Mouse Molecular Genetic Technologies: Title and subTitle BreakPromise for Psychiatric Research

It has become clear that genetic factors substantially influence susceptibility to psychiatric disorders, as well as the unique constellations of character traits that underlie human individuality. The recently completed sequencing of the 3 billion–base pair human genome paves the way for the identification of each of the 26 000 to 38 000 genes¹ with which humans are endowed. Rapid advances in mammalian genomics are providing new opportunities to identify genes that influence brain function and to determine their relevance to psychiatric disorders. An important approach for investigating genetic influences on brain function involves examining the consequences of experimental manipulations of genes on behavior. Because this work cannot be performed in humans, model organisms must be used for such studies. In particular, many investigators have turned to the mouse for the identification and characterization of genes that influence brain function and behavioral processes relevant to psychiatric disorders.

The recent explosion in the use of mice may be attributed in large part to dramatic advances in mouse genetics. Of all mammalian species, the mouse is currently the most amenable to genetic manipulation. Genetic regulation of behavior may be explored in the mouse by examining naturally occurring gene variants and by introducing genetic mutations. The effects of gene variants and induced mutations on a wide variety of behavioral traits (behavioral phenotypes) may be examined. Moreover, as fellow mammals, humans and mice possess remarkably similar genomes; with rare exceptions, a mouse ortholog exists for each human gene and vice versa.² Therefore, mouse genetic studies can provide insights into the actions of corresponding human genes, as well as animal models of clinical disorders.

The purpose of this review is to provide an overview of molecular genetic techniques in the mouse and their relevance to clinical psychiatry. These approaches may be divided into "transgenic/gene targeting" and "phenotype-based" strategies. In transgenic/gene targeting approaches, investigators study the function of known genes by manipulating them in mice and examining the resultant phenotypes. In phenotype-based strategies, investigators identify a phenotype of interest and attempt to determine the gene(s) that underlie the abnormality. Quantitative trait locus analysis and random mutagenesis strategies exemplify this approach. Examples highlighting the potential of mouse molecular genetic approaches to provide insights relevant to the etiology and treatment of neuropsychiatric disorders will be provided. Space limitations prohibit a detailed overview of important findings in this field.

How can we hope to gain insights into genetic influences on complex human character traits and psychiatric disease by studying a diminutive rodent? Importantly, a remarkable degree of similarity exists among mammalian species in central nervous system organization. The fundamental patterns of brain development, the organization of neural circuits, and neuronal signaling molecules are highly conserved between mouse and human. Accordingly, parallels also exist between the behavioral responses of mice and humans to psychoactive drugs. In this section, we will briefly discuss the relevance of a few commonly used mouse behavioral assays to psychiatric disorders.

A wide variety of behavioral assays have been used in rodents to study anxiety.^{3 - 4} In the most commonly used mouse paradigms, such as the open field, elevated plus maze, and light-dark box, animals are provided a choice between exposed areas and regions that provide cover. The time spent exploring the open and exposed regions is considered to inversely correlate with the anxiety state of the animal. Thus, anxiolytic drugs such as diazepam increase time spent exploring these regions, while anxiogenic agents have opposing effects. Accordingly, such assays have been extensively used to predict the anxiolytic potential of compounds in humans.³

With regard to behavioral paradigms that have been related to human depressive disorders, the 2 most commonly used assays have been termed behavioral despair tests because they measure the cessation of efforts to escape from aversive situations. In the forced swim assay, animals are placed in a container of water. After a period of active swimming, animals display increasing bouts of immobility (floating), a behavior interpreted as the cessation of attempts to escape. Analogously, in the tail suspension assay, animals are suspended by the tail and the time spent in a state of immobility is measured. The ability of a wide range of compounds to reduce immobility predicts their efficacy as antidepressants.^{5 - 7}

Given the limited prospects of modeling hallucinations, delusions, and thought disorders in mice, the application of mouse behavioral paradigms relevant to schizophrenia seems challenging. However, a few mouse behavioral assays have been considered relevant to components of schizophrenic disorders. Prepulse inhibition of the acoustic startle response assesses the ability of a sound stimulus to attenuate the startle response to a subsequent, more intense acoustic stimulus and is considered a measure of sensorimotor gating. This inhibitory response, which can be measured in both mice and humans, is reduced in schizophrenia (and in several other psychiatric disorders).⁸ Attempts have also been made to draw parallels between the social withdrawal observed in schizophrenia and measures of social interactions in mice, including assessment of the extent to which animals approach each other in dyadic encounters and huddle together during sleep.^{9 - 10}

Several behavioral paradigms are available for the assessment of learning and memory in the mouse. The most commonly used procedure, the Morris water task, is frequently used for the assessment of spatial learning.¹¹ Typically, animals receive multiple training sessions in which they are placed in a circular pool of opacified water. To escape, they must use spatial cues to navigate to a platform located just beneath the water surface. After training, a trial is performed with the escape platform removed, and the time spent by the animal in the region previously containing the platform is used as an indicator of memory retention. Variants of this procedure, in which the platform locations are regularly altered during training, may also be used to assess working memory. For a detailed discussion of these and other learning and memory paradigms, see Crawley.¹²

Mouse models of substance abuse are further developed than those of other psychiatric disorders, in part because responses to a known causative agent, ie, the abused drug, can be examined. A strong concordance exists in the predisposition of humans and mice to self-administer drugs. For some drugs, self-administration may be assessed in 2-bottle preference tests in which mice may choose to consume water or drug solutions (eg, see Berrettini et al¹³ ). In addition, operant conditioning paradigms may be used to determine the extent to which mice will work (lever press or nose poke) for oral or intravenous drug administration.^{14 - 15} Another procedure, conditioned place preference, measures the preference of animals for a chamber in which they had previously received a drug vs a chamber that had been paired with vehicle administration. Preference for the drug-paired chamber has been interpreted to provide a measure of drug-seeking behavior, or craving.¹⁶ Drug withdrawal syndromes may also be observed in mice and scored by standard behavioral observation methods.¹⁷

Limitations in the ability to study aspects of human psychiatric conditions in mice must be acknowledged. Because self-report data are not available in the mouse, many phenomenologic aspects of psychiatric disorders, such as abnormalities of self-esteem, guilt, obsessions, hallucinations, and delusions, may not be assessed. Instead, perturbations of psychological processes must be inferred from behavior. Available behavioral assays relating to clinical conditions such as anxiety and depression appear to have some predictive validity, ie, they predict the anxiolytic and antidepressant efficacy of compounds. However, their face validity (the degree to which the measured behaviors resemble these complex human disorders) and their construct validity (the degree to which the assays reproduce the cause and pathophysiology of human psychiatric disorders) are weak.¹⁸ This problem relates in part to limited knowledge of the pathophysiology of psychiatric disorders.¹⁹ With few exceptions, current mouse models may be most productively used to examine the neural bases of particular behavioral elements relevant to psychiatric disorders rather than as comprehensive models of psychiatric syndromes.

Molecular genetic procedures exist for introducing new genes, expressing elevated levels of endogenous genes, and eliminating or altering the function of identified target genes in the mouse. These techniques may be divided into 2 categories: (1) transgenic technologies, in which exogenous gene sequences are inserted into the mouse genome, and (2) gene targeting technologies, in which mutations are targeted to disrupt or otherwise modify an endogenous gene. Both approaches are in widespread use for examining gene function and for generating animal models of disease.

The term transgenic has been applied to mice in which exogenous DNA has been inserted into the genome.²⁰ The typical application of transgenic techniques results in the insertion of a gene of interest into a random location in the mouse genome. The procedures required for introducing "transgenes," exogenous DNA sequences, into the mouse genome have been adapted for widespread use.

Typically, transgenes are introduced into the genome by microinjection into fertilized mouse eggs (Figure 1A). The DNA constructs used as transgenes usually consist of a gene of interest linked to "promoter" sequences that regulate the anatomic distribution of transgene expression. Transgene DNA is introduced into the mouse genome by microinjection into the male pronucleus of a fertilized mouse egg. After microinjection, the eggs are surgically transferred into the oviducts of foster mothers. To determine whether the transgene has been incorporated into the genomes of offspring, DNA is isolated from mouse tail biopsy specimens and screened by Southern blotting and/or polymerase chain reaction techniques. Commonly, 20% to 40% of the mice that develop to term possess the transgene, and the remaining "wild-type" animals are devoid of induced mutations. Integration of multiple copies of the transgene usually occurs at a single random chromosomal location. At this stage, transgenic mice are heterozygous for transgene insertion. These "founder" mice are then bred to produce mice that are homozygous for the transgene.

Because of variable transgene expression in mice derived from different founders, multiple lines of mice are typically examined to identify lines that are optimal for further work. Several factors contribute to inconsistencies in transgene expression. For example, variable copy numbers of the transgene may be integrated into the genomes of different founders, leading to variable levels of transgene expression. In addition, the genomic site of transgene integration may affect the brain region–specific pattern and level of transgene expression. Moreover, transgenes may occasionally integrate into and disrupt native genes, leading to phenotypic abnormalities unrelated to the function of the transgene.²¹ This possibility can be assessed by determining whether similar phenotypes are present in animals derived from different founders because of the low likelihood that 2 founders would possess the same transgene integration site.

In recent years, transgenic mice have provided useful models for the study of human neurobehavioral disorders. For example, lines of transgenic mice have been generated that exhibit phenotypic characteristics resembling features of Alzheimer disease. One transgenic line was generated with neural expression of an abnormal form of human amyloid precursor protein associated with familial Alzheimer disease of early onset.²² The brains of these animals exhibited neuropathologic changes reminiscent of Alzheimer disease, such as amyloid deposits, neuritic plaques, synaptic loss, and gliosis. These transgenic mice also displayed cognitive impairments in the Morris water task. Moreover, both the neuropathologic and behavioral abnormalities progressed rapidly with increasing age. This and subsequent transgenic models have played important roles in furthering understanding of the etiology and treatment of Alzheimer disease.²³

Transgenic mice typically express their transgenes throughout development, which may complicate the interpretation of neurobehavioral phenotypes because of potential abnormalities of brain development. To address this problem, strategies have been developed that enable the inducible expression of transgenes in adulthood. A recent example of this is found in a line of transgenic mice with inducible expression of ΔFosB, a transcription factor proposed to influence the stimulant and rewarding effects of cocaine.²⁴ To test this hypothesis, transgenic mice were generated in which long-term treatment with a tetracycline analogue (doxycycline) during early development suppressed expression of ΔFosB. Subsequently, doxycycline was discontinued, leading to the induction of ΔFosB expression in the basal ganglia. In support of the study hypothesis, these animals exhibited increased locomotor responses to cocaine and enhancement of behavior associated with the rewarding properties of cocaine.

Gene targeting procedures enable the precise introduction of a mutation into a native gene of interest. Typically, mutations are designed to eliminate gene function, resulting in the generation of "knockout" or "null mutant" mice. Mutations that produce more subtle alterations in gene function may also be introduced. The initial step in the generation of gene-targeted mice is the targeted introduction of mutations into embryonic stem (ES) cells (Figure 1B). The ES cells are derived from mouse embryos and retain the ability to contribute to all of the tissues of the developing fetus. The ES cells are made permeable to engineered DNA by exposing them to an electrical current. The introduced targeting construct DNA typically consists of target gene sequences into which loss-of-function mutations have been engineered. Targeting constructs are designed to achieve homologous recombination, a process through which construct DNA precisely replaces its homologous native gene sequence.²⁵ In most ES cells, the exogenous DNA will incorporate at random locations throughout the genome at frequencies greatly exceeding homologous recombination. Therefore, drug selection strategies have been developed to select for those rare cells in which homologous recombination has occurred. After selection, ES cell clones are screened for homologous recombination by either polymerase chain reaction or Southern blot analysis.

The homologous recombinant cells, which are heterozygous for the introduced mutation, are then used to generate mice. The ES cells are typically microinjected into 2.5- to 3.5-day-old mouse embryos. The injected embryos are then surgically transferred into the uteri of foster mothers, which give birth to "chimeric" mice that are partly derived from the injected ES cells and partly derived from the host embryo. For example, ES cells derived from a brown strain of mice are often injected into embryos derived from a black strain, resulting in chimeras with coats containing black and brown patches. Chimeras are then mated with wild-type mice, and the progeny are screened for the presence of the mutation. The resulting heterozygous mice are bred to produce homozygous mutant mice completely lacking the normal gene product. Animals bearing "null" mutations that completely eliminate gene function are termed null mutant or knockout mice.

Studies of null mutant mice can disclose novel functional roles of neural genes and behavioral phenotypes relevant to human neuropsychiatric disorders. For example, marked behavioral disturbances were observed in a line of mice with a targeted null mutation of the serotonin (5-HT) 5-HT_1A receptor subtype.^{26 - 28} Behavioral phenotyping of these animals showed an enhancement of anxiety-related behaviors. Further studies demonstrated that animals lacking these receptors displayed increased frontal cortex 5-HT release when confronted with a mild stressor, suggesting a role for 5-HT_1A receptors in the serotonergic response to stress.²⁹ Thus, 5-HT_1A receptor mutant mice provide a useful animal model for examining the serotonergic regulation of anxiety.

Frequently, behavioral studies in null mutant mice show unanticipated neural roles for gene products. A good example of this is provided by a recent study of mice lacking the hypothalamic neuropeptide orexin (also known as hypocretin). Observations of homozygous mutant mice disclosed a dramatic behavioral syndrome.³⁰ The mutants exhibited frequent episodes of inactivity characterized by sudden collapse of the head and buckling of extremities. Electroencephalographic analysis showed these episodes to be extremely similar to narcoleptic attacks observed in humans and in a strain of narcoleptic dogs. Moreover, a mutation of an orexin receptor gene was found to underlie the canine syndrome,³¹ and recent clinical studies report orexin deficiencies in narcoleptic humans.^{32 - 33} Thus, studies of orexin null mutant mice demonstrated a novel role for orexin in sleep regulation and an important animal model for examining the pathophysiology and treatment of narcolepsy.

Gene targeting approaches may also be used to examine mechanisms underlying the behavioral effects of psychoactive drugs. For example, the nonselective serotonin receptor agonist m-chlorophenylpiperazine has been used as a probe of 5-HT function in clinical studies. The drug binds with high affinity to the 5-HT_2C receptor and to several other 5-HT receptor subtypes.³⁴ Although m-chlorophenylpiperazine typically reduces locomotor activity in rodents, it produced a paradoxical hyperlocomotor response in a line of 5-HT_2C receptor null mutant mice.³⁵ This response was blocked by pretreatment with a 5-HT_1B receptor antagonist, indicating that the absence of 5-HT_2C receptors unmasked a hyperlocomotor effect of m-chlorophenylpiperazine on 5-HT_1B receptors. These results illustrate how individual differences in genetic endowment may contribute to variable behavioral responses to drugs. Thus, when a drug alters the function of multiple gene products with opposing influences on behavior, mutations or allelic variation of these genes may lead to paradoxic responses.

Although gene targeting techniques are most commonly used to generate animals with null mutations, subtle mutations may also be introduced to alter, but not eliminate, gene function. For example, a single amino acid change was engineered in a gene encoding the α1 subunit of the γ-aminobutyric acid (GABA) GABA_A receptor, rendering GABA_A receptors containing this subunit insensitive to benzodiazepines, without affecting their responsiveness to endogenous GABA.^{36 - 37} The resulting animals exhibited reduced sensitivity to the sedative and amnestic effects of diazepam, but no change in sensitivity to the anxiolyticlike effects of this drug. These results indicate that benzodiazepine site ligands devoid of activity at α1 subunit–containing GABA_A receptors may retain anxiolytic properties, while lacking some of the side effects typically associated with benzodiazepines. This prediction has received support from a recent report of the behavioral effects of such a compound.³⁷

The standard application of gene-targeting technology has several inherent limitations. The mutations are typically present throughout embryonic and postnatal development. Therefore, potential developmental perturbations may complicate the interpretation of mutant phenotypes in adult animals. Another limitation of the standard gene targeting technology is that the mutations are ubiquitous, present in all of the cells of the animal. Thus, if a neural gene of interest is also expressed in peripheral tissues, then the absence of the gene product peripherally could lead to a lethal or altered phenotype, independent of its neural role. Moreover, for genes that are widely expressed in the central nervous system, it may also be difficult to anatomically localize the brain region(s) that underlies the mutant phenotypes. Although beyond the scope of this review, new strategies are under development to obviate these problems through the generation of targeted mutations that are inducible and brain region–specific.³⁸

In contrast to transgenic/gene targeting approaches, in which phenotypes arising from manipulations of known genes are examined, phenotype-based studies seek to uncover genes that underlie a particular phenotype of interest. Two approaches in common use are quantitative trait locus (QTL) and chemical mutagenesis strategies.

Individual differences among humans in psychological characteristics have both environmental and genetic causes. Recent heritability estimates for normal behavioral traits as well as psychopathologies in humans indicate that 40% to 90% of the phenotypic variation in a population is under the control of genetic factors.³⁹ One of the hallmarks of complex behavioral traits (eg, spatial learning ability and locomotor activity level) is that, unlike single-gene traits, gradations in the phenotype are observed such that a continuous distribution of phenotypic scores is exhibited in a population. A continuous distribution is consistent with polygenic regulation of a trait. Each of the genes regulating a complex trait may encode a protein that contributes a small portion of the variation in a population. To address the challenge this poses for gene mapping, Lander and Botstein⁴⁰ developed QTL analysis to allow for simultaneous mapping of multiple genes regulating a particular trait. A QTL is a chromosomal region that contains a gene, or genes, that regulates a portion of the genetic variation for a particular phenotype.

In QTL analysis, such genes are identified by means of a series of DNA marker sequences that are linked (in proximity) to the gene but are usually not within the gene itself. The DNA marker sequences that are different between 2 parents act as "shipping tags" that can be used to follow the transmission of genes. When there are multiple forms of such DNA marker sequences or genes, they are said to be polymorphic. The DNA polymorphisms that occur in repeats of DNA known as microsatellites are currently used in QTL analyses.⁴¹ These polymorphisms, termed simple sequence length polymorphisms or SSLPs, can be easily assayed⁴¹ and must be distributed throughout the entire genome to detect all potential QTLs reliably.

Most QTL experiments in rodents start with 2 parental strains or with selected lines that have been inbred. Inbreeding limits diversity, so that each animal from an inbred strain will have 2 identical copies of each gene (ie, homozygous at all autosomal loci). For QTL studies, 2 strains are selected that differ for a particular phenotype of interest (Figure 2). Animals from the 2 strains are crossed to create a generation of hybrid mice termed F1. In the F1 generation, all the genes that differ between the 2 inbred parents are found in the heterozygous state because these mice have received 1 chromosome (and 1 copy of each gene) from each of the 2 differing inbred parents. When 2 F1 animals are crossed, considerable recombination (physical shuffling of genes) occurs during gamete formation, producing new gene arrangements in the offspring (F2 generation).

In Figure 2A, representative chromosomes from each of 2 parental strains and the location of DNA marker sequences that detect SSLPs between these 2 strains are shown as A, B, and C. Because each parent is from an inbred strain, each is homozygous at all loci. Thus, their offspring may have only 2 possible polymorphisms at any locus in the genome. The polymorphisms for DNA markers A, B, and C are detectable by differential sizes of polymerase chain reaction–amplified products, which are visualized by gel electrophoresis. These SSLP markers (A, B, and C) are informative for the analysis because the 2 strains contain polymorphisms at these loci that result in different-size polymerase chain reaction products. Two different SSLPs are detected in the F1 population because one was inherited from each parental strain. In the F2 generation, each animal can be unique and will demonstrate varying patterns of the DNA polymorphisms when markers are analyzed on all chromosomes.

To perform the QTL analysis, a phenotypic score is ascertained for each F2 animal by means of a behavioral assay (see "Mouse Behavioral Phenotypes" section, above). Often, the F2 animals performing at both extremes of the phenotypic distribution provide the most information, and these are subjected to SSLP analysis. For the example in Figure 2, one would ask whether the behavioral scores of animals with bb and cc are different from those of animals with BB and CC, and whether these behavioral differences are associated with specific DNA marker polymorphisms throughout the genome. The QTL data analyses are then performed with the use of a variety of statistical techniques to test the probability that variation in the phenotype is associated with a particular mapped polymorphism. The association of a QTL with a phenotype is reported by LOD scores (log of the odds) or standard α levels (P values).⁴² With this approach, QTLs have been detected that account for as little as 4% to 5% of the phenotypic variation.

Once a particular QTL is implicated, additional genetic approaches are applied to confirm and further pinpoint the location of the QTL.^{43 - 45} For example, the chromosomal regions containing a QTL region from parental strain 2 that is thought to confer a low response (region between bb and cc in Figure 2) can be transferred by systematic breeding methods to the high-responding parental strain 1.²⁹ This "congenic" strain will retain all other portions of parental strain 1's genome except the bb to cc QTL region. When tested behaviorally, the congenic mice should now show a lower response compared with the original parental strain 1. Differential gene expression profiling using DNA microarrays may be used to elucidate the genes in the QTL region.

Ultimately, positional cloning of genes in the QTL region, sequencing of known candidate genes, or proof of differential gene expression is required to identify the gene bearing the functional polymorphism. Positional cloning involves the isolation of genomic clones that contain the gene of interest from yeast or bacterial artificial chromosome libraries by means of high-density SSLP marker information.^{46 - 47} At least 2 types of functional polymorphisms will contribute to phenotypic differences observed between the parental strains. The first is a polymorphism in the coding region of the gene that produces an amino acid change in a protein. A second type of polymorphism that may be more difficult to detect would be found in regions that regulate gene expression.

Analysis by QTL is useful for complex disorders and behaviors because it allows the localization of genes without any a priori knowledge of the genes themselves. Thus, QTL analyses are under way to identify genes contributing to emotionality,^{48 - 50} learning,^{11 ,51} seizure sensitivity,^{52 - 53} and sensorimotor gating.⁵⁴ Pharmacologic models of drug use and abuse⁵⁵ are being used to elucidate the genes regulating variation in sensitivity, tolerance, and withdrawal to barbiturates,⁵⁶ opiates,¹³ cocaine,⁵⁷ and alcohol.^{43 - 44} Recently, a polymorphism in the GABA receptor γ2 subunit gene has been identified in a QTL on mouse chromosome 11 that appears to be associated with variation in ethanol withdrawal seizures.⁵⁸

There are limitations to QTL analysis.⁵⁹ At the practical level, it may be impossible as well as prohibitively expensive to detect QTLs for traits that are regulated by a very large number of genes with minuscule effect sizes. Moreover, QTL analysis does not detect all genes that are essential to neurobiologic pathways regulating behavior, but rather only those that determine variation between the parents. In addition, examples do not yet exist of particular genetic polymorphisms that influence both mouse and human behavior.

Chemical mutagenesis strategies represent an alternative phenotype-based approach for uncovering genetic influences on behavior. Typically, chemical agents are used to produce large numbers of animals bearing randomly induced mutations. These animals are screened for mutations that alter physiologic functions of interest. An advantage of the chemical mutagenesis approach is that the entire genome may be screened for genes influencing a physiologic process, as opposed to the more limited set of genes that are polymorphic between 2 parental strains. By contrast with QTL strategies, the majority of mutations detected in chemical mutagenesis screens are unlikely to contribute to natural variation in the phenotype of interest.

Chemical mutagenesis techniques have been used effectively for gene discovery in invertebrates.⁶⁰ The invertebrate work illustrates the power of chemical mutagenesis strategies to uncover novel developmental, biochemical, and behavioral pathways. In Drosophila melanogaster and Caenorhabditis elegans, mutagenesis screens have led to the elucidation of multiple genes that regulate behavioral processes such as learning and memory,⁶¹ circadian rhythms,⁶² and responses to alcohol.⁶³ Recently, the feasibility of this approach has been demonstrated for the mouse, and several large-scale mutagenesis programs have been initiated in this species.^{64 - 67}

Typically, male mice are treated with the mutagen N-ethyl-N-nitrosourea (ENU) to induce single base-pair mutations in the spermatogonia⁶⁴ (Figure 2B). These mice are then mated with normal females and offspring (G₁ generation) are screened for phenotypes of interest. Because all mutations in G₁ offspring will be in the heterozygous state, phenotypic screening of these animals detects only dominant mutations. Additional crosses of G₁ mice would be required to generate offspring that are homozygous for induced mutations. This would enable detection of recessive mutations, ie, mutations that produce phenotypic abnormalities only in the homozygous state.

Although mouse mutagenesis strategies most often involve ENU treatment of mice from standard inbred strains, transgenic or gene-targeted lines of mice may also be the subjects of screens for genes that modify particular phenotypes of interest. Analogous "modifier screens" have been successfully used in invertebrate systems to identify novel genes that interact with known mutations to influence complex biological processes (eg, see Karim et al⁶⁸ ). In some instances, mutations that fail to produce robust behavioral effects on a standard inbred background may produce dramatic effects in mice with a preexisting dysregulation of behavior. Variations of these strategies can also be used to screen for mutations in specific chromosomal areas or alleles.^{64 - 67}

In typical ENU mutagenesis screens, mice are tested in a battery of developmental, physiological, and behavioral assays. The resulting behavioral data must be interpreted with care.^{69 - 70} For example, problems such as blindness, poor olfaction, or profound cognitive impairment could alter behavior in an assay intended to assess anxiety. In such cases, the primary mutation may be in a gene other than one controlling biochemical pathways important for anxiety. Therefore, tests of sensory function and the health of other organ systems are usually incorporated in the primary mutagenesis screen.

The reliability and the sensitivity of behavioral measures must be considered in establishing criteria for identification of mice with candidate mutations. Mice with putative mutations of interest should have behavioral scores near the extremes of the population distribution. Even with this selection, false-positive results may still occur because behavior is regulated by both genetic and environmental factors. Thus, a behavioral trait associated with a true-positive mutation must be transmittable between generations. Once identified, mutations are then localized to particular chromosomal regions by means of the same type of gene mapping methods used in QTL analyses. Ultimately, sequencing of the gene in the normal and mutant states is required to identify the actual mutation.

As with QTL analyses, numerous examples of the use of chemical mutagenesis to isolate mouse behavioral mutants do not yet exist. However, a successful application of the approach has been documented in the isolation of the Clock mutant, which displayed profound abnormalities of circadian rhythms. The responsible mutation was subsequently mapped, and the Clock gene was cloned, providing novel insight into neural mechanisms that underlie circadian rhythms.^{46 ,71} The potential power of chemical mutagenesis approaches in the mouse has been widely recognized, and several international centers have been created that are devoted to mutagenesis screens. Already the mutagenesis center at the Medical Research Council in Harwell, England, has posted on the World Wide Web a series of mutants with neurobehavioral phenotypes. The ongoing results from this and other centers are posted on the World Wide Web.⁶⁴ It is anticipated that current large-scale efforts will produce thousands of single-gene mutants that will increase our basic understanding of biochemical processes underlying complex behaviors during the next decade.

The application of mouse molecular genetic technologies to the study of complex behavior has begun to provide important insights into the biology of neural processes relevant to the pathophysiology and treatment of psychiatric disorders. Continuing technological advances in this field will further enhance the power of these approaches. For example, the continued development of "second-generation" transgenic technologies enabling the induction of mutations in adult animals will obviate concerns regarding the contributions of potential developmental abnormalities to mutant phenotypes (eg, see Kelz et al²⁴ ). Ongoing advances in procedures for the precise restriction of mutations to particular neuronal cell types and brain regions will facilitate studies of the roles of gene products in defined neuronal circuits (eg, see Tsien et al⁷² ). Moreover, the growing accessibility of sophisticated "gene chip" technologies will allow simultaneous examination of the influence of targeted mutations on the expression of thousands of neuronal genes.

The utility and ease of performing phenotype-based genetic screens will also be facilitated by ongoing technological advances. In the future, it is likely that QTL analyses will be performed by means of single nucleotide polymorphisms that have been discovered both in humans⁷³ and across inbred mouse strains.⁷⁴ The incorporation of gene chip technology will allow simultaneous examination of thousands of single nucleotide polymorphisms and should revolutionize gene mapping and QTL analyses. Moreover, following on the heels of the Human Genome Project, the Mouse Genome Project will provide a complete map of the mouse genome, further facilitating the process of mapping QTLs and ENU-induced mutations.

To maximize the benefits of mouse molecular genetics to neuropsychiatric research, more rapid progress must be made in establishing mouse behavioral assays with demonstrated relevance to additional features of psychiatric disorders. For example, advances are needed in assays relevant to attention, psychosis, impulsivity, compulsions, panic, social withdrawal, and feeding regulation. An even greater challenge will be the generation of models that more fully mimic the complex constellations of features characterizing human psychiatric disorders. Prospects of generating such models will improve with advances in the development and application of mouse molecular genetic technologies, combined with advances in understanding the neurobiology and genetics of psychiatric disorders. Although the application of mouse molecular genetic approaches to psychiatric disease is still in its infancy, developments to date and the rapid progress in this field promise dramatic insights into neural and genetic mechanisms underlying psychiatric disease etiology and treatment.

Accepted for publication April 3, 2001.

This study was supported by grants DA-00282 from the National Institute on Drug Abuse, Bethesda, Md, and MH-61624 and MH-01949 from the National Institute of Mental Health, Bethesda (Dr Tecott), and by grants MH-53668 from the National Institute of Mental Health and AA-11275, AA-03527, and AA-00141 from the Institute on Alcohol Abuse and Alcoholism, Bethesda (Dr Wehner).

We thank Adele Dorison for editorial assistance.

Corresponding author and reprints: Laurence H. Tecott, MD, PhD, Department of Psychiatry and Center for Neurobiology and Psychiatry, University of California, 401 Parnassus Ave, San Francisco, CA 94143-0984 (e-mail: tecott@itsa.ucsf.edu).