Drug and Alcohol Dependence, 4 (1979) 61 @ Elsevier Sequoia S.A., Lausanne -- Printed
GENETICS
GERALD Institute (U.S.A.)
AND ETHANOL
E. McCLEARN
61
in the Netherlands
TOLERANCE
and SALLY
for BeharJioral Genetics,
- 76
M. ANDERSON
University
of Colorado,
Boulder,
Colorado
80309
Summary This paper reviews some of the research on genetic bases of individual differences in ethanol tolerance of mice conducted at the Institute for Behavioral Genetics and at its predecessor laboratory at the University of California, Berkeley. Tolerance is, of course, a complex concept. Theoretical distinctions are made between tachyphylaxis and more slowly acquired tolerance and between dispositional and tissue tolerance. Pragmatically, a variety of measures (such as locomotor activity, sleep time, hypothermia) can be used to define these processes, and the different indices may yield quite different results even when they presumably indicate the same process. It is clear that an understanding of genetic influence in “ethanol tolerance” will require wide sampling of this complex domain. The work described here represents only a beginning, but it may illustrate the general approaches that are available for addressing the issue.
The relatively young literature of animal pharmacogenetics has been derived in large part from studies of inbred mouse strains. There are two principal reasons why this has been the case. First, inbred mice are easily and quickly available. Second, the methodology of strain comparisons is conceptually straightforward. As a rough approximation, it can be easily understood that all mice within a given strain are genetically alike and are different from members of any other inbred strain. Therefore, although the genetic information derived from strain comparisons is of a low order, the use of inbred strains does constitute a very effective means of controlling the characteristics of the animals employed in pharmacological investigations.
Inbred mouse strain differences
in alcohol-related
behavior
The work of our laboratory on the genetics of alcohol-related behavior and its mechanisms, like that of many other research groups, began with
0
I 35 ETHANOL
Fig. 1. Dose-response mouse strains.
curves
DOSAGE
2.70 (gmJkg.1
for influence
of ethanol
on arena activity
of several inbred
inbred mice. Our initial results revealed a large and stable strain difference in voluntary alcohol consumption between C57BL and DBA/B mice. The former display moderate to high ethanol ingestion in a two-bottle choice situation, while the latter are enthusiastic abstainers [l] . Several subsequent studies have characterized these differences in further detail [2] . However, of more direct relevance to the topic of this meeting is the research undertaken on strain differences in sensitivity to ethanol. We first approached the issue of first-dose, or naive, tolerance by recording changes in activity subseqi!+‘*. .‘?a,:bodinhalation of ethanol vapor. Just about every possible out,” j> e-1:the activity of C57BL mice was lowered; the activity of .lged, and the activity of C3H mice was increased. Using ns of ethanol, so as to be able to control dosage more ‘halation procedures, these strains and others were ., on activity of ethanol at 1.35 g/kg and 2.70 g/kg ,ults are shown in Fig. 1. It may be seen that the low dose tiivity in the BALB/c, C3H and A strains, but had no effect on L.. ./b;\ or C57BL mice. The high dose resulted in either no further change (in the case of the C3H and A strains) or a decrease in activity (in the BALB/ c, DBA and C57BL strains). These results clearly demonstrate the importance of the genotype of (7 ‘11 in pharmacological research. Not only does the dose-re1 :liffer quite markedly among strains, but the characteriza* s behaviorally depressing, exciting, or neutral, is also straindept ir4,t 1:. Stran. ilfferences have also been apparent in our research utilizing dura tion of loss of the righting response (“sleep time”) to investigate genetic dif-
63
ferences in the hypnotic effect of ethanol. C57BL and C3H/2 mice have much shorter sleep times than do DBA/B or BALB/c mice after an intraperitoneal injection of 3.4 g ethanol per kg [4]. From these and similar researches, in our own laboratory and others, the accumulated evidence makes it clear that the genotype of the research animal is a critical consideration. In brief, the paramount lesson is that investigators must be very modest indeed concerning the breadth of generality of results obtained with a particular group of animals. Instead of regarding a given outcome as general until proven idiosyncratic, we should regard results as idiosyncratic until proven general. At first glance, this attitude may appear to be quite negative. Where is scientific lawfulness if strains can be found to differ in response to some particular manipulation ? There is a positive aspect to the situation, however, which far outweighs the negative one. The very differences which prevent us from immediately assuming that we have tapped an eternal verity when we obtain a research outcome with a particular sample of animals offer us the tools to explore more deeply the mechanisms underlying those differences, where we might indeed find the lawful generalities to reside.
A polygenic
schema for pharmacology
With the accumulation of evidence of strain differences such as that just cited for a variety of drug-related responses, it has become increasingly clear that genetic influence is not confined to the special “abnormal” case but is ubiquitous and pervasive. Thus, while we may seek to characterize an average rate of uptake or distribution of alcohol, or an average rate of alcohol metabolism, or an average sensitivity of the central nervous system to alcohol, we must always bear in mind that such average figures actually characterize very few individuals in a natural population and that there is a spectrum of diversity surrounding each value. With respect to the genetic determinants of this diversity, we should not expect so simple a system that there is one gene for uptake, another for metabolism, and so on. With respect to individual differences existing among individuals in a natural population, we will, in many instances, have the much more complex situation in which a number of genes, and a variety of environmental factors, influence the trait of interest. An examination of a schema that addresses these multiple factor influences might be apposite at this point. Consider the statement that a single gene determines, or influences, a trait. Actually, such a statement must be taken to mean that a single gene pair determines or influences the trait. This situation can be represented by the following drawing, in which the paired circles symbolize a gene pair, with one member of the pair contributed by each parent, 0
o-p
and “P” refers to phenotype,
which is the geneticist’s
term for the measured
64
A
0
0
l
B
0
0
+
c
0
0
+
D
0
0
+
E
0
0
+
F
0
0
+
G
0
0
+
H
0
0
l
Fig. 2. Highly simplified thetical phenotypes.
schema illustrating
pathways
of genetic
influence
on three hypo-
trait or attribute. Obviously, what is really meant by this representation is that different antecedents on the left lead to different consequences on the right. The different antecedents arise from the fact that genes exist in alternative forms (called alleles) and that these may be arranged in different combinations. For example, if two different alleles of a gene are represented by l and o, the possible different combinations are 00, 00, and 00. The relative frequencies of individuals with these genotypes will depend upon the relative frequencies of l and o in the gametes of the reproducing members of the population. The rules relating the genotypic antecedents to the phenotypic consequences can be quite simple and linear: 0
0-p
0
l
0
e-P
-P
Sometimes, however, the rules become a bit more complicated in that two of the genotypic conditions give rise to the same phenotypic results: 0
0-p
0
0-P
l
.--+P
This, of course, is the classical situation of dominance-recessiveness. In the example, the allele represented by l would be described as dominant to the allele represented by 0. Two important extensions of this relationship are that (1) several gene pairs might influence a single trait, and (2) a given gene pair can influence several traits. The former case is called poZygenic inheritance; the latter, pleiotropy. Eight gene pairs, identified as A-H, are represented by the paired circles in Fig. 2. The short arrow from each gene pair to the asterisk-like symbol
65
B
0
c
0
D
0
E
0
F
0
Fig. 3. Schema showing phenotypes.
common
network
of genetic
influences
on two hypothetical
represents the primary gene action of establishing specificity of an enzyme. These enzymes promote the biochemical actions that influence the functions of the organ systems of the body. These organ systems - the endocrine systems, the receptor systems, the motor systems, the autonomic and central nervous systems, the liver, and any other aspect of the organism pertinent to the phenotypes of interest - are represented by the squares in the drawing. Three phenotypes (I, II and III) are symbolized at the far right. Thus,
q might
@ might represent the level of a particular neurotransmitter, represent density of receptors,
a
might represent the turnover
of NAD to NADH, and so on. The mechanism of pleiotropy is now more comprehensible. Although each gene pair has only one primary action in determining the specificity of an enzyme, the effects can spread out downstream in the causal channels so that more than one phenotype is affected. It can be seen, for example, that gene pair B has an influence on phenotype I through
q+
q + q , through q + q
/6.
The important
+
q
+
q , and
point with respect to pleiotropy
through
q
+
is that gene pair
B also influences phenotype II through the latter route. Herein lies the genetic contribution to correlations in drug effects. In fact, most of our hypothetical diagram does deal with a causal nexus shared by phenotypes I and II, as shown in Fig. 3. Only three of the gene pairs are not included in this network of common influences: Pair A influences only phenotype I; pair G influences only phenotype II; and pair H is shown as having a direct influence only on phenotype III, which has nothing genetic in common with the other two traits. Our diagram also shows that there are many ways of achieving a given level of any particular phenotype. The pattern of covariation, or commonality, with some other phenotype is therefore dependent upon the particular pathway through which the level of the first is attained. If, for example, a high level of phenotype I is determined by a particular genotype at gene pair B, the shared causal pathways are as shown in Fig. 4. If, on the other hand, the high level of phenotype I is attained by virtue of a particular genotype at gene pair F, the shared pathways are quite different (see Fig. 5). Because a given level of a phenotype might be generated genetically through several different routes, it appears to be folly to attempt to characterize “the” alcoholic or “the” heroin addict, etc.
66
Fig. 4. Schema ical phenotypes.
F
showing
pathways
of influence
of a particular
showing
pathways
of influence
of another
gene pair on two hypothet-
0
Fig. 5. Schema phenotypes.
gene pair on the same two
At this point, it may be appropriate to comment upon the capacity of the genetic system for generating diversity. If each gene pair is capable of existing as three alternative genotypes (00, 00, or lo), the number of different genotypic possibilities just among our eight hypothetical gene pairs is 38, or 6561. When the number of gene pairs actually segregating in humankind is considered, the number of possibilities becomes astronomical. Bodmer and Cavalli-Sforza [ 51 make the point well by noting that a single individual human being has the potential to produce about 103000 different types of eggs or sperm. They compare this huge number to an estimate of the total number of sperm ever produced by all human males, about 10z4. Clearly, with a system capable of generating that kind of variability, the view of individual differences as “error” is conceptually restricting. In addition to the variance generated by the reshuffling of genes from generation to generation, the whole realm of environmental influences must obviously be considered. In Fig. 6 environmental influences are shown as wavy arrows impinging upon the same causal pathways that convey the effects of the genes. With a schema of this sort, it is readily seen that the dichotomous, even antagonistic, view of nature uersus nurture is untenable. Environmental agencies, broadly understood to range from such things as adequacy of placental attachment through peer-group pressures, co-act with the enzymatic products of the genes in an intimate and inextricable manner. Environmental sources of variance can also lead to covariance, as would be the case for the environmental
influences
and
can also cause independent
q . Environmental
agencies
on m
and 181 and on variation
m of one
67
A0
0
00
0
co
0
DO
0
EO
0
F
0
0
to
0
HO
0
Fig. 6. Schema
phenotype
illustrating
joint
influences
alone, as in the influence
of genes and environment.
shown on
q . Furthermore,
environ-
mental variance can induce covariation where genetic factors alone would not. In our diagram, for example, the environmental agency impinging on 1111 and
115j would generate
a commonality
between
phenotypes
II and
III. Neither genetic nor environmental variance sources are distributed randomly within a population. Some environmental events are unique to an individual, while others affect all members of a family (or litter), and there are some that influence all members of a culture (or laboratory colony room). Similarly, genetic segregation confers genotypic uniqueness on individuals, while shared common ancestry within a family causes relatives to resemble each other more than they resemble unrelated individuals, and the gene pool of one population may differ from that of another in terms of the relative frequencies of different alleles. In terms of our schema, the progressive reduction of genetic variance that accompanies inbreeding has brought all of the gene pairs into like configuration in all members of an inbred strain (except, of course, that males and females differ with respect to sex-linked genes). In other words, the genetic influences through the causal network are the same in all strain members. When we make strain comparisons, we are exploiting genetic variability that exists between strains. Environmental sources of variance persist, however, even in the best controlled laboratories, and they give rise to the manifest within-strain variability. The primary virtue of inbred strains is their relative stability with respect to means, but a small amount of variability is also desirable for most research purposes. The issue of comparative variance size within inbred strains and within heterogeneous groups is quite complex [6] and cannot be explained in detail here.
68
Limitations
of inbred strains
A particularly important limitation of inbred strains arises with respect to hypotheses concerning correlated characters. Within a single inbred strain, any covariance between characters must arise from environmental sources. Possible genetic contributions to covariance have been eliminated. Comparison of two inbred strains that differ in attribute A to determine if they also differ in attribute B does not provide strong evidence concerning the causal relatedness of the two traits. Because the process of inbreeding causes fixation of all loci, animals of a particular inbred strain are stabilized genotypitally for all traits subject to genetic influence. Therefore, given a difference between two strains in trait A, the expectation is that they will show a significant mean difference with respect to a very large number of other traits that may have no connection whatsoever with trait A. Therefore, a finding that C57BL (alcohol-preferring) and DBA/2 (alcohol-avoiding) mice differ with respect to liver alcohol dehydrogenase activity [ 71 is no more definitive than the observation that they differ in tail length (C57BL mice have longer tails). It is permissive evidence, in the sense that it permits us to entertain further the hypothesis that alcohol preference might be related to a particular enzyme system involved in the metabolism of ethanol, but it is not per-
suasive . Another disadvantage of inbred strains is that the investigator must, in effect, take pot-luck. Inbreeding itself is a non-directional process, characterized only by the mating of relatives. Thus, one does not inbreed for something; the inbreeding is accomplished and the differences fall where they may. In reality, many of the standard inbred strains available today are the results of a mixture of inbreeding and selective breeding for particular characteristics. Nevertheless, when these strains are used in a new research area, the investigator must essentially sift and winnow to find strain differences with respect to the phenotype of his or her interest, and the differences found might not be as large as would be desired.
Selective
breeding
Selective breeding allows the a priori specification of a particular trait for which differences are sought. Insofar as the phenotypic variability for that trait is due to genotypic differences, systematic mating of like animals from the extremes of the distributions in successive generations will generate animals differing with respect to that trait. More specifically, the total variance in the population from which selected lines are derived may be regarded as composed of genetic components, environmental components, and a covariance between these genetic and environmental sources. Thus v,
= vo
+ Vz + cou,,
The last term, in carefully
controlled
laboratory
conditions,
can be substan-
69
Fig. 7. Phenotypic distributions for parent (top) and offspring (bottom) generations in a selective breeding study. Included in the illustration are the phenotypic mean of the entire parental generation (go), the phenotypic mean of those selected to reproduce (x,), the phenotypic mean of the offspring generation (_%I), the selection differential (S), and the selection response (R).
tially reduced; for present purposes, at least, it can be ignored. The VG term, referring to genetic variance, can itself be broken down into VA (a component related to the average effects of allelic differences) and V,, (a portion due to dominance and epistatic effects). The heritability of the phenotype is narrowly defined as
This index provides a means of predicting breeding by the relationship
the rate of progress in selective
R = h2S where R is the selection response, and S is the selection differential. These relationships are illustrated in Fig. 7, which shows a normally distributed phenotype, a sample mean (indicated by the vertical line in the center of the distribution), and a group (represented by the shaded upper tail) selected to be the parents of the next generation. S is defined as the difference between the mean of the total population (x0) and the mean of the group selected to be parents (rf,). The progeny of the selected animals will themselves form a phenotypic distribution such as the one shown at the bottom of the figure. Insofar as selection has been successful, the mean of the offspring distribution (X1) will differ from the mean of the parental generation (xc), and the difference between these means represents R (the selection response). As selection continues, with extreme animals selected to be progenitors of successive generations, there will be an increasing change in the phenotypic mean. However, selection “uses up” VA. Therefore, as selection proceeds, the basic raw material declines and the selection response R will reach a plateau.
70
GENERATlONS
Fig. 8. Highly study.
idealized
outcome
in the early
generations
of a bidirectional
selection
Because it is desirable for most research purposes to have both a high and a low line, selection is usually bidirectional. The same considerations, of course, apply to selection in either direction. A highly idealized outcome of selection would be as shown in Fig. 8. Departures from symmetry of response may reflect characteristics of the genetic structure of the trait - for example, with respect to dominance of the alleles at the relevant loci. Environmental effects may also intrude to give a rather saw-toothed or ragged appearance to the selection results. It is important to begin a selection study with a foundation stock as genetically heterogeneous as feasible. An Fz generation derived from inbred strains is a genetically segregating population and can serve as a foundation stock. However, a segregating population derived from just two strains contains only the genetic differences between those strains. Greater genetic variability may obviously be generated by intercrossing four or eight strains, for example. Our HS stock, derived by intercrossing the A, BALB/c, C3H/2, C57BL, DBA/B, AKR, Is, and RI11 strains, is an example of a genetically heterogeneous group derived from an eight-way cross [ 81.
The Colorado
long-sleep
and short-sleep
lines
Using this HS stock as a foundation population, we initiated a selective breeding study [9] for sensitivity to a hypnotic dose of intraperitoneally administered ethanol. The divergence of the long-sleep (LS) and short-sleep (SS) lines over generations is shown in Fig. 9. The interruption in the line plots is due to a fertility problem in generations 6 - 8, probably ensuing from a change in laboratory location. At this time, each line went through a severe genetic “bottleneck”, and the response to selection from that time forward is probably less than it would have been otherwise. In any case, the divergence of the lines has increased to the point that there was no overlap between the two populations in generation 18 (see Fig. 9). From the response to selection pressure, it is possible to estimate the heritability of the phenotype in the foundation population. This realized heritability estimate for
71
CLING
SLEEP
2500
0
0
2
4
6
0
IO
I2
14
I6
I8
GENERATION
Fig. 9. Response to 18 generations of selection as measured by duration of loss of the righting
for sensitivity response.
to hypnotic
dose of ethanol
sleep time under the conditions of this study (computed over the first five generations) was 0.18. Thus, even when only 18% of the total phenotypic variance is due to additive genetic factors, selective breeding can generate lines that are dramatically different. When the selection process has generated lines of widely different phenotypes, they become available as a powerful tool for exploring mechanisms underlying the trait for which they differ. A particularly useful design is to compare the lines with respect to attributes hypothesized to be part of the causal mechanism for the selected trait or related characters. Such twogroup comparisons, described earlier as being particularly hazardous with inbred animals, are less hazardous with selected lines because, while such lines may become homozygous for all pertinent loci, they will be segregating for all non-pertinent loci. The LS and SS mice have found employment in studies of ethanol-induced hypothermia [lo] , voluntary ethanol ingestion [ 111, blood alcohol elimination rates [ 121, effects of fetal alcohol exposure [ 131, withdrawal reactions [ 141, low alcohol dose sensitivity [ 151, and nervous system sensitivity to alcohol [ 161 . Indicating the breadth of utility of selected lines is the fact that the LS and SS animals have also been used in studies of catecholamine turnover rate [ 171 and of sensitivity to pentobarbital [ 151, to trichloroethanol and paraldehyde [ 181, to pentobarbital, chloral hydrate, trichloroethanol, and paraldehyde [ 191, to gamma-butyrolactone [ 201, to salsolinol [21], to flurothyl-induced seizures [22], and to apomorphine, haloperidol, pilocarpine, scopolamine, and bicuculline [22] . They have also been used as a model for hyperkinesis [ 231. The successful selection for sleep time and for voluntary alcohol consumption [24] in our own laboratories, and the success of the Finnish study [25. 261, the Rutgers’ study [27], and the Indiana study [28], make it clear
72
that genetic differences contribute to a wide variety of alcohol-related phenotypes. It is also clear that a wide variety of relevant animal models can be generated by selective breeding. To this point, efforts at selective breeding have been univariate, but the time has now come for the initiation of attempts to generate more complex models. It may prove possible to derive somewhat more complex, multidimensional models from the selected lines already in existence. Consider that genetic variability with respect to ease of acquiring dependence or severity of withdrawal symptoms or any other attribute (insofar as it is not related to sleep time) should remain high in the LS and SS lines. Therefore, by initiating a new selective breeding program with these lines as foundation populations, it might be possible to generate, for example, long-sleep/severe-withdrawal, long-sleep/light-withdrawal, shortsleep/severe-withdrawal, and short-sleep/light-withdrawal lines. This type of selection is referred to as “tandem”, and is the least efficient of the three methods available for selecting for multiple attributes [ 29, 301. Another approach would be to use independent culling levels. According to this procedure, individuals selected for breeding would be required to have phenotypes exceeding a certain minimum value in each character chosen for selection. This procedure is generally more desirable than tandem selection, but less desirable than index selection. With the latter procedure, a composite score derived from multiplying individual phenotypic scores by appropriate coefficients is utilized as the selection criterion. Consideration of complex phenotypes for selection raises the issue of the factor structures of domains studied by pharmacologists. As is true of any empirical science, pharmacology must utilize operational definitions of its concepts. For example, consider the phenomena of tolerance and dependence. Definitions of tolerance often include such phrases as “unusual resistance” to an “ordinary” dose of a drug, or state that increasing doses must effect. Clearly, these are terms that be taken to produce a “characteristic” require further explication. What does “unusual” mean ? What is “ordinary” ? What is a “characteristic” effect ? How is resistance measured ? The ultimate meaning of such definitions resides in the operations that are described to explain the terms. Operationally, one might measure tolerance by observing sleep times of a mouse after intraperitoneal injection of 4.0 g ethanol per kg on two occasions separated by a 24-hour period. However, there are many other operations that might yield equally useful measures of tolerance. Similarly, dependence is often assessed in terms of withdrawal symptoms. The actual data may be observations of muscle contractions of a particular kind (seizures) that are used to index a hypothetical process or state of dependence. Many such indices of dependence are currently used, and an indefinitely large number of other ones could be contrived. While there may be differences among measures in convenience, or in sensitivity, or in susceptibility to error, there is absolutely no way to determine which one of the many possible indices of tolerance or dependence is the most valid measure. Situations of this kind exist in many fields of scientific endeavor, and many highly developed multivariate analytical procedures are now available
1
moment
correlation
Anderson
[31].
0.10 -
0.00 0.65 -
3 0.36 0.22 0.12 -
4
between
0.04 0.55 0.69 0.22 -
5 -0.21 0.00 -0.03 -0.48 0.50 0.73 0.62 -
0.13 0.33 0.15 0.10 0.09 -0.01 -0.13 -0.08 -
9 0.04 0.39 0.06 0.09 0.10 0.07 -0.15 -0.01 0.35 0.48 -
11 0.11 0.43 0.02 0.11 0.07 0.06 -0.20 -0.04 0.44 0.39 0.89 -
12
traits in HS mice*
-0.09 0.02 -0.11 -0.03 -0.03 0.05 -0.05 0.05 --0.56 -
10
of various
from zero at the 0.05 level.
-0.02 0.01 0.38 0.11 0.83 0.78 -
8
measurements 7
different
0.04 0.16 0.04 0.22 0.73 -
6
experimental
of 0.10 are significantly
12
coefficients
Values in excess
Body weight Liver weight mg protein per ml homogenate Total liver ADH ADH/g liver ADH/g body weight Specific activity Sleep time Blood ethanol at “recovery” Blood ethanol - 1 hour Blood ethanol - 2 hours Blood ethanol - 3 hours Ethanol elimination rate
*From
2. 3. 4. 5. 6. ‘7. 8. 9. 10. 11. 12. 13. 14.
1. Age
Pearson product
TABLE
0.14 0.37 -0.04 0.10 0.02 0.04 -0.22 -0.06 0.43 0.21 0.57 0.88 -
13
-0.11 0.06 0.12 0.00 0.11 0.05 0.09 0.07 -0.05 0.33 0.53 0.10 -0.38
14
Fig. 10. Three factors identified in HS mice (see Table 1).
from
the pattern
of intercorrelations
among
several traits
to deal with them. Consider, for example, the results of a study in which Anderson [ 311 found a pattern of intercorrelations among a number of different traits in a sample of the HS population from which the LS and SS lines were derived (see Table 1). Factor analysis permits us to imagine some number of latent factors (less than the number of variables) that might “account for” the pattern of reiationships. The results of such a factor analysis are diagrammed in Fig. 10. The three factors may be regarded as hypothetical entities whose nature is revealed by the pattern of correlations between each factor and the separate measures. Thus, Factor 1 can be described as a size factor. Modestly related to age, it is principally identified by high “loadings” (correlations with the factor) of body weight, liver weight, mg protein per ml homogenate, and total liver ADH. This size factor has very little relationship to specific activity, to blood ethanol concentration at time of recovery of the righting response, or to ethanol elimination rate. It has only a modest relationship to sleep time. Factor 2 is an enzyme activity factor, identified by high loadings of liver ADH, ADH per g liver, and specific activity. It has no relationship to sleep time. Factor 3 might be interpreted as a sensitivity factor, since it is characterized by sleep time, by blood ethanol concentration at recovery, and (somewhat less markedly) by ethanol elimination rate. It has no significant correlation with any of the ADH measures or with the size factor. From the relationships among the variables that define Factor 3, we may conclude that animals with shorter sleep times tend to have higher blood ethanol concentrations at recovery and higher ethanol elimination rates than do animals with longer sleep times. Recall that we used only the sleep time index of sensitivity in the selection study which generated the LS and SS lines. By adding information concerning blood ethanol levels at
75
recovery and ethanol elimination rates, we now have a much better assessment of the hypothetical “sensitivity” factor. From the genetic point of view, we might have been able to select more precisely had we employed a composite index using information regarding all of the variables that load on the factor. From the pharmacological viewpoint, there are perhaps few surprises in this particular outcome, but it does illustrate a method that might be useful in clarifying relationships among different indices of particular pharmacological processes and among the processes themselves. With the methods of selective breeding, and with attention to the complexities of the processes involved, the pharmacogeneticist should be able to facilitate pharmacological research by providing animal models of increasing power and utility.
Acknowledgment GEM wishes to acknowledge the support of a faculty fellowship from the University of Colorado Council on Research and Creative Work. We also thank Rebecca G. Miles for her expert assistance in the preparation of the manuscript.
References
5 6 7 8
9 10 11 12
13 14 15 16 17 18 19
G. E. McClearn and D. A. Rodgers, Quart. J. Stud. Ale., 20 (1959) 691 - 695. G. E. McClearn, Proc. 6th Int. Cong. Pharmacol., 3 (1975) 59 - 66. G. E. McClearn and D. L. Shern, Behav. Genet., 5 (1975) 103. R. Y. Kakihana, unpublished Doctoral Dissertation, University of California, Berkeley, 1965. W. F. Bodmer and L. L. Cavalli-Sforza, Genetics, Evolution and Man, Freeman, San Francisco, 1976. G. E. McClearn, J. Toxicol. Environ. Hlth., (1978) in press. G. E. McClearn et al., Nature, 203 (1964) 793 - 794. G. E. McClearn, J. R. Wilson and W. Meredith, in G. Lindzey and D. D. Thiessen (eds.), Contributions to Behavior-Genetic Analysis: The Mouse as a Prototype, Appleton-Century-Crofts, New York, 1970, pp. 3 - 22. G. E. McClearn and R. Kakihana, Behav. Genet., 3 (1973) 409 - 410. R. Kakihana and J. Moore, Behav. Genet., 7 (1977) 71. J. L. Fuller and A. C. Church, Behav. Genet., 7 (1977) 59. W. Heston and V. G. Erwin, Some biochemical parameters of alcohol metabolism in selected strains of mice, Paper presented at Behavior Genetics Association Meeting, 1972. D. S. Baer and D. W. Crumpacker, Behav. Genet., 7 (1977) 95 - 103. D. B. Goldstein and R. Kakihana, Life Sci., 17 (1975) 981 - 986. B. Sanders, J. Comp. Physiol. Psychol., 90 (1976) 394 - 398. J. W. MacInnes and R. P. Damjanovich, Behav. Genet., 3 (1972) 408. A. C. Collins and R. A. Deitrich, Behav. Genet., 3 (1973) 398. B. Sanders et al., Psychopharmacology, 56 (1978) 185 - 190. V. G. Erwin, W. D. W. Heston and G. E. M&learn, Pharmacol. Biochem. Behav., 4 (1976) 679 - 683.
16 20 21 22 23 24 25 26 27 28 29 30 31
B. C. Dudek, Behav. Genet., 8 (1978) 91. A. C. Church, J. L. Fuller and B. C. Dudek, Psychopharmacology, C. A. Greer and H. P. Alpern, Life Sci., 21 (1977) 385 - 392. H. P, Alpern and C. A. Greer, Life Sci., 21 (1977) 93 - 98. G. E. McClearn and S. M. Anderson, Behav. Genet., in press. K. Eriksson, Science, 159 (1968) 739 - 741. K. Eriksson, Ann. N. Y. Acad. Sci., 197 (1972) 32 - 41. E. P. Riley et al., J. Stud. Ale., 38 (1977) 1705 - 1717. T.-K. Li et al., Drug Ale. Depend., 4 (1979) 45. D. S. Falconer, Introduction to Quantitative Genetics, Ronald, F. Pirchner, Population Genetics in Animal Breeding, Freeman, S. M. Anderson, unpublished Doctoral Dissertation, University
47 (1976)
49
- 52.
New York, 1960. San Francisco, 1969. of Colorado, 1975.
GENETICS
GERALD Institute (U.S.A.)
AND ETHANOL
E. McCLEARN
61
in the Netherlands
TOLERANCE
and SALLY
for BeharJioral Genetics,
- 76
M. ANDERSON
University
of Colorado,
Boulder,
Colorado
80309
Summary This paper reviews some of the research on genetic bases of individual differences in ethanol tolerance of mice conducted at the Institute for Behavioral Genetics and at its predecessor laboratory at the University of California, Berkeley. Tolerance is, of course, a complex concept. Theoretical distinctions are made between tachyphylaxis and more slowly acquired tolerance and between dispositional and tissue tolerance. Pragmatically, a variety of measures (such as locomotor activity, sleep time, hypothermia) can be used to define these processes, and the different indices may yield quite different results even when they presumably indicate the same process. It is clear that an understanding of genetic influence in “ethanol tolerance” will require wide sampling of this complex domain. The work described here represents only a beginning, but it may illustrate the general approaches that are available for addressing the issue.
The relatively young literature of animal pharmacogenetics has been derived in large part from studies of inbred mouse strains. There are two principal reasons why this has been the case. First, inbred mice are easily and quickly available. Second, the methodology of strain comparisons is conceptually straightforward. As a rough approximation, it can be easily understood that all mice within a given strain are genetically alike and are different from members of any other inbred strain. Therefore, although the genetic information derived from strain comparisons is of a low order, the use of inbred strains does constitute a very effective means of controlling the characteristics of the animals employed in pharmacological investigations.
Inbred mouse strain differences
in alcohol-related
behavior
The work of our laboratory on the genetics of alcohol-related behavior and its mechanisms, like that of many other research groups, began with
0
I 35 ETHANOL
Fig. 1. Dose-response mouse strains.
curves
DOSAGE
2.70 (gmJkg.1
for influence
of ethanol
on arena activity
of several inbred
inbred mice. Our initial results revealed a large and stable strain difference in voluntary alcohol consumption between C57BL and DBA/B mice. The former display moderate to high ethanol ingestion in a two-bottle choice situation, while the latter are enthusiastic abstainers [l] . Several subsequent studies have characterized these differences in further detail [2] . However, of more direct relevance to the topic of this meeting is the research undertaken on strain differences in sensitivity to ethanol. We first approached the issue of first-dose, or naive, tolerance by recording changes in activity subseqi!+‘*. .‘?a,:bodinhalation of ethanol vapor. Just about every possible out,” j> e-1:the activity of C57BL mice was lowered; the activity of .lged, and the activity of C3H mice was increased. Using ns of ethanol, so as to be able to control dosage more ‘halation procedures, these strains and others were ., on activity of ethanol at 1.35 g/kg and 2.70 g/kg ,ults are shown in Fig. 1. It may be seen that the low dose tiivity in the BALB/c, C3H and A strains, but had no effect on L.. ./b;\ or C57BL mice. The high dose resulted in either no further change (in the case of the C3H and A strains) or a decrease in activity (in the BALB/ c, DBA and C57BL strains). These results clearly demonstrate the importance of the genotype of (7 ‘11 in pharmacological research. Not only does the dose-re1 :liffer quite markedly among strains, but the characteriza* s behaviorally depressing, exciting, or neutral, is also straindept ir4,t 1:. Stran. ilfferences have also been apparent in our research utilizing dura tion of loss of the righting response (“sleep time”) to investigate genetic dif-
63
ferences in the hypnotic effect of ethanol. C57BL and C3H/2 mice have much shorter sleep times than do DBA/B or BALB/c mice after an intraperitoneal injection of 3.4 g ethanol per kg [4]. From these and similar researches, in our own laboratory and others, the accumulated evidence makes it clear that the genotype of the research animal is a critical consideration. In brief, the paramount lesson is that investigators must be very modest indeed concerning the breadth of generality of results obtained with a particular group of animals. Instead of regarding a given outcome as general until proven idiosyncratic, we should regard results as idiosyncratic until proven general. At first glance, this attitude may appear to be quite negative. Where is scientific lawfulness if strains can be found to differ in response to some particular manipulation ? There is a positive aspect to the situation, however, which far outweighs the negative one. The very differences which prevent us from immediately assuming that we have tapped an eternal verity when we obtain a research outcome with a particular sample of animals offer us the tools to explore more deeply the mechanisms underlying those differences, where we might indeed find the lawful generalities to reside.
A polygenic
schema for pharmacology
With the accumulation of evidence of strain differences such as that just cited for a variety of drug-related responses, it has become increasingly clear that genetic influence is not confined to the special “abnormal” case but is ubiquitous and pervasive. Thus, while we may seek to characterize an average rate of uptake or distribution of alcohol, or an average rate of alcohol metabolism, or an average sensitivity of the central nervous system to alcohol, we must always bear in mind that such average figures actually characterize very few individuals in a natural population and that there is a spectrum of diversity surrounding each value. With respect to the genetic determinants of this diversity, we should not expect so simple a system that there is one gene for uptake, another for metabolism, and so on. With respect to individual differences existing among individuals in a natural population, we will, in many instances, have the much more complex situation in which a number of genes, and a variety of environmental factors, influence the trait of interest. An examination of a schema that addresses these multiple factor influences might be apposite at this point. Consider the statement that a single gene determines, or influences, a trait. Actually, such a statement must be taken to mean that a single gene pair determines or influences the trait. This situation can be represented by the following drawing, in which the paired circles symbolize a gene pair, with one member of the pair contributed by each parent, 0
o-p
and “P” refers to phenotype,
which is the geneticist’s
term for the measured
64
A
0
0
l
B
0
0
+
c
0
0
+
D
0
0
+
E
0
0
+
F
0
0
+
G
0
0
+
H
0
0
l
Fig. 2. Highly simplified thetical phenotypes.
schema illustrating
pathways
of genetic
influence
on three hypo-
trait or attribute. Obviously, what is really meant by this representation is that different antecedents on the left lead to different consequences on the right. The different antecedents arise from the fact that genes exist in alternative forms (called alleles) and that these may be arranged in different combinations. For example, if two different alleles of a gene are represented by l and o, the possible different combinations are 00, 00, and 00. The relative frequencies of individuals with these genotypes will depend upon the relative frequencies of l and o in the gametes of the reproducing members of the population. The rules relating the genotypic antecedents to the phenotypic consequences can be quite simple and linear: 0
0-p
0
l
0
e-P
-P
Sometimes, however, the rules become a bit more complicated in that two of the genotypic conditions give rise to the same phenotypic results: 0
0-p
0
0-P
l
.--+P
This, of course, is the classical situation of dominance-recessiveness. In the example, the allele represented by l would be described as dominant to the allele represented by 0. Two important extensions of this relationship are that (1) several gene pairs might influence a single trait, and (2) a given gene pair can influence several traits. The former case is called poZygenic inheritance; the latter, pleiotropy. Eight gene pairs, identified as A-H, are represented by the paired circles in Fig. 2. The short arrow from each gene pair to the asterisk-like symbol
65
B
0
c
0
D
0
E
0
F
0
Fig. 3. Schema showing phenotypes.
common
network
of genetic
influences
on two hypothetical
represents the primary gene action of establishing specificity of an enzyme. These enzymes promote the biochemical actions that influence the functions of the organ systems of the body. These organ systems - the endocrine systems, the receptor systems, the motor systems, the autonomic and central nervous systems, the liver, and any other aspect of the organism pertinent to the phenotypes of interest - are represented by the squares in the drawing. Three phenotypes (I, II and III) are symbolized at the far right. Thus,
q might
@ might represent the level of a particular neurotransmitter, represent density of receptors,
a
might represent the turnover
of NAD to NADH, and so on. The mechanism of pleiotropy is now more comprehensible. Although each gene pair has only one primary action in determining the specificity of an enzyme, the effects can spread out downstream in the causal channels so that more than one phenotype is affected. It can be seen, for example, that gene pair B has an influence on phenotype I through
q+
q + q , through q + q
/6.
The important
+
q
+
q , and
point with respect to pleiotropy
through
q
+
is that gene pair
B also influences phenotype II through the latter route. Herein lies the genetic contribution to correlations in drug effects. In fact, most of our hypothetical diagram does deal with a causal nexus shared by phenotypes I and II, as shown in Fig. 3. Only three of the gene pairs are not included in this network of common influences: Pair A influences only phenotype I; pair G influences only phenotype II; and pair H is shown as having a direct influence only on phenotype III, which has nothing genetic in common with the other two traits. Our diagram also shows that there are many ways of achieving a given level of any particular phenotype. The pattern of covariation, or commonality, with some other phenotype is therefore dependent upon the particular pathway through which the level of the first is attained. If, for example, a high level of phenotype I is determined by a particular genotype at gene pair B, the shared causal pathways are as shown in Fig. 4. If, on the other hand, the high level of phenotype I is attained by virtue of a particular genotype at gene pair F, the shared pathways are quite different (see Fig. 5). Because a given level of a phenotype might be generated genetically through several different routes, it appears to be folly to attempt to characterize “the” alcoholic or “the” heroin addict, etc.
66
Fig. 4. Schema ical phenotypes.
F
showing
pathways
of influence
of a particular
showing
pathways
of influence
of another
gene pair on two hypothet-
0
Fig. 5. Schema phenotypes.
gene pair on the same two
At this point, it may be appropriate to comment upon the capacity of the genetic system for generating diversity. If each gene pair is capable of existing as three alternative genotypes (00, 00, or lo), the number of different genotypic possibilities just among our eight hypothetical gene pairs is 38, or 6561. When the number of gene pairs actually segregating in humankind is considered, the number of possibilities becomes astronomical. Bodmer and Cavalli-Sforza [ 51 make the point well by noting that a single individual human being has the potential to produce about 103000 different types of eggs or sperm. They compare this huge number to an estimate of the total number of sperm ever produced by all human males, about 10z4. Clearly, with a system capable of generating that kind of variability, the view of individual differences as “error” is conceptually restricting. In addition to the variance generated by the reshuffling of genes from generation to generation, the whole realm of environmental influences must obviously be considered. In Fig. 6 environmental influences are shown as wavy arrows impinging upon the same causal pathways that convey the effects of the genes. With a schema of this sort, it is readily seen that the dichotomous, even antagonistic, view of nature uersus nurture is untenable. Environmental agencies, broadly understood to range from such things as adequacy of placental attachment through peer-group pressures, co-act with the enzymatic products of the genes in an intimate and inextricable manner. Environmental sources of variance can also lead to covariance, as would be the case for the environmental
influences
and
can also cause independent
q . Environmental
agencies
on m
and 181 and on variation
m of one
67
A0
0
00
0
co
0
DO
0
EO
0
F
0
0
to
0
HO
0
Fig. 6. Schema
phenotype
illustrating
joint
influences
alone, as in the influence
of genes and environment.
shown on
q . Furthermore,
environ-
mental variance can induce covariation where genetic factors alone would not. In our diagram, for example, the environmental agency impinging on 1111 and
115j would generate
a commonality
between
phenotypes
II and
III. Neither genetic nor environmental variance sources are distributed randomly within a population. Some environmental events are unique to an individual, while others affect all members of a family (or litter), and there are some that influence all members of a culture (or laboratory colony room). Similarly, genetic segregation confers genotypic uniqueness on individuals, while shared common ancestry within a family causes relatives to resemble each other more than they resemble unrelated individuals, and the gene pool of one population may differ from that of another in terms of the relative frequencies of different alleles. In terms of our schema, the progressive reduction of genetic variance that accompanies inbreeding has brought all of the gene pairs into like configuration in all members of an inbred strain (except, of course, that males and females differ with respect to sex-linked genes). In other words, the genetic influences through the causal network are the same in all strain members. When we make strain comparisons, we are exploiting genetic variability that exists between strains. Environmental sources of variance persist, however, even in the best controlled laboratories, and they give rise to the manifest within-strain variability. The primary virtue of inbred strains is their relative stability with respect to means, but a small amount of variability is also desirable for most research purposes. The issue of comparative variance size within inbred strains and within heterogeneous groups is quite complex [6] and cannot be explained in detail here.
68
Limitations
of inbred strains
A particularly important limitation of inbred strains arises with respect to hypotheses concerning correlated characters. Within a single inbred strain, any covariance between characters must arise from environmental sources. Possible genetic contributions to covariance have been eliminated. Comparison of two inbred strains that differ in attribute A to determine if they also differ in attribute B does not provide strong evidence concerning the causal relatedness of the two traits. Because the process of inbreeding causes fixation of all loci, animals of a particular inbred strain are stabilized genotypitally for all traits subject to genetic influence. Therefore, given a difference between two strains in trait A, the expectation is that they will show a significant mean difference with respect to a very large number of other traits that may have no connection whatsoever with trait A. Therefore, a finding that C57BL (alcohol-preferring) and DBA/2 (alcohol-avoiding) mice differ with respect to liver alcohol dehydrogenase activity [ 71 is no more definitive than the observation that they differ in tail length (C57BL mice have longer tails). It is permissive evidence, in the sense that it permits us to entertain further the hypothesis that alcohol preference might be related to a particular enzyme system involved in the metabolism of ethanol, but it is not per-
suasive . Another disadvantage of inbred strains is that the investigator must, in effect, take pot-luck. Inbreeding itself is a non-directional process, characterized only by the mating of relatives. Thus, one does not inbreed for something; the inbreeding is accomplished and the differences fall where they may. In reality, many of the standard inbred strains available today are the results of a mixture of inbreeding and selective breeding for particular characteristics. Nevertheless, when these strains are used in a new research area, the investigator must essentially sift and winnow to find strain differences with respect to the phenotype of his or her interest, and the differences found might not be as large as would be desired.
Selective
breeding
Selective breeding allows the a priori specification of a particular trait for which differences are sought. Insofar as the phenotypic variability for that trait is due to genotypic differences, systematic mating of like animals from the extremes of the distributions in successive generations will generate animals differing with respect to that trait. More specifically, the total variance in the population from which selected lines are derived may be regarded as composed of genetic components, environmental components, and a covariance between these genetic and environmental sources. Thus v,
= vo
+ Vz + cou,,
The last term, in carefully
controlled
laboratory
conditions,
can be substan-
69
Fig. 7. Phenotypic distributions for parent (top) and offspring (bottom) generations in a selective breeding study. Included in the illustration are the phenotypic mean of the entire parental generation (go), the phenotypic mean of those selected to reproduce (x,), the phenotypic mean of the offspring generation (_%I), the selection differential (S), and the selection response (R).
tially reduced; for present purposes, at least, it can be ignored. The VG term, referring to genetic variance, can itself be broken down into VA (a component related to the average effects of allelic differences) and V,, (a portion due to dominance and epistatic effects). The heritability of the phenotype is narrowly defined as
This index provides a means of predicting breeding by the relationship
the rate of progress in selective
R = h2S where R is the selection response, and S is the selection differential. These relationships are illustrated in Fig. 7, which shows a normally distributed phenotype, a sample mean (indicated by the vertical line in the center of the distribution), and a group (represented by the shaded upper tail) selected to be the parents of the next generation. S is defined as the difference between the mean of the total population (x0) and the mean of the group selected to be parents (rf,). The progeny of the selected animals will themselves form a phenotypic distribution such as the one shown at the bottom of the figure. Insofar as selection has been successful, the mean of the offspring distribution (X1) will differ from the mean of the parental generation (xc), and the difference between these means represents R (the selection response). As selection continues, with extreme animals selected to be progenitors of successive generations, there will be an increasing change in the phenotypic mean. However, selection “uses up” VA. Therefore, as selection proceeds, the basic raw material declines and the selection response R will reach a plateau.
70
GENERATlONS
Fig. 8. Highly study.
idealized
outcome
in the early
generations
of a bidirectional
selection
Because it is desirable for most research purposes to have both a high and a low line, selection is usually bidirectional. The same considerations, of course, apply to selection in either direction. A highly idealized outcome of selection would be as shown in Fig. 8. Departures from symmetry of response may reflect characteristics of the genetic structure of the trait - for example, with respect to dominance of the alleles at the relevant loci. Environmental effects may also intrude to give a rather saw-toothed or ragged appearance to the selection results. It is important to begin a selection study with a foundation stock as genetically heterogeneous as feasible. An Fz generation derived from inbred strains is a genetically segregating population and can serve as a foundation stock. However, a segregating population derived from just two strains contains only the genetic differences between those strains. Greater genetic variability may obviously be generated by intercrossing four or eight strains, for example. Our HS stock, derived by intercrossing the A, BALB/c, C3H/2, C57BL, DBA/B, AKR, Is, and RI11 strains, is an example of a genetically heterogeneous group derived from an eight-way cross [ 81.
The Colorado
long-sleep
and short-sleep
lines
Using this HS stock as a foundation population, we initiated a selective breeding study [9] for sensitivity to a hypnotic dose of intraperitoneally administered ethanol. The divergence of the long-sleep (LS) and short-sleep (SS) lines over generations is shown in Fig. 9. The interruption in the line plots is due to a fertility problem in generations 6 - 8, probably ensuing from a change in laboratory location. At this time, each line went through a severe genetic “bottleneck”, and the response to selection from that time forward is probably less than it would have been otherwise. In any case, the divergence of the lines has increased to the point that there was no overlap between the two populations in generation 18 (see Fig. 9). From the response to selection pressure, it is possible to estimate the heritability of the phenotype in the foundation population. This realized heritability estimate for
71
CLING
SLEEP
2500
0
0
2
4
6
0
IO
I2
14
I6
I8
GENERATION
Fig. 9. Response to 18 generations of selection as measured by duration of loss of the righting
for sensitivity response.
to hypnotic
dose of ethanol
sleep time under the conditions of this study (computed over the first five generations) was 0.18. Thus, even when only 18% of the total phenotypic variance is due to additive genetic factors, selective breeding can generate lines that are dramatically different. When the selection process has generated lines of widely different phenotypes, they become available as a powerful tool for exploring mechanisms underlying the trait for which they differ. A particularly useful design is to compare the lines with respect to attributes hypothesized to be part of the causal mechanism for the selected trait or related characters. Such twogroup comparisons, described earlier as being particularly hazardous with inbred animals, are less hazardous with selected lines because, while such lines may become homozygous for all pertinent loci, they will be segregating for all non-pertinent loci. The LS and SS mice have found employment in studies of ethanol-induced hypothermia [lo] , voluntary ethanol ingestion [ 111, blood alcohol elimination rates [ 121, effects of fetal alcohol exposure [ 131, withdrawal reactions [ 141, low alcohol dose sensitivity [ 151, and nervous system sensitivity to alcohol [ 161 . Indicating the breadth of utility of selected lines is the fact that the LS and SS animals have also been used in studies of catecholamine turnover rate [ 171 and of sensitivity to pentobarbital [ 151, to trichloroethanol and paraldehyde [ 181, to pentobarbital, chloral hydrate, trichloroethanol, and paraldehyde [ 191, to gamma-butyrolactone [ 201, to salsolinol [21], to flurothyl-induced seizures [22], and to apomorphine, haloperidol, pilocarpine, scopolamine, and bicuculline [22] . They have also been used as a model for hyperkinesis [ 231. The successful selection for sleep time and for voluntary alcohol consumption [24] in our own laboratories, and the success of the Finnish study [25. 261, the Rutgers’ study [27], and the Indiana study [28], make it clear
72
that genetic differences contribute to a wide variety of alcohol-related phenotypes. It is also clear that a wide variety of relevant animal models can be generated by selective breeding. To this point, efforts at selective breeding have been univariate, but the time has now come for the initiation of attempts to generate more complex models. It may prove possible to derive somewhat more complex, multidimensional models from the selected lines already in existence. Consider that genetic variability with respect to ease of acquiring dependence or severity of withdrawal symptoms or any other attribute (insofar as it is not related to sleep time) should remain high in the LS and SS lines. Therefore, by initiating a new selective breeding program with these lines as foundation populations, it might be possible to generate, for example, long-sleep/severe-withdrawal, long-sleep/light-withdrawal, shortsleep/severe-withdrawal, and short-sleep/light-withdrawal lines. This type of selection is referred to as “tandem”, and is the least efficient of the three methods available for selecting for multiple attributes [ 29, 301. Another approach would be to use independent culling levels. According to this procedure, individuals selected for breeding would be required to have phenotypes exceeding a certain minimum value in each character chosen for selection. This procedure is generally more desirable than tandem selection, but less desirable than index selection. With the latter procedure, a composite score derived from multiplying individual phenotypic scores by appropriate coefficients is utilized as the selection criterion. Consideration of complex phenotypes for selection raises the issue of the factor structures of domains studied by pharmacologists. As is true of any empirical science, pharmacology must utilize operational definitions of its concepts. For example, consider the phenomena of tolerance and dependence. Definitions of tolerance often include such phrases as “unusual resistance” to an “ordinary” dose of a drug, or state that increasing doses must effect. Clearly, these are terms that be taken to produce a “characteristic” require further explication. What does “unusual” mean ? What is “ordinary” ? What is a “characteristic” effect ? How is resistance measured ? The ultimate meaning of such definitions resides in the operations that are described to explain the terms. Operationally, one might measure tolerance by observing sleep times of a mouse after intraperitoneal injection of 4.0 g ethanol per kg on two occasions separated by a 24-hour period. However, there are many other operations that might yield equally useful measures of tolerance. Similarly, dependence is often assessed in terms of withdrawal symptoms. The actual data may be observations of muscle contractions of a particular kind (seizures) that are used to index a hypothetical process or state of dependence. Many such indices of dependence are currently used, and an indefinitely large number of other ones could be contrived. While there may be differences among measures in convenience, or in sensitivity, or in susceptibility to error, there is absolutely no way to determine which one of the many possible indices of tolerance or dependence is the most valid measure. Situations of this kind exist in many fields of scientific endeavor, and many highly developed multivariate analytical procedures are now available
1
moment
correlation
Anderson
[31].
0.10 -
0.00 0.65 -
3 0.36 0.22 0.12 -
4
between
0.04 0.55 0.69 0.22 -
5 -0.21 0.00 -0.03 -0.48 0.50 0.73 0.62 -
0.13 0.33 0.15 0.10 0.09 -0.01 -0.13 -0.08 -
9 0.04 0.39 0.06 0.09 0.10 0.07 -0.15 -0.01 0.35 0.48 -
11 0.11 0.43 0.02 0.11 0.07 0.06 -0.20 -0.04 0.44 0.39 0.89 -
12
traits in HS mice*
-0.09 0.02 -0.11 -0.03 -0.03 0.05 -0.05 0.05 --0.56 -
10
of various
from zero at the 0.05 level.
-0.02 0.01 0.38 0.11 0.83 0.78 -
8
measurements 7
different
0.04 0.16 0.04 0.22 0.73 -
6
experimental
of 0.10 are significantly
12
coefficients
Values in excess
Body weight Liver weight mg protein per ml homogenate Total liver ADH ADH/g liver ADH/g body weight Specific activity Sleep time Blood ethanol at “recovery” Blood ethanol - 1 hour Blood ethanol - 2 hours Blood ethanol - 3 hours Ethanol elimination rate
*From
2. 3. 4. 5. 6. ‘7. 8. 9. 10. 11. 12. 13. 14.
1. Age
Pearson product
TABLE
0.14 0.37 -0.04 0.10 0.02 0.04 -0.22 -0.06 0.43 0.21 0.57 0.88 -
13
-0.11 0.06 0.12 0.00 0.11 0.05 0.09 0.07 -0.05 0.33 0.53 0.10 -0.38
14
Fig. 10. Three factors identified in HS mice (see Table 1).
from
the pattern
of intercorrelations
among
several traits
to deal with them. Consider, for example, the results of a study in which Anderson [ 311 found a pattern of intercorrelations among a number of different traits in a sample of the HS population from which the LS and SS lines were derived (see Table 1). Factor analysis permits us to imagine some number of latent factors (less than the number of variables) that might “account for” the pattern of reiationships. The results of such a factor analysis are diagrammed in Fig. 10. The three factors may be regarded as hypothetical entities whose nature is revealed by the pattern of correlations between each factor and the separate measures. Thus, Factor 1 can be described as a size factor. Modestly related to age, it is principally identified by high “loadings” (correlations with the factor) of body weight, liver weight, mg protein per ml homogenate, and total liver ADH. This size factor has very little relationship to specific activity, to blood ethanol concentration at time of recovery of the righting response, or to ethanol elimination rate. It has only a modest relationship to sleep time. Factor 2 is an enzyme activity factor, identified by high loadings of liver ADH, ADH per g liver, and specific activity. It has no relationship to sleep time. Factor 3 might be interpreted as a sensitivity factor, since it is characterized by sleep time, by blood ethanol concentration at recovery, and (somewhat less markedly) by ethanol elimination rate. It has no significant correlation with any of the ADH measures or with the size factor. From the relationships among the variables that define Factor 3, we may conclude that animals with shorter sleep times tend to have higher blood ethanol concentrations at recovery and higher ethanol elimination rates than do animals with longer sleep times. Recall that we used only the sleep time index of sensitivity in the selection study which generated the LS and SS lines. By adding information concerning blood ethanol levels at
75
recovery and ethanol elimination rates, we now have a much better assessment of the hypothetical “sensitivity” factor. From the genetic point of view, we might have been able to select more precisely had we employed a composite index using information regarding all of the variables that load on the factor. From the pharmacological viewpoint, there are perhaps few surprises in this particular outcome, but it does illustrate a method that might be useful in clarifying relationships among different indices of particular pharmacological processes and among the processes themselves. With the methods of selective breeding, and with attention to the complexities of the processes involved, the pharmacogeneticist should be able to facilitate pharmacological research by providing animal models of increasing power and utility.
Acknowledgment GEM wishes to acknowledge the support of a faculty fellowship from the University of Colorado Council on Research and Creative Work. We also thank Rebecca G. Miles for her expert assistance in the preparation of the manuscript.
References
5 6 7 8
9 10 11 12
13 14 15 16 17 18 19
G. E. McClearn and D. A. Rodgers, Quart. J. Stud. Ale., 20 (1959) 691 - 695. G. E. McClearn, Proc. 6th Int. Cong. Pharmacol., 3 (1975) 59 - 66. G. E. McClearn and D. L. Shern, Behav. Genet., 5 (1975) 103. R. Y. Kakihana, unpublished Doctoral Dissertation, University of California, Berkeley, 1965. W. F. Bodmer and L. L. Cavalli-Sforza, Genetics, Evolution and Man, Freeman, San Francisco, 1976. G. E. McClearn, J. Toxicol. Environ. Hlth., (1978) in press. G. E. McClearn et al., Nature, 203 (1964) 793 - 794. G. E. McClearn, J. R. Wilson and W. Meredith, in G. Lindzey and D. D. Thiessen (eds.), Contributions to Behavior-Genetic Analysis: The Mouse as a Prototype, Appleton-Century-Crofts, New York, 1970, pp. 3 - 22. G. E. McClearn and R. Kakihana, Behav. Genet., 3 (1973) 409 - 410. R. Kakihana and J. Moore, Behav. Genet., 7 (1977) 71. J. L. Fuller and A. C. Church, Behav. Genet., 7 (1977) 59. W. Heston and V. G. Erwin, Some biochemical parameters of alcohol metabolism in selected strains of mice, Paper presented at Behavior Genetics Association Meeting, 1972. D. S. Baer and D. W. Crumpacker, Behav. Genet., 7 (1977) 95 - 103. D. B. Goldstein and R. Kakihana, Life Sci., 17 (1975) 981 - 986. B. Sanders, J. Comp. Physiol. Psychol., 90 (1976) 394 - 398. J. W. MacInnes and R. P. Damjanovich, Behav. Genet., 3 (1972) 408. A. C. Collins and R. A. Deitrich, Behav. Genet., 3 (1973) 398. B. Sanders et al., Psychopharmacology, 56 (1978) 185 - 190. V. G. Erwin, W. D. W. Heston and G. E. M&learn, Pharmacol. Biochem. Behav., 4 (1976) 679 - 683.
16 20 21 22 23 24 25 26 27 28 29 30 31
B. C. Dudek, Behav. Genet., 8 (1978) 91. A. C. Church, J. L. Fuller and B. C. Dudek, Psychopharmacology, C. A. Greer and H. P. Alpern, Life Sci., 21 (1977) 385 - 392. H. P, Alpern and C. A. Greer, Life Sci., 21 (1977) 93 - 98. G. E. McClearn and S. M. Anderson, Behav. Genet., in press. K. Eriksson, Science, 159 (1968) 739 - 741. K. Eriksson, Ann. N. Y. Acad. Sci., 197 (1972) 32 - 41. E. P. Riley et al., J. Stud. Ale., 38 (1977) 1705 - 1717. T.-K. Li et al., Drug Ale. Depend., 4 (1979) 45. D. S. Falconer, Introduction to Quantitative Genetics, Ronald, F. Pirchner, Population Genetics in Animal Breeding, Freeman, S. M. Anderson, unpublished Doctoral Dissertation, University
47 (1976)
49
- 52.
New York, 1960. San Francisco, 1969. of Colorado, 1975.
