllscle~ in the head, which appear to arise from cephalic mes(xlerm (reviewed in Ref. 2). The somites are tran!'>ient ~tru(.1ures that form in a rostrocaudal progres~10n by segmentation of the paraxial mesoderm adjacent to till' neural tuhe. Newly fonned somlte~ appear as paired epithelial spheres that become compartmentalized Into a sheet of dor!'>al epithelial cells, known as the dennamyotome, which produces muscle precurS()rs that migrdte to the limbs and body wall. Immediately heneath the de~lrnyotome is the

TIG SEPTEMBER 1997 VOl .. 13 NO.9

(.'~lynghl ~ lW~ cht"\Il'r " .... n«· Ltd ... 11 flghl, R" .... ·,,,~I Ol('H·9'il5'9-:~I~m I'll '>('lnH·')'il'j''l7'OlI71·Z

364

REVIEWS

26 Wolfe. K H. and Shields. D.C (}9<)7) .Yat/lre 3~r'. 70H-7U 27 Lundin. LO (1lJ'-)j) GellomicsI6.1-19 28 Garda-Femandl'z. J. and Holland. I' W. ( 1(94) Naill re

:\70. '56~'56() 29 Storm. E.E. el al (1994) ,Yallln! ~,(~. 639-64.3 30 LI. X. and :-'011.1\1. (1994) ,\·(lllIr(':\67. H,3--8"" 31 Dc Beer. G. (1910) Emln)'os alld Allceslors,

Cl.lrl'ndon Pre~~ 32 Haffter. P. el til. (19%) ["JeI>e/opmell/1.B, 1-:16 33 Joyner, A (1996) 1h'I/(ls (;ell£'t 12, 1-;""'20 34 I.\'on~, K 1\1 ,J()nl'~, C\I and Hogan. IH.M (1991)

'f'rl'lItis Gellel 7. 40H-i 12 35 Sld()\\. A. (1992) Pmc. ,\'all Acad. So. (' S A. H9, '509B-'i102 36 \X'ur~t. W. Allerharh. A B and Joyner. A.L. (1994)

["JeI'e/opmeIll120.206-;.....2075 37 L:.unb. T M. et til (199:1) SCience 262. 71j-71R 38 Sa~;II. Y. 1'1 (/1 (199'1) .\'alllr£'376 . . ~:\j-.:t~6 39 Hellllll,ltl-i3manloli. A . Kelly. (Hi and Melton. 0 A

( 19<)1) Cell ....... 2R'}-29S

T ran~criptional regulation of tis1>ue-spedfic genes is controlled by cell-type-specific promoters or enhancers that interpret unique combinations of tr.lm.cription factors in different cell types. Studies of gene regulation in cul­tured cells have led to the identification of numerou~ tissue-spedfic promoters and enhancers. However, it i~ becoming Increasingly clear that the regulatory DNA sequences that are Important for tmnscription in tissue culture are often different from those that direct correct ti~sue ~pecificity and temporospatial re!,'lliation in l'i11(). Recent studies of gene regulation in the skeletal, cardiac and ~mooth muscle cell lineages of transgenic mice have revealed that the complete developmental expression pat­terns of individual muscle genes arc frequently depend­ent on composites of independent cis-acting regulatory region!'>, or mexlules, each of which directs a portion of the expression pattern of a gene. 111US, a regulatory mod­ule that directs trJn~cription in it ~uhset of mu!'>cle cells at a specific time and place in the embryo might be com­pletely silent in other muscle cells of the same type. The strict temporospatial specificity of myogenic regulatory modules reveals surprising molecular heterogeneity among muscle cells within the ~ame lineage, and sug­gests the existence of myogenic subprogmms of gene ex­presS)(ln that arc established through comhinatorial inter­acti()n~ between muscle-~pecific, positionally restricted and widely expressed tmnscription factor~.

Here, we review the results of several recent analyses of muscle transgene expression and consider the impli­cations of these studies for understanding the molecular mechafllsms that generate muscle cell diversity. Thb type of modularity of Cis-acting regulatory ~cqllences provides a means of generating a multitude of patterns of gene expression from a finite number of regulatory elements and is emerging as a common theme in the developmental control of gene expre~sion in other cell types in organisms ranging from Drosophila to humans l .

Diversity of muscle cell types

40 Patel. K et at 09<)(}) [Jel'. BIoi 178. 327-.-\42 41 Nakamllf'J, T .. Sugino. K .. Titani. K. and Sligino. H. (1991>

] Bioi Chem 266. 194.32-19437 42 Zimmennan. L 13 . De Jesll~-EM'Uhar. J.M. and Harland, R.M.

<19%) eel/B(,. '5l)<H)06 43 Piccolo, s .. Sa~ai, B.L and De R()hertl~. E.M. <ll)96) Cell

H6, ;8<)-;9H 44 Bieh~, B. Francob. Y and Bier, E (1996) Gelles Del' 10.

2922-2934

J. Cooke is 111 the Dit'isioll of Del'elopmelltat Nellrobiologv, Sa/lOlllll IlIslillllefor Jfedical Research. The RitiRell'ay. Mrlllfill. LOlldol/. eK i\lF71M M.A. Nowak alld M. Boerlljst are m Ihe lJepartment of ZOO/OlD'. SO/llh Parks Road. Oxford. UK OX I 3P

J. Maynard-Smith IS ill /he Sch(Jol oj BIOI{)J,1iclIl SCIences. r:llil'ersIZI' o/SUSS(!x. Falmer. Bnp,h/oll. {'K B: .. /1 9QG

Modular regulation of muscle gene transcription: a mechanism for muscle cell diversity

AN11IONY B. FlRUUJ (firuUi@utsw.swmed.edu)

ERIC N. OlSON (eolson@ha.mon.swmed.edu)

Skeleta~ cardiac and smooth muscle ceUs express overlapping sets 0/ muscle-specific genes, such that some muscle genes are expressed in only a single muscle ceQ type, whereas others are expressed in mulliple muscle ceQ Uneages. Recent studies in transgenic mice bave ret'ealed that, in many cases, mulliple, independent ds-regulatory regions, or modules, are required to direct 'he complete developmental pattern 0/ expression o/indlvidual muscle-specific genes, even within a single muscle ceQ type. The temporospattal speciflcity o/these myogenic regulatory modules is established by unique combinations o/transcription/actors and bas ret'ealed unanticipated diversity In tbe regulatory programs tbat control muscle gene expression. This type of composite regUlation 0/ muscle gene expression appears to reflect a general strategy for 'he control 0/ ceU-specific gene expression.

Skeletal, cardiac and smooth muscle cells originate during embryogenesIs from different mesodermal precur­sor cell populations. All skeletal muscle in vertebrates is

deriv<..>d from the S()mites, except for S()me mllscle~ in the head, which appear to arise from cephalic mes(xlerm (reviewed in Ref. 2). The somites are tran!'>ient ~tru(.1ures that form in a rostrocaudal progres~10n by segmentation of the paraxial mesoderm adjacent to till' neural tuhe. Newly fonned somlte~ appear as paired epithelial spheres that become compartmentalized Into a sheet of dor!'>al epithelial cells, known as the dennamyotome, which produces muscle precurS()rs that migrdte to the limbs and body wall. Immediately heneath the de~lrnyotome is the

TIG SEPTEMBER 1997 VOl .. 13 NO.9

(.'~lynghl ~ lW~ cht"\Il'r " .... n«· Ltd ... 11 flghl, R" .... ·,,,~I Ol('H·9'il5'9-:~I~m I'll '>('lnH·')'il'j''l7'OlI71·Z

364