Gene-targeting studies of mammalian behavior: is it the mutation or the background genotype?

Gene-targeting studies of mammalian behavior: is it the mutation or the background genotype?

DEBATE Gene-targeting studies of mammalian behavior: is it the mutation or the background genotype? Robert Gerlai Genetargetingto createnullmutationsi...

1MB Sizes 0 Downloads 19 Views

DEBATE Gene-targeting studies of mammalian behavior: is it the mutation or the background genotype? Robert Gerlai Genetargetingto createnullmutationsin miceisa powerfulnewtool in biologywhichwillallow the molecular dissectionof complex phenotypessuch as mammalian brain function, and learningand memory.However,the attempt to interpretthe phenotypicalchangeswhicharise in null-mutantmiceissubjectto severalcaveats. For example,the abilityto disrupta singlegene in a targetedmanner might leadone to overlookthe effectsof other genes.Ignoringthe genetic backgroundmight lead to misinterpretationof results:the presentarticle will demonstrate that the phenotypicalabnormalitiesattributed to the null mutation in severalmolecula~ neurobiologicalstudiescouldsimplyresultfrom the effectsof backgroundgenes. TrendsNeurosci.(1996) 19, 177-181

w

ITHGENETARGETING,one can practicallyknock out a gene in vivo and create a mutant organism that completelylacksthe geneproduct.Thus,the promise of gene targetinghas been to revealthe in vivo function of the gene of interestl-3.Forexample,Grant et al.z wrote: ‘targeteddisruptionof genesprovidesa powerful tool for examining the role of specific proteins in the functionof the brain’.However,the functionalrelevance of gene targeting has been questioned&Gbecause the mutation might lead to an avalancheof compensatory processes(up-or downregulationof gene products)and resultingsecondaryphenotypicalchanges.Clearly,a nullmutant organismmight not only lack the productof a singlegenebut might alsopossessa numberof developmental, physiological,or evenbehavioralprocessesthat havebeen alteredto compensateforthe effectof the null mutation.Therefore,one might expect an arrayof complex phenotypical changes that might not be directly relatedto the function of the gene of interest. Teasing out the primary and secondary changes will require co-ordinated efforts of scientists from severalfields of biology. However,these efforts might be conductedin vain if the effects of genes other than those of the one targetedhave not been ruled out with certainty. in the geneticbackgroundmight Polymorphism makethe resultsof gene-targetingstudiesdifficult to interpret

The compensatorychangestriggeredby the disruption of the targetedgenewillnot onlydependon the targeted geneitselfandits involvementin certainmolecularpathways,but alsoon the backgroundgenotype(seealsothe concept of modifiergenes’).Targeteddisruptionof gene a, for example,might leadto a differentialexpressionof allelesb andB of gene13.Furthermore,a similarregulatory change of gene p might lead to differentphenotypical effects dependingon which allele (b or B) waspresent in the organismwith the null mutation in the a gene. Consequently,polymorphismin the geneticbackground willnot allowone to concludewith certaintythat a particular phenotypicalchange observedin a null-mutant Copyright 01996,

Elsevier Science Ltd. All rights reserved. 0166-

2236/96/$15.00

animal was indeed due to the null mutation or to the geneticbackground.This issueis especiallyproblematic if the geneticbackgroundof the null-mutantanimalsis differentfrom that of their wild-type(control) counterparts.Asdiscussedbelow,this latter possibilityhas unfortunatelybeen the case in a large number of recent, molecularneurobiologystudies.To appreciatethe problem, consider how gene targeting is performed. Null-mutantmiceof gene-targetingstudiesare often hybridsof two mousestrains

Mostgenetargetin~is currentlycarriedout in cultured embryonic stem (ES) cells derived from the mouse strain ‘129’ (seeTable 1). The 129-typeEScells carrying the targetedmutation are introducedinto a blastocyststage embryo and the survivingchimeric embryos are allowedto develop to term, raised to adulthood, and matedto ‘wild-type’,that is, non-mutated,mice. In the caseof successfulgermline transmission,these matings producean offspringgenerationin which heterozygous null-mutantmice arefound.Problemsarise,however,if the geneticbackgroundof the EScell and of the mice to which the chimeras are mated are not identical. In a number of studies (see Table 1) the ES cells were derivedfrom the mouse strain 129 but the chimeric mice werematedto mice from a differentstrain,for example, C57BL/6(BL6). The offspringof such matings (the F1 generation),therefore,arenot only heterozygousfor the null-mutant allele, but have one set of chromosomes fromstrain 129 and another from strainBL6 (seeFig.1). Theseheterozygousmice,whenmatedwiththeirsiblings, will produce a segregating F2 population in which, accordingto Mendel’sLaw,homozygousnull-mutant, heterozygousnull-mutantandwild-typemicearefound. RobertGdai is at GENENTECHInc.,

Hybridnull-mutant micearegenetically different NeuroscienceDept, fromtheircontrolIittermates notonly at the locus 460 Point San of the targetedgenebut at other locias well BrunoBoulevard,

Comparisonof the homozygousmutant,heterozygous SouthSan mutant and wild-typelittermates of an F2 population Francisco,CA appears to be an ideal way to reveal phenotypical 94080-4990, USA. PI1: S0166-2236(96)20020-7

TINS Vol. 19, No. 5, 1996

177

DEBATE

Gene-targeting studies of mammalian behavior

TABLE 1.Examples of gene-targeting studies of mammalian behavior Targeted gene

ES

ESorigin

Chimeras crossed to

En-2

D3

129

BL6, CD 1, 129

En-2

D3

129

5-ht/B

na

fin

Genotype of analyzed mice

--- -. . --- --

Behavioral abnormality

129–BL6, 12Y–CIJI hybrids, inbred 129

None reported

129

Inbred 129

Impaired motor-learning performance

129

129

Inbred 129

Enhancedaggressivebehavior

na

129

129and BL6

Inbred 129,and 129–BL6hybrid

Impaired spatial learning

d2 (clopamine receptor)

PI

129

BL6

BL6-129 hybrid

Decreased Iocomotory activities, impaired motor co-ordination

el (NMDA receptor)

TT2

BL6 x CBA

BL6

BL6–CBA hybrid

Impaired spatial learning

Glu receptor 82 locus

TT2

BL6 x CBA

BL6

BL&CBA hybrid

Decreased Iocomotory activities, impaired motor co-ordination

f12 (nACh-receptor subunit)

HMI

129

BL6 x DBA/2

129–BL6-DBN2 hybrid

Impaired passive-avoidance learning

l-Adenylate cyclase

ABI

129

BL6

129–BL6hybrid

Impaired spatial learning

mgluI

D3

129

BL6

129-BL6 hybrid

Motor impairment, impaired eye-blink conditioning

mgluI

HMI

129

BL6

129–BL6hybrid

Decreased Iocomotory activities, impaired spatial learning

NCAM

E14

129

not given

129– ?

Impaired spatial learning

PKCy

E14

129

BL6

129–BL6hybrid

Impaired spatial and passiveavoidancelearning

aCaMKll

E14

129

BALBIC

129–BALB/Chybrid

Impaired spatial learning

Ref.

of the EScell line is differentfrom the mousestrainthat the ES-cellchimerawas crossedto. In this case,the generated mutant mice are not pure bred and are subjectto problemsassociatedwith geneticlinkageand geneticvariability.Note, that in severalof the cited studies,the behavioraldefectsof the null-mutantanimalswere similarto those seenin the mousestrain (strain 129) from whichthe EScellswere obtained.These observationsare in line with the suggestionthat the phenotypicaldifferencesobservedbetween wild-type and null-mutantanimals might,in fact, be causedby 129-typegeneslinkedto the targeted locus,and not by the mutatedgene of interest. BL6 denotesmousestrain CS7BU6 and 129 denotesmousestrain 129.Abbreviations:CaMKH,Ca2+–calmodulin kinasetype H;ES,embryonicstem cell;GIu, glutamate;mglu,metabotropic glutamatereceptor; na, not available;NCAM, neural-celladhesionmolecule;PKC, protein kinaseC. *indicates studies where the origin

changesbroughtaboutby the null mutation. However, it is important to remember that such a segregating population constitutes mice with recombinant genotypes derived from the two, parental mouse strains (see Fig. 1). The difficultiesarisingfrom this are threefold. First, the recombination pattern (that is, which locus contains alleles from strain 129 and which locus contains allelesfrom strain BL6, and whether these are in a homozygousor heterozygousform) might be different between littermates. This implies that not even the wild-typelittermates of their mutant counterparts represent a good control population as their alleles might be different from those of the mutants, not only at the locus of the gene of interest but also at other loci. This might lead to false-positive results. Second, owing to the genetic variation resulting from the hybrid background,detecting significant effects of the mutant gene of interest might be difficult. Therefore, one might find false negativeresults.These 178

TINS Vol. 19, No. 5, 1996

two problems might be alleviated if one measures a largernumber of animalsand thus increasesthe power of statisticalcomparisonsanddecreasesthe possibilityof samplingerror associatedwith recombination-pattern differencesbetween littermates. Increasingthe sample size, however,will not solve the third problem, which is associatedwith genetic linkage (see below). The allelesof genesthat surroundthe targeted locuswill be of 129-typein the null-mutantmice, andof BL6-typein the wild-typemice

If the targetedmutagenesisis made in ES cells from strain 129, the chromosomewith the targetedlocuswill carryallelesof genesof 129-type.Becausethe probability of geneticrecombinationis generallyinverselyrelatedto the distancebetweenthe loci of the genes,the 129-type alleles of the geneswhose loci are close to the locus of the mutatedgenewillremaintogetherwith the mutated allele of the gene of interest (see Fig. 1). That is, any

DEBATE

Gene-targeting studies of mammalian behavior

time the mutation is detectedin a mouse, for example, by Southernblotting, there is a high probabilityof that particular animal also carrying the linked 129-type genes. Conversely,a non-mutant control animal will, most probably,not carrythese 129-typeallelesand will have BL6-type alleles instead (if the 129-type ES cell chimera wascrossedto a BL6 strain). In effect, the mutation can be seen as a marker for the 129-type genes linkedto the locus of the targetedgene. Consequently, any phenotypicaldifferencesobservedbetweenmutant and control Iittermates of the hybrid genetic origin might be due either to the introduced null mutation or to the backgroundgeneslinkedto the targetedlocus. Thus, one might find false-positiveresults, a problem that has been ignored in a number of recent genetargeting studies (see Table 1).

‘Wild-type’ offspring

II

The behavioralalterationsseenin null-mutantmice might simplybe due to the geneticbackground

Empirical evidence implies that one might not be able to dismissthe problem of the genetic background or to arguethat the potential effect of linkedgenes,and the effectof the geneticvariation,in relationto the effect of the mutation is negligible. For example, recent gene-targeting studies investigating developmental consequences of the disruption of the receptor for epidermal growth factor have demonstrated nullmutation effects in mice that are dependent on the background genotype’oz’.The literature on behavior genetics also providesample evidencefor largegenetic differences between inbred mouse strains at the behavioral and neurobiology levels22.This suggeststhat the potential effect of linked backgroundgenescannot be ignoredin gene-targetingstudies.Furthermore,recent comparisons between strains have revealed that, unluckily, strain 129 (which has been the choice of gene-targetingstudies in ES cells) is one of the most unique strains in terms of behavior and neuroanatomyZW. These animals are severely impaired in spatial-learningtasks,a behavioralparadigmthat is frequentlyusedin molecular-neurobiology studies’-3131c,1’. They are also considered passive24)a behavioral trait that can be a seriousconfoundingproblemin a number Fig. 1. Chromosomolcorrsdtution of mice generated by gene targeting. Embryonicstem (ES) of other behavioraltests, includingthe context-depend- cellsoriginating from mouse strain 129 carry one chromosome (grey) with the disrupted a//e/e ent fear conditioning (in which a freezingresponseis (white lesion) of the targeted gene. If these EScellspapu}ate the germ line in the chimeric mice, measured),the open-field exploration test12,and tests the mutation will be transmitted when the chimera is mated. (A) A crossbetween a germ /ineof motor function such as the rotorod14’17’18. Moreover, transmitting chimera and a mause fram strain BL6 (black chromosomes) wi// produce an FI mice from strain 129 sufferfrom dysgenesisof the cor- population (B), in which 50% of the anima/s have one copy afthe mutant a//e/e (heterozygous pus callosum and possibly possessa number of other mutants) and 50% of them do not contain the mutant a//e/e (wi/d-type animals) at the neuroanatomicalpeculiarities23-25. Interestingly,in sev- targeted locus. Using Southern blatting or PCR techniques, ane can detect the presence af the eral gene-targetingstudies13’12’15-18 in which the mice mutant a//e/e and thus identify the heterazygous mutant anima/s. /f these animals are mated with each other, according to Mende/’s law homozygous mutant (two mutant a//e/es), originated from a cross between a 129-type ES cell heterozygaus mutant (ane mutant and one wild-type allele) and wild-type (two wild-type chimera and another strain, the null-mutant animals alleles) anima/s will be obtained. Ffawever, it is aka important to remember how genes at loci sufferedfrom behavioral defects similar to those seen other than the targeted ane will be inherited. Crass-over events during the meiotic process of in strain 129 (see Table 1 and Fig. 2). It is therefore gametagenesis will ‘shuffle’ the alleles af these background genes and will create recombinant possiblethat the differencesobservedbetweenmutant chromosomes (C), which characterize the genotype af the sperm and the egg of the F7 mice. and control mice were, in fact, due to the genetic dif- The genotype af an F2 individual/,therefare, wi// be represented by a pair of such recombinant ferences (in the linkedbackgroundgenes)betweenthe chromosomes. For example, a hamozygaus null-mutant mouse might have chromosomes inbredstrainsusedin the generationof the null-mutant a and b, a and c, or b and c; a heterozygous mouse might have one of the recombinant chromosomes with the lesion (a, b, or c) and another without the lesion (d, e, or ~; whereas animals and not to the null mutation. How couldone avoidthe confoundingeffectof linkedgenes?

A classical solution, one might suggest,wouldbe to decrease the probability of contribution of background genes by backcrossingthe mutant hybrid animals severaltimes to the strain of choice, for example,

a wild-type control mouse might have chromosomes d and e, d and ~ or e and f. (C) shows that the null-mutant allele af the targeted gene will be surrounded by 729-type genes, however, the wild-type allele of the gene will be surrounded by BL6-type genes. This linkage disequilibrium is simply due to the fact that the null-mutant a//e/e came from a genetic background from strain 129. In an F2 animal af the abave origin, the null-mutant allele could only be surrounded by BL6 genes it during the meiotic processes of gametogenesis, crass overs occurredpreciselyon both sidesflanking the targeted gene, events whose cambinedprobability is infinitesimally small.

TINSVOI. 19, NO. 5,1996

179

DEBATE

Gene-targeting studies of mammalian behavior

A

c

example,one couldusehomologous recombination (gene targeting) to 200, insert a small DNAmarkerflanking ~ 80 the gene of interest, without dis1$ 160----. +---+ rupting it. The knock-in mice gen0 g 70 VI erated this way would have a fully c K 1200 functional targeted gene and they .~ 60 5 would have, on average,the same al E 80; 50 0 linked genes as the knockout aniv mals in which the gene of interest ~ 40~ 40 has actually been disrupted.There‘ ~ + fore, the knock-in animals would OL — 30 represent the ideal control for the 1 2 3 knockoutnull mutants. Perhapsthe Days of training Strain of mouse Strain of mouse most elegant solution to the backFig. 2. Examples of behavioral differences between mice from strains 129 and BL6. (A) shows spatial-learning per- ground gene problem would be to formance in the water mazd’. Mice were trained in a cylindrical pool (diameter 1.5 m) filled with water. Each mouse use inducibleknockout techniques. received a training of six trials every day for three days, and was required to locate and climb onto a hidden platform These methods, which are being that was submerged 5 mm below the surface of the water. Mice were allowed to search for the platform fora maximum developed in several molecular duration of 120s, and the latency to find the platform (escape latency) was recorded. Data were analysed using genetic laboratories, would allow Student’s t-test. Although mice from both strain 129 (open squares; n =47) and strain BL6(black circles;n = 61) started the investigatorto switch the gene with the same level of performance [day 1: t= 1.22, df (degrees of freedom)= 100, P>O.20], mice from strain of interest on or off at particular BL6 achieved a significantly better performance with training (day2: t=3.66, df= 100, P<0.001; day3: t= 6.54, df= 100, P
B

90,

147

180

TINS Vol. 19, No. 5, 1996

DEBATE

Gene-targeting studies of mammalian behavior

variable, and are most probably influenced by a large number of genes as well as environmental factors. In orderto dissectsuch traits and to understandthe complex web of interactions among the underlying biological mechanisms, it is crucial to provide adequate controls for as many variablesas possible.Eliminating the confounding effects of background genes, therefore, is an important step forwardthat will facilitate our understanding of gene action in mammalian behavior. Selectedreferences 1 Abeliovich, A. et aL (1993) Cell 75, 1263-1271 Grant, S.G.N. et aL (1992) Science258, 1903-1910 3 Silva, A.J. et aL (1992) Science257, 206-211 4 Rose, S. (1995) Nature373, 38*383 5 Morris,R.G.M.and Kennedy,M.B. (1992) Cur. BioL2, 511-514 6 Routtenberg, A. (1995) Nature, 374, 314-31S 7 MacPhee, M. et aL (1995) Cell81, 957-966 8 Joyner, A.L. (1993) in Gene Targeting. A Practical Approach, p. 234, OxfordUniversityPress 2

9 Joyner, A.L. et aL (1991) Science 251, 1239-1243 10 Gerlai, R. et al. (1996) Behav. Neurmci. 110, 126-133 11 Saudou, F. et aL (1994) Science265, 1875-1878 12 Balk, J-H. et al. (1995) Nature 377, 424428 13 Sakimura, K. et aL (1995) Nature 373, 151-155 14 Kashiwabuchi,N. et aZ.(1995) Cell81, 245-252 15 Picciotto, M.R. et al. (1995) Nature374, 65-67 16 Wu, Z-L. et al. (1995) Proc. Natl Acad. Sci. USA92, 220-224 17 Afba,A. et al. (1994) Cell 79, 377-388 18 Conquet, F. et al. (1994) Nature 372, 237-243 19 Cremer, H. et aL (1994) Nature 367, 455-459 20 Threadgill, D.W. et aL (1995) Science 269, 230-234 21 Sibilia,M. and Wagner, E. (1995) Science 269, 234-237 22 Plomin, R., DeFries,J.C. and McClearn,G.E. (1990) in Behavioral Genetics.A Primer(2nd edn), pp. 262–295, Freeman& Company 23 Lipp, H-P., Bozicevic-Stagliar, M. and Wolfer, D.P. (1995) Behav. Genet. 25, 275 24 Wolfer, D.P., Bozicevic-Stagliar, M. and Lipp, H-P. (1995) Soc.Neurosci.Abstr. 21, 1227 25 Wahlsten, D. and Ozaki, H.S. (1994) in Callosal Agenesis (Lassonde,M. andJeeves,M.A., eds),pp. 119-155, PlenumPress 26 Festing, M.F.W. (1992) in Techniques for the Genetic Analysis of Brain andBehavior: Focas on theMouse(Goldowiz,D., Wahlsten,D. and Wimer, R.E., eds), pp. 17-38, Elsevier

Acknowledgements Thanks are due to ].G. cu~otti (Toronto), L. KarkowskiShuman (Richmond), J.C. Roder (Toronto),and J Rossant(Toronto) for theircomments on the manuscript. Supportedby MedicalResearch Council(Canada) and Ciba-Geigy.

Unusual behavioral phenotypes of inbred mouse strains Trends Neurosci. (1996) 19, 181-182

Gerlailpresentsa cogent pIeafor considerationof backgroundgenes,in the context of the excitingnewstudies of knockout and transgenic mice. Unusualbehavioral traits in the parental strains can seriouslycompromise the interpretationof unusualphenotypesin knockouts and transgenics. A major issueis the compensationby other genesfor the missingor overexpressedgenein knockoutandtransgenic mice. The techniques for developingmice with mutationsin genesof interestrequirethat the mutation alreadybe present in the first stagesof embryogenesis. The gene is missing throughout development. Since nature abhors a vacuum, another gene, coding for a product in a related biosynthetic pathway,might take overthe function of the missinggeneof interest.Therefore,in caseswherea knockoutmouseappearsto be normal in regardto traits linked to the missing gene, it is likelythat a compensatoryprocessduringdevelopment has maskedthe functional outcome of the mutation. From the point of view of developmentalbiology, the compensatory process is a fascinating study unto itself. In cases where the knockout mice appearto be phenotypicallynormal,wehavethe opportunityto learn a great deal about genetic redundancyand alternative biochemical pathways. Such an example arose in our work on knockout models of Tay-Sachs disease and Sandhoff disease2,3. Tay-Sachsdiseaseis a human sYndrome causedby a mutation in the HEXA gene for the alpha subunit of the heterodimer lysosomal storage enzyme, f3-hexosaminidaseA (cx,~),which results in accumulation of GM2 gawlioside in the CNS, motor deterioration, mental retardation, and early deathz. However,knockout mice lacking &hexosaminidaseA show normal motor behaviors, normal performance on a memory task, and minimal levels of ganglioside accumulationin the brain and spinalcord, ascompared to the levels of ganglioside accumulation that are Copyright @ 1996, Elsevier Science Ltd. All rights reserved. 0166-

2236/96/$15.00

common in Tay-Sachsdiseasez.Sandhoff disease,a relatedhuman syndromethat is alsocharacterizedby G~2 accumulation in the CNS, is the result of a mutation in the HEXBgenefor the &subunitof ~-hexosaminidase A (ct,~)and of the homodimerjl-hexosaminidaseB (PJ3). Knockingout HEXBproducedmice with severemotor deficitson rotorodbehavior,a paradigmwhichmeasures motor co-ordination mediatedby the cerebellumand spinal cord3.Concomitant with the highly significant motor deficits are high levels of gangliosideaccumulation in the brain3.Thus,in humans,a mutationin the HEXA gene producesa deleteriousphenotype,while in mice compensation appearsto occur for the missing HEXA gene. The two enzymes, ~-hexosaminidaseA and ~-hexosaminidaseB, appearto substitutefor each other to some degreein mice. These findings illustrate the ability of a related enzyme to compensate for the other, when one is missing during development. Furthermore, mice and humans appearto differ in their mechanismsof compensationin the ~-hexosaminidase pathways.Findingsof compensatorygenesin one species might not hold true for another species. Gerlai makes the excellent suggestionthat phenotypic characterizationof a new knockout or transgenic mouse requires a comparison of the homozygous mutant, the heterozygousmutant, and the wild-type littermates, in the F2 population. F2 populations of knockout and transgenicmice are generallycreatedby using two available parental strains: (a) embryonic stem (ES)cells that contain the mutation, are derived from genomic clones from the 129/SVstrain library, and injected into the blastocoel cavity of (b) a parental strain that is a good breeder, often C57BL/6J.The F1 offspringare then mated to each other to producethe F2 generation, in which the functional consequences of the mutation are studied.The genetic background of the parentalstrains,at the chromosomal locus to be PII: S0166-2236(96)20021-9

TINS Vol. 19, NO. 5, 1996

Commerrtaryby JacquelineN. CrawZey Sectionon Behavioral NeurophawnaCology Experimental Therapeutics Branch, National Instituteof Mental Health, Bethesda, MD20892-1380, USA.

181