Wim E. Crusio

Institut de Transgénose

CNRS UPR 9074, Génétique, Neurogénétique et Comportement

3b rue de la Férollerie

45071 Orléans Cedex 2

France

Key words: gene targeting, behavioral neurogenetics, behavior, hippocampus, inbred mouse strains, neurological mutants

Send correspondence to: Dr. Wim E. Crusio at the above address.

Tel. 19 33 1 42 86 22 06; Fax. 19 33 1 42 86 22 50

e-mail: crusio@cnrs-orleans.fr

I have chosen from the wealth of information presented at this meeting, that was attended by over 200 participants, a few topics illustrating the major lines along which research in this field is proceeding.

One of the oldest approaches to study genes influencing behavior is by treating animals with mutagenic substances and subsequently screening their descendants for any deviations in phenotypes of interest. Fittingly, the Honorary Keynote Lecture with which the meeting was opened was delivered by Seymour Benzer (Pasadena, CA, USA), a pioneer in this field and by now in his mid eighties. During his long career, he has worked with many organisms, but above all with that icon of genetics, the fruitfly (Drosophila melanogaster). In a fascinating talk, Prof. Benzer presented results from his search for life extension mutants in order to identify individual genes that regulate biological aging. Several such genes have been described for the nematode Caenorhabditis elegans². By generating a set of P-element insertion lines, the mutant methuselah (mth) was isolated. This mutant lived on the average 35% longer than the parent strain, both at 25 and at 29°C. As was the case with the mutations that had earlier been isolated in C. elegans, the life extension properties of this mutation were associated with elevated resistance to various forms of stress, such as starvation, high temperature, and dietary paraquat³. In both worms and flies, the underlying mechanisms appear to be related with increased resistance to the damaging effects of free radicals.

In recent years the mutagenesis approach has also been applied to that other icon of genetics, the mouse. Screening strategies employing the alkylating agent N-ethyl-N-nitrosourea (ENU) have been used by Joseph Takahashi (Evanston, IL, USA) to identify a semidominant, autosomal mutation called Clock that lengthens the circadian period by 1 hour in heterozygotes and by 4 hours in homozygotes. In constant darkness, homozygous mutants completely lose persistent circadian rhythms. The gene in question has been identified by positional cloning and it was subsequently shown that the regulation of the expression of this gene is by way of a feedback loop in which the CLOCK protein induces the transcription of its own inhibitors per and tim⁴. It appears that fly and mammalian circadian clocks likely share a conserved molecular mechanism⁵.

The current rapid progress obtained with the Clock mutation and, as reviewed by Wayne Frankel (Bar Harbor, ME, USA), also with numerous other neurological mutations of the mouse, are finally providing behavioral neurogeneticists with the tools they need to explore not only the (now answered) question of whether genes influence behavior, but also exactly how they do it.

Susumu Tonegawa (Cambridge, MA, USA) reviewed strategies to uncover molecular, cellular, and neuronal ensemble mechanisms underlying various cognitive functions. Traditionally, this has been done by pharmacological manipulations, but these have the disadvantage of often insufficient knowledge about the specificity of administered compounds or the area of the brain affected by them. As an alternative technology, the induction of specific mutations in targeted genes has become available in recent years. Generally, these mutations involve a complete loss-of-function (knockout), throughout the organism and throughout its life span. Conditional and inducible knockout strategies that do not suffer from this drawback because the induction of the null-mutation is limited both anatomically and temporally, are gradually becoming available. Knockout studies are already producing exciting results, starting to provide answers to important questions in physiological psychology and psychopharmacology.

An alternative approach to random mutagenesis is to look for genes that are naturally polymorphic and underlie normal, non-pathological individual variation. In the last decade, mapping of so-called Quantitative Trait Loci (QTL) has become increasingly fashionable. John Belknap (Portland, OR, USA) discussed some of the methods used to localize QTLs, such as Recombinant Inbred Strains and F₂ crosses, and results from his work on drug-related behaviors. Jeanne Wehner (Boulder, CO, USA) mapped QTLs for cued and contextual fear conditioning to chromosomes 1, 2, 3, 10, and 16. However, there was a notable lack of progress in identifying the genes responsible for QTL effects. In fact, current QTL methods suffer from several problems such as large confidence intervals (generally to the order of 10-15 cM) and a lack of replicability. Indeed, as yet not a single gene has been identified using these methods, despite all the resources and best efforts poured into this endeavor by many different groups. A promising new technique is the use of a heterogeneous stock (HS), derived more than 60 generations ago from a cross between 8 standard inbred strains. This method makes use of the large number of meioses that has occurred in this stock, a strategy also used with Advanced Intercross Lines (AILS)^7,8. Talbot and colleagues (Oxford, UK) presented a poster reporting the mapping of a QTL for mouse "emotionality" to a region of less than 1 cM on chromosome 1 using HS mice. Together, these new approaches will perhaps finally allow the QTL emperor to obtain some new clothes that have, this time, some real substance.

Another important problem in the field is the fact that different laboratories often report different results, despite using similar mouse stocks and, nominally, similar behavioral test protocols. John Crabbe (Portland, OR, USA) found that mutant mice lacking the 5-HT_1B receptor gene voluntarily drank twice as much ethanol as wild-types and were less sensitive on one test of ethanol-induced ataxia. However, on other ataxia sensitivity tests, no differences were found between mutants and wild-types. It therefore appears that the effect of a single gene manipulation can be highly specific even when tests are limited to a presumably limited domain such as ataxia, underscoring the need for cross-validation studies of these tests.

To address this problem, Crawley (Bethesda, MD, USA) has proposed to establish a standardized test battery, which could subsequently be used by different laboratories and would hopefully lead to a better replicability of results. This approach was also taken by Richard Brown (Halifax, NS, Canada), who presented a poster advocating the use of a battery of tests, named "Mouse IQ", to evaluate learning and memory. Although there is a definite need for more standardization of tests, I would argue that too much standardization also could have unwanted consequences, for two reasons.

First, in my own talk I presented data showing that rather minor changes of protocol could dramatically alter results obtained with a battery of inbred strains of mice in spatial 8-arm radial maze tasks. At first sight, this of course argues for more stringent standardization, but by systematically manipulating several features of the task protocol the learning behavior of mice in this task could be dissected rather precisely. It appears that, at least initially, mice use spatial strategies to solve such tasks. Only at later stages of learning do they eventually switch to nonspatial strategies. Spatial memory as measured in these tasks appeared to be unitary and distinct from nonspatial memory, of which there appeared to exist multiple forms. No support could be obtained for differentiating between the constructs of working and reference memory⁹. As an aside it may be noted that radial-maze tasks, when suitably designed, are less stressful for mice than water-maze tasks¹⁰. It is therefore somewhat surprising that the great majority of studies of learning and memory in mutant mice use water navigation tasks. In any case, these radial-maze data showed that if we were to standardize behavioral protocols too rigidly, we might miss important phenomena.

Traditionally, the goals of the field of behavioral genetics have been seen as the elucidation of causation, which has two aspects: a proximate, phenogenetic and an ultimate, phylogenetic one¹¹. The former relates to the physiological mechanisms underlying a trait, whereas the latter concerns questions about its adaptive value. Most current research, and all of the work described above, obviously addresses phenogenetic problems. It appears that current methods of analysis would add another, orthogonal dimension to this dichotomy: that of naturally polymorphic vs. monomorphic genes (Table I). When studying knockout mice, and mutants in general, we are mainly dealing with genes that are normally not polymorphic. Indeed, most null mutations are not found to occur spontaneously in natural populations. When investigating the effects of such mutations, we will generally deal with underlying mechanisms common to most if not all members of a species. In contrast, we are investigating mechanisms underlying individual differences when studying natural genetic variation. In short, whereas one type of question addresses, for example, how animals store information, the other type of question asks why some individuals perform better in a given task than others. Mutational analysis will only rarely identify genes causing individuals to differ among eachother.

Another observation to be made at this point concerns whether single-gene analysis, be it by means of mutational analysis or by investigating the genes underlying QTL effects (once it will finally become feasible to identify those), will ever help us to elucidate all aspects of, say, learning behavior. I and others¹² would argue that it will not. Sooner or later, single-gene analysis will certainly help us clarify basic cellular mechanisms of information storage and there is very clearly a great potential for the development of new therapeutic tools. However, defining the function of the hippocampus, say, or to explain the existence of multiple memory systems would be a daunting task indeed if it were to be done by single-gene analysis only and would take reductionism a bridge too far. This appears akin to trying to deduce the orbit of the earth around the sun using only knowledge about subatomic particles.

The preparation of this article greatly benefited from discussions with John Crabbe (Portland, OR, USA) and Douglas Wahlsten (Edmonton, AB, Canada). WEC was supported by the Centre National de la Recherche Scientifique (UPR 9074), Ministry for Research and Technology, Région Centre, and Préfecture de la Région Centre. UPR 9074 is affiliated with INSERM and the University of Orléans.