Online Issues
<< All Back-issues
<< This Issue's Table of Contents
ILAR Journal V41(3) 2000
Mouse Behavioral Models in Biomedical Research
Gerald E. McClearn and David J. Vandenbergh
| Gerald E. McClearn, Ph.D., is Director of the Center for Developmental and Health Genetics and Evan Pugh Professor of Health and Human Development at the Pennsylvania State University, University Park, Pennsylvania. David J. Vandenbergh, Ph.D., is Assistant Professor in the Department of Biobehavioral Health at the Pennsylvania State University Center for Developmental and Health Genetics. |
Some animal research is comparative by intent, with interest in different species in their own rights. Much nonhuman animal research, however, is concerned with the use of the animal as a model system to describe, or test hypotheses about, some phenomenon of importance to human beings.
The fundamental rationale for use of animal models derives from the phyletic relatedness of living things. The branches and subbranches of the phyletic tree can be considered as general themes and variations on these themes. When we use some species other than our own, there is a hope that in spite of the variation, we share enough of the pertinent theme that information from the model system will illuminate something about ourselves. As the molecular exploration of the human genome and that of selected animal species has proceeded, there has been increasing evidence of genes shared between various species and humankind. This realization has strengthened enormously the logical base for expectations of successful application of animal models to complex human phenotypes.
There is a general expectation that the closer the phyletic relationship, the more informative will be the model. A species from the same family as Homo sapiens might be preferred to another from a different family from the same order, and so on. However, some themes might be so pervasive and fundamental that almost any species from the same phylum will be informative. For example, even the phylogenetically distant eukaryote Saccharomyces cerevisiae produces matches of up to 30% of human disease-related genes (Foury 1997). In the case of mice, the total increases to 81% (Bassett et al. 1997). Each organism brings with it unique characteristics that make it better suited for some lines of research than for others; we cannot expect there to be a gold standard reference species suitable for all purposes. Thus, from Foury's example (Foury 1997), the genes involved in nucleotide excision repair are clearly homologous between yeast and human beings. The power, not to mention speed, with which genetic experiments can be conducted in S. cerevisiae makes it an obvious favorite-candidate organism for studying excision repair. However, for an extreme and obvious contrary example, S. cerevisiae would likely be less useful in examining the role of specific neural receptors in alcohol consumption.
For myriad reasons, different research domains have come to focus on a few "species of choice." To a considerable extent, differential rates of progress in different research domains may be due, in fact, to the suitability of the species that have become the disciplinary favorites. In the early development of many biomedical and biobehavioral research fields, rodents were tried as convenient, easily accessible model systems. As information accumulated about these species, they became ever more logical as model choices because of the value of the increasing database that provided a context for interpretation of new results. It is arguable (Austad 1993) that from a phyletic perspective, rats and mice may not have been optimal initial choices for many of the phenomena in these research domains. However, the total cumulative knowledge about anatomy, behavior, biochemistry, endocrinology, genetics, immunology, pharmacology, physiology, and so on, as well as about husbandry of these animals, places them high on any list of species of current choice. Particularly in view of the burgeoning knowledge of the mouse genotype, it appears that there will be heavy reliance on this species for the discernible future. Of course, until the entire sequences of the mouse and human are known, it will not be clear exactly how extensive the genetic similarities are, and it is still an open question for any particular behavioral domain whether the mouse model is suitably homologic. It is clear, however, that the ability to compare the genomes of the mouse and human base by base will yield insights into many aspects of the genetic contributions to human health and behavior. This article is concerned with the use of the mouse in investigating alcohol-related processes that are presumed constituents of the immense human problems of alcohol use, alcohol abuse, and alcoholism.
Indicator Variables and the Total Model System
The term model has been variously defined. Sattler (1986), for example, observes that it is increasingly used to denote biologic generalizations over the entire range from theories, through laws and rules, to hypotheses. Models can also range from an abstract mathematical expression to the practical experimental system. It is the latter type of model we address herein. Common to all of these definitions of models is the understanding that the essence of models is a simplification or abstraction of the phenomenon being modeled. The sometimes-encountered plea for a complete and comprehensive model involves a contradiction in terms.
Following decades of biochemical, pharmacologic, physiologic, behavioral, and social research, the problems of alcohol use and abuse and alcoholism can properly be described as comprising a complex system. Much attention has been devoted recently to properties of systems and to emergent phenomena that may not be derivable from knowledge about the elements of the system. Among the attributes of complex systems from which our models are abstracted, several have important implications for the information that we can derive from the particular variables that we choose. In the first place, such systems are often overdetermined and replete with feedback mechanisms, alternate pathways, and compensatory mechanisms. In such circumstances, Newtonian notions of unidirectional causality can be deceiving, and analyses appropriate to circular or network causal fields become appropriate. It is also characteristic of open complex systems that they change over time.
The implications of the structure and dynamics of complex systems are becoming increasingly evident in the biologic, biosocial, biobehavioral, and biomedical sciences. (A noteworthy recent example is the establishment by the Journal of Molecular and Cellular Biology of a new section, "Mammalian Genetic Models with Minimal or Complex Phenotypes.' In part, this section will advance understanding of the compensatory mechanisms that complicate interpretation of the consequences of genetic manipulations such as "knockout" genes with "no" phenotype.) Some implications of these system characteristics for the measurements we make in alcohol-related animal model research are explored in the remainder of this article (see also McClearn 1993).
There is much more to the specification of a model system than the species chosen to serve as a stand-in for humanity. The establishment of an experimental model is an attempt to isolate a part of a complex system to study the effects of a particular element or elements of the system. The process can be characterized in the familiar terms of experimental design: One or a few elements are selected to be the manipulated, independent variables; other variables are subjected to control by fixation or randomization; one or more others are identified as the outcome or dependent variable(s) to be measured; and still others are assessed as covariates important to the final interpretation of results.
Thus, the "animal model" consists not only in the species employed but also in the totality of the assessment situation, with all of its manipulated, controlled, and measured variables (and all of the overlooked and ignored ones as well). As for most complex phenotypes, the investigator has numerous choices with respect to all of the variable types, and the particular assemblage chosen may yield idiosyncratic results relative to others that differ only in some detail.
A striking illustration of the embeddedness of a variable in its context has been provided by Crabbe et al. (1999), who measured a panel of variables in three different laboratories. The measures included indices of locomotor activity, anxiety, coordination, learning, effects of cocaine, and alcohol preference. Extraordinary efforts were made to standardize the conditions of testing, including the use of the same eight genotypically specified groups of mice. The effect of genotype was predominant in accounting for variance for most of these measures. However, significant effects were also attributable to the different laboratories for most variables, despite the high degree of environmental control. It appears that subtle environmental nuances that can influence these measurements remain unknown, or their specification is imprecise.
With specific reference to modeling an alcohol-related phenotype (e.g., voluntary alcohol consumption), a decision must be made about the sort of response that will be required-a simple selection of liquids from bottles offered in a cafeteria mode or pressing of a bar to gain access to a liquid cylinder or drinking pan. If the former response is chosen, the following information must be determined: (1) number of bottles to be presented and in what pattern; (2) concentrations of alcohol; (3) whether there should be a water-only control or one or several solutions of alcohol; (4) whether all bottles should be refilled simultaneously regardless of remaining level; (5) whether the positions of the various bottles should be shifted; (6) how often observations should be made; (7) husbandry conditions; (8) food to be supplied and whether it should be available ad libitum; (9) type of bedding; (10) laboratory and colony temperature, humidity, lighting conditions, decibel level, air turnover frequency, and level of permissible traffic of laboratory personnel; (11) schedule for cage changing; (12) required general health conditions; (13) precautions against any liver-damaging pathogens; (14) age of animals; and (15) whether one or both sexes should be examined.
It is no surprise that many varied measures of voluntary alcohol consumption have been employed in the mouse. We cite here some examples from two different measures: alcohol preference and alcohol acceptance. The version of alcohol preference in use in the authors' laboratories offers the animals a choice between two 25-mL graduated cylinders, one containing tap water and the other containing a 10% solution of ethanol (v/v) in tap water. Observations are made over a period of 14 or 15 consecutive days, with the position of the cylinders rotated on a regular basis to obviate position preferences. This basic design has been used widely and has provided highly reproducible differences among inbred strains in numerous laboratories. A measure of less widespread use is that of "alcohol acceptance," in which consumption from a single cylinder containing 10% solution is recorded for 24 hr after 24 hr of water deprivation.
In addition to choices about these experimental, husbandry, and organismal variables, crucial decisions must be made about the gene pool to be sampled from the chosen species. The range of choices extends from wild-trapped animals, through systematically generated genetically heterogeneous stocks, to inbred strains. In selecting animal groups for a research program, the central issues are representativeness and reproducibility. Briefly, the relatively uniform and constant genotypes of inbred strains make possible the standardization of genotypes across laboratories and across time. This valuable attribute is purchased at the cost of reduced generalizability: The genotype of an inbred strain is only one of uncounted possible genotypes of that species. Systematically generated and maintained genetically heterogeneous stocks offer enlarged scope for generalization, but no single genotype within such a group is replicable (although the genetic parameters of the group can be replicated quite well). These issues are amplified in recent discussions by McClearn and Hofer (1999a,b) and Miller et al. (1999).
Complexity of the Alcohol Domain
The various disciplines that have turned their attention to alcohol problems have utilized a huge range of concepts. Pharmacologists, for example, deal with uptake, distribution, disposition, sensitivity, tolerance, and dependence. Physiologists may measure impairment of physiologic processes under different loads of alcohol. Pathologists may be concerned with evidence of cirrhosis. Behavioral and social scientists provide other terms: peer groups, family values, religious affiliation, reinforcing properties, hedonic tone, elasticity of alcohol as an economic product, role of support groups in therapy, effects of pattern of availability on use and abuse measures, impairment of motor skills and of judgment under the influence of alcohol, and so forth. It is clear that the domain of alcohol studies does indeed contend with very complex systems, and no single measure or model can capture all of the aspects of the latent processes in which we are interested. The old but apt simile of the blind men assessing the elephant should be constantly in our minds. The use of any of our models, human or animal, is like grasping the elephant at one point only. Much groping is required to obtain an adequate representation.
Criteria of Model Systems
Among other criteria, there are two that are critical for evaluating the adequacy of a model system: validity and reliability.
Validity
In very nontechnical terms, validity is the degree to which the model system actually measures what it was designed to measure. As we have noted, we cannot require that it assess the totality of the target phenomenon. Such total assessment would be impossible and would not then be a model. However, the model should assess some features of the targeted phenomenon. We need to be assured that it is actually touching the elephant somewhere. Validity is thus at the very core of the relevance of the model.
Evaluation of validity can be accomplished in a variety of ways (e.g., see McClearn 1988). In the unlikely event that an accepted gold standard does exist, and the model system is sought because it is easier, cheaper, or quicker to measure, then a deliberate study can assess the degree to which the model correlates with the gold standard. In most cases, such a standard will not exist. Indeed, our identification of the target phenomenon will gradually change and (hopefully) improve--as different model systems are tried out--and validity can become a bootstrap concept as we relate the different models to each other and a consensus coalesces. Ghiselli et al. (1981) noted as a general principle, "The process of defining our variables and devising operations that we can use in the description of individual differences is a never-ending one. Not only is there an interaction between the definition and the devising of operations, but also the results obtained from our operations give us new insights into the nature of our variable so that we redefine it and modify our operations. The development of ways for describing individuals, then, is a dynamic process" (p. 15).
Just so do our ideas of the structure of the problem of alcoholism grow and change. However, it is clear that the nosology of alcohol-related processes in human beings, as well as in the animal model systems, is still a work in progress with researchers and clinicians struggling to find more precise and functional definitions of alcohol problems. This topic is one of the most important in the alcohol research arena. Nosologic categories presumably reflect differential etiology or pathogenesis, and the identification of taxonomic subgroups within alcoholism is an essential step toward differential intervention. The importance of tailoring treatments or preventive measures to the subtype is obvious to all. The point can be illustrated by a moment's reflection that the effective treatment for phenylketonuria would be not only ineffective, but actually detrimental, for other subtypes of mental retardation.
There have been continuing attempts at differential definitions in alcoholism. Particularly significant was the work of Jellinek (1960), whose influential categorization of alcoholism into subtypes, or "species" as he labeled them, was based both on severity of drinking as well as the severity of the consequences at a societal level. Thus, the three predominant species were alpha alcoholism, caused only by psychological dependence on alcohol to ameliorate emotional or bodily pain; beta alcoholism, in which there are significant internal manifestations such as polyneuropathy, gastritis, or cirrhosis of the liver, even in the absence of psychological or physical dependence; and gamma alcoholism, with evidence of physical adaptation to prolonged exposure to alcohol including acquired tissue tolerance, withdrawal or craving symptoms, and loss of control (Jellinek 1960).
The underlying motivation of Jellinek's classification was the laudable, pragmatic one of providing a basis for differential treatment based on observable medical and social measures; such a nosology may not be optimal for assessing basic biologic mechanisms. Recently, Cloninger (1987) presented a more salient classification in which genetic information is important. This scheme separated alcoholism into two types: type 1, characterized by adult onset without associated antisocial behavior and with little evidence of heritability; and type 2, teenage onset associated with antisocial behavior and a stronger genetic "loading." The characterization of alcoholism into subtypes with potentially different biologically active mechanisms represents a refinement that may aid in understanding the root causes of alcoholism, as opposed to its outcomes.
To summarize, the issue of the validity of any chosen measure is intimately entwined with our conceptualization of the thing or process we want to measure. At any particular time, we will be able to obtain some level of validity by comparing our measure with the consensual concept of the moment; but the concept itself will be evolving as different measures contribute to the total body of empirical information.
Reliability
The concept of reliability relates to accuracy of measurement. Briefly stated, the concept posits the existence of a true score but assumes that there are inherent sources of error in the operations of measurement so that the value of the measurement on any given occasion is drawn from an array of potential measures distributed, presumably, around the true score value. There is thus an aura of uncertainty about any measured value, and the more reliable measures are those with the narrowest error distribution. Accumulated experience in any scientific domain generates the lore of good experimental practice to be followed in research in the area--the acceptable conditions of the model system that minimize the error variance. Inevitably, of course, different measures will have different reliabilities.
Temporal Stability
It is traditional to view variability from time to time as a manifestation of error, as noted above. However, complex systems are maintained dynamically, not statically (Yates et al. 1972). Thus, we may expect there to be temporal fluctuations in values of measured variables and in their relationships to other variables. There is increasing interest in considering intraindividual variability of this kind as fluctuation due not to error but instead, to biologically lawful processes that change the "true score" from day to day, hour to hour, or even moment to moment (McClearn and Hofer 1999a; Nesselroade 1991). There is at present a dearth of alcohol-related data on fluctuance, but illustrations are given in Figure 1 that show consecutive alcohol intake scores of several samples of mice of an F2 between C57BL/6 and DBA/2 (data from study of Tarantino 1998) selected to display the wide variability among the animals in their intra-individual variability. The record for four animals is shown in Figure 1A. Two animals (47 and 121) have intermediate overall levels of consumption but display quite wide fluctuations over the observation period, and two (56 and 90) are relatively stable with low levels of consumption. In Figure 1B, another two animals (1 and 20) are portrayed with low and relatively stable consumption levels: one (4) that increases for approximately 1 week and remains fairly stable thereafter, and another (15) that consistently increases over the observation period. In this group, early observations are clearly not highly predictive of later consumption levels. The same point is illustrated in Figure 1C, where highly diverse initial levels converge, with differing fluctuance (e.g., compare 99 and 113) on the same general intake level late in the measurement period. These individual differences in trajectories, and in variability around the trajectories, indicate that any selection of a few days of observation from these series would not be completely representative. Of course, we do not really know the optimal duration of an observation period (e.g., 1 month or 1 year). The only answer is a pragmatic one.
The processes that lead to the increases and decreases of voluntary alcohol intake are, of course, not revealed in these data, but a reasonable conjecture is that feedback processes from previous consumption may have some influence. That such (presumed) processes can affect alcohol consumption was shown by McClearn (1972), who examined the characteristics of the "ethanol intake control system" in C57BL/6 mice by altering concentrations of the alcohol solutions offered (intraperitoneal injections bypassing the normal ingestional route) and forced deprivation or overfeeding of alcohol. All results from these manipulations are consistent with a negative feedback model that operates to maintain intake levels to a set-point (or set-zone). An outcome pertinent to the present topic is the reduction in voluntary intake in C57BL/6 mice after a 6-day period in which only ethanol solution was available (Figure 2). During these 6 days, the animals consumed nearly twice the volume of alcohol solution as they did during a two-bottle choice period. When two-bottle choice was reinstated, voluntary consumption dropped to approximately one-half the normal choice level, with gradual recovery to this normal level over several subsequent 3-day periods. A similar dynamic adjusting process may well be influencing day-to-day differences in alcohol intake measures of individual mice.
If the logic of a study requires an estimate of the average true score over some period of time, a single measurement occasion might be inadequate for those variables with a high fluctuance level. Observe that these variables are not necessarily "bad" variables. Fluctuance might better be interpreted as revealing a measure's sensitivity to environmental nuances or to subtle organismic changes rather than its proneness to error. In any case, fluctuance will emerge in the error term, involving the usual considerations of required sample size. Obtaining the mean of repeated measurements on each animal may offer an attractive solution in the case of some variables, but this recourse is not possible, of course, for variables requiring sacrifice or if one measurement seriously biases later ones.
Indeed, a possible provocative point can be advanced: If the phenomenon being modeled has been of much significance in natural selection, then a homeostatic feedback control system is likely to exist. The default expectation should be that fluctuance will be present, and it is incumbent on the experimenter to demonstrate whether its magnitude is sufficiently small that a single measure will suffice for the research purposes at hand.
Development
Biologic systems are particularly characterized by developmental processes. An implication for present purposes is that the set of variables being employed in a model system may be acting and interacting in quite different ways at different life stages. In effect, this means that any chosen variable can have a different "meaning" at different ages. As an example, Kakihana and McClearn' s (1963) developmental studies demonstrated continuous differentiation among several inbred mouse strains in hypnotic dose sensitivity from 4 to 16 weeks of age and a rather abrupt decline in alcohol preference in BALB/c mice at about 9 weeks of age. The latter result is shown in Figure 3. Although the consumption level fluctuates widely at 8 to 9 weeks, the difference between the levels before and after that period is evident.
In these examples, only polygenic complexes are implicated. The prospects of examining individuated loci in the developmental context are now promising. One pertinent example is the demonstration that a quantitative trait locus (QTL1) on mouse chromosome 15 influences alcohol acceptance at about 100 days of age but has no detectable influence at about 300 days of age (McClearn et al. 1998). These studies suggest that marked differences in alcohol consumption among strains may be influenced by one set of genes at one particular age and with a second, partially overlapping set of genes at another age. The recognition that these changes are occurring in adulthood is a departure from a common presumption that all development takes place early in life.
Multivariate Orientation
The fact that single measures are generally incapable of capturing all of the meaning of a complex latent variable implies clearly that researchers must be prepared to contemplate multiple, partially overlapping indicator variables. Such complex situations are the province of multivariate statistical techniques.
In the absence of an all-encompassing gold standard, the meaning of any empirically measurable indicator variable will be defined ultimately by its relationships to other variables. Thus, a central research objective should be to examine these relationships systematically. It is obviously not required that each and every animal model research utilize multiple independent, control, and outcome variables. However, researchers should be receptive to the general perspective of such research that assumes a complex associational structure among variables. Many experimentalists view computing a correlation as a weak method because a high correlation does not imply a cause-and-effect association between variables. This view may be false in a complex system. Nevertheless, if an association cannot be shown between variables measured in a relatively intact complex system, there may be little point in pursuing additional experiments. The old adage that "correlation does not imply causation" is true, but if there is causation, there will be correlation.
Perhaps most important is that no single variable should be expected to indicate everything about a system; its meaning will ultimately be determined by its relationships to other variables. One example is provided by the early attempt by one of us (G.E.M.) to devise a test of voluntary alcohol consumption shorter than the 14- or 15-day "alcohol preference" test sessions with choices between tap water and 10% ethanol solution. "Alcohol acceptance" was measured as the consumption of a 10% ethanol solution during a 24-hr period after a previous 24-hr period when no liquid was available to the animals. Alcohol acceptance thus assessed "willingness" to consume ethanol when thirsty. Initial results (McClearn 1968) revealed essentially the same ordinal relationships among six inbred strains on the alcohol preference and alcohol acceptance measures. The conclusion that the tests were interchangeable was wrong, however. Sometime later, we assessed the relationship using the more appropriate procedure of correlations within a genetically heterogeneous F2 (derived from C57BL and C3H strains) (Anderson et al. 1979). The correlation was a mere 0.27. Thus, with only about 9% shared variance, it is clear that these measures are probing quite different parts of the elephant. Tarantino (1998) found a similar level of association (0.25 in males and 0.20 in females) between the two measures in an F2 between C57BL/6 and DBA/2 strains. Rodriguez et al. (1994) found a correlation among recombinant inbred strain means for these measures of 0.24, which indicates that there is a partial overlap of the gene sets influencing the two phenotypes.
Another example can be taken from the work of Erwin et al. (1980) who sought to determine the interrelationships among measures of alcohol preference, sensitivity to alcohol, and biochemical values. A selection from the results of two independent samples of HS mice (a genetically heterogeneous stock, of choice for correlational studies; see McClearn and Hofer 1999b) is shown in Table 1. As described in the table footnote, significant correlations obtained from study 1 involving 20 HS males and females appear below the diagonal, and results from study 2 on 48 HS males and females appear above the diagonal.
Although there are some differences in results between the two studies, likely attributable to the small sample size of the first study, there is solid agreement that alcohol preference (as here measured) is positively related to alcohol tolerance (as here measured) and negatively related to soluble brain protein (as here measured). The indicated relationships are robust, indicating that about one quarter of the variances of alcohol preference and acute tolerance are shared, with a similar result for alcohol preference and soluble brain protein. Results of this sort provide glimpses of the structure of the system and the relationships within it.
These examples are within the category of quantitative genetics; recent advances in genome mapping have made possible the identification of chromosomal location of genes (QTLs) with a detectable effect on phenotypes that are influenced by polygenic systems. Scientists have been quick to exploit these new opportunities in alcohol studies, and several examples are pertinent to the present topic.
Quantitative Trait Loci
The genetic components of several alcohol-related behaviors have been measured extensively in mice as a means to begin the identification of genes related to the biochemical and pharmacologic effects of ethanol. It is important first to confirm the presence of a true QTL in a region of a chromosome and then to narrow the size of the QTL region to clone the gene. An example of a well-established QTL is for alcohol acceptance on chromosome 2 (30-47 cM) detected in recombinant inbred strains (Rodriguez et al. 1995), in F2s (Tarantino 1998), and by genotypic selection (Tarantino 1998).
Although a unique QTL pattern has emerged for each alcohol-related phenotype that has been studied, there is clear accumulating evidence for QTLs that are associated with more than one phenotype. For example, several studies have indicated evidence for such a QTL in a region of chromosome 15 for alcohol preference (Belknap et al. 1997; Gehle and Erwin 1998; Rodriguez et al. 1995), alcohol acceptance (Rodriguez et al. 1995; Tarantino et al. 1998), alcohol-induced loss of righting response (Markle et al. 1997), and ethanol withdrawal (Crabbe 1998; Tarantino 1998). One or several genes at this locus that contribute to differences in multiple aspects of alcohol-related behaviors may enable us to understand the basic biochemical pathways of alcohol's effects on the body. It is clear that the capability of individuating QTLs from polygenic sets will provide increasingly powerful tools for describing the architecture of the complex systems of alcohol-related processes.
The complex relationship between measures of a behavior and the genes underlying the behavior is further illustrated by the results of QTL analysis of cocaine-related behaviors (Jones et al. 1999). The locomotor effects of different doses of cocaine are related to different QTLs, implying that as the dose of cocaine increased, new genes are involved and others diminish in their effect on the behavior. We know that locomotion increases with an increasing dose of cocaine until a plateau is reached and then diminishes as the dose is increased further, generating an inverted "U" shaped curve when plotting dose versus locomotion. However, the doses used in the study by Jones and colleagues (5-30 mg/kg body weight) were on the increasing slope of the curve, implying that genetic changes are evident without massive doses of cocaine where nonspecific pharmacologic effects are seen. This area of research has not yet been explored in alcohol-related behaviors, in which similar complex effects such as increased locomotion at low dose and decreased locomotion at high dose have been detected (Crabbe et al. 1982).
Summary and Conclusions
Problems with alcohol use, alcohol abuse, and alcoholism clearly constitute complex systems, with numerous influential variables identified with the basic genetic, physiologic, biochemical attributes of the individual and the broad gamut of environmental circumstances in which the individual lives. We believe that the adequacy of marker variables in animal models of complex systems such as this can be evaluated only in the total context of the model system being employed. The results of Crabbe et al. (1999) highlight the sensitivity to environmental context. The differing effects of knocked-out genes on different genetic backgrounds illustrate the critical importance of the genetic milieu.
Several issues of generalizability are pertinent to this evaluation: (1) the generalizability from the particular configuration of the particular model system to the process being modeled in the animal; (2) the generalizability of results obtained from animals of the particular gene pool sampled to their species at large; and (3) the extent to which the processes of the mouse are homologic to the human processes being modeled. Because no single marker variable is adequate to characterize such a system, it is desirable to gather information about as many partially overlapping probe measures as possible. In the absence of natural gold standard measures, our concepts of alcohol problems will emerge only from the convergence of information derived from many models. To facilitate this convergence, associational research is required. Systematic exploration of interrelationships among popular marker variables will begin to define the dimensions/architecture/structure of the complex system under consideration so our grasp on the elephant can be expanded.
For many biobehavioral and biomedical purposes, the target phenomenon of the model will have a complex genetic architecture, with a large number of genes affecting the phenotype. For these circumstances, the quantitative genetic model is the apposite one, yielding assessments of the relative contributions of genetic and environmental factors to the variance of the phenotype (Falconer and Mackay 1996; Plomin et a1.1997) and providing the logical underpinnings of the many programs that have generated alcohol-relevant models through selective breeding (Crabbe and Harris 1991). Similarly, molecular genetics has made stupendous strides in the identification and molecular characterization of single genes that have individually detectable effects on a wide array of medical and other phenotypes. Furthermore, the generation of marker genotypes now permits the chromosomal localization of some of the hitherto anonymous genes (QTLs) of polygenic systems. Thus, with both the quantitative and molecular genetic tools, identified or located individual loci and collectivities of anonymous loci can be utilized as key variables in the total model systems employed in the study of alcohol problems.
Although the general tenor of this discussion has been a cautionary one, the intent has not been to lament the limitations of the variables we deploy in alcohol research. Indeed, most of the differences among inbred strains or selectively bred lines in alcohol-related phenotypes compare favorably with other research domains in terms of replicability across laboratories. (Alcohol preference was one of the two measures in the study of Crabbe et al. [1999] not to show a significant influence of laboratory site.) Rather, our objective has been to remark on the expanded opportunities that emerge from a multivariate systems perspective.
Acknowledgments
Dr. Lisa Tarantino kindly gave permission for use of her data for Figure 1. The authors' current alcohol research is supported by grant AA08125 from the National Institute on Alcohol Abuse & Alcoholism.
References
Anderson SM, McClearn GE, Erwin VG. 1979. Ethanol consumption and hepatic enzyme activity. Pharmacol Biochem Behav 11:83-88.
Austad SN. 1993. The comparative perspective and choice of animal models in aging research. Aging Clin Exp Res 5:259-267.
Bassett DE Jr, Boguski MS, Spencer F, Reeves R, Kim S, Weaver T, Hieter P. 1997. Genome cross-referencing and XREFdb: Implications for the identification and analysis of genes mutated in human disease. Nature Genet 15:339-344.
Belknap JK, Richards SP, O'Toole LA, Helms ML, Phillips TJ. 1997. Short-term selective breeding as a tool for QTL mapping: Ethanol preference drinking in mice. Behav Genet 27:55-66.
Cloninger, CR. 1987. Neurogenetic adaptive mechanisms in alcoholism. Science 236:410-416.
Crabbe JC. 1998. Provisional mapping of quantitative trait loci for chronic ethanol withdrawal severity in BXD recombinant inbred mice. Pharmacol Exp Therapeut 286:263-271.
Crabbe JC, Harris RA, eds. 1991. The Genetic Basis of Alcohol and Drug Actions. New York: Plenum Press.
Crabbe JC Jr, Johnson NA, Gray DK, Kosobud A, Young ER. 1982. Biphasic effects of ethanol on open-field activity: Sensitivity and tolerance in C57BL/6N and DBA/2N mice. J Comp Physiol Psycho196:440-451.
Crabbe JC, Wahlsten D, Dudek BC. 1999. Genetics of mouse behavior: Interactions with laboratory environment. Science 284:1670-1672.
Erwin VG, McClearn GE, Kuse A. 1980. Interrelationships of alcohol consumption, actions of alcohol, and biochemical traits. Pharmacol Biochem Behav 13:297-302.
Falconer DS, Mackay TFC. 1996. Introduction to Quantitative Genetics. 4th ed. Essex, UK: Longman Group, Ltd.
Foury F. 1997. Human genetic diseases: A cross-talk between man and yeast. Gene 195:1-10.
Gehle VM, Erwin VG. 1998. Common quantitative trait loci for alcohol-related behaviors and CNS neurotensin measures: Voluntary ethanol consumption. Alcohol Clin Exp Res 22:401-408.
Ghiselli EE, Campbell JP, Zedeck S. 1981. Measurement Theory for the Behavioral Sciences. San Francisco: Freeman.
Jellinek EM. 1960. The Disease Concept of Alcoholism. New Haven, CT: College and University Press. p 36-38.
Jones BC, Tarantino LM, Rodriguez LA, Reed CL, McClearn GE, Plomin R, Erwin VG. 1999. Quantitative-trait loci analysis of cocaine-related behaviours and neurochemistry. Pharmacogenetics 9:607-617.
Kakihana R, McClearn G. 1963. Development of alcohol preference in BALB/c mice. Nature 199:511-512.
McClearn, GE. 1968. The use of strain rank orders in assessing equivalence of techniques. Behav Res Methods Instru 1:49-51.
McClearn, GE. 1972. The genetics of alcohol preference. In: Proceedings of the International Symposium on Biological Aspects of Alcohol Consumption. Helsinki, Finland: The Finnish Foundation for Alcohol Studies. p 113-119.
McClearn GE. 1988. Strategies for biomarker research: Experimental and methodological design. Exp Gerontol 23:245-255.
McClearn GE. 1993. Genetics, systems, and alcohol. Behav Genet 23:223-230.
McClearn GE, Hofer SM. 1999a. Genes as gerontological variables: Uniform genotypes. Neurobiol Aging 20:95-104.
McClearn GE, Hofer SM. 1999b. Genes as gerontological variables: Uses of genetically heterogeneous stocks. Neurobiol Aging 20:147-156.
McClearn GE, Tarantino LM, Hofer SM, Jones B, Plomin R. 1998. Developmental loss of effect of a chromosome 15 QTL on alcohol acceptance. Mamm Genome 9:991-994.
Markle PD, Bennett B, Beeson M, Gordon L, Johnson TE. 1997. Confirmation of quantitative trait loci for ethanol sensitivity in long-sleep and short-sleep mice. Genome Res 7:92-99.
Miller RA, Austad S, Burke D, Chrisp C, Dysko R, Galecki A, Jackson A, Monnier, V. 1999. Exotic mice as models for aging research: Polemic and prospectus. Neurobiol Aging 20:217-231.
Nesselroade JR. 199 I. Interindividual differences in intraindividual change. In: Collins LM, Horn JL, eds. Best Methods for Analysis of Change. Washington, DC: American Psychological Association. p 92-105.
Plomin R, DeFries JC, McClearn GE, Rutter M. 1997. Behavioral Genetics. 3rd ed. New York: Freeman.
Rodriguez LA, Plomin R, Blizard DA, Jones BC, McClearn GE. 1994. Alcohol acceptance, preference, and sensitivity in mice: I. Quantitative genetic analysis using BXD recombinant inbred strains. Alcohol Clin Exp Res 18:1416-1422.
Rodriguez LA, Plomin R, Blizard DA, Jones BC, McClearn GE. 1995. Alcohol acceptance, preference and sensitivity in mice. II. Quantitative trait loci mapping analysis using BXD recombinant inbred strains. Alcohol Clin Exp Res 19:367-379.
Sattler RF. 1986. Biophilosophy. Berlin: Springer-Verlag.
Tarantino, LM. 1998. Quantitative Trait Loci for Alcohol-Related Behaviors in Mice. Unpublished doctoral dissertation. University Park, PA: The Pennsylvania State University.
Tarantino LM, McClearn GE, Rodriguez LA, Plomin R. 1998. Confirmation of quantitative trait loci for alcohol preference in mice. Alcohol Clin Exp Res 22:1099-1105.
Yates FE, Marsh DJ, lberall AS. 1972. Integration of the whole organ-ism--A foundation for a theoretical biology. In: Behnke JA, ed. Challenging Biological Problems: Directions Toward Their Solution. Fairlawn, NY: Oxford University Press. p 110-132.
1Abbreviation used in this article: QTL, quantitative trait locus.
Table 1 Correlations among selected alcohol-related phenotypesa
| 1 | 2 | 3 | 4 | 5 | |
| 1 Preference | x | 0.51 | -0.46 | ||
| 2 Acute tolerance | 0.47 | x | -0.34 | ||
| 3 CNSb sensitivity | x | ||||
| 4 Brain aldehyde reductase | -0.42 | -0.51 | x | ||
| 5 Soluble brain protein | -0.67 | -0.63 | x |
aValues are partial correlation coefficients with the effects of sex, age, and body weight removed. Only values with p < 0.05 are shown. Significant correlations obtained from study 1 involving 20 HS males and females appear below the diagonal, and results from study 2 on 48 HS males and females appear above the diagonal. Alcohol preference was measured as the ratio of 10% alcohol solution to total liquid consumption over a 15-day test period. Acute tolerance was assessed as the change in central nervous system sensitivity on a second trial immediately after the first. Central nervous system sensitivity was measured by ability to remain on a fixed wooden rod after intraperitoneal injection of 2 g/kg ethanol. For the total model systems, see Erwin VG, McClearn GE, Kuse A. 1980. Interrelationships of alcohol consumption, actions of alcohol, and biochemical traits. Pharmacol Biochem Behav 13:297-302. bCNS, central nervous system.
Figure 1 Ethanol intake scores (mL ethanol solution consumed) of four F2 mice (from C57BL/6 and DBA/2 progenitors) with consumption patterns (A) high and fluctuating or low and consistent, (B) slowly diverging, or (C) slowly converging. These patterns illustrate the wide range of interindividual differences in intra-individual variability. Observations were made over 15 days of an animal's intake from a bottle containing 10% alcohol in the presence of a second bottle containing only water. All observations are for 24-hr periods, except observations 5 and 10, which are means of three 24-hr periods (Friday-Monday).
Figure 2 Mean daily alcohol consumption (ml of ethanol) from a 10% ethanol solution by 15 C57BL/6 mice. During days 1-3, a choice was given by providing a second bottle containing only water. The water-only bottle was removed during days 4-9 and returned on day 10. Note the decreased consumption during the second choice period. (Redrawn from data shown in McClearn 1972.)
Figure 3 Age-related changes in the mean weekly alcohol preference (volume of 10% ethanol consumed/volume of water consumed) in cross-sectional samples of BALB/c mice. Each point represents the mean of a 2-wk test beginning during the week indicated on the x-axis. Successive sample sizes were 12, 10, 10, 11, 11, 10, 11, 10, 11, 10, 19, 10, 3, and 3. (Redrawn from data shown in Kakihana and McClearn 1963.)
Copyright © 2008. National Academy of Sciences.
All rights reserved.
500 Fifth St. N.W., Washington, D.C. 20001.
Terms of Use and Privacy Statement