behavior. This chapter presents a brief introduction to the neurogenetic approach and discusses its potential use and feasibility in understanding the neural mechanisms of interval timing.


Forward genetics identifies novel genes: it is most useful when one has no prior knowledge or wants to make no assumptions about the genetic basis of the trait being studied. It starts with a random mutagenesis of the genome by exposure to x-ray radiation or chemical mutagens. The mutated genes are transmitted to the next generation when the germline cells are affected during mutagenesis. The progeny is screened for modifications in the behavioral trait of interest, and when such mutant phenotypes are isolated, research proceeds to map and clone the genes underlying the phenotypic defect.

Forward genetic research requires substantial investments of time, space, and funds. The feasibility of forward genetics depends on multiple factors that determine the efficiency of obtaining and isolating mutant phenotypes and the practicality of positional cloning. First of all, the behavioral tests that are used to screen the phenotypes should be appropriate for screening literally thousands of animals over a reasonable period of time. Obviously, the model organism should have a short generation time and low maintenance costs. Further, high-density linkage maps and genomic libraries of the model organism should be available for positional cloning. These factors impose considerable limitations on the choice of the model organism. As unlikely a model of complex behavioral traits as it seems to be, the fruit fly Drosophila melanogaster has been the favorite species of behavior geneticists because it is rather unique for fulfilling the criteria for the feasibility of forward genetic analysis. Forward genetic analysis of behavior in the mouse has been taken up only very recently: for example, the circadian period gene of Drosophila (Konopka and Benzer, 1971) was identified 23 years before the Clock gene of the mouse (King et al., 1994; Vitaterna et al., 1994).

11.1.1 Genome-Wide Saturation Mutagenesis Requires Screening Thousands of Phenotypes

The first step in forward genetic analysis is the random mutagenesis of the genome. Genome-wide saturation mutagenesis refers to mutating every locus in the genome at least once. Theoretically, approaching saturation mutagenesis should allow the detection of a set of genes, each of which produces a phenotypic modification in the behavioral trait of interest when mutated. Conversely, the failure to saturate the genome might result in a failure to mutate and thereby detect the genes that affect the behavioral trait being studied.

Several effective germline mutagens are available for both Drosophila and mice, each associated with a different type (e.g., small or large deletions, insertions, point mutations) or rate of mutation. When the purpose is to produce mutant behavioral phenotypes, what is intended in most cases is the point mutation of a gene, resulting in either a gain or loss of function for that gene that would produce a detectable novel phenotype for the behavioral trait being studied. This can be contrasted with small or large deletions in the genome with widespread effects that can considerably reduce viability or fertility, in addition to a modification in the behavioral trait of interest.

How many phenotypes need to be screened to reach saturation mutagenesis? Let me give an example. The alkylating agent ethylnitrosourea (ENU) can produce point mutations at a rate of 0.0015 mutations per locus per gamete in the mouse (Takahashi et al., 1994). In the case of a dominant mutation where each first-generation offspring represents one mutagenized gamete, approximately 2600 gametes are required for an 87.5% chance of mutating any locus. However, because a mutagenized gamete is not always passed onto the progeny, or it might fail to be recognized or recovered after being passed on, even higher numbers of gametes are screened to make sure that a mutation can be isolated when it occurs (Justice, 2000). In practice, approximately 3000 animals are screened to approach saturation mutagenesis in the mouse following ENU treatment (Takahashi et al., 1994). That is a large number. Due to the substantial investment of time, space, and funds it requires, the forward genetic approach to behavior using the mouse as the model organism had not been considered feasible until recently (e.g., Capecchi, 1989).

11.1.2 Mutant Screens Should Be Both Task Relevant and Efficient

Behaviorally screening 3000 animals to isolate mutant phenotypes is a daunting task, and it is only the initial step in identifying genes. Assuming that one animal is screened per day, it would take approximately 10 years of working 6 days a week to reach saturation mutagenesis. Clearly, mutant screens should be extremely rapid and efficient.

Experimental psychologists observe behavior in detail, which takes time. The majority of the procedures developed and used most productively over the years by experimental psychologists take too long and are too inefficient for mutant screening purposes. In the case of interval timing, for example, a variety of temporal discrimination, production, and estimation procedures are widely used to study temporally controlled behavior in rats, pigeons, humans, and other mostly vertebrate species. Most interval timing research is aimed at understanding the quantitative properties of behavior at the steady state when behavior is no longer changing appreciably. Depending on the timing procedure used, it might take 2 to several weeks for behavior to reach the steady state. Assuming it takes 2 weeks to reach the steady state, and subjects are tested in groups of 50, it would take 2.3 years to screen 3000 animals. Obviously, experimental procedures for faster behavioral testing are necessary for using a forward genetic approach to identify genes involved in interval timing. Nonassociative Learning Procedures

Interval timing procedures with extensive associative learning requirements are not useful as mutant screens because they are too long and labor-intensive. Nonassocia-tive learning procedures such as habituation are more appropriate. Habituation refers to the reduction in the amplitude of a response as the result of the repeated presentation of the stimulus that triggers the response. Temporal variables have well-established effects on the rate of habituation: the longer the interstimulus interval, the slower the habituation (e.g., Leaton, 1976; see also Leaton and Tighe, 1976). If the mechanisms of interval timing under associative conditioning procedures overlap with those that control the effects of temporal variables under nonassociative learning procedures, habituation can be used to screen for interval timing mutants (see Sasaki et al., 2001). For example, a mutation that affects the well-established inverse relation between the duration of the interstimulus intervals and the rate of habituation is also expected to affect temporally controlled behavior under a variety of time-based associative conditioning procedures. However, one should be cautious not to select the mutants that fail to habituate since this behavioral phenotype might result from a learning deficit as well as a timing deficit (see Section

Habituation is indeed an appropriate protocol for a mutant screen: First of all, the neural basis of habituation seems to have been conserved across phyla, and it has been subject to extensive research, pioneered by Kandel and associates (e.g., Castelluci et al., 1970; Kandel, 1985; also see Thompson and Spencer, 1966). Second, unlike most interval timing procedures that involve associative learning, habituation does not require extensive training; it can be obtained within minutes using automated apparatus for a variety of behavioral reflexes. Rate of habituation for a given behavioral response depends on multiple factors, including the frequency of stimulus presentation and the intensity of stimulus. The values of these parameters in a habituation protocol can easily be adjusted to ensure rapid testing. The Temporal Conditioning Procedure

The temporal conditioning procedure (also known as fixed-time procedure) might also be useful as a rapid behavioral screen. In this procedure, a reinforcer is presented at fixed temporal intervals regardless of the subject's behavior at the time of rein-forcer delivery. The subjects learn to anticipate the reinforcer right before its occurrence under the temporal conditioning procedure. Notice that the experimental protocol is quite similar to that of habituation in that both procedures basically involve periodic presentation of the same stimulus. However, when the stimulus is a biologically important one (e.g., food for a hungry animal), anticipatory responding rather than habituation is observed. Temporal conditioning as a mutant screen can isolate those mutant phenotypes that show changes in the accuracy and precision in the timing of the anticipatory responses (see Section Once again, the selection of the mutants that fail to show anticipatory responses should be avoided, because this behavioral phenotype can possibly be produced by a learning deficit rather than a timing one. Screening Phenotypes en Masse

One strategy to increase the efficiency of mutant screens is testing subjects as a group. For example, Diptera had long been recognized to exhibit associative learning (for reviews, see McGuire, 1984; Tully, 1984), but the early experimental procedures that were used to demonstrate conditioned responding in bees and flies were too long and effortful to serve as mutant screens. It is not a coincidence that dunce, the first Drosophila memory mutant (Dudai et al., 1976), was isolated shortly after the development of an olfactory avoidance learning procedure that allowed the training and testing of flies en masse (Quinn et al., 1974). A highly efficient and automated Pavlovian olfactory discrimination procedure that allows the testing of approximately 100 flies at a time was later developed (Tully and Quinn, 1985), which led to the expedited discovery of new learning mutants. Group testing of flies has been revolutionary in the genetic dissection of learning and memory (Dubnau and Tully, 1998), in spite of being controversial on the grounds that it fails to show whether individual flies learn (Holliday and Hirsch, 1986).

11.1.3 The Mutated Gene Should Produce a Clear Behavioral Phenotype

The choice of the behavioral domain is critical in whether attempts to dissect the genetic basis of the behavioral trait yield a success story. Behavioral mutations can be isolated if they result in a clear, detectable modification in the phenotype. If the behavioral trait of interest is well defined and quantifiable in the wild type, a mutation-induced deviation from the wild-type phenotype will also be clear and detectable. Such is the case for two behavioral domains, namely, circadian rhythms and learning and memory that have been subject to extensive genetic research. For example, mutations that affect circadian rhythms disrupt the phenotypic expression of 24-h rhythms or modify the intrinsic 24-h period by either shortening or lengthening it (Hall, 1998; Low-Zeddies and Takahashi, 2000; Ralph and Menaker, 1989). It is straightforward to measure both the absence of a 24-h rhythm and a change in the period of rhythms. Similarly, in the realm of learning and memory, mutations that affect the acquisition, storage, or retrieval of information can be detected in the form of an inability (or an enhanced ability) to perform a learned task (Dubnau and Tully, 1998; Silva et al., 1997).

The behavioral criteria for isolating interval timing mutants are important for minimizing the false positive and false negative errors in the identification of interval timing mutants. Below is a discussion of phenotypic changes that can be produced by mutations of single genes that are involved in interval timing and whether these behavioral modifications are appropriate for mutant screening purposes. Loss-of-Timing Mutations

Let us first consider the issue of whether we should expect to find genes (or neural mechanisms that involve the products of these genes) that are dedicated to timing such that when mutated, these genes affect solely the timing of stimuli or events, sparing other cognitive and perceptual processes. If learning the changeable temporal relationships in the animal's current environment has a significant effect on its survival, there might have been selection for evolution of genes that control the potential to learn such temporal regularities. But even if such genes exist, should we expect them to be any different than the genes that are involved in learning, or neural plasticity in general? Because everything happens in real time, everything that the animal learns has a temporal dimension that ought to be part of what is learned (e.g., Gallistel and Gibbon, 2000; Miller and Barnet, 1993). So the mechanisms of interval timing should be intricately bound to the real-time dynamics of the learning process, although it has been suggested that the timing of stimuli and events is underlain by an independent mechanism (e.g., Malapani et al., 1998, Matell and Meck, 2000; Matell et al., this volume).

Notice that even if the mechanisms of learning and timing are independent, screening for loss of interval timing mutations can result in misidentification of learning mutants as timing mutants. Unlike the circadian timing mechanism, the interval timing mechanism does not have an intrinsic period. The experimental contingency based on the timed interval is learned under interval timing procedures. That interval timing is reflected in learned behavior makes it very difficult to distinguish a loss-of-timing mutation from a loss-of-learning mutation because both would produce a phenotype that fails to learn a time-based contingency. For example, in a peak-interval procedure, the subjects are reinforced for their first response after a fixed interval (e.g., 30 sec) elapses on some trials. On other trials, the reinforcement is omitted, and the subject's responses are recorded over a period longer than the fixed interval. In these trials, response rate increases smoothly as time elapses, reaches a peak around the time when reinforcement is usually delivered, and decreases fairly symmetrically afterwards. Animals that fail to learn the temporal contingencies in this procedure respond at a constant, undifferentiated rate throughout the trial. This is exactly how mice that express a defective form of the cyclic AMP responsive element binding transcription factor (CREB) perform under the peak-interval procedure (Carvalho et al., 2000). Now, is the failure of the CREB-deficient mice to learn temporal contingency under the peak-interval procedure due to a timing deficit, a learning deficit, or both? CREB plays a well-established role in the formation of long-term memories, and animals that express a deficient form of CREB have impaired performance under other memory tests that do not involve time-based contingencies (Silva et al., 1998). So their impaired performance under the peak-interval procedure might be underlain by a deficit in learning. But because a deficiency in interval timing would produce exactly the same performance profile, it is hard to infer whether these animals have a timing problem as well. Clearly, in screening for interval timing mutants, the better strategy would be to look for mutation-induced modifications in well-established qualitative and quantitative properties of temporally controlled behavior instead of a loss-of-timing mutation. Mutations That Change the Speed of Timing

Let us assume for the sake of argument that timing is controlled by a single polymorphic gene such that individuals that express different allelic forms of the gene fall into discrete phenotypic categories with respect to timing speed. How would the behavior of the phenotypes be different? One would be tempted to say that individuals that have the fast-timing allele would overestimate the duration of intervals under time-based schedules, and vice versa. However, discrete phenotypic categories with respect to timing speed would not produce discrete phenotypic categories of behavior because individuals would learn to estimate durations correctly in relation to their own timing speed. The effects of a change in timing speed will be reflected in behavior only when the subjective speed of time changes for a given individual by an intervention such as the administration of psychoactive drugs (Meck, 1983), which does not provide any information about individual differences in timing speed (see Section 11.2.5). Therefore, speed of timing is not a good behavioral trait to start looking for single-gene effects due to the difficulty of measuring individual differences with respect to this trait. Mutations That Affect the Accuracy or Precision of Timing

What other aspects of temporally controlled behavior are likely to be controlled by single genes with measurable phenotypic effects? Accuracy and precision of interval timing are good candidates. Accuracy of timing refers to the coincidence of the average subjective estimates of duration with the actual duration of timed intervals. Precision of timing refers to the extent of variability in the subjective estimates of duration (Malapani et al., 1998). Both people and animals show individual differences in accuracy and precision of timing, although the heritability of neither measure has yet been reported.* Given that animals are raised and tested under the same conditions show variability with respect to accuracy or precision of timing, individual differences in these measures are likely to have a genetic basis. Whether accuracy or precision would be affected by single-gene mutations is an open experimental question.

Extensive research shows that both accuracy and precision of temporally controlled behavior are remarkably systematic and predictable under both Pavlovian and operant time-based schedules (e.g., Gibbon, 1977), which satisfies the criterion of a well-defined phenotypic trait for mutant screening purposes. A thorough review of the quantitative properties of temporally controlled behavior is beyond the scope of this chapter, but one such property, timescale invariance (Gallistel and Gibbon, 2000), is worth mentioning. Timescale invariance refers to the observation that when a time-based experimental protocol is repeated using different absolute time intervals, data from these experiments superimpose if plotted on relative rather than absolute timescales. In addition to temporal perception being controlled by relative rather than absolute durations, superimposition of curves also requires that the variability in the estimates of time increase in proportion with the length of the temporal intervals being estimated. Timescale invariance is the manifestation of a strong form of Weber's law, which states that a difference of equal ratio (but not of equal absolute amount) in a continuous variable (e.g., brightness, temperature, weight, etc.) yields equal perceptual discrimination. That temporally controlled behavior obeys Weber's law shows that both accuracy and precision of interval timing change systematically under interval timing schedules.

Accuracy and precision of timing are appropriate behavioral criteria to be used in mutant screens to isolate interval timing mutants. First of all, it is possible to detect individual differences in both accuracy and precision of timing. Second,

* This experiment is currently in progress in my lab.

although there are exceptions, Weber's law applies for temporally controlled behavior under most interval timing schedules (e.g., Gibbon, 1977; Platt, 1979; Stubbs, 1979). Mutation-induced deviations from Weber's law would be straightforward to detect and measure as long as one uses a time-based schedule under which the behavior of the wild-type controls is shown to obey Weber's law. Further, Weber's law in interval timing holds true across species (Green et al., 1999), suggesting that the mechanisms of accuracy and precision in interval timing might be evolu-tionarily conserved. In fact, because it applies to the perception of all continuous variables, Weber's law might reflect a constraint common to all sensory-perceptual processing.

11.1.4 Why Not Use the Rat?

So far, the feasibility of the forward genetic approach has been discussed in relation to the availability of rapid behavioral screens and existence of detectable phenotypes. But when research moves from the phenotype to the gene, even the existence of a clear, detectable phenotype, and the availability of efficient and valid behavioral tests to screen for that phenotype, does not ensure that the underlying molecular defect can be found easily.

How can novel genes be discovered if the mutant phenotype is not associated with a known molecular defect? The answer lies in positional cloning for which no information about gene function is necessary. Positional cloning starts with linkage analysis to map the location of the novel gene on a chromosome. When the gene is located on a narrow enough region on the chromosome through iterative linkage analysis, the DNA from that region is cloned, and candidate genes on the clone are identified.

Very briefly, here is how it works: The lower the distance between a pair of genes on a chromosome, the higher the probability that they will be inherited together. Linkage maps that depict the proximity and order of genes on a chromosome can be constructed by arranging crosses between parental strains that are polymorphic for a set of genes and then analyzing the offspring for the pattern of segregation between genes. As a result of the widespread use of molecular techniques that allow genotypic detection, not only the genes with visible phenotypic effects, but also other types of molecular markers in noncoding regions of DNA that exhibit polymorphisms can now be used to construct high-resolution whole-genome maps. These maps are used in positional cloning to locate the novel gene by determining its proximity to known markers using linkage analysis.

Note that the marker alleles used in the construction of high-density genome maps should be homozygous for a given inbred strain and polymorphic between strains for the experimental crosses between inbred strains to be informative. This requires that genotypic information on several inbred strains of a species be available. One of the early goals of the Human Genome Project was to construct high-density linkage maps of increasing resolution for model organisms, including the fruit fly Drosophila melanogaster and the mouse Mus musculus. The higher the resolution of a high-density linkage map, the easier and quicker it is to map and clone the novel genes using positional cloning. Unfortunately, the resolution of the linkage map for the rat, the favorite species of experimental psychologists, is not nearly as high as that of the mouse. One can obtain and isolate behavioral mutants in the rat, but given the low resolution of the linkage maps, positional cloning of the genes that produce the mutant phenotype would take several years. This is the main reason why behavior geneticists do not use the rat as a model organism in spite of the abundance of data on its behavioral profile. Mouse and Drosophila linkage maps have been developed over the years with the cumulative effort of several labs, and such an extensive effort is necessary for the rat before it becomes a popular model organism for genetic analysis.


The reverse genetic approach (also known as from gene to phenotype, or gene targeting approach) is used to study the function of genes that have already been identified and sequenced. This approach is based on producing targeted mutations in genes to alter their function. Several techniques are available to change the level of expression of a gene, each appropriate for a different type of experimental question. A knockout animal is produced if the targeted mutation prevents the expression of the gene, eliminating its function. A knockdown animal is produced if the targeted mutation reduces the gene function without completely eliminating it. Animals that have multiple copies of a particular gene are used to observe the effects of overexpression of that gene. A transgenic animal expresses a gene of another organism; these animals are often used as animal models of human diseases that have been linked to single genes (e.g., Huntington's disease).

The rationale behind using knockout animals is to understand the function of the targeted gene by observing the phenotypic defects that are produced in its absence. It should be emphasized that knockout animals do not provide direct information about gene function. Therefore, one should be rather conservative and cautious in making inferences about gene function based on the phenotypic changes produced by the absence of genes observed in the knockout animals.

11.2.1 Producing Gene Knockouts

Although gene targeting is possible in other species as well, the mouse is the favorite animal model for a reverse genetic approach to behavior. A knockout mouse strain can be produced as follows: once a gene has been identified and sequenced, its fragments can be introduced into mouse embryonic stem (ES) cells where they would insert themselves into the chromosome by undergoing homologous recombination with the endogenous DNA, causing a mutation at the locus they replace. In order to select the ES cells that have taken up the mutant gene, a selective marker such as the neomycin resistance gene is inserted into the targeting vector along with the homologous sequence. Then the ES cells are plated on neomycin, and only the ones that express the resistance gene, which are also those that express the mutant form of the targeted gene, survive. The transfected ES cells are then implanted into normal blastocysts, which in turn are implanted into pseudopregnant females. The resulting mouse pups will be chimeric, i.e., some of their cells will have originated from transfected ES cells, whereas others will have originated from the normal ES cells of the recipient blastula. If the transfected ES cells develop into germline cells, the mutation will be passed on to the offspring, which are bred to produce a mutant strain (Hasty et al., 2000).

11.2.2 Strain Differences

The genetic background is anything but trivial in assessing the effects of a targeted gene mutation because a targeted gene might have different phenotypic effects, depending on the genetic background (Choi, 1997; Dubnau and Tully, 1998; Silver, 1995). This is because genes do not always affect the phenotype on their own. Rather, interactions between genes (i.e., epistasis) or between the two copies of the same gene (i.e., dominance) play important roles in determining phenotypic expression. For example, it is possible for a targeted mutation to produce a behavioral effect in one strain, but no phenotypic difference in another due to epistatic interactions (Crusio, 2002). Not surprisingly, different strains of mouse show considerable differences with respect to several behavioral traits, including learning and memory (Wehner and Silva, 1996). Therefore, it is advisable to test the effects of a knockout on at least two different backgrounds.

In most cases, the ES cells that are used to produce knockouts are derived from the 129 inbred mouse strain. However, 129 is not the best genetic background to study for the behavioral effects of a modified gene because certain substrains of 129, such as the 129/J, fail to develop a normal corpus callosum (Wahlsten, 1982), and they have impaired memory (e.g., Montkowski et al., 1997). Behavioral pheno-typing of knockout strains is usually carried out after the 129-derived ES line is expressed in either a C57BL/6 or BALB/c background by repeatedly backcrossing the mice that carry the knockout gene to a C57BL/6 or BALB/c background (for breeding strategies, see Wolfer et al., 2002).

Behavioral testing should be carried out with the homozygous--, heterozygous"17-, and wild-type littermates obtained from breeding heterozygous"- parents. Note that the genetic background might still be segregating for the three genotypes after an insufficient number of backcrosses, which itself can produce behavioral differences, or prevent the expression of the potential behavioral difference produced by the targeted gene. Similarly, one should avoid breeding the homozygous-/- and the wildtype animals amongst themselves for the sake of efficiency of obtaining each genotype because this breeding strategy too can result in the segregation of the backgrounds, such that the homozygous-- genotype will be expressed in a 129 background from which the ES cells were derived, and the wild-type alleles will be expressed in another (e.g., C57BL/6) background.

11.2.3 Inducible and Tissue-Specific Knockouts

Conventional knockout animals lack a specific gene, and hence the protein product of that gene. They are not good models if one would like to understand how a gene is involved in the manifestation of behavior as it occurs in the adult organism. This is because genes that are involved in controlling the manifestation of a particular behavior during adulthood might also serve other functions in the central nervous system or other parts of the body during development. The absence of such genes might affect the viability, development, and general health of the organism to various degrees. Conversely, the behavioral phenotype of the knockout animals might appear to be normal due to the compensatory effects of other genes that have taken over during development (Crusio, 2002).

Inducible and tissue-specific mutations provide a good solution to the above problems by limiting the genetic lesions in time and space. In inducible mutations, the transcription of the mutant gene is controlled by the administration of a specific substance (e.g., the antibiotic tetracycline) or exposure to a specific event (e.g., heat shock). Tissue-specific mutations are obtained by putting the expression of the mutant gene under the transcriptional control of another gene that is expressed exclusively in the tissue of interest. Mutations can be both inducible and tissue specific, which limits the effects of the mutation in both time and space as required by the experimenter. In essence, then, genetic lesions produced by inducible and tissue-specific mutations are the equivalent of extremely selective drugs that temporarily affect the tissue of interest only.

11.2.4 Behavioral Phenotyping of Knockouts

The targeted mutation of a gene can affect a behavioral trait if the gene is involved in either the development of the neural substrates for behavior or the actual manifestation of behavior. In both instances of behavioral control by single genes, it is reasonable to expect that the gene be expressed in the tissues that serve the behavioral function (see Baker et al., 2001). On the other hand, mutations of genes that are involved in sensory or motor processes can also produce a modified phenotype with respect to the behavioral trait, and one should be cautious not to ascribe specific behavioral functions to such genes. Detailed descriptions of the test batteries that are designed to assess the general health and the motor and sensory-perceptual abilities of the knockout mice, as well as specific behavioral traits such as learning, memory, social behaviors, and emotion, are available (e.g., Crawley, 2000).

If the targeted gene plays a critical role in the normal development of the embryo, the homozygous-'- or the heterozygous"1"'- animals might not be as healthy as the wild-type littermates, making it difficult to obtain behavioral measures from all three phenotypes. In some cases, the differential viability of the phenotypes requires that the homozygous-'- or the heterozygous"'- animals be maintained under special conditions. For example, mice that lack dopamine transporters (DAT-KO) have an unusually high metabolism, so they are healthier when fed a high-fat diet. However, the high-fat diet causes the wild-type littermates to gain extra weight. Under these conditions, either the type or the amount of food has to be different for the homozy-gous and the wild-type littermates, which is a concern when food reinforcement is used during behavioral testing (see Section 11.2.5). Because maintenance conditions in the home cage environment might also affect behavior, caution should be addressed in treating the littermates of all genotypes as equally as possible.

Finally, the availability of valid, task-relevant behavioral tests can be a major problem, depending on the behavioral domain. The majority of the behavioral tests currently used were originally designed for rats and later adapted for mice. It seems to be an underappreciated fact that rats and mice are not as similar in behavior as they are in sight. A behavioral procedure that has not been appropriately adjusted for the mice might lead to erroneous or biased conclusions about this organism.

11.2.5 An Example: Interval Timing in Dopamine-Transporter Knockout Mice

A pilot interval timing experiment carried out with dopamine transporter knockout (DAT-KO) mice is presented here to provide an example of the reverse genetic approach to interval timing.**

The neurotransmitter dopamine is suggested in the timing of temporal intervals in the seconds-to-minutes range because a disruption or imbalance of dopaminergic activity alters the perception of temporal intervals. For example, following self-administration of amphetamines, humans report that "the world looks like it is in slow motion." Similarly, rats overestimate short temporal intervals following systemic injections of methamphetamine (e.g., Meck, 1983), suggesting that they too perceive time as being stretched out under amphetamine.

Amphetamine reverses the action of the dopamine transporter by causing it to release dopamine from the presynaptic terminals instead of clearing dopamine from the synapse. The net result of this process is increased and prolonged availability of dopamine in the synaptic cleft, causing a nonselective activation of different classes of dopamine receptors (Jones et al., 1998). Although all monoamine transporters in the central nervous system are substrates for amphetamine and its derivatives, amphetamine has higher affinity to the dopamine transporter relative to the serotonin transporter, and it does not release norepinephrine as much as dopamine. Therefore, at low to moderate doses, most of the behavioral and perceptual effects of amphetamine are due to its action on the dopamine system (Feldman et al., 1997).

A targeted mutation of the dopamine transporter (DAT), much like an acute amphetamine administration, prolongs the availability of dopamine in the synaptic cleft in mice that are homozygous (DAT-/-) or heterozygous (DAT+/-) for the mutation because the excess dopamine cannot be cleared from the synaptic cleft in the absence of functional dopamine transporters. On the other hand, the absence of the dopamine transporter also causes compensatory changes during development: the amount of dopamine released per electrical pulse is lower, dopamine synthesis is reduced (Jones et al., 1998), and the sensitivity of both pre- and postsynaptic receptors is altered in complex ways (Jones et al., 1999). Nevertheless, in spite of the compensatory changes, the net effect of a dopamine transporter knockout is functional hyperdopam-inergia in both DAT/- and DAT+/- mice (Gainetdinov et. al., 1999a).

The behavioral effects of the absence of the dopamine transporter are also quite similar to those of moderate doses of amphetamine. DAT/- mice exhibit marked

** These data were collected at Duke University when the author was a post-doctoral fellow working with Warren Meck.

hyperactivity along with an impairment in learning, which has led to the suggestion of the phenotype as a model for human attention deficit hyperactivity disorder (ADHD) (Gainetdinov et al., 1999b). In support of this view, an association between ADHD and polymorphisms in the noncoding regions of the human DAT gene has been suggested (Cook et al., 1995; Gill et al., 1997). Interestingly, ADHD patients have been reported to have altered temporal perception. In fact, it has been suggested that the perception of time being stretched out, or passing slowly, might contribute to the frequently reported boredom of ADHD patients, and that the impulsivity of these individuals can possibly be controlled by preventing this boredom (although see Goddard, 2000).

The purpose of the present experiment was to study temporal discrimination in dopamine transporter knockout mice. Male C57BL/129SvJ mice homozygous (DAT-/-) or heterozygous (DAT+/-) for DAT deletion and their wild-type (WT) littermates were used as subjects. They were generated by crossing DAT+/- parents and 3-month-old mice at the beginning of the experiment. In previous studies that reported the behavioral and physiological profile of the DAT knockouts, the mice had been backcrossed to a C57BL/6 background for seven to nine generations over an approximately 2-year period (e.g., Gainetdinov et al., 1999a). In the present experiment, knockout strains had not been backcrossed to a C57BL/6 background for more than two generations, so it is possible that the genetic background was still segregating for DAT/-, DAT+/-, and WT mice. Although the physical appearance and the behavioral profile of the DAT/- or DAT+/- mice used in this experiment were similar to those reported previously (e.g., Gainetdinov et al., 1999b; Spielowoy et al., 2000), a segregation in the genetic background might possibly have produced behavioral differences between the three genotypes in addition to the DAT deletion.

Although all three genotypes had originally been planned to be included in the experiment, DAT/- animals were later dropped. The present experiment involved administration of approximately 80 standard pellets (of which about 65 were collected) daily as food reinforcement. This required that the subjects be kept at their 85% body weight. DAT/- mice could not tolerate this food deprivation regimen because they are dwarf with very low-fat reserves, and they have an unusually high basal metabolism. In previous behavioral studies that involved food reinforcement, food was removed from the home cages 5 h before the beginning of daily experimental sessions, which seems to be the most that the DAT/- mice can tolerate (e.g., Gainetdinov et al., 1999b). When we tried the same 5-h food deprivation schedule, neither the DAT/- mice nor the others were hungry enough to eat 50 pellets during the experiment. Therefore, we decided to continue the experiment with DAT+/- and WT mice only. Notice that because the DAT+/- animals have one functional copy of the DAT gene, a comparison of DAT+/- and WT mice reveals the effects of a 50% reduction in the dosage rather than the absence of the gene. Also, using DAT+/- mice avoids confounders introduced by the absence of the gene during development. Although the DAT+/- mice too have functional hyperdopaminergia, they are indistinguishable in general appearance and health from their WT littermates. There were six subjects in each group at the beginning of the experiment, but one of the DAT+/-

mice died due to a brain tumor, and a WT mouse failed to learn the task. So the experiment continued with five subjects per group.***

The experiment started with training the subjects to press levers for food. Once a subject learned to press the levers for food, intertrial intervals (ITIs) of progressively longer duration were inserted between lever presentations until the ITIs were 45 sec long on average. Lever press training lasted until the subjects learned to press both the left and right levers reliably. During the two-signal training phase, each trial started with the onset of a light located approximately an inch above the food hopper. The light remained on for either 2.52 (short) or 12.7 (long) sec. The levers were presented immediately after the light was off, and the first response on either lever caused both levers to be retracted. A food pellet was delivered immediately if the left lever was pressed following the short signal, or the right lever following the long signal. If the mice failed to press either lever, the levers were retracted after 5 sec, and a null response was recorded. Trials were separated by ITIs that were 45 sec long on average. All lights, including the houselight, were off during the ITIs. Each group was run in two short daily sessions because the mice got satiated and stopped responding if they were run in a single long session. Each session consisted of 40 trials, so the animals were exposed to 80 trials daily. Half of the trials involved the presentation of the short signal, and the other half involved the presentation of the long signal.

A correction procedure was used to avoid preferences for either lever. When the subjects made an incorrect response, the same trial was repeated until a correct response was made. If the subject failed to make the correct response over two consecutive trial presentations, only the correct lever was presented during the third presentation of the same trial. Animals that were performing poorly were exposed to a cued training procedure where only the correct lever was presented to them following signal presentation. However, a cued training procedure, if used extensively, can retard rather than enhance learning because it prevents the animal from making a discrimination by presenting it with the correct response alternative only. So cued training was used only for those mice who had stopped responding during two-signal correction training, and it was terminated as soon as the animal started responding reliably. Finally, if a subject continued to have a position bias regardless of the correction training, it was exposed to intermittent sessions that involved the presentation of either the short or the long signal only.

Eight-signal training started after 40 days of two-signal training (including the cued and correction training). In order to prevent position biases, the first daily session was eight-signal training, and the second daily session involved two-signal correction training for each group. Eight-signal training involved the presentation of six additional signal durations (3.2, 4.0, 5.04, 6.4, 8.0, and 10.8 sec), spaced at equal geometric intervals between the short (2.52 sec) and the long (12.7 sec) signals. Each signal was presented five times within a session. A left lever press was reinforced following the presentation of the first four signals (2.52, 3.2, 4.0, and 5.04 sec), and a right lever press was reinforced following the presentation of the

*** In general, it is advisable to use at least 10 subjects per group (Crawley, 2000), but because this was a pilot study, we used smaller groups.

The ADHD Success Formula

The ADHD Success Formula

This is an audio and guide that will help you battle through ADHD and Accomplish Twice As Much In Half The Time. Learn more by download your very own copy today.

Get My Free Ebook

Post a comment