You are on page 1of 23

Psychological Bulielin 2000, Vol 12ft, No.

6, 806-828

Copyright 2000 by the American Psychological Association, Inc. 0033-2909/00/$5.00 DOI: ~ " 10.1037//0033-2909.126.6.806 -

DNA
RoberJ Plomin
King's College London

John Crabbe
Portland Alcohol Research Center and Oregon Health Sciences University

The authors predict that in a few years, many areas of psychology will be awash in specific genes responsible for the widespread influence of genetics on behavior. As the focus shifts from finding genes (genomics) to understanding how genes affect behavior (behavioral genomics), it is important for the future of psychology as a science that pathways between genes and behavior be examined not only at the molecular biological level of cells or the neuroscience level of the brain but also at the psychological level of analysis. After a brief overview of quantitative genetic research, the authors describe how genes that influence complex traits like behavioral dimensions and disorders in human and nonhuman animals are being found. Finally, the authors discuss behavioral genomics and predict that DNA will revolutionize psychological research and treatment early in the 21st century.

A first draft of the entire sequence of 3 billion nucleotide bases of DNA in the human genome was reported this year, several years ahead of schedule, 99.9% of these DNA sequences are the same for all people. Identifying the 0.1% of the DNA sequences that differ is one of the next goals of the Human Genome Project because these 3 million DNA sequences are responsible for the genetic differences among human beings behaviorally as well as biologically. In this so-called post-genotnics world, when all the DNA sequences and their variants are known, research will switch from finding genes to understanding how these genes work (functioned genomics). The term functional genomics usually connotes molecular biological and biochemical research that identifies gene products (proteins) and investigates their function at a cellular level. However, higher levels of analysis are increasingly needed to understand the pathways between genes and behavior. For example, the brain will undoubtedly be a major target for functional genomic research in neuroscience using neuroimaging techniques for both human and nonhuman research (Kosslyn & Plomin, in press) and more precise techniques such as the injection of antisense oligodeoxynucleotides (short strings of bases complementary to a specific messenger RNA [mRNA]) into the brain to turn off the production of specific genes for animal model research. The psychological level of analysis focuses on behavioral functions of the whole organism. To highlight its importance to this level of analysis, we refer to it as behavioral genomics.

A huge research effort is currently focused on finding genes associated with behavioral disorders. Such research is difficult and expensive, and relatively few psychologists are likely to join the hunt for genes. However, we predict that within a few years, psychology will be awash with genes associated with behavioral disorders as well as genes associated with variation in the normal range. As discussed later, psychologists will use information about these genes in their research for three reasons: (a) DNA-based information is becoming increasingly inexpensive and easy to gather, (b) behavioral genomics can make important contributions toward understanding the functions of genes, and (c) DNA opens up new scientific horizons for understanding behavior. We also predict that DNA will change clinical psychology, leading to gene-based diagnoses and treatment programs. These advances will need to be integrated with corresponding advances in understanding the multiple, interactive environmental influences as well. We discuss the ethical implications of such predictions at the end of this article. For these reasons, it is crucial that psychologists be prepared to take advantage of the exciting developments in molecular genetics. In the same way that computer literacy is now an essential goal to be achieved during elementary and secondary education, students in psychology must be taught about genetics to prepare them for this future. Otherwise, this opportunity for psychology will slip away by default to geneticists, and genetics is much too important a topic to be left to geneticists! The basic concepts of genetics and evolution should be incorporated into a solid grounding in neuroscience that is also crucial for psychological training. Clinical

Robert Plomin, Social, Genetic and Developmental Psychiatry Research Centre, Institute of Psychiatry, King's College London, United Kingdom; John Crabbe, Portland Alcohol Research Center, Portland, OR, and Department of Veterans Affairs Medical Center and Department of Behavioral Neuroscience, Oregon Health Sciences University. Correspondence concerning this article should be addressed to Robert Plomin, Social, Genetic and Developmental Psychiatry Research Centre, Institute of Psychiatry, King's College London, 111 Denmark Hill, London SE5 8AF United Kingdom. Electronic mail may be sent to r.plomin@iop. kcl.ac.uk.

psychologists use the acronym DNA to note "did not attend"it is critical to the future of psychology as a science that DNA means deoxyribonucleic acid rather than did not attend, In this article, we describe how genes associated with behavior are identified in human beings and in animal models and how these genes can be put to work in psychological research and treatment. To put this discussion in context, we begin with a brief overview of quantitative genetic studies. However, technical details about quantitative genetic methods, results, and qualifications pertinent

806

SPECIAL ISSUE: DNA

807

to interpretation of data are beyond the scope of this article (for discussion of these topics, see, e.g., Plomin, DeFries, McCleam, & McGuffin, 2001). Background references are also provided in our discussion of DNA in case we have missed the mark in trying to balance breadth and depth.

Quantitative Genetics The fundamental accomplishment of behavioral genetic research in psychology has been to demonstrate the ubiquitous importance of genetics throughout psychology. Historically, this evidence has comprised inbred strain and selection studies of nonhuman animal behavior and, for human behavior, twin studies that compared the similarity of identical and nonidentical twins and adoption studies that considered, for example, the resemblance of adopted-away children to their biological parents (Plomin et al., 2001). These methods and the theory that underlies them are called quantitative genetics in contrast to molecular genetic studies that attempt to identify specific genes. Behavioral genetics includes both quantitative and molecular genetic approaches to investigating genetic influences on individual differences in behavior. Behavioral genetics focuses on questions of why individuals within a species differ in behavior (e.g., why children differ in rates of language acquisition) as opposed to species-typical behavior (e.g., when, on average, children use two-word sentences). Genetics influences both species-typical behavior and individual differences within a species, in different ways. For example, human evolutionary history written in DNA code accounts for the fact that human beings use two-word sentences at the average age of 18 months, but this does not mean that genetics is the reason why some children do not use two-word sentences until much later in development. The controversy that swirled around human behavioral genetics research on individual differences in psychology during the 1970s has largely faded, despite a recent resurrection of the controversial issues of group differences between races and between social classes (Herrnstein & Murray, 1994). It is now more generally accepted that the genetic contributions to individual differences within racial groups do not necessarily explain between-group differences, which can have their own genetic and environmental contributing factors. During the 1980s and especially the 1990s, psychology became much more accepting of genetic influence on individual differences, as can be seen in the increasing number of behavioral genetic articles in mainstream psychology journals and the number of research grants. One symbol of this change was the 1992 Centennial Conference of the American Psychological Association, in preparation for which a committee selected two themes that best represented the past, present, and future of psychology. One of the two themes chosen was behavioral genetics (Plomin & McCleam, 1993). Indeed, the wave of acceptance of genetic influence throughout psychology threatens to engulf a second, equally important message coming from behavioral genetic research, namely, that individual differences in complex psychological traits are due at least as much to environmental influences as they are to genetic influences. In some areas of psychology, especially psychopathology, the pendulum representing the accepted view may be swinging too far from environmental determinism to genetic determinism.

Nonetheless, in part because of the far-reaching implications of its findings, behavioral genetics still provokes controversy. For example, one area of current controversy involves two findings about the environment called nonshared environment and the nature of nurture. The first finding is that environmental influences tend to make children growing up in the same family different, not similar. Because environmental influences that affect psychological development are not shared by children in the same family, they are called nonshared environment (Dunn & Plomin, 1990; Plomin & Daniels, 1987; Turkheimer & Waldron, 2000). The second finding is that measures of the environment, especially of the family environment, show genetic influence when embedded in genetically sensitive designs and also show genetic mediation of associations between environmental measures and behavioral outcomes (Plomin, 1994; Reiss, Neiderhiser, Hetherington, & Plomin, 2000). Taking these arguments to the extreme, a presidential address of the Society for Research in Child Development (Scarr, 1992) and two recent books (Harris, 1998; Rowe, 1994) have concluded that socialization research is fundamentally flawed because it has not considered the role of genetics. These attacks have met with stiff resistance from developmental psychologists (W. A. Collins, Maccoby, Steinberg, Hetherington, & Bornstein, 2000; Maccoby, 2000; Vandell, in press). Both sides agree that progress toward resolving these differences depends on better articulation and measurement of specific environmental influences. The Example of Schizophrenia Until the 1960s, schizophrenia was thought to be environmental in origin, with theories putting the blame on poor parenting to account for the fact that schizophrenia clearly runs in families. Although a minority view representing the nascent discipline of biological psychiatry offered biochemically based etiologies, the idea that schizophrenia could run in families for genetic reasons was not seriously considered by most American psychiatrists and psychologists. Despite limits to their interpretation not discussed here, twin and adoption studies changed this view. Twin studies showed that identical twins are much more similar than nonidentical twins, which suggests genetic influence. If one member of an identical twin pair is schizophrenic, the chances are 45% that the other twin is also schizophrenic. For nonidentical twins, the chances are 17%. Adoption studies showed that the risk of schizophrenia is just as great when children are adopted away from their schizophrenic parents at birth as when children are reared by their schizophrenic parents, which provides dramatic evidence for genetic transmission. There are now intense efforts to identify some of the specific genes responsible for genetic influence on schizophrenia. In the 1960s, when schizophrenia was thought to be caused environmentally, it was important to emphasize the evidence for genetic influence such as the concordance of 45% for identical twins. Now that evidence for the importance of genetic influence throughout psychology has largely been accepted, it is important to make sure that the pendulum stays in the middle, between nature and nurture. We need to emphasize that identical twins are only 45% concordant for schizophrenia, which means that in half of the cases, these pairs of genetically identical clones are discordant for schizophrenia. This discordance cannot be explained genetically

808

PLOMIN AND CRABBE

it must be due to environmental factors. It should be noted that the word environment in genetic research really means the nongenetic environment, which is a much broader definition of environment than is usually encountered in psychology. That is, environment denotes all nonheritable factors, including possible biological events such as prenatal and postnatal illnesses, not just the psychosocial factors that are usually considered in psychology. The point is that genetics can often explain half of the variance of psychological traits, but this means that the other half of the variance is not due to genetic factors. Genetics, Environment, and Heritability For nearly every area of psychology that has been studied, quantitative genetic research has shown genetic as well as environmental influence (Plomin et at, 2001). For example, genetic research has consistently shown genetic influence in many traditional areas of psychological research such as psychopathology, personality, cognitive disabilities and abilities, and substance use and abuse. Some areas showing strong genetic influence are more surprising, such as school achievement, self-esteem, interests, and attitudes. Nonetheless, it is important to remember that these partitions of the genetic and environmental sources of individual differences pertain to the specific populations and social/environmental conditions in which they are ascertained. Any changes to the underlying assumptions can lead to different conclusions about the relative importance of genetic and environmental influences, but few now seriously question that both influences are important. Most importantly, genetic research in psychology is moving beyond heritability. The questions whether and how much genetic factors affect psychological dimensions and disorders represent important first steps in understanding the origins of individual differences, but these questions about heritability are only first steps. The next steps involve the question how, that is, the mechanisms by which genes have their effects. Examples of these directions for genetic research in psychology include genetic change as well as continuity during development, genetic links between dimensions and disorders, multivariate genetic analysis, and the interplay between genetics and environment (Plomin & Rutter, 1998). Such quantitative genetic research will become increasingly important as it guides molecular genetic research toward the most heritable components and constellations of disorders and dimensions as they interact and correlate with the environment throughout development. Conversely, as discussed later, finding specific genes associated with behavior will greatly enhance psychologists' ability to address quantitative genetic issues such as those just mentioned. Identifying DNA Associated With Behavior Now that the contribution of genetic factors to individual differences in psychology is widely accepted, molecular genetic techniques that can identify some of the genes responsible for this genetic variation are becoming available. The heritability of complex traits is likely to be due to multiple genes of varying but small effect size rather than to one gene or a few genes with major effect. Genes in such multiple-gene systems are inherited in the same way as any other gene, but they have been given a different name

quantitative trait loci (QTLs)to highlight some important distinctions. Unlike single-gene effects that are necessary and sufficient for the development of a disorder, QTLs contribute interchangeably and additively, analogous to probabilistic risk factors. If there are multiple genes that affect a trait, it is likely that the trait is distributed quantitatively as a dimension rather than qualitatively as a disorder, this was the essence of R. A. Fisher's classic 1918 paper on quantitative genetics (R. A. Fisher, 1918). Of course, it is also possible that categorical disorders are etiologically distinct from dimensions, and molecular genetic information could help resolve these alternatives. For example, different genes could be involved in abnormal disorders versus extremes of normal behavior that appear to be similar; alternatively, such behaviors could be due to particular combinations of genes (Lykken, 1982). Such interactions among genes, called epistasis, are discussed later. From a QTL perspective, most disorders are just the extremes of quantitative traits caused by the same genetic and environmental factors responsible for variation throughout the dimension. In other words, the QTL perspective predicts that genes found to be associated with complex disorders will also be associated with normal variation on the same dimension and vice versa (Deater-Deckard, Reiss, Hetherington, & Plomin, 1997; Plomin, Owen, & McGuffin, 1994). Although the QTL perspective has some specific implications for design and analysis of molecular genetic studies, its general importance is conceptual. At the most general conceptual level, a common mistake is to think that humans are all basically the same genetically except for a few rogue mutations that lead to disorders. In contrast, the QTL perspective suggests that genetic variation is normal. Many genes affect most complex traits, and, together with environmental variation, these QTLs are responsible for normal variation as well as for the abnormal extremes of these quantitative traits. This QTL perspective has some implications for thinking about mental illness because it blurs the etiological boundaries between the normal and the abnormal. That is, human beings all have many alleles that contribute to mental illness, but some are unluckier in the hand that they draw at conception from their parents' genetic decks of cards. A more subtle conceptual advantage of a QTL perspective is that it frees psychologists to think about both ends of the normal distributionthe positive end as well as the problem end, abilities as well as disabilities, and resilience as well as vulnerability. It has been proposed that psychologists move away from an exclusive focus on pathology toward considering positive traits that improve the quality of life and perhaps prevent pathology (Seligman & Csikszentmihalyi, 2000). The QTL perspective is the molecular genetic version of the quantitative genetic perspective that assumes that genetic variance on complex traits is due to many genes of varying effect size. The goal is not to find the one gene for a particular trait but rather some of the many genes that make contributions of varying effect sizes to the variance of the trait. Perhaps 1 gene will be found that accounts for 5% of the variance, 5 other genes that might each account for 2% of the variance, and 10 other genes that might each account for 1% of the variance. If the effects of these QTLs are independent, these QTLs would together account for 25% of the total variance. If the heritability of the trait is 50%, they would thus account for half of the heritable variance. If, as is likely, they are to some degree interactive, they would in sum control less of the

SPECIAL ISSUE: DNA

809

heritable variance. All of the genes that contribute to the heritability of the trait are unlikely to be identified because some of their effects may be too small or complicated to detect. All attempts to identify genes rely on genetic markers that are stretches of DNA that differ among individuals. These different forms of DNA at a place (called a locus) on a chromosome are alleles. For example, A, B, and O are three alleles at a locus on chromosome 9 that code for the different forms of the blood products called the ABO blood system. Classical genetic markers like the ABO blood system were limited to a few dozen markers for single-gene traits, often measured in the blood. Systematic gene mapping was made possible beginning in 1980 when genetic markers were developed that involve DNA itself rather than the products of genes. Individual differences in DNA, usually a single base-pair difference, occur on average about one in every thousand nucleotide base pairs of DNA, which means that there are about 3.5 million potential DNA markers throughout the 3.5 billion nucleotide base pairs in the human genome. There are also potentially more than 50,000 DNA markers that involve short sequences of DNA that repeat for unknown reasons. The number of repeats varies greatly across individuals and is inherited. Such short-sequence repeat markers have been most often used in gene hunting to date. However, interest is now turning to developing hundreds of thousands of another type of DNA marker called single nucleotide polymorphisms (SNPs, pronounced snips). As their name implies, SNPs involve a difference in a single base pair of DNA. SNPs are especially interesting in that they are more likely to be responsible for functional DNA differences because changes in the coding sequence of DNA usually involve a single base pair of DNA (F. S. Collins, 1999). The following sections describe the methods used to identify genes associated with behavior and provide examples of replicated results using these methods. We begin with nonhuman animal models. Animal models provide powerful means to identify genes because the genotype can be manipulated through breeding and because genes themselves can also be manipulated. Moreover, animal models arc more than models when it comes to DNA: Nearly any gene found in mice or even fruitflies can also be found in the human species although some of the gene's DNA sequences differ across species and the gene may have different effects across species. Thus, gene-behavior associations found in nonhuman animals are reasonable candidates for the human species. Even identifying a chromosomal region (locus) rather than a specific gene can be valuable because large chunks of chromosomes in mice and man have the same genes in the same order (called syntenic conservation or synteny). In addition, studying animals enables researchers to use much more powerful methods for exploring underlying neurobiological mechanisms. Finally, interactions between genes and environment can be studied more powerfully using animal models because the environment can be controlled and manipulated. Identifying Genes for Nonhuman Behavior Through Induced and Naturally Occurring Mutations Research in multiple species will become much more interactive and mutually informative because of the great similarity of human and rodent genomes. Mice and humans shared a common ancestor as recently as 60 million years ago, and large chunks of genomes

of the two species have been conserved with nearly identical linear organization of bases and, therefore, genes (Battey, Jordan, Cox, & Dove, 1999; Silver, 1995). Thus, finding a gene region in mouse leads the investigator more than 80% of the time to a precise location in the human genome. Long before DNA markers became available in the 1980s, associations were found between genes and behavior in animals. The first example, discovered in 1915, is a single gene that alters eye color in the fruit fly Drosophila and also affects its mating behavior (Sturtevant, 1915). Another well-known example involves the single gene that causes albinism and also affects openfield activity in mice. The Jackson Laboratory maintains hundreds of mutations that have occurred over the years, many as frozen embryos that can be reconstituted on order. These mutants include genes affecting a wide variety of behaviors including neurological functions such as gait, balance, and seizures. Chemical- and Radiation-Induced Mutants In addition to studying naturally occurring genetic variation, geneticists have used X-irradiation or chemicals such as ethylnitrosourea to create mutations that enable them to dissect the behavioral effects of genes. During the past 50 years, hundreds of behavioral mutants have been created in organisms as diverse as worms, fruit flies, mice, and single-celled organisms such as bacteria and paramecia (Plomin et al.. 2001). This research illustrates the principle that most normal behavior is influenced by many genes. Although any one of many single-gene mutations can seriously disrupt a given behavior, normal development is orchestrated by many genes working together. For example, bacteria move toward and away from many kinds of chemicals by rotating their propeller-like flagella. Since the first behavioral mutant in bacteria was isolated in 1966, the dozens of mutants that have been created indicate that many genes are involved in rotating the flagella and controlling the duration of the rotation. Hundreds of induced behavioral mutants have been isolated in paramecia, including at least 20 involved in backing up and swimming forward in a new direction to avoid certain chemicals and heat. The nematode (roundworm) Caenorhabditis elegans has provided examples of several mutants with great longevity (Duhon, Murakami, & Johnson, 1996) and will be an increasingly important animal model for genetic analysis of behavior because the development of each of its 959 cells, the wiring diagram of its 302 neurons, and its entire 100 million base pairs of DNA are known. Hundreds of behavioral mutants, for example, for courtship and memory, have also been identified for the fruit fly Drosophila, one of the most studied organisms in genetic research (Weiner, 1999). Perhaps the best known project of this sort in neuroscience is the screening of the offspring of chemically mutated mice for circadian rhythms in activity. A single mutant mouse with a long circadian rhythm was identified in a screen and used in a systematic program to locate the clock gene and prove its responsibility for the trait. Mice with a mutant form of this single gene had abnormal patterns of daily activity and rest (Antoch et al., 1997; King et al., 1997). Further studies have shown that the suprachiasmatic nucleus (SCN) of the anterior hypothalamus acts like a pacemaker to control circadian rhythms. SCN neurons isolated from clock mutant mice have arrhythmic firing patterns, suggesting that some mechanisms within this nucleus synchronize neuro-

810

PLOMIN AND CRABBE other neurohormones, these knockouts were pituitary dwarves as well (Bosse et al., 1997). As a result, the knockouts exhibited a fascinating complex of compensations throughout the dopaminergic cascade, ranging from increased synthesis of dopamine to reduced levels of the enzyme tyrosine hydroxylase, which transforms the amino acid tyrosine to DOPA, which is then transformed to dopamine (S. R. Jones et al., 1998). These comparisons lead to improved, but not completely normal, functioning of the dopaminergic systems compromised by the knockout. However, in most instances, compensations for the loss of gene function are invisible to the experimenter, and caution must be taken to avoid attributing other changes in the animals to the gene itself. This problem will not be overcome until the second generation of knockout technology enters wide usage. Rather than completely disrupting the transcription of the gene from conception, it is possible to alter parts of a gene that increase or decrease rates of the gene's transcription, to add regulators that act as a switch turning the gene on or off, and to change the expression of the gene in a specific area of the brain. Several approaches to accomplishing this goal have been reviewed (Crusio, 1999). When the technology is fully developed, conditional knockouts will bear DNA constructs that include the gene of interest as well as sequences that allow the experimenter to turn expression of the gene on or off at will at any time during the animal's life span. An example of a conditional knockout involves a receptor for W-methyl-D-aspartate (NMDA). This receptor affects neurotransmission by way of the excitatory neurotransmitter glutamate that plays an important role in memory and in long-term potentiation (LTP), a common cellular electrophysiological model of learning. LTP involves structural and functional changes in the synapse. The NMDA receptor serves as a switch for memory formation by detecting coincident firing of different neurons and affects the intracellular second messenger signaling protein cyclic adenosine monophosphate among other systems. Overexpressing one particular NMDA receptor gene (NMDA receptor 2B) enhanced memory in various tasks as well as LTP (Tang et al., 1999). A conditional knockout of the NMDARZB gene was used to limit the mutation to a particular area of the brain, in this case, the forebrain. Normally, expression of this gene slows during adulthood, which may contribute to decreased memory in adults. In this research, the gene was altered so that it continued to be expressed in adulthood, and this resulted in enhanced memory in several tasks. This report generated much media interest, as the finding was interpreted by some as showing that a so-called intelligence gene had been found, with obvious social implications. However, an interesting feature of this research is that there was little evidence that learning was enhanced: Rather, the conditional knockouts retained their memories longer, so the relevance to intelligence remains to be seen. In an earlier research effort, mice were created with a conditional knockout of the gene encoding the NMDA 1 receptor (Tsien et al., 1996; Tsien, Huerta, & Tonegawa, 1996). These mice developed to adulthood with normal levels of expression of the gene throughout most of the brain. However, the transgenic mutants were constructed such that beginning during the immediate postnatal period, gene expression was gradually turned off over a period of about a month, but only in certain cells in the hippocampus. These tissue-specific, conditional knockout mice were deficient in LTP generated where the NMDA 1 receptor was turned off but showed normal LTP generated in cells where NMDA 1 recep-

nal firing patterns that eventually regulate the activity of the mouse (Herzog, Takahashi, & Block, 1998). The current focus of this research is on understanding how the clock gene regulates SCN neuronal firing rates to control the complex set of behaviors that follow a circadian pattern and on understanding the roles of other, related genes (Vitaterna et al., 1999). There are major new initiatives in the United States and the United Kingdom directed toward using chemical mutagenesis on a large scale to develop new mutants. These initiatives are targeted at behavioral and neural development and depend on sophisticated and sensitive behavioral assays for detecting which mice bear a mutation. In addition, there is interest on the part of pharmaceutical manufacturers in such efforts, as there is the possibility that novel genes affecting the targeted traits could be discovered, which in turn could lead to novel phannacotherapies.

Targeted Gene Mutation: Transgenics and Knockouts


Mouse is also the species most used for targeted gene mutation (gene targeting), a technique hi which mutations that "knock out" the transcription of a gene entirely or alter its regulation in order to underexpress or overexpress the gene are created (Capecchi, 1994). The mutated gene is then transferred to mouse embryos, and the mice are called transgenics when the mutated gene is from another species. This method is discussed later in relation to behavioral genomics because it is used primarily to understand how a gene of a priori interest affects a behavior rather than to find which genes are associated with a behavior. Nonetheless, gene targeting provides an important, although not conclusive, confirmation of a gene's effect on behavior. Knockouts (also called null mutants} of various genes have been shown to affect several types of learning (Wehner, Bowers, & Paylor, 1996), aggression (Nelson et al., 1995; Saudou et al., 1994), alcohol preference (Crabbe et al., 1996), nicotine effects on pain (Marubio, Arroyo-Jimencx, Cordero-Erausquin, Lina, & Novhre, 1999), and general sensitivity to abused substances (Rocha et al., 1998; Rubinstein et al., 1997). Knockout studies are in progress for hundreds of genes, and many will have multiple effects on behavior (Brandon, Idzerda, & McKnight, 1995). For example, a recent summary lists 22 genes shown to affect learning and memory (Wahlsten, 1999), and a compendium of knockouts affecting various behaviors has recently appeared (Nelson & Young, 1998; see also Silva & Mayford, 1998).

Beyond Knockouts: Conditional Gene Expression


Gene-targeting strategies are not without their limitations (Gerlai, 1996). The most obvious problem with knockout mice is that the gene is inactivated throughout the animal's life span. Particularly during development, the organism copes with the loss of the gene's function by compensating wherever possible. In some instances, this can be very informative. For example, deletion of a gene coding for a dopamine transporter protein (which is responsible for inactivating dopaminergic neurons by transferring the neurotransmitter back into the presynaptic terminal) resulted in mice with extraordinarily high activity in a novel environment (Giros, Jaber, Jones, Wightman, & Caron, 1996). The animals failed to habituate even after several hours. Probably because dopamine is an important regulator of secretion of prolactin and

SPECIAL ISSUE: DNA

811

tors were expressed at normal levels. They also were deficient in spatial learning, which is known to depend on the hippocampus; electrophysiological activity of hippocampal cells specialized for spatial learning (place cells) was disrupted as well. These studies represent first forays into the second generation of targeted gene deletion studies. The ability to turn genes on or off at will in specific brain areas will go a long way toward overcoming the current limitations of these model systems, as the compensations occurring throughout development can be largely avoided. However, to attain full power, conditional knockout technology still must overcome other major hurdles. For example, better control is needed over the location of transgene insertion in the genome and the precise timing of the suppression or augmentation of gene function. Because so little is known about why the precise location of inserted genes is important to gene function, the current technologies for producing conditional mutants amount to making several such mutants and laboriously characterizing each of them to see where in the brain and to what degree the introduced gene construct is expressed. In many cases, multiple copies of the transgene are incorporated into the DNA. but it is often unclear whether more copies equal higher levels of gene expression (i.e., more gene product produced). Thus, the experimenter finally chooses the best available mutant strain from the array of choices. For example, in the study described above (Tsien, Chen, et al., 1996; Tsien, Huerta, & Tonegawa, 1996), 11 mutant lines were tested to select the mutant that expressed the conditional mutation only in the desired hippocampal region.

Identifying Genes for Naturally Occurring Genetic Variation in Nonhuman Behavior The mouse has also been the key organism studied in research on naturally occurring genetic variation (Silver, 1995). As indicated earlier, complex quantitative traits, whether biological or behavioral, are likely to be influenced by QTLs. Because their breeding can be controlled and because their small size makes them an affordable mammalian species, mouse models have been valuable in detecting QTLs for complex traits. The chromosomal location of a QTL affecting a trait is identified through a method called linkage mapping in which the coinheritance of the trait and a particular DNA marker is traced. Linkage is essentially a violation of Mendel's second law of heredity called independent assortment, which states that the inheritance of one gene is not affected by the inheritance of another gene. Genes do not assort independently if they happen to be close together on the same chromosome, which is called linkage. Within families, recombination is likely to separate a particular marker allele from an allele affecting a trait if the two loci are far apart on the chromosome. Recombination occurs during the formation of eggs and sperm (meiosis) when chromosomes cross over and exchange parts of maternal and paternal chromosomes (see Figure 1). There is one cross-over per chromosome per meiosis on average. The extent of recombination can be used to estimate the distance between the marker and the QTL for the trait.

Inbred and Recombinant Inbred Strains One widely used strategy involves crosses between inbred strains. Inbred strains of mice are created by mating brothers and sisters for at least 20 generations, which makes animals within the inbred strain genetically identical (and homozygous at each gene, i.e., with two identical copies of a single allele at each locus). When strains are crossed, the first filial (Fj) generation is heterozygous at all loci that differ in the parental strains (see Figure 2). Crossing F, mice produces an F2 generation in which heterozygous alleles segregate so that each individual is genetically unique. QTLs can be identified by correlating specific alleles of DNA markers with quantitative scores of F2 individuals. In an F, population, each allele is known to be derived from one or the other progenitor strain and has a 50% frequency. If a particular marker allele is more frequent in high- or low-scoring mice, it can be inferred that a nearby QTL, linked to the marker, affects the trait. In outbred populations, which include wild mice and F2 crosses of inbred strains, as well as the human species, recombination separates alleles for loci on the same chromosome, and more generations of outbreeding increase the number of recombinations. A conceptually similar approach, called recombinant inbred (RI) strains, reinbreeds from a single F2 population (see Figure 2). Unlike an F2 population in which each individual is genetically unique and not replicable, RI strains provide replicable genotypes that are fixed homozygously at each gene for a single allele inherited from one or the other progenitor inbred strain. On each chromosome, each RI strain displays a mosaic of chromosomal segments derived from one or the other of the parental inbred strains. This means that once RI strains are genotyped for DNA markers, they never need to be genotyped again because the genotypes of the inbred strains remain unchanged. Thus, RI strains

Antisense Oligodeoxynucleotides Another method uses antisense DNA to "knock down" gene function. Antisense DNA is a DNA sequence typically 18-25 base pairs long that is complementary to a specific mRNA sequence. By binding with mRNA, antisense DNA prevents the mRNA from being translated into protein. Injected in the brain, antisense DNA has the advantage of high temporal and spatial resolution (Ogawa & Pfaff, 1996). One behavioral study used antisense DNA against the CREB gene and confirmed the involvement of this gene in memory formation (Guzowski & McGaugh, 1997). Antisense DNA is being widely used in psychopharmacology, for example, to block drug effects by preventing the synthesis of receptor molecules in specific brain regions (Pasternak, 2000). Antisense DNA knockdowns have been shown to affect behavioral responses for dozens of drugs (Buck, Crabbe, & Belknap, 2000). Antisense DNA has often proven effective in cases where there is a dearth of agonists or antagonists with high selectivity. The principal limitations of antisense technology currently are its unpredictable efficacy and a tendency to produce general toxicity. Targeted mutations and antisense DNA reflect the complexity of brain systems for learning and memory. For example, none of the genes and signaling molecules in flies and mice found to be involved in learning and memory are specific to learning processes. They are involved in many basic cell functions, which raises the question whether they exert their effects on memory by modulating the cellular background in which memories are encoded (Mayford & Kandel, 1999).

812

PLOMIN AND CRABBE

nn n
a b c
A B A B C C

PPPP
a b c a b
A B
C

A B C

BXD-101

BXD-102

BXD-103

Figure I. Recombination. A pair of a single chromosome is depicted for an F, hybrid animal during five stages of meiosis. The chromosome homolog inherited maternally is depicted as unshaded, and the locations of three genes are shown, as well as the specific alleles for each gene (a, b, and c). The chromosome homolog derived paternally is shaded and shows three alternate alleles (A, B, and C). In meiosis, the precursor cells of the sperm or ova must multiply (i.e., double, as shown in the second stage) and can then physically exchange material at a point of random contact. This is termed a crossing-over and occurs in the third depicted stage, between Genes B and C. The result is a recombination, depicted in the fourth stage. When the final stage of gamete production occurs, the chromosome number is reduced to the normal diploid number, and individual recombinant gametes, as well as nonrecombinant gametes, give rise to germ cells. From Behavioral Genetics (p. 16), by R. Plomin, J. C. DeFries, G. E. McClearn, and M. Rutter, 1997, New York: Freeman. Copyright 1997 by W. H. Freeman. Adapted with permission.

can be used for QTL analysis of behavioral traits without any additional genotyping (McClearn, Plomin, Gora-Maslak, & Crabbe, 1991). For example, the most frequently used set of RI strains, the BXD RI strains, were derived from a cross between C57BL/6J (B6) and DBA/2J (D2) progenitor inbred strains that differ markedly in many behavioral traits, especially in their responses to many drugs (Crabbe & Harris, 1991). The BXD RT

Figure 2, Construction of a set of recombinant inbred (RI) strains from C57BL/6 (B6) and DBA/2 (D2) progenitors. A single pair of chromosome homologs (e.g., chromosome 2) is followed through the process of RI strain development. Genomic intervals derived from B6 are shaded and intervals from D2 are indicated without shading. The map positions of five fictitious loci (15) are indicated in the progenitor generation. In the F, generation, all animals are genetically identical and are heterozygotes (as shown) at each locus. In the F2 generation, segregation and independent assortment occur because of crossing over. The genotypes at the five loci for six individual mice are shown as shaded signifying the B6-derived allele or unshaded signifying the D2-derived allele. The six mice are shown as three breeding pairs, one for each of the RI strains to be developed. The two circled regions represent segments of the genome that are already fixed for one progenitor type within both members of the breeding pair. At the F20 generation (i.e., after 20 generations of brother-sister mating, or inbreeding), all genetic variability has been eliminated within each RI strain. Four individuals are shown for each strain (fictitiously called BXD-101, BXD102, and BXD-103). Each new RI strain displays a mosaic of chromosomal segments derived from one or the other of the parental inbred strains, shuffled through 20 generations of recombination, and fixed through inbreeding. Although the patterns of chromosomal material shown for the six F2 mice depict crossing over only at the borders between the five genes, it can occur anywhere. Additional crossovers occur with each generation of inbreeding, and the result (seen in the F2C fictitious BXD RI strains) is smaller regions of chromosomal material preserved for all subsequent generations. From Mouse Genetics: Concepts and Applications (p. 209), by L. M. Silver, 1995, Oxford, England: Oxford University Press. Copyright 1995 by the Oxford University Press. Reprinted with permission.

SPECIAL ISSUE: DNA

813

strains have been genotyped systematically for more than 1,500 markers throughout the mouse genome, with each marker positioned precisely within a chromosomal region. Most QTL work in mice using these methods has been in the area of pharmacogenetics, which denotes genetic effects on responses to drugs. At least 24 QTLs have been definitively mapped for drug responses such as alcohol drinking, alcohol-induced loss of righting reflex, acute alcohol and pentobarbital withdrawal, cocaine seizures, and morphine preference and analgesia (Crabbe, Phillips, Buck, Cunningham, & Belknap, 1999). This represents considerable progress from the first effort to summarize QTLs for drug responses 5 years earlier (Crabbe, Belknap, & Buck, 1994). A specific example is a QTL affecting alcohol withdrawal severity, first provisionally mapped in BXD RI strains (McClearn et al., 1991). Follow-up studies used the RI strains, F2 mice, and selectively bred lines to verify rigorously that there was indeed a QTL in this region (Buck, Metten, Belknap, & Crabbe, 1997). Another example is emotionality as assessed by a battery of measures such as activity in a brightly lit open field and exploration in a Y maze (Flint et al., 1995). Using the most and least emotional F2 mice, researchers found three highly significant QTL regions. One of these QTLs was specific to the open field, whereas the other two QTLs were related to all of the measures of emotionality. Beyond QTLs: Candidate Genes, High-Resolution Mapping, and Positional Cloning In some instances, the location of a mapped QTL is close enough to a previously mapped gene of known function (candidate gene) to make studies of that gene informative. For example, several groups have mapped QTLs for alcohol preference drinking in mice to the middle of mouse chromosome 9 (T. J. Phillips, Belknap, Buck, & Cunningham, 1998), a region that includes the gene coding for the dopamine D2 receptor subtype. Studies with D2 receptor knockout mice revealed that they showed reduced alcohol preference drinking (T. J. Phillips, Brown, et al., 1998). Although this does not prove that the D2 gene is the basis for the QTL association, it fails to disprove the hypothesis, encouraging further investigation. As more QTLs for complex behaviors are localized in mice, there are enticing signs that certain regions of the mouse genome seem to represent hot spots where multiple QTLs are found. This clustering was noted in one of the first explorations of this technique (McClearn et al., 1991). Figure 3 is a cartoon showing the location of QTLs for sensitivity to several behavioral effects of alcohol and the sedative drug pentobarbital (Nembutal). For example, the close association of QTLs for acute alcohol withdrawal, acute pentobarbital withdrawal, and chronic alcohol withdrawal near the end of mouse chromosome 1 suggests (but cannot prove) that a single gene in this region may be influencing all three responses. If this is true, the gene on mouse chromosome 1 is virtually certain to map to the q21-q23 region of human chromosome 1 because linked groups of genes homologous to mice and man have been found for this chromosomal region. Figure 3 does not show the confidence intervals surrounding the map locations of the QTLs. The QTL locations depicted are "neighborhoods," and the approximate 95% confidence interval surrounding each depicted location is about 20 million base pairs of DNA, about one fifth the size of the chromosomea region that

contains more than 1,000 genes. To identify the gene for each QTL would require positional cloning, that is, laborious sequencing of virtually every base pair in the DNA region surrounding the QTL. Given the current size of QTL confidence intervals, this region needs to be reduced in size by at least an order of magnitude, and scientists engaged in QTL mapping are currently using classical genetic techniques to achieve this higher resolution mapping so that a "street address" can be achieved for each underlying gene (Darvasi, 1998). One method for achieving higher resolution mapping of QTLs is to look at advanced intercross lines of mice. To pursue the earlier examples of inbred strain crosses, this means crossing F2 mice to obtain F3s, crossing F3s to obtain F4s, and so on. After many generations of intercrossing (i.e., mating unrelated individuals, not mating siblings as in the inbreeding strategy for creating RI strains), only marker alleles that are essentially within the QTL itself or very close to it remain closely linked to the gene affecting the trait. Such markers are said to be in linkage disequilibrium with the QTL affecting the trait (see Figure 4). Because an F2 population has only one generation for recombination, markers show association (linkage disequilibrium) with a QTL even if they are much further away on the chromosome. As a result, about 150 DNA markers are sufficient to scan the entire genome for QTL associations in F2s as compared with thousands of DNA markers needed for highly outbred populations. For example, a follow-up study used outbred mice (a genetically heterogeneous stock derived from a cross of eight inbred strains) to increase the resolution of the Flint et al. (1995) QTL analysis for emotionality QTLs (Talbot et al., 1999). Examination of linkage disequilibrium in this population enabled the investigators to reduce the size of the QTLs on chromosome 1 and 12 to a region containing fewer than 100 genes, although the third QTL, on chromosome 15, was not replicated. Although the rat is widely used in behavioral as well as physiological and pharmacological studies, quantitative and molecular genetic research on the rat has lagged far behind the mouse. However, the rat is beginning to catch up. For example, a map of more than 5,000 DNA markers is now available (Watanabe et al., 1999). Convergence and Divergence of Genetic Influences: Multigenic Complexity, Epistasis, and Pleiotropism Behavior is nothing if not complex, and complex traits are the final frontier for genetic analyses. To the extent that genetic influences are important in addition to environmental factors, behaviors are typically multigenic. Complex traits are multigenic in the sense that the nature or degree of a behavior expressed in an individual is not influenced by a single gene but rather by multiple genes. Thus, each gene contributes only a relatively small amount of influence to that trait. As discussed earlier, the small effects of individual genes make them difficult to detect and map genetically, but in the aggregate, knowledge about the multiple genes with converging influence on a behavior can provide a great deal of predictive power. A clear example of multigenicity can be drawn from the animal genetics literature. One classical approach to animal behavioral genetics is selective breeding, where individuals with extreme scores on a behavior are mated. After many generations, nearly all

814
Distance from Centromere

PLOM1N AND CRABBE

Mouse Gene

Homologous Human Gene

Human C'hr.

50 cM -

Acrd, Acrg Ethanol-induced loss of righting reflex

CHRND CHRNG

60 cM Htr5b Atp2b4 HTR5B

2q36-q37 2q21.1-q21.3 2q37 5ql4 18q21-22 2qll-ql3 Iq32

70 cM
Alcohol preference Cchral

Iq31
Iq25 Iq23

80 cM -

Alcohol conditioned taste aversion

Atplbl

90 cM -

Acute alcohol withdrawal ATP1A2 Acute pentobarbital withdrawal

Iq21-q23

lOOcM -

Chronic alcohol withdrawal

Iq41

Figure 3. Quantitative trait loci (QTLs) for sensitivity to several behavioral effects of alcohol and the sedative drug pentobarbital on distal mouse chromosome 1. Each QTL is represented by a circle on the chromosome at the point of highest association. Distances are in centimorgans (cM) from the centromere, a linkage estimate of relative distance. Plausible candidate genes mapped near these QTLs are listed. Known regions of syntenic conservation between the mouse and human chromosome are indicated by the human chromosome numbers and regions. In these syntenic regions, human genes homologous to the mouse genes are noted. From "Alcohol and Genetics: New Animal Models," by K. E. Browraan & J. C. Crabbe, 1999, Molecular Medicine Today, 5, p. 316. Copyright 1999 by Elsevier. Reprinted with permission.

genes favoring the selected trait will have been collected in the selected line. The classic selection studies in psychology were conducted by Tolman (1924) and Tryon (1940), who selected for maze-bright and maze-dull rats. In an attempt to identify the genes affecting a simple behavior in mice, sensitivity to the sedative effects (loss of righting reflex) of a hypnotic dose of alcohol, researchers used selectively bred lines of mice, recombinant inbred strains, and a number of other specialized genetic methods to map five QTLs (Markel, Bennet, Beeson, Gordon, & Johnson, 1997). The progenitor selected lines of mice used in this project were bred for alcohol-induced sedation. Following a standard dose of alcohol, long sleep (LS) and short sleep (SS) mice were sedated for approximately 180 min or 10 min, respectively. The basis for this 170-min differential sensitivity is almost entirely genetic. Each QTL conferred a difference in sleep time of between 19 and 25 min. That is, an average individual mouse possessing the LS allele at one of these loci would be sedated for only about 20 min longer than an individual with the SS allele. However, if an individual

possessed all five LS alleles, its genotype could predict 130 min of the total of 170 min in response difference between the LS and SS mice. Although methods currently available are not sensitive enough to test the hypothesis directly, it is highly likely that the remaining 40 min of unexplained genetic difference between LS and SS mice can be traced to two places. First, it is very likely that more than five genes contribute to this response to alcohol, and some of those genes probably have effect sizes too small to detect unless hundreds or even thousands of mice are tested. The more important source of the unexplained genetic variance, however, is likely to be in gene-gene interactions, a phenomenon termed epistasis. That is, gene interactions may not be additive: Possession of alleles of two QTLs, each of which would contribute 20 min of sedation independently, may confer 50 min (or 30 min) of sedation when an individual has both alleles. Another example can be drawn from the study of a QTL on mouse chromosome 11 for acute pentobarbital withdrawal severity (Buck, Metten, Belknap, & Crabbe,

SPECIAL ISSUE: DNA

815

of chromosome 1 (indicated in Figure 3). Mice with the C57BI76J (vs. a DBA/2J) allele at the chromosome 1 QTL show no evidence of a difference in withdrawal when they have the C57BL/6J markers for the chromosome 11 QTL (Hood, Crabbe, Belknap, & Buck, 2000). This sort of statistical interaction is familiar to psychologists accustomed to ANOVA models, and there are many examples of physiological interactions, some of which have been traced to genetic epistasis. Detection of genetic interactions of this sort is difficult, and for QTLs in particular, behavioral geneticists need to be aware that the models in current use still have limited statistical power. Nonetheless, the complex interactions surrounding gene effects are a very important area for future research. Psychologists' fearlessness in the face of statistics (and interactions in particular) is an asset that will facilitate their contributions to future behavioral genomic analyses. In the same way that any behavior can be traced to multiple genes, any single gene can be shown to influence multiple behaviors, a radiation of influence termed pleiotrapism by geneticists. One of the growth industries for psychologists who study rodent behavior is the characterization of pleiotropic effects of gene knockouts and other transgenic manipulations in mice. The training that psychologists receive in experimental design, statistical analysis, and attention to interpreting the subtleties of behavior are all valuable assets for experiments with single-gene animal models. In the future, it will be useful to add to that arsenal an understanding of linkage, recombination, epistasis, and pleiotropism as well.

FQ:

Gene-Environment Interplay
In addition to questions about how genes interact with each other, an even larger set of questions involves the interplay between genes and environment as seen in gene-environment correlations and interactions. Gene-environment correlation refers to genetic differences in exposure to environments, literally, a correlation between genes and environment. That is, genetic dispositions may be correlated with experiential dispositions. For example, children with the chromosome 6 gene who find it hard to learn to read might avoid reading or, more fundamentally, might not be interested in reading or being read to. The topic of geneenvironment correlation goes beyond describing and explaining a correlation between genotype and environment. The larger issue involves the environmental mechanisms by which gene-behavior associations develop (Rutter et al., 1997). Gene-environment interaction refers to genetic differences in sensitivity to experiences. A general form of interaction assumed in psychopathology is called the diathesis-stress model: Individuals who are at genetic risk or predisposition are most sensitive to environmental risk (Paris, 1999). For example, in an adoption study of criminal behavior, being reared in an adoptive home in which a parent had been convicted of a crime did not increase the likelihood that the adoptee would have a criminal record unless the adoptee had a biological parent who had been convicted of a crime (Medniek, Gabrielli, & Hutchings, 1987). A similar type of interaction was found in which negative adoptive home environments had especially deleterious effects on antisocial behavior in adolescent adoptees whose biological parents were antisocial (Cadoret, Yates, Troughton, Woodworth, & Stewart, 1995).

Figure 4.

Linkage disequilibrium. F^: Both chromosome homologs are

shown for an Fj hybrid individual derived from crossing a B6 and a D2 inbred parent. Alleles for nine genes are shown. The shaded homolog represents the aileles inherited from the B6 parent. The unshaded homolog represents the aileles inherited from the D2 parent. A mutation causing a behavioral deficit is shown in gene 6, inherited from D2. F2: The haploid genotypes (i.e., only one of the two chromosomes of the homologous pair) are shown for seven individual F2 mice. Five of the seven show evidence of a recombination due to crossing over during meiosis. Genes 2, 3, 4, 5, and 7 all show evidence of linkage disequilibrium between D2 aileles and the mutant allele. The first individual excludes linkage with gene 9, the second excludes linkage with gene 1, and the fifth excludes linkage with gene 8. F9: Haploid genotypes for nine individuals after nine generations of intercrossing. After many more recombinations in F3-F9 (not indicated), only genes 5 and 7 remain in linkage disequilibrium with the mutant gene. This can be seen by the fact that any individual with the mutant allele also has D2 aileles for genes 5 and 7.

1999). Mice possessing markers from the C57BL/6J strain have significantly higher withdrawal scores than those possessing markers from the DBA/23 strain. However, this is only true if they have also inherited the DBA/2} allele for a second QTL on the distal end

816

PLOMIN AND CRABBE profit from the knowledge gained by genotyping children for specific genes associated with relevant behavioral traits. This added knowledge about the individual will represent a better articulated phenotype, whose responsiveness to efforts to counteract the effects of risk-promoting genes (and likelihood of benefiting from the effects of protective genes) can better be gauged. We also predict that psychologists in the 21st century will examine the interactions and correlations between family environment and genes more closely. The costs and benefits of the knowledge added by DNA are already the subject of much ethical debate, to which we return in the final section.

A different type of genotype-environment interaction emerged from a study of Scandinavian men exploring two variants of alcoholism, each of which was heritable. Rearing environments were classified as being risk-promoting or protective. One form of alcoholism (Type I, characterized by relatively mild abuse, minimal criminality, and passive-dependent personality variables) showed pronounced evidence of dependence on the rearing environment: Only those with both genetic and environmental risk factors showed higher rates of Type I alcoholism. For the other form (Type II, characterized by early onset, violence, criminality, and being largely limited to males), the rearing environment had little effect on genetic risk (Cloninger, Bohman, & Sigvardsson, 1981). This example probably reflects genetic heterogeneity of the trait alcoholism; that is, the underlying predisposing genes are probably different for the two types. Nonetheless, it clearly illustrates the concept that the same environmental risk factors can play a different role depending on an individual's genotype. Yet another type of genotype-environment interaction has been reported for alcoholism: The heritability of alcohol consumption is greater for unmarried women than for married women (Heath, Jardine, & Martin, 1989). There are many other ways of thinking about genotypeenvironment interaction (Kendler & Eaves, 1986). This is another area in which animal models will be especially important because both genotype and environment can be manipulated experimentally. For example, a recent study tested several inbred strains and one null mutant simultaneously in three laboratories on a battery of six behaviors (Crabbe, Wahlsten, & Dudek, 1999). As many variables as possible (e.g., apparatus, test protocols, and many environmental variables) were rigorously equated. The strains differed remarkably in all behaviors, but for some tests, there were significant Strain X Laboratory interactions. In general, the weaker the overall genetic influence on a trait, the more likely there was to be a Genotype X Environment interaction. One implication of this study is that in experiments characterizing null mutants, results may prove subsequently to be idiosyncratic to a particular laboratory. However, one environmental manipulation that is frequently asserted to be an important determinant of mouse behavior (so-called shipping stress) was also varied in this experiment. Half the animals were bred in the testing location, and half were snipped as adults from commercial breeders. Shipping status interacted with genetic differences only for escape latency in a water maze but had no influence on anxiety, activity, or alcohol preference drinking. For human behavior, replicable Genotype X Environment interactions are especially difficult to find (Wachs & Plomin, 1991). In part, this is because of a lack of the degree of experimental control that is possible in animals, such as making rather extreme environmental manipulations, and in part, simply because detection of interactions in analysis of variance designs with reasonable statistical power requires far more subjects than detection of main effects (Wahlsten, 1990). Modeling genotypeenvironment interactions in animals where specific genes can be studied might offer insight for the more difficult studies with humans. We predict that tracing the developmental pathways between specific genes and behavior through correlations and interactions with environmental mechanisms is likely to be one of the most important advances that emerges from applications of specific genes associated with behavior. Moreover, psychologists will

Identifying Genes for Single-Gene Effects in Humans Using Traditional Linkage Designs
Until the past decade, attempts to find genes in the human species were limited to rare disorders caused by a single gene necessary and sufficient to cause the disorder. In a traditional linkage design using a large pedigree of many individuals across several generations, linkage can easily be detected for a marker that is 10 million base pairs (about one tenth the average length of a chromosome) from the gene responsible for the disorder. As a result, 350 markers evenly spaced throughout the chromosomes can systematically scan the genome for linkage. In other words, linkage is farsighted in that it can detect distant mountains (singlegene effects). Once linkage finds the chromosomal neighborhood of a gene, it is more difficult to locate the gene's specific address because few recombinations occur between markers and a gene that live in the same neighborhood.

Triplet Repeat Disorders


In 1983, Huntington's disease was the first single-gene disorder linked to a chromosomal region (the tip of chromosome 4) using DNA markers (Gusella et al., 1983). Huntington's disease is a rare disorder (affecting 1 in 20,000 individuals) caused by a single dominant gene whose effects are first seen in middle adulthood in personality changes, forgetfulness, and involuntary movements. Slowly over 20 years, the disorder leads to a complete loss of motor control and intellectual function. The disease was linked to chromosome 4 using hundreds of individuals in a five-generation pedigree. Locating the chromosomal region through linkage led in 1993 to identification of the specific sequence of DNA responsible for the disorder. In the case of Huntington' s disease and more than a dozen other single-gene disorders, the genetic problem is a repeating sequence of three nucleotide bases of DNA. Normal alleles at this locus contain between 11 and 34 copies of the triplet repeat, but alleles that cause Huntington's disease have more than 40 copies. It is not yet known exactly how this form of the gene causes neurodegeneration or how to intervene to prevent the disease. Identifying the DNA responsible for the disorder has made it possible to diagnose with great accuracy whether an individual with an affected parent (and thus at 50% risk for the disorder) does in fact carry the gene, but there is some possibility that the mutation is not fully penetrant: Not all individuals with triplet repeats develop the disease (Nance, 1997; Rubinzstein et al., 1996). A similar approach has been used to locate the genes responsible for hundreds of other rare single-gene disorders. In 1991, a single

SPECIAL ISSUE: DNA gene on the X chromosome that is the most common cause of mental retardation after Down's syndrome was identified, although its effects are variable, with some so-called fragile X individuals having normal intelligence. The gene contributes to the excess mental retardation in males versus females because males with their single X chromosome always express the gene, whereas females express the gene only if they inherit copies on both of their X chromosomes. Nonetheless, the disorder is still relatively rare in the population, with a frequency of about one in several thousand males and half as many females. The disorder is called fragile X because the X chromosome carrying the fragile X allele lends to break when cells that carry it are grown on a special medium. The fragile X gene is especially interesting because, like Huntington's disease, it involves a triplet repeat. For fragile X, the triple repeat is unstable and gets longer as it passes through several generations until its length becomes a liability (a phenomenon called anticipation). Parents who inherit X chromosomes with a normal repeat number for this triplet repeat (654 repeats) sometimes produce eggs or sperm with an expanded number of repeats (up to 200 repeats), called a premutation. This premutation does not cause retardation in the offspring, but it is unstable and often leads to greater expansions (200 or more repeats) in the next generation, which does cause retardation. Protective Genes and Environmental Modulation Not all mutations lead to dysfunctionin one instance, possession of a mutant allele has been shown to exert a protective effect against development of alcoholism, a prevalent disorder that is clearly heritable. Alcohol is metabolized to the highly toxic compound acetaldehyde by the enzyme alcohol dehydrogenase. Acetaldehyde in turn is rapidly metabolized by the enzyme aldehyde dehydrogenase (ALDH), which has many variant alleles. Populations of Asian ancestry have a high prevalence of the allele ALDH2*2, a variant that is very inefficient at converting acetaldehyde to the innocuous end-products acetate and water. In individuals with two copies of the ALDH2*2 allele, ingestion of alcohol is therefore followed by a facial flushing reaction and feelings of nausea and dysphoria that can be intense. In these populations, possession of ALDH2*2 alleles protects against the development of alcoholism, as well as the related quantitative trait of heavy drinking (Harada, Agarwal, Goedde, Tagaki, & Ishikawa, 1982; Higuchi et al., 1994). Furthermore, in heterozygotes who do drink, the increased consumption leads to increased risk for such adverse biomedical consequences as colon cancer (Murata et al., 1999). The protection is apparently so pronounced that ALDH2*2/*2 homozygotes seem almost never to become alcoholic. In a recent study of 409 alcoholics, none were homozygotes (vs. 31 of 461 controls), and only 12.7% of the alcoholics were ALDH2*l/*2 heterozygotes (vs. 35.1% of controls; Murayama, Matsushita, Muramatsu, & Higuchi, 1998). However, one ALDH2*2/*2 homozygote has been identified in a sample of 420 Han Chinese alcoholics, so the protection is not absolute (Chen et al., 2000). The adverse reaction to acetaldehyde is the basis for an alcoholism therapy using the drug disulfiram (Antabuse), which inhibits ALDH activity and leads to the same nausea and dysphoria when alcohol is consumed. Even if a deleterious gene has been inherited, its effects in some cases may be modulated environmentally. One of the earliest and

817

most well-known single-gene examples of environmental modulation is phenylketonuria (PKU), caused by a gene on chromosome 12. The PKU disease allele produces an enzyme that does not work properly to metabolize phenylalanine, which comes from food, especially red meats. If phenylalanine cannot be broken down, its metabolic products build up and damage the developing brain. Although PKU is diagnosed with a simple biochemical test, it serves as a useful reminder that an environmental interventiona diet low in phenylalaninecan successfully prevent the development of mental retardation for a single-gene recessive disorder that occurs in about 1 in 10,000 births and previously accounted for about 1% of severely retarded individuals in institutions. Identifying Human QTLs Using Linkage Designs Although the traditional linkage design is an effective technique for locating the general chromosomal region of the gene responsible for rare single-gene disorders, it is less applicable to common complex traits influenced by multiple genes as well as by multiple environmental factors. Attempts in the 1980s to use this approach to investigate linkage for schizophrenia and bipolar manicdepressive psychosis led to well-publicized reports of linkage, but these reports were later retracted (Moldin, 1997). The main problem is that if multiple genes are responsible for the heritability of a trait, any particular gene is likely to account for only a small amount of variance, which greatly increases the difficulty of detecting it. In addition, many psychological traits are continuous quantitative dimensions rather than discontinuous qualitative disorders. Although there are thousands of single-gene mutations that act like sledgehammers wreaking havoc on development, it is generally accepted that normal development of complex traits is orchestrated by a large symphony of genes working together. Genetic variation in these genes is thought to be largely responsible for the ubiquitous genetic influence found for psychological disorders and dimensions. These require a QTL approach. Indeed, as mentioned earlier, the QTL perspective suggests the radical view that the genetic contribution to disorders is quantitative rather than qualitative. Newer linkage designs use just a few family members in many families rather than many family members in a few large pedigrees (Burmeister, 1999). For example, the affected sib-pair design (Blackwelder & Elston, 1985; Suarez & Van Eerdewegh, 1984) selects families in which two siblings reach diagnostic criteria for a disorder. If a particular DNA marker is linked to a gene that influences a disorder, affected siblings, who presumably share some of the same genes for the disorder, would be more likely to share the same alleles at each such risk-promoting gene from their parents for that DNA marker. That is, siblings can share zero, one, or two alleles from each of their parents. Given that their mother and father each have two alleles, siblings on average are expected to share one allele for a marker. However, if a DNA marker is linked to a gene for the disorder, affected siblings are more likely to share both alleles for that marker. The affected sib-pair design is more compatible with a QTL perspective than the traditional large-pedigree linkage design because the selected sibs in which both members of a pair are affected can be viewed as being at the extreme of a quantitative dimension. The sib-pair QTL linkage design explicitly assesses a

818

PLOMIN AND CRABBE

quantitative dimension rather than a qualitative disorder (Fulker & Cherny, 1996). The approach correlates degree of allele sharing (i.e., zero, one, or two) for each DNA marker with quantitative trait differences within sibling pairs. Although the method was first proposed for unselected siblings (Haseman & Elston, 1972), the method's power comes from selecting sibling pairs in which at least one sibling is at the high or low extreme of the quantitative trait. Even so, the method is unlikely to detect a gene that accounts for less than 10% of the variance of the trait. This seems like a small effect to those who are used to thinking about single-gene disorders, but it seems like a large effect to those who think that complex and common disorders and especially quantitative traits are influenced by very many genes. Sib-pair QTL linkage was first applied to reading disability and yielded a significant QTL linkage on the short arm of chromosome 6 in two samples of siblings (Cardon et al., 1994). That is, the quantitative reading scores of siblings of reading-disabled individuals were worse when the siblings shared alleles in this region. This linkage has been confirmed in three subsequent studies (S. E. Fisher et al., 1999; Gayan et al., 1999; Grigorenko et al., 1997) and appears to apply to diverse components of reading disability. Because this QTL region was identified using linkage, which has power to detect only relatively large effects, these findings imply that this region on chromosome 6 harbors a gene (or genes) accounting for at least 10% of the average reading difference between reading-disabled individuals and the rest of the population. An implication of the QTL approach is that this QTL is not specific to reading disability but rather affects reading throughout the normal range as well, although this has not yet been demonstrated. The specific gene on chromosome 6 that contributes genetic risk for reading disability has not yet been identified. The QTL approach can also be used in other types of linkage analyses, such as more traditional pedigrees. Recent methodological advances have enabled investigators to use all phenotypic and genetic data in multiple families, regardless of the genetic relationship among individuals screened. These methods rely on variance partitions rather than simply analyzing allele sharing, and they allow greater power to detect QTLs, as well as explicit analyses of Genotype X Environment interaction and epistatic interactions (Almasy & Blangero, 1998). For example, joint consideration of each individual's score on the Novelty-Seeking subscale of the Tridimensional Personality Questionnaire and a diagnosis of alcoholism led to improved ability to detect a QTL on chromosome 4 using this method (Czerwinski, Mahaney, Williams, Almasy, & Blangero. 1999). These and other advances will continue to increase the power of linkage analyses for QTLs. Identifying Human QTLs Using Association An alternative QTL strategy is allelic association, which can detect QTLs of small effect size (Plomin et al., 1994; Risch & Merikangas, 1996). Allelic association refers to a correlation between alleles of a DNA marker and trait scores across unrelated individuals. That is, allelic association occurs when individuals with a particular allele for the marker have higher scores on the trait, which can occur for three reasons. The DNA marker itself may be the genetic factor that directly affects the trait. Second, the DNA marker may be close enough to the QTL on the chromosome to reflect the effect of another stretch of DNA that is actually

responsible for the effect (called linkage disequilibriumsee Figure 4). The third reason is artifactualassociations can appear between traits that differ among population subgroups and alleles whose frequency also differs among those subgroups (called population stratification). We discuss the implications of, and possible solutions for, this problem at the end of this section. One problem with finding QTLs for complex traits is that linkage is systematic but not powerful and allelic association is powerful but not systematic. As mentioned earlier, linkage is farsighted in that it can detect distant mountains (single-gene effects) but cannot see nearby hills (QTLs of small effect size). In contrast, allelic association is nearsighted in the sense that it can see nearby QTLs but not more distant single-gene effects. That is, if a marker is to be detected by allelic association, it must either be the functional QTL or be very near to it, namely, within a hundred thousand base pairs. For this reason, the street address of a gene that has been mapped by linkage to a specific neighborhood of a chromosome is located using allelic association with known genes in the neighborhood. The main advantage of allelic association is that it is powerful because it correlates alleles with traits in unrelated individuals, and power can be increased simply by increasing the sample size. In addition, allelic association can be just as easily applied to quantitative traits as to qualitative disorders. Using the same logic that drives animal QTL researchers to test candidate genes, human gene-hunting studies have used association with DNA markers in or near genes that seem relevant to the trait under investigation; these genes are called candidate genes (Malhotra & Goldman, 1999). A problem with this approach is that any of the 30,000 or so genes expressed in the brain could conceivably be considered as candidate genes for most behaviors. In other words, association has not been systematic in the way that linkage can be used to conduct a systematic scan of the genome (but see below). The best example of a QTL association was found in 1993 between the late-onset dementia of Alzheimer's disease and a gene on chromosome 19 called apolipoprotein E (Corder et al., 1993). One allele (allele 4) for this gene is associated with Alzheimer's disease and is the only known predictor of this common disorder of later life affecting as many as 15% of individuals over 80 years of age. As replicated in scores of studies, the frequency of this allele is about 20% in the population, whereas for individuals with Alzheimer's disease, the allelic frequency is about 40%. This allele is neither a necessary nor a sufficient cause of Alzheimer's disease because more than half of Alzheimer's disease patients do not have this allele and many people with this allele do not have Alzheimer's disease. There is some evidence that allele 2 of the apolipoprotein E gene may play a protective role (Corder et al., 1994). Finding QTLs that protect rather than increase risk for a disorder is an important direction for genetic research. In 1992, one of the genes on chromosome 14 was identified as accounting for most cases of a rare (1 in 10,000) single-gene type of Alzheimer's disease that appears before 65 years of age (St. George-Hyslop et al., 1992). In 1995, the specific offending gene, called presenilin-1, was identified (Sherrington et al., 1995), although it is not yet known how the gene causes early-onset Alzheimer's disease. Multiple susceptibility genes are now known to exist, and there are transgenic mouse models with alterations in both amyloid precur-

SPECIAL ISSUE: DNA sor protein and presenilin genes (Guenette & Tanzi, 1999; Lippa, 1999).

819

agreed that a complete scan would require maps many times denser. Genotyping so many markers seems an impossible task. With 200 subjects each in groups of cases and controls, each marker would require 400 genotypings, which means that 3,500 markers would require 1.4 million genotypings. However, a new technique called DNA pooling makes this prospect less daunting (Daniels, Holmans, Plomin, McGuffin, & Owen, 1998). DNA pooling combines DNA from cases and compares it with pooled DNA for controls. The pooled DNA for the two groups can be genotyped and compared as if they were just two individuals. In terms of genoryping effort in the example just given, this means that 3,500 markers needed for a systematic genome screen require only 7,000 genotypings. DNA pooling is relevant only when selected groups are investigated, such as cases and controls or groups high and low on a quantitative trait. Unselected samples require individual genotyping because there are no groups to compare. DNA pooling is being used to scan the genome for allelic association for general cognitive ability (Plomin, in press). Although the approach is expected to detect only some of the largest QTLs associated with general cognitive ability, several replicated QTLs have been reported (Chomey et al., 1998; P. J. Fisher et al., 1999; Hill et al., 1999). DNA chips, another technical advance that will facilitate genomewide scans using association, are discussed later.

Neurotransmitter Candidate Genes


Genes in the dopamine and serotonin neurotransmitter systems have been widely used to investigate associations with behavior. Although allelic association has been primarily used to compare diagnosed cases versus controls, as in Alzheimer's disease or drug abuse, it can also be used to correlate the presence of a particular allele with scores on a quantitative trait. For example, in 1996, a dopamine receptor gene (DRD4), which is a gene largely expressed in the brain limbic system, was reported to be associated with the personality trait of novelty seeking in three analyses (Benjamin et al., 1996; Ebstein et al., 1996). In the initial three studies, individuals with the long-repeat alleles had higher noveltyseeking scores. The marker involves a sequence of 48 base pairs that repeat from 2 to 8 times. It is postulated that receptors with longer DRD4 repeats (6-8 repeats) are less efficient at binding dopamine and thus that individuals with the long-repeat alleles are dopamine deficient and seek novelty to increase dopamine release. Although the association has not been consistently replicated, there is a trend toward replication, and methodological issues may account for some of the failures to replicate (Plomin & Caspi, 1998; Wahlsten, 1999). In reviewing several subsequent studies, the original authors noted that the QTL appears to increase novelty-seeking scores by only about 5%, an effect size that is difficult to detect in many experiments (Ebstein & Belmaker, 1997). DRD4 has also been reported in several studies to be associated with attention-deficit hyperactivity disorder (ADHD; Thapar, Holmes, Poulton, & Harrington, 1999) and with heroin addiction (Kotler et al., 1997; U et al., 1997). Another dopamine gene (DAT1) shows even more consistent associations with ADHD (Thapar et al., 1999). A recent study replicated the DAT1 association and also reported associations with two other dopamine genes (DBH and DRDS: Daly, Hawi, Fitzgerald, & Gill, 1999). Another recent study also replicated the DAT1 association and suggested that the association primarily involves the hyperactiveimpulsive component of ADHD rather than the inattentive component (Waldman et al., in press). Genes involved in serotonin function have also been reported to be associated with traits related to anxiety (Ebstein, Nemanov, Klotz, Gritsenko, & Belmaker, 1997; Katsuragi et al., 1999; Lesch, Greenberg, & Murphy, in press; Lesch et al., 1996; Osher, Hamer, & Benjamin, 2000), although several failures to replicate have been reported (Flory et al., 1999). Several other genes have been reported to be associated with personality (Benjamin, Ebstein, & Belmaker, in press; Hamer & Copeland, replicated. Research on candidate genes such as dopamine-related genes has dominated allelic association research. To some degree, this represents the art of the possible. Because allelic association is nearsighted, a systematic scan of the genome comparable to a scan using linkage would require thousands of DNA markers. For example, 3,500 evenly spaced markers would provide a marker approximately every 1 million base pairs in the 3.5 billion basepair genome, which means that no QTL would be more than 500,000 base pairs away from a marker. However, it is generally 1998), but none has as yet been consistently

Replicability of Association Studies: Population Stratification and Other Methodological Issues


From the qualifications raised in the preceding examples, it should be clear that association studies for complex traits have generally been difficult to replicate. An excellent discussion of the reasons why has recently appeared (Malhotra & Goldman, 1999), and we have touched on some of those reasons. One of the most widely publicized examples began as a report of an association between a gene marker in the dopamine D2 receptor gene and alcoholism. In 1990, a particular allele in the D2 receptor gene was reported to have a higher frequency of occurrence in brain samples from alcoholics than in those from controls (Blum et al., 1990). Dozens of studies have been performed since to test this hypothesis, with very mixed success. The statistical validity of the association was questioned in surveys of the existing literature, and these surveys also disagreed (Gelerntner, Goldman, & Risch, 1993; Pato, Macciardi, Pato, Verga, & Kennedy, 1993). Because the "finding" was already highly localized in the genome, a careful linkage study was performed using several polymorphisms in the gene: This study failed to find any evidence that individual haplotypes (particular groups of linked alleles) occurred more frequently in alcoholics (Suarez et al., 1994). Looking back from early in the year 2000, we conclude that there may be some relative increase in frequency of the Taq 1 A allele of the dopamine D2 receptor gene in alcoholics, but it is very clear that this will never constitute a risk marker with useful prognostic power. The original D2 story received a great deal of publicity, most of which tended to foster the misconception that the "alcoholism gene" had been identified. The subsequent failures to reproduce these results led to an equally uninformed backlash that has damaged the credibility of linkage and association mapping efforts for all complex traits. Why are such findings difficult to replicate? In the case of the D2 result, the most likely explana-

820

PLOMIN AND CRABBE

tion is that population stratification underlies the original finding and some of the subsequent positive findings. Ethnic groups often yield different allelic frequencies for DNA markers. Any such markers would appear to show associations wilh any phenotypic differences between the ethnic groups. For example, in samples that include French and Finns, associations with speaking French would be found for any DNA marker that showed allelic frequency differences between French and Finns because French are more likely to speak that language than Finns. Because ethnicity does not vary systematically within families even when one parent is French and the other is Finnish (for each gene, each offspring is just as likely to receive an allele from the French parent as from the Finnish parent), associations found within families avoid this possible bias. An added complication is that even within ethnic groups, there are often vast differences in allele frequencies for different subgroups (e.g., African American describes a large and diverse collection of allelic groups). This artifact can be attenuated by studying very well-matched cases and controls, and it can be eliminated by testing associations within families (e.g., by comparing siblings who differ genotypically or phenotypically) because ethnicity does not vary within families. (This is why the failure to find within-family association has convinced many that the D2 association is spurious.) A widely used design involves family triads of a proband and two parents. One version of this test is called the transmission disequilibrium test (TDT; Spielman, McGinnis, & Ewens, 1993). The TDT usually uses trios consisting of affected individuals and their biological parents. Each affected individual must have received the susceptibility alleles from his or her parents. These alleles transmitted from parents to affected individuals can be viewed as a group of case alleles. What about controls? Each parent transmits only one of two alleles at each gene to an offspring (see Figure 1). The alleles from the parents that are not transmitted can he considered as control alleles. In other words, the TDT needs only affected individuals and their parents (who do not need to be assessed phenotypically)no control group of individuals is required. The TDT rests on testing departures from the expected equal frequency of transmitted and nontransmitted alleles (i.e., linkage disequilibrium; see Figure 4). For example, a recent TDT analysis of ADHD confirms the casecontrol reports of association in that the same DRD4 allele was transmitted from parents to ADHD children in 61% of the families rather than in 50%, a highly significant difference for the 199 families in the sample. Other nontransmitted alleles appeared at equal frequency in parents and their offspring (Sunohara et al., in press). The reason for creating such a complicated group of transmitted versus nontransmitted alleles within families rather than comparing cases and controls is that investigating association in this way removes the possibility that associations found in the usual case-control design are due to ethnic differences. However, ethnicity can often be controlled more simply by matching cases and controls, as long as this is done with care. Another issue relevant to failures to replicate relates to statistical powerthat is, the balance between false positives and false negatives. When many markers are examined, the possibility of false-positive results must be taken into account. One way to address this issue is to lower alpha values to provide genomewide protection against finding even one false-positive result (Type 1 error). However, when hunting for QTLs of small effect size, the

danger of false negatives (Type II errors) must also be considered to avoid throwing out babies with the bath water. Multistage screening can be used to balance the two types of error. This approach starts with more lenient initial alpha levels and then replicates initial results in a different population to discriminate true associations from false-positive associations (Belknap et al., 1993). In the end, replication of QTL results across laboratories will be the ultimate arbiter. Behavioral Genomics: Using DNA Associated With Behavior It is important that psychologists be prepared to use genes as they are found. The pace of molecular genetics research leads us to predict that genes associated with behavior will be available for psychologists to use in their research and clinics to assess genetic risk. This is a safe prediction because it is already happening in research on cognitive decline in the elderly, where it has quickly become standard practice for research to take advantage of the genetic risk information provided by the DNA marker for apolipoprotein E. This has happened even for researchers who are primarily interested in psychosocial risk mechanisms because they are now able to study gene-environment interaction and correlation using measured genotypes. Because the contribution of apolipoprotein E to diagnostic specificity is only modest and because no therapeutic interventions are now available for individuals with this genetic risk, genotyping for apolipoprotein E to predict cognitive decline is not yet approved for use in clinical settings (Mayeux et al., 1998). However, a sign of things to come is that, because this genetic risk may be exacerbated by head injuries and especially by boxing (Mayeux, Ottman, Maestre, Ngai, & Tang, 1995; Mayeux et al., 1998), there are calls for boxers in the United States to be genotyped for apolipoprotein E (Jordan et al., 1997). This is an example of gene-environment interaction in that the specific environmental insult of head injury exacerbates the risk for dementia in genetically susceptible individuals. It is also an example of ways in which finding specific genes associated with behavior can greatly sharpen analyses of quantitative genetic issues that have been notoriously difficult to broach, such as genotypeenvironment interaction. As mentioned in. the introduction, the emphasis in genetic research is shifting from finding genes to finding out how they work, functional genomics. The term functional genomics is variously defined but generally is used in the context of molecular goals, that is, to identify the polymorphism's DNA difference, trace its effect on the gene's protein product, and delineate the protein's effect at a molecular level of analysis. Thus, functional genomics asks how the gene works and what its protein does. However, analysis at the highest level of organismic integrationbehavioris also appropriate for understanding how genes work. We refer to the analysis of contributions of individual genes and groups of interacting genes to behavioral functions of the whole organism as behavioral genomics. The different levels of explanation addressed by functional and behavioral genomics do not represent choices between right and wrong methods; rather, they are more or less useful for addressing different questions. To an extent, functional genomics and behavioral genomics represent the familiar distinction between bottom-up and top-down approaches to understanding relationships between genes and be-

SPECIAL ISSUE: DNA

821

havior. Functional genomics is intrinsically bottom-up, beginning with molecules to ask how genes work and hoping eventually to trace pathways from the gene through the brain and finally on to behavior. This is a long road to travel, especially for complex traits (P. C. Phillips, 1999). Top-down (behavioral genomics) analyses of the effects of genes on behavior broach questions about the functioning of the whole organism such as cognitive processes involved in learning and memory. Thus, behavioral genomics asks how the behavior works and how its genetic determinants are important for understanding its relationship to other behaviors and to the environment. It then traces the chain of accompanying biological events at a more cellular/molecular level of analysis, often using behaviorally distinct genetic animal models as starting points. The behavioral level of analysis may be the most appropriate level of understanding for evolution, which focuses on the functioning of the whole organism. It may also yield a quicker return on investment for society in terms of prevention and treatment, as discussed later. Of course, distinctions between behavioral and functional genomics are somewhat artificial. For example, the gene-mapping approaches discussed earlier (e.g., QTL analysis) leapfrog from behavior to the genome. Similarly, experiments with targeted gene deletion mutants often attempt to jump straight to the behavioral level of analysis from a candidate gene suggested by a QTL analysis. Leapfrogging back and forth between QTLs and the whole animal to ask questions about the effects of genes at the behavioral level of analysis may provide a useful perspective on how genes work, and it will certainly advance the understanding of behavior. A key role for psychologists in the future will be to articulate the behavioral domains for studies at both the behavioral and functional genomics levels of analysis. We begin this discussion of behavioral genomics by mentioning several developments that will make it easier for DNA to be used in psychological research and practice. We then discuss our view of the emerging nature of behavioral genomic analysis. Technical Advances That Will Facilitate Behavioral Genomics Although it is difficult and expensive to find genes associated with complex traits, it is relatively easy and inexpensive to use information about these genes once they have been identified. One example is that blood is no longer needed to obtain DNA. Simply rubbing a cotton swab on the inside of the mouth provides enough DNA to genotype thousands of DNA markers. This means that DNA can be obtained by mail and that parents can collect it from their children (Freeman et al., 1997). Costs for materials and labor in extracting DNA are surprisingly low. We have estimated costs at about $5 per participant, which means that the one-time cosl for extracting DNA in a study of 400 individuals would be about $2,000. Extracted DNA can be preserved indefinitely. Obtaining DNA for any valuable sample is a worthwhile investment because it puts researchers in a position to analyze their data in relation to any relevant genes as soon as they are identified. The cost of genotyping a single DNA marker could be roughly estimated as no more than $3 per participant, or $1,200 for 400 participants, although several markers could be genotyped for the same price. If a commercial company is used, these costs could be many times higher; on the other hand, collaboration with a molecular genetics

colleague could reduce the costs considerably. The point is that using DNA is less expensive than many other psychological procedures such as videotape ratings, cortisol assays, and especially neuroimaging. Although some psychology departments already have the capability to extract and genotype DNA, most psychologists will get DNA work done through collaborations or contracts. Automated DNA-sequencing machines and robotics instruments, which permit thousands of genotypings per day, can cost hundreds of thousands of dollars. Such machines create an economy of scale and thus make the use of large commercial operations more cost effective than doing it oneself, especially when psychologists are using a few genes that have been found to be associated with behavior rather than trying to find genes. Genotyping will be made quicker and more efficient by DNA chip technology in which a microarray of DNA can be used to genotype thousands of DNA. When genes of interest are identified, it will often be of interest to determine how active they are. DNA chips can also be used to assess the degree to which a gene is expressed (Watson & Akil, 1999). Thousands of short DNA sequences (oligonucleou'des), each uniquely representing a gene, are applied to a chip, and a sample of DNA is also applied to the chip. Only those genes actively expressed at the time the DNA sample was collected bind to their complementary, chip-embedded oligonucleotide because, in an actively expressed gene, the DNA double helix is uncoiled to allow the complementary sequences of base pairs in the oligonucleotide and the gene's DNA to bind. The analysis of a broad spectrum of expressed genes using DNA chips or other expression array technology is sometimes also called functional genomics (Brownstein, Trent, & Boguski, 1998) and is beginning to be used in studies comparing diseased versus healthy cell lines, for example. A widespread effort is underway to develop applications to the study of DNA expression in humans and nonhuman animal models. A fascinating recent example used an expression array constructed of oligonucleotides for 6,347 genes. When DNA from skeletal muscle was compared between aged and young adult mice, only 113 genes showed significant differences in their expression. These genes were classified by function, and most showing increased expression in aging could be classified as stressresponse genes, whereas those showing decreased expression were metabolic or biosynthetic genes. For most genes whose expression changed during aging, caloric restriction, an environmental manipulation known to prolong longevity in rodents, prevented or attenuated the age-related changes (Lee, Klopp, Weindruch, & Prolla, 1999). As noted earlier, animal models are valuable for finding genes associated with behavior. Animal models will become even more valuable as the emphasis of research shifts toward behavioral genomics because both genes and environments can be manipulated experimentally in nonhuman animal species. The mouse has become the leading animal model for functional genomic analyses (Battey et al., 1999), and learning and memory in mice provide one good example of work of this type (Silva, Smith, & Giese, 1997). As discussed earlier, conditional knockouts that permit turning the expression of a gene on or off during development in specific areas of the brain will increase the value of mouse models even more (Silva & Mayford, 1998).

822
Understanding Behavior: The Key to Behavioral Genomics

PLOMIN AND CRABBE

Behavioral genomics can add to the understanding of how genes work even when little is known about the molecular biology or neurobiology of the gene. However, it will be increasingly important for psychology as a science to emphasize the important role that behavioral genomics has to play in functional genomics. Ask any competent geneticist what the key to understanding is when attempting to detect a genetic influence on a trait, and the answer will be careful and reliable ascertainment and description of the phenotype. In much the same way that a computer can produce nothing of value if the input data are flawed, genomic analyses, whether molecular or behavioral, are worthless if the phenotype is not carefully measured. Thus, the key to behavioral genomics is behavior.

disorder before realizing its expression in drug dependence. Cases like this are more likely to be the rule than the exception, and disentangling gene-environment correlation will be a future area of active research. Identification of specific genes will greatly facilitate analyses of this type. Specifying Behavioral and Temporal Domains

Heterogeneity and Comorbidity An important set of questions for behavioral genomics is heterogeneity and coinorbidity. Do reading-disabled children with the chromosome 6 genetic risk factor differ in cognitive processes from other reading-disabled children? A question of this type that is already being asked is whether the linkage is specific to orthographic or phonological disabilities. The answer so far is that the linkage appears to be general in its effect across a broad range of reading processes (Gayan et al., 1999). Are the effects of the gene broader still, contributing to the comorbidity between reading disability and mild mental retardation or between reading disability and behavior problems? In a future where genes' increasing risk for each of these problems is known, it will be easier to address this question, and ameliorative strategies may be better chosen to suit the task. We predict that finding genes related to behavior will lead to a gene-based classification of disorders that will bear little resemblance to our current symptom-based diagnostic systems. Indeed, from a QTL perspective, we predict that many common disorders are the quantitative extreme of normal genetic variation rather than being qualitatively different. That is, the QTL on chromosome 6 for reading disability may not be a gene for reading disability but rather a gene that contributes to variation in reading ability throughout the distribution, not just at the low end. An example from the human literature is the comorbidity among several personality disorders and alcoholism and/or substance abuse. Careful description of the common and unique aspects of these behavioral traits has allowed investigators to determine that there is a degree of genetic commonality among these disorders (Pickens, Svikis, McGue, & LaBuda, 1995). However, there is also a degree of environmental similarity shared by subjects in different groups, and more powerful experiments are needed to discriminate among genetic, environmental, and interacting paths underlying comorbid risk (van den Bree, Svikis, & Pickens, 1998). A similar orientation to these issues led to the suggestion that there are two distinguishable genetic paths to increased risk for drug abuse (Cadoret et al., 1995). One path directly reflects inheritance from a biological parent's alcoholism. The other route to risk is more complex, starting with a biological parent with antisocial personality disorder and proceeding through intervening variables of adoptee aggressivity, conduct disorder, and antisocial personality

In the animal literature, researchers are forced to model the most crucial aspects of human behavior, and any animal model at best represents only a reduced part of the complex whole. Because most animal researchers do not talk to their animal subjects, they must infer animal cognition and feelings from their behavior, and the inevitable anthromorphism accompanying animal model design can easily lead one astray. Two simple examples can represent the issues. In a study of a knockout mouse with the serotonin IB receptor gene deleted, knockouts drank twice as much alcohol as controls and did not show differential preference for sucrose or saccharin solutions (Crabbe et al., 1996). This suggested strongly that the serotonin IB receptor gene in some way decreased alcohol's appeal. On the other hand, when alcohol was used in three other tests that putatively tap a drug's reinforcing efficacy (conditioned place preference, conditioned taste aversion, and oral operant self-administration), the two genotypes showed every possible sort of difference, including none at all (Risinger, Bormann, & Oakes, 1996; Risinger, Doan, & Vickrey, 1999). Clearly, not all of these behavioral assays tap the same domain in mice. In an analogous study, knockout and wild-type mice were tested after alcohol injection using eight different assays of drug-induced motor incoordination and found to differ on two tests, but not on the other six (Boehm, Schafer, Phillips, Browman, & Crabbe, 2000). The conclusion to be drawn is that the behavioral domain envisioned by the experimenter may not be perceived in the same way by the mouse. Motor incoordination in these tasks appears to involve task-specific combinations of muscle strength, balance, precisely patterned gait, and intact proprioceptive feedback, only some of which are sensitive to the serotonin IB gene deletion. Part of understanding how any gene affects behavior lies in understanding when the gene's effects on behavior come on line during development, the specificity or generality of the effects, and the extent to which the gene's effects differ in different environments. Developmental questions focus on the early origins of the gene association. For example, does the gene on chromosome 6 linked to reading disability show up earlier in development as language problems in infancy, before children learn to read? The developmental sequelae are also interesting. For example, does the genetic risk continue to affect cognitive processing in adulthood even after reading problems are ameliorated? We have already discussed the crucial aspects of developmental compensations that may obfuscate (or elucidate) the actions of genes manipulated in animal models. Here, the animal literature is likely to lead the way in understanding developmental effects because of the relative ease with which environments can be manipulated and controlled and because control of the timing of gene action will soon be achieved. Facing DNA How will psychologists use information about genes when gene-behavior associations are firmly identified? One obvious

SPECIAL ISSUE: DNA

823

answer is that knowledge of the genes associated with behavior can be incorporated into any psychological analysis. For example, when the gene underlying the linkage on chromosome 6 for reading disability is identified, the major research questions in the field will be reexamined with the greater precision of a specific gene, even if that gene is only one of many genes involved in reading disability. This is another example in which molecular genetic analyses can sharpen quantitative genetic analyses that rely on familial resemblance to address such issues as genetic links between the normal and abnormal and heterogeneity and comorbidity. Important applied questions involve treatment and prevention. Do children with the genetic risk for reading disability respond better to a particular treatment than other children with reading disability who do not have the genetic risk? One of the most important benefits of DNA is that genetic risk can be assessed early in life, which creates the possibility for secondary prevention. Currently, intervention must wait until reading disability is diagnosed at school or alcoholism is diagnosed in adulthood, by which time such problems are often resistant to change and lead to other life-course problems. Research using genes as risk indicators will possibly permit behavioral interventions that prevent problems before they occur. Educational Needs of and Opportunities for Future Psychologists We have already mentioned that we believe that training in genetics will be important to the psychologist of the future. A textbook (Plomin et al., 2001) and edited volumes (Crusio & Gerlai, 1999; B. C. Jones & Mormede, 1999) contain chapters for nonspecialists that can guide the interested through the basics. In addition to the biological basics, familiarity with the statistical concepts underlying human genetics; the potentials, methods, and limitations of the Human Genome Project; gene expression array profiling; and work with genetically engineered animals will be helpful. One domain that contains parts of all aspects of behavioral and functional genomics discussed in this article is the emerging field of bioinformatics (Boguski, 1998). Psychologists gravitate toward the study of complexity, and bioinformatics is a field that attempts to deal with the multiple levels of complexity required to move from genes to behavior. A multivariate-minded psychologist who is not afraid of DNA sequences and the biology of genetics will be a highly prized specialist during the next 10 years. Here is what the future might look like for clinical psychologists. DNA will be routinely collected. The most powerful potential for DNA is to predict risk so that genes can be used to aid in diagnosis and to plan treatment programs. Such secondary prevention for complex psychological traits is not likely to take the form of high-tech genetic engineering because many genes are involved. Even in the case of simple single-gene disorders, environmental interventions are more likely than genetic engineering. For example, as described earlier, PKU has been largely prevented not by high-tech solutions such as correcting the mutant DNA or by eugenic programs but rather by a change in diet that prevents the mutant DNA from having its damaging effects on the developing brain. Once the damage has been done, there is no way to ameliorate the profound retardation it causes. For this reason, newborns have been screened for decades for this genetic disorder to

determine which of the 1 in 10,000 babies has the disorder so that their diet can be changed. Clinical psychologists with even rudimentary sophistication in behavioral genomics will find their therapeutic task becoming easier in the future as behavioral disorders become better understood through the addition of genomic information.

Ethical Implications of Behavioral Genomics This picture of the future will confirm some people's worst fears about DNA. They fear that finding out about genetic risk, especially when no prevention or cure is available, will serve only to label people in ways that might lead to discrimination for education, insurance, and employment. Knowing about genetic risk might also become a self-fulfilling prophecy, for example, if a child is labeled as at risk for learning disorders. Parents using in vitro fertilization might select so-called designer babies with fewer genetic risks and more genetic strengths. These are serious problems, but it should be noted that most advances in science create new problems as well as new potential for doing good. For example, when amniocentesis was developed 2 decades ago to screen for genetic problems prenatally, it led to a minefield of problems involving abortion, especially the spectre that society might force this procedure on women at risk. However, mothers voted with their feet by choosing amniocentesis when they had specific genetic risks in their family or when they were older than 40, which increases the risk for chromosomal abnormalities such as the most common form of mental retardation, Down's syndrome (trisomy 21). There have been abuses of amniocentesis as well. In India, for example, where sons are considered more valuable than daughters, some parents have used amniocentesis to abort female fetuses. Some of the fears derive from misunderstandings about what genetics can and cannot do (Rutter & Plomin, 1997). The main misunderstanding is to think that genes determine outcomes in a hardwired, there's-nothing-anyone-can-do-about-it way. For many rare single-gene disorders, such as the gene on chromosome 4 that causes Huntington's disease, it is the case that genes determine outcomes. If one inherits the Huntington's disease allele, one will die from the disease regardless of one's other genes or environment. However, behavioral dimensions (and their extremes and disorders) are complex traits influenced by many genes as well as by many environmental factors. Thus, increased knowledge of the specific genes can, in theory, aid the psychologist in steering individuals toward desired outcomes through environmental treatments. For complex traits, genetic risk for undesired outcomes is often probabilistic. More importantly, as we have shown for several examples, the efficacy of a therapeutic intervention may differ depending on an individual's genotype at some crucial genes. Genetic engineering for genes that fine-tune behavior is still rather remote, but knowledge of genotypes is here now. hi many senses, the ethical dilemmas posed by the knowledge derived from DNA are novel, but in some senses, they represent familiar domains long considered by bioethicists (Crabbe & Belknap, 1997; Parker, 1995). A component of the Human Genome Project known as the Ethical, Legal and Social Implications program is devoted to the study of issues raised by advances in human genomics, and current advances as well as numerous links can be found at their website address (http://www.ornl.gov/hgmis/elsi/elsi.html). To gain knowledge

824

PLOMIN AND CRABBE

about someone's DNA requires his or her informed consent, but the consequences of that knowledge have an impact not only on the individual but on immediate family members including minor children. Many are concerned about the use of such information by employers and medical care and insurance providers to restrict access or increase costs to individuals identified as at increased risk. No reasonable person wants to see a return to the excesses of the eugenics movement, which is not so far in the past (for one horrifying example, sec Wahlsten, 1997). Several states have passed Genetic Privacy Acts declaring an individual's DNA to be his or her private property, and the pace of legislation will no doubt outstrip understanding of the issues! Clearly, society has some serious thinking to do to capitalize on the potential of the DNA revolution while circumventing the hazards. Although it is important to consider the potential problems, it is also timely to celebrate the potential benefits of DNA. Psychologists' expertise in investigating behavior can contribute immensely to the appropriate development of behavioral genomics.

Browman, K. E., & Crabbe, J. C. (1999). Alcohol and genetics: New animal models. Molecular Medicine Today, 5, 310-31 S. Brownstehl, M. J., Trent, J. M., & Boguski, M. (1998). In Trends guide to bioinformatics (pp. 27-29). Amsterdam: Elsevier Science. Buck, K. J., Crabbe, J. C., & Belknap. J. K. (2000). Alcohol and other abused drags. In D. W. Pfaff, W. H. Berrettini, T. H. Jon, & S. C. Maxson (Eds.), Genetic influences on neural and behavioral functions (pp. 159-183). Boca Raton, FL: CRC Press. Buck, K. J., Metten, P., Belknap, J. K., & Crabbe, J. C. (1997). Quantitative trait loci involved in genetic predisposition to acute alcohol withdrawal in mice. Journal of Neuroscience, 17, 3946-3955. Buck, K.., Metten, P., Belknap, J., & Crabbe, J. C. (1999). Quantitative trait loci affecting risk for pentobarbital withdrawal map near alcohol withdrawal loci on mouse chromosomes 1, 4, and 11. Mammalian Genome, 10, 431-437. Burmeistcr, M. (1999). Basic concepts in the study of diseases with complex genetics. Biological Psychiatry, 45, 522-532. Cadoret. R. J., Yates, W. R., Troughton, E., Woodworm, G., & Stewart, M. A. (1995), Adoption study demonstrating two genetic pathways to drug abuse. Archives of General Psychiatry; 52, 42-52. Capecchi. M. R. (1994, March). Targeted gene replacement. Scientific American, 52-59. Cardon, L. R., Smith, S. D., Fulker, D. W., Kimberling, W. J., Pennington, B. F., & DeFries, J. C. (1994). Quantitative trait locus for reading disability on chromosome 6. Science, 266, 276-279. Chen, Y.-C, Lu, R.-B., Peng, G.-S., Wang, M., Wang, H.-K., Ko, H.-C., Chang, Y.-C., Jang, J.-L., Ting, K., & Yin, S.-J. (2000). Alcohol metabolism and cardiovascular response in an alcoholic patient homozygous for the ALDH2*2 variant gene allele. Alcoholism: Clinical and Experimental Research, 23, 1853-1860. Chorney, M. J., Chorney, K., Seese, N., Owen, M. J., Daniels, J., McGuffin. P., Thompson, L. A., Detterman, D. K., Benbow, C. P., Lubinski, D., Eley, T. C., & Plomin, R. (1998). A quantitative trait locus (QTL) associated with cognitive ability in children. Psychological Science, 9,

References
Almasy, L., & Blangero, J. (1998). Multipoint quantitative-trait linkage analysis in general pedigrees. American Journal of Human Genetics, 62, 1198-1211. Antoch, M. P., Song, E. J., Chang, A. M., Vitaterna, M. H., Zhao, Y., Wilsbacher. L. D., Sangoram, A. M., King, D. P., Pinto, L. H., & Takahashi, J. S. (1997). Functional identification of the mouse circadian clock gene by rransgenic BAC rescue. Cell, 89, 655-667. Battey, J., Jordan, E., Cox, D., & Dove, W. (1999). An action plan for mouse genomics. Nature Genetics, 21, 73-75. Belknap, J. K., Metten, P., Helms, M. L., O'Toole, L. A., Angeli-Gade, S., Crabbe, J. C., & Phillips, T. J. (1993). QTL applications to substances of abuse: Physical dependence studies with nitrous oxide and ethanol. Behavior Genetics, 23, 213-222. Benjamin, J., Ebstein, R., & Belmaker, R. H. (in press). Molecular genetics of human personality. Washington, DC: American Psychiatric Press. Benjamin, J., Li, L., Patterson, C., Grcenburg, B. D., Murphy, D. L., & Hamer, D. H. (1996). Population and familial association between the D4 dopamine receptor gene and measures of novelty seeking. Nature Genetics, 12, 81-84. Blackwelder, W. C., & Elston, R. C. (1985). A comparison of sib-pair linkage tests for disease susceptibility loci. Genetic Epidemiology, 2, 85-97. Blum, K., Noble, E. P., Sheridan, P. J., Montgomery, A., Ritchie, T., Jagadeeswaran, P., Nogami, H., Briggs, A. H., & Cohn, J. B. (1990). Allelic association of human dopamine D2 receptor gene in alcoholism. JAMA, 263. 2055-2060. Boehm, S. L., II, Schafer, G. L., Phillips, T. J., Browman, K. E., & Crabbe, J. C. (2000). Sensitivity to ethanol-induced motor incoordination in 5-HTIB receptor null mutant mice is task-dependent: Implications for behavioral assessment of genetically altered mice. Behavioral Neuroscience, 114, 401-409. Boguski, M. S. (1998). Bioinformatics: A new era. In Trends guide to bwmfnrmatics (pp. 1-3). Amsterdam: Elsevier Science. Bosse, R., Fumagalli, F., Jaber, M., Giros, B., Gainctdinov, R. R., Wetsel, W. C., Missale, C., & Caron, M. G. (1997). Anterior pituitary hypoplasia and dwarfism in mice lacking the dopamine transporter. Neuron, 19, 127-138. Brandon, E. P., Idzerda, R. L., & McKnight. G. S. (1995). Targeting the mouse genome: A compendium of knockouts (Part I). Current Biology, 5, 625-634.

1-8.
Cloninger, C. R., Bohman, M., & Sigvardsson, S. (1981). Inheritance of alcohol abuse: Cross-fostering analysis of adopted men. Archive?; of General Psychiatry, 38, 861-868. Collins, F. S. (1999). Shattuck lectureMedical and societal consequences of the human genome project. New England Journal of Medicine., 341, 28-37. Collins, W. A., Maccoby. E. E., Steinberg, L., Hetherington, E. M., & Bomstein, M. (2000). Contemporary research in parenting: The case for nature and nurture. American Psychologist, 55, 218-232. Corder, E. H., Saunders, A. M., Risen, N. J.. Strittmatter, W. J., Shmechel, D. E., Gaskell, P. C., Jr., Rimmler, J. B., Locke, P. A.. Conneally, P. M., Schmader, K. E., & et al. (1994). Protective effect of apolipoprotein E type 2 allele for late onset Alzheimer disease. Nature Genetics, 7, 180-184. Curder, E. H., Saunders, A. M., Strittmatter, W. J., Schmechel, D. E., Gaskell, P. C., Small, G. W., Roses, A. D., Haines, J. L., & Pcricak Vance, M. A. (1993). Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer's disease in late onset families. Science, 261, 921-923. Crabbe, J. C., & Belknap, J. K. (1997). Ethical consequences of mapping QTLs for complex human traits. In A. H. Paterson (Ed.), Molecular dissection of cample* traits (pp. 279-285). Boca Raton, FL: CRC Press. Crabbe, J. C., Belknap, J. K., & Buck, K. J. (1994). Genetic animal models of alcohol and drug abuse. Science, 264, 1715-1723. Crabbe, J. C., & Harris, R. A. (Eds.). (1991). The genetic basis of alcohol and drug fictions. New York: Plenum. Crabbe, J. C., Phillips, T. J., Buck, K. J., Cunningham, C. L., & Belknap, J. K. (1999). Identifying genes for alcohol and drug sensitivity: Recent progress and future directions. Trends in Neurosciences, 22, 173-179. Crabbe, 1. C., Phillips, T. J., Feller, D. J., Hen, R., Wenger, C. D., Lessov,

SPECIAL ISSUE: DNA C. N., & Schafer, G. L. (1996). Elevated alcohol consumption in null mutant mice lacking 5-HTIB serotonin receptors. Nature Genetics, 14, 98-101. Crabbe, J. C., Wahlsteri, D., & Dudek, B. C. (1999). Genetics of mouse behavior: Interactions with laboratory environment. Science, 284, 16701672. Crusio, W. E. (1999). Using spontaneous and induced mutations to dissect brain and behavior genetically. Trends in Neumsciences, 22, 100-102. Crusio, W. E., & Gerlai, R. T. (Eds.). (1999). Handbook of moleculargenetic techniques for Elsevier. brain and behavior research. Amsterdam:

825

transporter (5-HTTLPR) polymorphism. Molecular Psychiatry, 4, 251-

257.
Freeman, B., Powell. J., Ball, D. M., Hill, L., Craig, I. W., & Plomin, R. (1997). DNA by mail: An inexpensive and noninvasive method for collecting DNA samples from widely dispersed populations. Behavior Genetics, 27, 251-257. Fulker, D. W., & Chemy, S. S. (1996). An improved multipoint sib-pair analysis of quantitative traits. Behavior Genetics, 26, 527-532. Gayan, J., Smith, S. D., Chemy, S. S., Cardon, L. R., Fulker, D. W., Brower, A. W., Olson, R. K., Pennington, B. F., & DeFries, J. C. (1999). Quantitative-trait locus for specific language and reading deficits on Chromosome 6p. American Journal of Human Genetics, 64, 157-164. Gelemtner, I, Goldman, D., & Risch, N. (1993). The A, allele at the D2 dopamine receptor gene and alcoholism: A reappraisal. JAMA, 1673-1677. Gerlai, R. (1996). Molecular genetic analysis of mammalian behavior and brain processes: Caveats and perspectives. Seminars in the Neurosciences, 8, 153-161. Giros, fl., Jaber, M., Jones, S. R., Wightman, R. M., & Caron, M. G. (1996). Hyperlocomotion and indifference to cocaine and amphetamine in mice lacking the dopamine transporter. Nature, 379, 606-612. Grigorenko, E. L., Wood, F. B., Meyer, M. S., Hart, L. A., Speed, W. C., Sinister, A., & Pauls, D. L. (1997). Susceptibility loci for distinct components of developmental dyslexia on chromosomes 6 and 35. American Journal of Human Genetics, 60, 27-39. Guenette, S. Y., & Tanzi, R. E. (1999). Progress toward valid transgenic mouse models for Alzheimer's disease. Neurobiology of Aging, 20, 201-211. Gusella, J. F., Wexler, N. S., Conneally, P. M., Naylor, S. L., Anderson, M. A., & Tanzi, R. E. (1983). A polymorphic DNA marker genetically linked to Huntington's disease. Nature, 306, 234-238. Guzowski, J. P., & McGaugh, J. L. (1997). Antisense oligodeoxynucleotide-mediated disruption of hippocampal cAMP response element binding protein levels impairs consolidation of memory for water maze training. Proceedings of the National Academy of Sciences of the United States of America, 94, 2693-2698. Hamer, D., & Copeland, P. (1998). Living with our genes. New York: Doubleday. Harada, S., Agarwal, D. P., Goedde, H. W., Tagaki, S., & Ishikawa, B. (1982). Possible protective role against alcoholism for aldehyde dehydrogenase deficiency in Japan. Lancet, 2, 827, Harris, J. R. (1998). The nurture assumption: Why children turn out the way they do. New York: Free Press. Haseman, J. K., & Elston, R. C. (1972). The investigation of linkage between a quantitative trait and a marker locus. Behavior Genetics, 2, 3-19. Heath, A. C., Jardine. R., & Martin, N. G. (1989). Interactive effects of genotype and social environment on alcohol consumption. Journal of . Studies on Alcohol, 50, 38-48. Hermstein, R. J., & Murray, C. (1994). The bell curve: Intelligence and class structure in American life. New York: Free Press. Herzog, E. D., Takahashi, J. S., & Block, G. D. (1998). Clock controls circadian period in isolated suprachiasmatic nucleus neurons. Nature. Neuroscience, 1, 708-713. Higuchi, S., Matsushita, S., Imazeki, H., Kinoshita, T., Takagi, S., & Kono, H. (1994). Aldehyde dehydrogenase genotypes in Japanese alcoholics. Lancet, 343, 741-742. Hill, L., Craig, I. W., Ball, D. M., Eley, T. C., Ninomiya, T., Fisher, P. J., McGuffin, P., Owen, M. J., Chorney, K., Chorney, M. J., Benbow, C. P., Lubinski, D., Thompson, L. A., & Plomin, R. (1999). DNA pooling and dense marker maps: A systematic search for genes for cognitive ability. NeuroReport, 10, 843-848. Hood, H. M., Crabbe, 1. C., Belknap, J. K., & Buck, K. J. (2000). Epistatic interaction between quantitative trait loci affects the genetic predispo169,

Czerwinski, S. A., Mahaney, M. C., Williams, J. T., Almasy, L., Blangero, J. (1999). Genetic analysis of personality traits and alcoholism using a mixed discrete continuous trait variance component model. Genetic Epidemiology, 17, S121-S126. Daly, G., Hawi, Z., Fitzgerald, M., & Gill, M. (1999). Mapping susceptibility loci in attention deficit hyperactivity disorder: Preferential transmission of parental alleles at DAT1, DBH and DRD5 to affected children. Molecular Psychiatry, 4, 192-196. Daniels, J., Holmans, P., Plomin, R., McGuffin, P., & Owen, M. J. (1998). A simple method for analyzing microsatellite allele image patterns generated from DNA pools and its application to allelic association studies. American Journal of Human Genetics, 62, 1189-1197. Darvasi, A. (1998). Experimental strategies for the genetic dissection of complex traits in animal models. Nature Genetics, 18, 19-24. Deater-Deckard, K., Reiss, D., Hetherington, E. M., & Plomin, R. (1997). Dimensions and disorders of adolescent adjustment: A quantitative genetic analysis of unselected samples and selected extremes. Journal of Child Psychology and Psychiatry, 38, 515-525. Duhon, S. A., Murakami, S., & Johnson, T. E. (1996). Direct isolation of longevity mutants in the nematode Caenorhabditis elegans. Developmental Genetics, 18, 144-153. Dunn, J. R, & Plomin, R. (1990). Separate lives: Why siblings are so different. New York: Basic Books. Ebstein, R. P., & flelmaker, R. H. (1997). Saga of an adventure gene: Novelty seeking, substance abuse and the dopamine D4 receptor (D4DR) exon III repeat polymorphism. Molecular Psychiatry, 2, 381-384. Ebstein, R. P., Nemanov, L., Klotz, I., Gritsenko, L, & Belmaker, R. H. (1997). Additional evidence for an association between the dopamine D4 receptor (D4DR) exon III repeat polymorphism and the human personality trait of novelty seeking. Molecular Psychiatry, 2, 472-477. Ebstein, R.. Novick, O., Umansky, R., Priel, B., Osher, Y., Blaine, D., Bennett, E. R., Nemanov, L., Katz, M., & Belmaker, R. H. (1996). Dopamine D4 receptor (D4DR) exon III polymorphism associated with the human personality trait of novelty seeking. Nature Genetics, 12, 78-80. Fisher, P. I., Turic, D., McGuffin, P., Asherson, P. J., Ball, D. M., Craig, I. W., Eley, T. C., Hill, L., Chorney, K., Chomey, M. I., Benbow, C. P., Lubinski, D., Plomin, R., & Owen, M. J. (1999). DNA pooling identifies QTLs for general cognitive ability in children on chromosome 4. Human Molecular Genetics, 8, 915-922. Fisher, R. A. (1918). The correlation between relatives on the supposition of Mendelian inheritance. Transactions of the Royal Society of Edinburgh, 52, 399-433. Fisher, S. E., Marlow, A. J., Lamb, J., Maestrini, E., Williams, D. F., Richardson, A. J., Weeks, D. E., Stein, I. F., & Monaco, A. P. (1999). A quantitative-trait locus on Chromosome 6p influences different aspects of developmental dyslexia. American Journal of Human Genetics, 64, 146-156. Hint, I., Corley, R., DeFries, J. C., Fulker, D. W., Gray, J. A., Miller, S., & Collins, A. C. (1995). A simple genetic basis for a complex psychological trait in laboratory mice. Science, 269, 1432-1435. Flory, J. D., Manuck, S. B., Ferrell, R. E., Dent, K. M., Peters, D. G., & Muldoon, M. F. (1999). Neuroticism is not associated with the serotonin

826

PLOM1N AND CRABBE Synergistic effects of traumatic head injury and apolipoprotein-epsilon 4 in patients with Alzheimer's disease. Neurology, 45, 555-557. Mayeux, R., Saunders, A. M., Shea, S., Mirra, S., Evans, D., Roses, D., Hyman, B. T., Grain, B., Tang, M. X., & Phelps, C. H. (1998). Utility of the apolipoprotein E genotype in the diagnosis of Alzheimer's disease. New England Journal of Medicine, 338, 506-511. Mayford, M., & Kandel, E. R. (1999). Genetic approaches to memory storage. Trends in Genetics, 15, 463-470. McCleam, G. E., Plomin, R., Gora-Maslak, G., &Crabbe, J. C. (1991). The gene chase in behavioral science. Psychological Science, 2, 222-229. Mednick, S. A., Gabrielli, W. F., Jr., & Hutchings, B. (1987). Genetic factors in the etiology of criminal behavior. In S. A. Mednick, W. F. Gabrielli, Jr., & B. Hutchings (Eds.), The causes of crime: New biological approaches (pp. 74-91). Cambridge, England: Cambridge University Press. Moldin, S. O. (1997). The maddening hunt for madness genes. Science, 17, 127-129. Murata, M., Tagawa, M., Watanabe, S., Kimura, H., Takeshita, T., & Morimoto, K. (1999). Genotype difference of aldehyde dehydrogenase 2 gene in alcohol drinkers influences the incidence of Japanese colorectal cancer patients. Japanese Journal of Cancer Research, 90, 711-719. Murayama, M., Matsushita, S., Muramatsu, T., & Higuchi, S. (1998). Clinical characteristics and disease course of alcoholics with inactive aldehyde dehydrogenase-2. Alcoholism, Clinical and Experimental Research, 22, 524-527. Nance, M. A. (1997). Genetic testing of children at risk for Huntington's disease. Neurology, 49, 1048-1053. Nelson, R. J., Demas, G. E., Huang, P. L., Fishman, M. C., Dawson, V. L., Dawson, T. M., & Snyder, S. H. (1995). Behavioural abnormalities in male mice lacking neuronal nitric oxide synthase. Nature, 378, 383-386. Nelson, R. J., & Young, K. A. (1998). Behavior in mice with targeted disruption of single genes. Neuroscience Binhehavior Review, 22, 453
462.

silion for physical dependence on pentabarbital in mice. Manuscript submitted for publication. Jones, B, C., & Mormede, P. (Eds.). (1999). Neuwbehavioral genetics: Methods and applications. Boca Raton, FL: CRC Press. Jones, S. R., Gainetdinov, R. R., Jaber, M., Giros, B., Wightman, R. M., & Caron, M. G. (1998). Profound neuronal plasticity in response to inactivation of the dopamine transporter. Proceedings of the National Academy of Sciences of the United States of America, 95, 4029-4034. Jordan, B. D., Relkin, N. R., Ravdin, L. D., Jacobs, A. R., Bennett, A., & Gandy, S. (1997). Apolipoprotein E epsilon 4 associated with chronic traumatic brain injury in boxing. JAMA, 278, 136-140. Katsuragi, S., Kunugi, H., Sano, A., Tsutsumi, T., [sogawa, K., Nanko, S., & Akiyoshi, J. (1999). Association between serotonin transporter gene polymorphism and anxiety-related traits. Biological Psychiatry, 45, 368-370. Kendler, K. S., & Eaves, L. J. (1986). Models for the joint effects of genotype and environment on liability to psychiatric illness. American Journal of Psychiatry, 143, 279-289. King, D. P., Zhao, Y., Sangoram, A. M., Wilsbacher, L. D., Tanaka, M., Antoch, M. P., Steeves, T. D., Vitaterna, M. H., Kornhauser, J. M., Lowrey, P. L., Turek, F. W., & Takahashi, J. S. (1997). Positional cloning of the mouse circadian clock gene. Cell, 89, 641-653. Kosslyn, S., & Plomin, R. (in press). Towards a neuro-cognitivc genetics: Goals and issues. In D. Dougherty, S. L. Rauch, & J. F. Rosenbaum (Eds.), Psychiatric neuroimagiftg strategies: Research and clinical applications. Washington, DC: American Psychiatric Press. Kotler, M., Cohen, H., Segman, R., Gritsenko, I., Nemanov, L., Lerer, B., Kramer, F., Zer-Zion, M., Kletz, I., & Ebstein, R. P. (1997). Excess dopamine D4 receptor (D4DR) exon III seven repeat allele in opioiddependent subjects. Molecular Psychiatry, 2, 251-254. Lee, C. K., Klopp, R. G., Weindrach, R., & Prolla, T. A. (1999). Gene expression profile of aging and its retardation by caloric restriction. Science, 285, 1390-1393. Lesch, K.-P., Bengel, D., Heils, A., Zhang Sabol, S., Greenburg, B. D., Petri, S., Benjamin, J., Muller, C. R., Hamer, D. H., & Murphy, D. L. (1996). Association of anxiety-related traits with a polymorphism in the serotonin transporter gene regulatory region. Science, 274, 1527-1531. Lesch, K.-P., Greenberg, B., & Murply, D. L. (in press). The serotonin transporter, human anxiety, and affective disorders. In J. Benjamin, R. Ebstein, & R. H. Belmaker (Eds.), Molecular genetics of human personality. Washington, DC: American Psychiatric Press. Li, T., Xu, K., Deng, H., Cai, G., Liu, J., Liu, X., Wang, R., Xiang, X., Zhao, J., Murray, R. M., Sham, P. C., & Collier, D. A. (1997). Association analysis of the dopamine D4 gene exon III VNTR and heroin abuse in Chinese subjects. Molecular Psychiatry', 2, 413-416. Lippa, C. F. (1999). Familial Alzheimer's disease: Genetic influences on the disease process [Review]. International Journal of Molecular Medicine, 4, 529-536. Lykken, D. T. (1982). Research with twins: The concept of emergenesis. Psychophysiology, 19, 361-373. Maccoby, E. E. (2000). Parenting and its effects on children: On reading and misreading behavior genetics. Annual Review of Psychology, 51,
1-27.

Ogawa, S.. & Pfaff, D. W. (1996). Application of antisense DNA method for the study of molecular bases of brain function and behavior. Behavior Genetics. 26, 279-292. Osher, Y., Hamer, D., & Benjamin, J. (2000). Association and linkage of anxiety-related traits with a functional polymorphism of the serotonin transporter gene regulatory region in Israeli sibling pairs. Molecular Psychiatry, 5, 216-219. Paris, J. (1999). Genetics and psychopathology: Predisposition-stress interactions. Washington. DC: American Psychiatric Press, Parker, L. S. (1995). Ethical concerns in the research and treatment of complex disease. Trends in Genetics, II, 520-523. Pasternak, G. W. (2000). Genetics and opioid pharmacology. In D. W. Pfaff, W. H. Berrettini, T. H. Joh, & S. C. Maxson (Eds.), Genetic influences on neural and behavioral functions (pp. 13-29). Boca Raton, FL: CRC Press. Pato, C. N., Macciardi, F., Pato, M. T., Verga, M., & Kennedy, J. L. (1993). Review of the putative association of dopamine D2 receptor and alcoholism: a meta-analysis. American Journal of Medical Genetics (Neuropsychiatric Genetics), 48, 78-82. Phillips, P. C. (1999). From complex traits to complex alleles. Trends in Genetics, 15, 6-8. Phillips, T. J., Belknap, J. K., Buck, K. J., & Cunningham, C. L. (1998). Genes on mouse chromosomes 2 and 9 determine variation in ethanol consumption. Mammalian Genome, 9, 936-941. Phillips, T. J., Brown, K. J., Burkhart-Kasch, S., Wenger, C. D., Kelly, M. A., Rubinstein, M., Grandy, D. K., & Low, M. J. (1998). Alcohol preference and sensitivity are markedly reduced in mice lacking dopamine D2 receptors. Nature Neuroscience. I, 610-615. Pickens, R. W., Svikis, D. S., McGue, M., & LaBuda, M. C. (1995).

Malhotra. A. K., & Goldman, D. (1999). Benefits and pitfalls encountered in psychiatric genetic association studies. Biological Psychiatry, 45, 544-550. Markel, P. D., Bennet, B., Beeson, M., Gordon, L., & Johnson, T. E. (1997). Confirmation of quantitative trait loci for ethanol. Genome Research, 7, 92-99. Marubio, L. M., Arroyo-Jimenex, M. M., Cordero-Erausquin, M, Lina, C., & Novhre, N. L. (1999). Reduced antinociception in mice lacking neuronal nicotinic receptor subunits. Nature, 398, 805-810. Mayeux, R., Oilman, R., Maestre, G., Ngai, C., & Tang, M. X. (1995).

SPECIAL ISSUE: DNA Common genetic mechanisms in alcohol, drug, and mental disorder consumption. Drug and Alcohol Dependence, 9, 936-941. Plomin, R. (1994). Genetics and experience: The interplay between nature and nurture. Thousand Oaks, CA: Sage. Plomin, R. (in press). Quantitative trait loci (QTLs) and general cognitive ability ("g"). In J. Benjamin, R. Ebstein, & R. H, Belmaker (Eds.), Molecular genetics of human personality. Washington, DC: American Psychiatric Press. Plomin, R., & Caspi, A. (1998). DNA and personality. European Journal of Personality, 12. 387-407. Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from each other? Behavioral and Brain Sciences, 10, 1-16. Plomin, R., DeFries, ]. C., McClearn, G. E., & McGuffln, P. (2001). Behavioral genetics (4th ed.). New York: Freeman. Plomin, R., DeFries, J. C, McClearn. G. E., & Rutter, M. (1997). Behavioral genetics (3rd ed.). New York: Freeman. Plomin, R., & McClearn, G. E. (1993). Nature, nurture, and psychology. Washington, DC: American Psychological Association. Plomin, R., Owen, M. J., & McGuffin, P. (1994). The genetic basis of complex human behaviors. Science, 264, 1733-1739. Plomin, R., & Rutter, M. (1998). Child development, molecular genetics, and what to do with genes once they are found. Child Development, 69, 1221-1240. Reiss, D., Neiderhiser. J. M., Hetherington, E. M., & Plomin, R. (2000). The relationship code: Deciphering genetic and social patterns in adolescent development. Cambridge, MA: Harvard University Press. Risen, N., & Merikangas, K. R. (1996). The future of genetic studies of complex human diseases. Science, 273, 1516-1517. Risinger, F. O., Bormann, N. M., & Oakes, R. A. (1996). Reduced sensitivity to ethanol reward, but not ethanol aversion, in mice lacking 5-HT|B receptors. Alcoholism: Clinical and Experimental Research, 20, 1401-1405. Risinger, F. O., Doan, A. M., & Vickrey, A. C. (1999). Oral operant self-administration in 5-HTIB knockout mice. Behavioral Brain Research, 102, 211-215. Rocha, B. L., Scearee-Levie, K., Lucas, J. J., Hiroi, N., Castanon, N., Crabbe, J. C., Nestler, E. J., & Hen, R. (1998). Increased vulnerability to cocaine in mice lacking the serotonin-IB receptor. Nature, 393, 175-

827

Seligman. M. E. P., & Csikszentmihalyi, M. (2000). Positive psychology: An Introduction. American Psychologist, 55, 5-14. Sherrington, R., Rogaev, E. I., Liang, Y., Rogaeva, E. A., Levesque, G., Dceda, M., Chi, H.. Lin, C., Li, G., Holman, K., & et al. (1995). Cloning of a gene bearing mis-sense mutation in early-onset familial Alzheimer's disease. Nature, 375, 754-760. Silva, A. J., & Mayford, M. (Eds.). (1998). Transgenic/knockout approaches in neurobiology [Special issue]. Learning Memory, 4-5, Silva. A. J., Smith, A. M., & Giese, K. P. (1997). Gene targeting and the biology of learning and memory. Annual Review of Genetics, 31, 527546. Silver, L. M. (1995). Mouse genetics: Concepts and applications. Oxford, England: Oxford University Press. Spielman, R. S., McGinnis, R. E., & Ewens, W. J. (1993). Transmission test for linkage disequilibrium: The insulin gene region and insulindependent diabetes inellitus (IDDM). American Journal of Human Genetics, 52, 506-516. St. George-Hyslop, P., Haines, J., Rogaev, E., Mortilla, M., Vaula, G., Pericak-Vance, M., Foncin, J.-F., Montesi, M., Bruni, A., Sorbi, S., & et al. (1992). Genetic evidence for a novel familial Alzheimer's disease locus on chromosome 14. Nature Genetics, 2, 330-334. Sturtevant, A. H. (1915). Experiments on sex recognition and the problem of sexual selection in Dt'osophila. Journal of Animal Behavior, 5, 351-366. Suarez, B. K., Parsian, A., Hampe, C. L., Todd, R. D., Reich, T., & Cloninger, C. R. (1994). Linkage disequilibria at the D2 dopamine receptor locus (DRD2) in alcoholics and controls. Genomics, 19, 12-20. Suarez, B. K., & Van Eerdewegh, P. (1984). A comparison of three affected-sib-pair scoring methods to detect HLA-linked disease susceptibility genes. American Journal of Medical Genetics, IS, 135146. Sunohara, G. A., Roberts, W., Malone, M., Schachar, R., Tannock, R., Basile, V., Wigal, T., Wigal, S. B., Chuck, S., Moriarity, J., Swanson, J., Kennedy, J. L., & Barr, C. L. (in press). Linkage of the dopamine D4 receptor gene and attention .deficit hyperactivity disorder. American Journal of Psychiatry. Talbot, C. J., Nicod, A., Cherny, S. S., Fulker, D. W., Collins, A. C., & Flint, 1. (1999). High-resolution mapping of quantitative trait loci in outbred mice. Nature Genetics, 21, 305-308. Tang, Y.-P., Shimizu, E., Dube, G. R., Rampon, C., Kerchner, G. A., Zhuo, M., Liu, G., & Tsien, J. Z. (1999). Genetic enhancement of learning and memory in mice. Nature, 401, 63-69. Thapar, A., Holmes, J., Poulton, K., & Harrington, R. (1999). Genetic basis of attention deficit and hyperactivity. British Journal of Psychiatry, 174, 105-111. Tolman, E. (1924). Genetic differences in maze-learning ability in rats. In The 39th yearbook of the National Society for the Study of Education (Part I, pp. 111-119). Chicago: University of Chicago Press. Tryon, R. C. (1940). The inheritance of maze-learning ability in rats. Journal of Comparative Psychology, 4, 1-8. Tsien, J. Z., Chen, D. F., Gerber, D., Tom, C., Mercer, E., Anderson, D. J., Mayford, M., & Kandel, E. R. (1996). Subregion- and cell typerestricted gene knockout in mouse brain. Cell, 87, 1317-1326. Tsien, J. Z., Huerta, P. T., & Tonegawa, S. (1996). The essential role of the hippocampal CA1 NMDA receptor-dependent synaptic plasticity in spatial learning. Cell, 87, 1327-1338. Turkheimer, E., & Waldron, M. (2000). Nonshared environment: A theoretical, methodological, and quantitative review. Psychological Bulletin, 126, 78-108. Vandell, D. (in press). Parents, peers, and others. Developmental Psychol-

178.
Rowe, D. C. (1994). The limits of family influence: Genes, experience, and behavior. New York: Guilford Press. Rubinstein, M., Phillips, T. J., Bunzow, J. R., Falzone, T. L., Dziewczapolski, G., Zhang, G., Fang, Y., Larson, J. L., McDougall. J. A., Chester, J. A., Saez, C., Pugsley, T. A., Gershanik, O,, Low, M. J., & Grandy, D. K. (1997). Mice lacking dopamine D4 receptors are supersensitive to ethanol, cocaine, and methamphetamine. Cell, 90, 991-1001. Rubinzstein, D. C., Leggo, J., Coles, R., Almqvist, E.. Biancalana, V., Cassiman, J. J., Chotai, K., & et al. (1996). Phenotypic characterization of individuals with 30-40 CAG repeats in the Huntington disease (HD) gene reveals HD cases with 36 repeats and apparently normal elderly individuals with 36-39 repeats. American Journal of Human Genetics, 59, 16-22. Rutter, M., Dunn, J. F., Plomin. R., Simonoff, E., Pickles, A., Maughan, B., Ormel, J., Meyer, J., & Eaves, L. (1997). Integrating nature and nurture: Implications of person-environment correlations and interactions for developmental psychopathology. Development and Psychopathology, 9, 335-364. Rutter, M., & Plomin, R. (1997). Opportunities for psychiatry from genetic findings. British Journal of Psychiatry, 171, 209-219. Saudou, F., Amara, D. A., Dierich, A., LeMeur, M., Ramboz, S., Segu, L., Buhot, M. C., & Hen, R. (1994). Enhanced aggressive behavior in mice lacking 5-HT1B receptor. Science, 265, 1875-1878. Scarr, S. (1992). Developmental theories for the 1990s: Development and individual differences. Child Development, 63, 1-19.

ogy. van den flree, M. B., Svikis, D. S., & Pickens, R. W. (1998). Genetic
influences in antisocial personality and drug use disorders. Drug and Alcohol Dependence, 49, 177-187.

828

PLOMIN AND CRABBE porter gene (>A77) and attention deficit hyperactivity disorder in children: Heterogeneity due to diagnostic subtype and severity. American Journal of Human Genetics. Watanabe, T. K., Bihoreau, M. T,, McCarthy, L. C., Kiguwa, S. L., Hishigaki, H., & et al. (1999). A radiation hybrid map of the rat genome containing 5,255 markers. Nature Genetics, 22, 27-36. Watson, S. J., & AMI, H. (1999). Gene chips and arrays revealed: A primer on their power and their uses. Biotogical Psychiatry, 45, 533-543. Wehner, J. M., Bowers, B. J., & Paylor, R. (1996). The use of null mutant mice to study complex learning and memory processes. Behavior Genetics, 26, 301-312. Weiner, J. (1999). Time, love and memory: A great biologist and his quest for the origins of behavior. New York: Knopf.

Vitatema, M. H., Selby, C. P., Todo, T., Nima, H., Thompson, C., Fruechte, E. M., Hitomi, K., Thresher, R. ]., Ishikawa, T., Miyazaki, )., Takahashi, J. S., & Sancar, A. (1999). Differential regulation of mammalian period genes and circadian rhythmicity by cryptochromes 1 and 2. Proceedings of the National Academy of Sciences of the United States of America, 96, 12114-12119. Wachs, T. D., & Plomin, R. (1991). Conceptualization and measurement of organismenvironment interaction. Washington, DC: American Psychological Association, Wahlsten, D. (1990). Insemitivity of the analysis of variance to heredityenvironment interaction. Behavioral and Brain Sciences, 13, 109-161. Wahlsten, D. (1997). Leilani Muir versus the philosopher king: Eugenics on trial in Alberta. Genetica, 99, 185-198. Wahlsten, D. (1999). Single-gene influences on brain and behavior. Annual Review of Psychology, 50, 599-624. Waldman, I. D., Rowe, D. C., Abramowitz, A., Kozel, S. T., Mohr, J. H., Sherman, S. L., Cleveland, H. H., Sanders, M. L., Card, J. M. C., & Stever, C. (in press). Association and linkage of the dopamine trans-

Received June 9, 1999 Revision received June 15, 2000 Accepted June 26, 2000

Low Publication Prices for APA Members and Affiliates


Keeping you up-to-date. All APAFellows, Members, Associates, and Student Affiliates receiveas part of their annual duessubscriptions to the American Psychologist and APA Monitor, High School Teacher and International Affiliates receive subscriptions to the APA Monitor, and they may subscribe to the American Psychologist at a significantly reduced rate. In addition, all Members and Student Affiliates are eligible for savings of up to 60% (plus a journal credit) on all other APA journals, as well as significant discounts on subscriptions from cooperating societies and publishers (e.g., the American Association for Counseling and Development, Academic Press, and Human Sciences Press). Essential resources. APA members and affiliates receive special rates for purchases of APA books, including the Publication Manual of the American Psychological Association, and on dozens of new topical books each year. Other benefits Of membership. Membership in APA also provides eligibility for competitive insurance plans, continuing education programs, reduced APA convention fees, and specialty divisions. More information. Write to American Psychological Association, Membership Services, 750 First Street, NE, Washington, DC 20002-4242.

You might also like