* Corresponding author: C. Downing, Institute for Behavioral Genetics, University of Colorado, Campus Box 447, Boulder, CO 80309–0447, USA. E-mail: email@example.com
Genetic influences on the psychomotor stimulant effect of ethanol may be a key feature of abuse liability. While earlier work has shown the activational effects of ethanol to be under the influence of a relatively uncomplicated additive genetic system, preliminary data from our laboratory suggested the possibility of nonadditive genetic variance. In the present study, a full Mendelian cross was conducted to further characterize gene action and search for quantitative trait loci (QTL) influencing the psychomotor stimulant properties of ethanol. We tested 3062 mice of the six Mendelian cross genotypes (P1, P2, F1, F2, BC1 and BC2) derived from a cross between the C57BL/6J (B6) and C3H/HeJ (C3H) inbred strains of mice. On day 1, mice were injected with saline, put in a holding cage for 5 min, then placed in an activity monitor for 5 min. On day 2, mice were injected with 1.5 g/kg ethanol, and activity again monitored for 5 min. Analysis showed the expected activation in the C3H strain and little activation in the B6 strain, with no effect of sex. Biometrical genetic analysis showed a best-fit model that included the mean (m), additive effect (a), and an epistatic parameter (i = homozygote by homozygote interaction). Analysis showed good evidence for QTL on chromosomes 1 (logarithm of odds (LOD) 3.4–7.5, 88–100 cM), 6 (LOD 9.1–10.4, 46–50 cM) and 15 (LOD 7.3–8.8, 28–32 cM). While the regions on chromosomes 1 and 6 have previously been implicated in several different ethanol-related phenotypes, this is the first report of a QTL influencing the psychomotor stimulant properties of ethanol on chromosome 15. Other studies have identified QTL in this region of chromosome 15 mediating locomotor activation caused by other psychostimulants, including cocaine, amphetamine and phencyclidine.
Human twin and adoption studies have consistently reported that alcoholism (alcohol abuse and dependence) is a heritable trait (Cadoret et al. 1980; Cadoret et al. 1995; Heath et al. 1997; Kendler et al. 1992, 1997; Prescott & Kendler 1999). Heritability estimates range from 0.24 to 0.73, which shows that genetic variation can have a significant influence on the development of alcohol-related problems. While the mode of inheritance is likely polygenic, it remains unclear precisely what characteristics are inherited. One possibility is an initial low level of response or decreased reaction to alcohol. Schuckit and colleagues have shown that in sons of alcoholics, an initial low level of response to alcohol is associated with a four-fold greater likelihood for developing alcohol-related problems (Schuckit 1988, 1994; Schuckit & Smith 1997).
Animal studies provide further evidence for a genetic influence on a number of alcohol-related measures, including dimensions thought to reflect acute ethanol sensitivity. Many measures of ethanol sensitivity have been reported in rodents (mice), including loss of the righting reflex, hypothermia, ataxia and locomotor activation and sedation (Crabbe 1983; Crabbe et al. 1994a,b,c; Dudek & Phillips 1990, 1991; Phillips et al. 1994). Psychomotor activation is a particularly compelling phenotype, given its role in models of addiction (Newlin & Thomson 1991; Wise & Bozarth 1987). The ability of low doses of alcohol to produce behavioral disinhibition and psychomotor arousal may also be an important component for some alcoholism typologies (Cloninger 1987).
The psychomotor stimulant properties of ethanol have been well studied in mouse models. Crabbe and colleagues have selectively bred for sensitivity and insensitivity to the locomotor activating effects of a low dose of ethanol (FAST and SLOW mice), which conclusively demonstrates genetic mediation of this phenotype (Crabbe et al. 1987; Phillips et al. 1991; Shen et al. 1995). Further support for a genetic influence comes from the finding that inbred strains of mice differ in their activational response to ethanol. Many studies have shown that while variability exists among inbred strains in response to an acute, low dose of ethanol, most strains become activated. (Crabbe 1983, 1986; Crabbe et al. 1980, 1982; Dudek et al. 1991; Tritto & Dudek 1994). One notable exception is the C57BL/6J (B6) strain, which seemingly lacks the activational limb of the biphasic alcohol dose–response curve. Studies have consistently reported either very little locomotor activation, or decreased activity, in B6 mice following a low dose of ethanol (Crabbe 1983, 1986; Crabbe et al. 1980, 1982; Dudek & Tritto 1994; Tritto & Dudek 1994). This anomalous B6 strain has proven to be a powerful tool for investigating the underlying genetic architecture of ethanol-induced psychomotor activation.
Earlier work has suggested that genetic control of the psychomotor stimulant properties of ethanol may be relatively straightforward. This phenotype has been characterized as predominantly additive, with no complicating factors such as sex linkage, maternal or sex effects, dominance or epistasis; the number of relevant loci is small, perhaps between 3 and 8 (Dudek & Tritto 1994; Dudek et al. 1991). However, a recent study from our laboratory (Caldarone 1997) found evidence for nonadditive genetic variance in the psychomotor stimulant response to ethanol, in a cross between B6 and C3H/HeJ (C3H) mice. As all possible crosses of the full Mendelian design were not conducted in this preliminary study, no firm conclusions could be drawn. Thus, in the present study, a full Mendelian cross was undertaken using B6 and C3H mice, in order to better characterize the mode of gene action in this system and identify quantitative trait loci (QTL) mediating ethanol-induced locomotor activation.
Materials and methods
Male and female mice from the B6 and C3H inbred strains were purchased from Jackson Laboratory, Bar Harbor, ME. Offspring from these original breeder pairs were used to produce F1, F2 and backcross populations of mice in our colony, using standard breeding methods. Both reciprocal F1 genotypes were produced, as well as the 12 reciprocal types of F2, backcross of F1 to B6 (BcB6) and backcross of F1 to C3H (BcC3H). Sample sizes for each genotype were as follows: B6 = 74, C3H = 75, F1 = 217, F2 = 1204, BcB6 = 794 and BcC3H = 680. Approximately equal numbers of males and females were tested for each genotype. Mice were maintained in our colony with a 16-h light/8-h dark cycle, lights on at 06.00. Temperature was maintained at 21°C (± 3°); food (Prolab RMH 2000 and 3000; Agway, Syracuse, NY) and water were available ad libitum at all times. All animals were between 55 and 75 days old-when tested, with testing done between 09.00 and 19.00; greater than 90% of the mice were tested between 12.00 and 17.00. Within litters, all mice were tested on the same days.
All mice were tested on four days, designated days 1, 2, 8 and 9. On day 1, all animals were treated with saline (0.9% NaCl), administered intraperitoneally (i.p). Immediately following injection, animals were placed in a holding cage for 5 min, then placed into an automated activity monitor for 5 min. Animals were then promptly removed and returned to their home cage. The procedure on day 2 was identical except that animals were injected i.p. with 1.5 g/kg ethanol (7.5%) at a constant injection volume of 20 ml/kg. The 5 min wait between injection and testing allowed for distribution and absorption of ethanol; previous studies from our laboratory have shown that many inbred strains of mice achieve peak blood ethanol levels (BEL) at 5 min following a 1.5-g/kg dose of ethanol (Dudek et al. 1999). Animals were not tested for the following 5 days. On days 8 and 9, animals were again tested using the same protocol. We chose to test animals one week later to examine the possibility of sensitization or tolerance to the psychomotor stimulant effects of ethanol. After testing was finished, animals were killed, and tail and spleen were removed for DNA extraction.
The activity monitoring system (digiscan: Accuscan Instruments Inc., Columbus, OH) consisted of a 16 beam photocell apparatus interfaced to a PC. Mice were placed into a 40-cm2 chamber with a 30.5-cm ceiling, and each monitor was enclosed in a larger sound-attenuating chamber. All testing was done in complete darkness, with a ventilating fan providing a masking background noise. While locomotor activity is typically tested in darkness in our laboratory, testing mice in complete darkness in the present study also compensated for the fact that C3H mice possess the retinal degeneration mutation and are essentially blind by about 30 days of age. The monitors provided counts of total photocell beam breaks, both vertical and horizontal. Computer software recorded a number of other indices, including total distance traveled (cm), distance traveled/time spent in the center and margins of the chamber and rest time. Two additional indices were derived which further clarified the nature of the behavioral activation. These were an average speed index (distance/movement time; cm/second) and an average movement length (distance/number of movements; cm/movement). Of these indices, we chose to focus on the total horizontal distance traveled measure (cm), as previous work from this laboratory and others (Dudek & Tritto 1994; Phillips et al. 1992; Tritto & Dudek 1994) has indicated the utility of this variable in characterizing ethanol-induced behavioral activation. The 1.5 g/kg dose of ethanol was chosen based on previous work from our laboratory, which indicates this dose is at the peak of the activational limb of the biphasic alcohol dose–response curve for many inbred strains of mice (Dudek et al. 1991; Dudek et al. 1999). Previous work has also established the appropriateness of a two-day procedure with saline administered on day 1 and ethanol administered on day 2. Such a procedure has yielded genotype differences similar in magnitude to between-groups procedures (Dudek & Tritto 1994; Tritto & Dudek 1994).
Genomic DNA was isolated from mouse tails using a DNeasy tissue kit (Qiagen, Valencia, CA) following the manufacturers instructions. Briefly, 0.6–1.2 cm mouse tail was placed into a centrifuge tube, to which 180 µl lysis buffer and 20 µg Proteinase K were added; the sample was then incubated in a rocking platform at 55°C overnight. After incubation, 400 µl of an additional lysis buffer was added to the tissue digest, centrifuged, and the tissue digest then centrifuged (twice) with 500 µl of a wash buffer. The tissue digest was then extracted twice with an elution buffer. Purity and concentration of the final samples were determined by ultraviolet spectroscopy.
Markers and genotyping
All genotyping was done using microsatellite markers (di-nucleotide repeats from the Massachusetts Institute of Technology (MIT) series) and polymerase chain reaction (PCR) amplification. Genomic DNA (5 ng) was amplified with 0.2 µl primer (20 mM), 1.2 µl Taq polymerase, 1.6 µl dNTPs (4 mM) and 4 µl of 5X RDA buffer (1 ml: 335 µl Tris (ph 8.8), 564 µl distilled H2O, 80 µl NH4SO4, 10 µl bovine serum albumin, 7.5 µl MgCL2 and 3.5 µl 2-mercapto ethanol). Reactions were performed using a Thermal Cycler (MJ Research, Waltham, MA) as follows: 5 min at 95°C, followed by 35 cycles at 95°C for 30 seconds, 1 min at 55°C and 1.5 min at 72°C. At the end of the amplification cycles, there was an extension cycle of 10 min at 72°C, and samples were then held at 4°C. PCR products were separated by electrophoresis in 0.5X TBE buffer on 3% or 4% agarose gels (NuSieve and Gibco BRL, Invitrogen Corp., Carlsbad, CA). Bands were visualized with ethidium bromide staining.
The behavioral indices derived from the activity monitors were submitted to analysis of variance (Statistical Package for Social Sciences (SPSS) manova procedure) with genotype and sex as independent variables. The sex factor did not interact with genotype for any behavioral measure, and thus all data reported are collapsed on sex. Analysis was done using change scores, which were calculated as day 2 (ETOH) – day 1 (SAL); for this initial report we present only data from days 1 and 2.
Quantitative genetic analysis of the difference score was performed using a model fitting approach that evaluated the contribution of additive, dominance and digenic epistatic interaction parameters. This was done with the use of joint-scaling tests, where parameters were generated from regression equations, and evaluated with the chi-square (χ2) goodness-of-fit test applied to genotypic means (Mather & Jinks 1982). The joint-scaling tests use theoretically defined contrast vectors as predictors (for additivity, dominance and epistasis) in multiple regression equations with the difference score measure of behavioral activation as the dependent variable.
We report here the first stage of QTL mapping, in which we selectively genotyped the top 5% and bottom 5% of the extreme responders from the F2 generation (134 mice) for 79 markers, spaced at an average of 17 cM. Selective genotyping is an efficient method for performing whole genome scans (Lander & Botstein 1989). In this preliminary stage of mapping, the focus was on three measures: (i) total distance traveled, measured as a difference score (ii) regression residuals, derived from regression of day 2 ethanol scores on day 1 saline scores and (iii) baseline activity, using day 1 scores.
We chose to perform QTL mapping on regression residuals, in addition to difference scores, because it has been argued that regression residuals are a better measure of individual differences in drug sensitivity (Nagoshi et al. 1986; Smolen et al. 1994). As discussed by Cohen and Cohen (1983), the problem with the use of change scores lies in their dependence on prescores (baseline activity in our case). If the correlation between prescore and change score is zero (which it generally is not), then one's change score measures what it purports to measure, change. However, if the correlation is not zero, then the change score contains some variance wholly due to prescore, which will distort the change score measure. Thus, individual differences measured by the change score may simply reflect regression toward the mean rather than treatment effects. The use of regression residuals, calculated as deviations of the observed postdrug scores from the value predicted by the regression of postdrug score (day 2 ethanol) on predrug baseline (day 1 saline), is attractive because they are by definition independent of baseline. The finding of similar QTL for difference scores and regression residual would support the utility of using difference scores to measure individual differences in response to drugs.
The 10% extreme responders for our difference score amounted to 120 mice; 106 of these 120 extreme responders were also in the 10% extreme for our regression residual measure, so an additional 14 animals with extreme regression residuals were genotyped to complete the 10%. Genotyping the 10% extreme responders for the day 8/9 measure, or the average of the two, will require much additional genotyping; very few F2 mice were in the extremes for both day 1/2 and day 8/9. This suggests that these are two somewhat genetically distinct phenotypes. Therefore, given that many previous studies reporting QTL mapping for ethanol-activation have used a 2-day paradigm, we chose to focus our initial mapping efforts on day 1/2 data.
Data were subjected to interval mapping using mapmaker qtl (Lander & Botstein 1989; http://www-genome.wi.mit.edu/ftp/distribution/software/mapmaker3) and mapmanager qtx version b17 (Manly & Olson 1999; http://mapmgr.roswellpark.org/mmQTX.html) to determine peak LOD scores and 1 LOD support intervals. mapmaker exp was used to create the genetic linkage map. For analyses using mapmaker qtl (mapmaker), significant linkage was determined using the guidelines recommended by Lander and Kruglyak (1995). For an F2 intercross, in which two parameters are estimated (additive and dominance effects; the ‘free’ or unconstrained model in mapmaker), the associated degrees of freedom are two; in this situation, LOD scores 2.8 or higher show suggestive linkage, while LOD scores 4.3 or higher show significant linkage. For analyses using mapmanager qtx (mapmanager) significance was determined using 5000 permutation tests, an option available in this mapping program.
We looked for interactions using the interaction function in mapmanager; this program searches for digenic epistasis by testing all pairs of marker loci (all chromosomes) for both main effects and interaction effects using regression. Pairs of loci must pass two tests in order to be reported as having a significant interaction effect: first, the total effect of the two loci must have a P-value < 10−5 (our criterion), and second, the interaction itself must have a P-value < 0.01. For each significant two–locus interaction, output from mapmanager includes likelihood ratio statistics (LRS, which we converted to LOD scores) for each main effect, the interaction effect and the total effect. In addition, mapmanager allows one to run permutation tests for two–locus interactions for a given dataset to determine threshold LRS statistics for declaring significance; we used 5000 permutation tests of our dataset for declaring a significant two–locus interaction.
Because sex did not interact with genotype for any of the behavioral indices, all means reported are collapsed across the gender variable. Reciprocal F1 differences were also not significant for any dependent variable except rest time (P < 0.05), thus effectively ruling out any maternal or sex linkage effects. Therefore, data from reciprocal groupings of F1, F2 and backcross genotypes were also collapsed. As expected, there were large differences in baseline activity between the parental strains, with B6 mice traveling more than twice the total distance of C3H mice on day 1 following saline administration (Fig. 1). All other genotypes were intermediate and in the expected direction based on gene dosage (i.e., F1 and F2 mice were intermediate and backcross generations resembled their respective parental genotypes; Fig. 1). This was verified by anova on day 1 saline scores, which showed that genotype accounted for 24% of the variance. Conversely, anova on day 2 ethanol scores indicated that genotype accounted for only 0.7% of the variance. Indeed, day 2 ethanol scores had a very narrow range, with a low of 2139 cm in B6 mice to a high of 2426 cm in BcC3H mice.
Ethanol-induced locomotor activation was analyzed using total distance traveled (cm), measured as difference scores; results are depicted in Fig. 2 as histograms. Again there was a large difference in the parental strains, with C3H mice significantly more activated than B6 mice (t-test, P < 0.001). Figure 2 shows that while there was not a great deal of overlap between the parental strains, there was significant variation within each parental strain, which illustrates an environmental effect on ethanol-activation. The F1 and F2 populations were intermediate, while each backcross tended to be in the direction of the respective parental strain, which gives the impression of a largely additive genetic system.
A one-sample t-test showed that the B6 mean for total distance traveled was significantly different from zero (P < 0.001), demonstrating activation in the B6 strain following a low dose of ethanol. Two recent studies from our laboratory (Caldarone 1997; C. Downing and B. C. Dudek, unpublished data) also detected a small degree of activation in the B6 strain. It should be noted, however, that while ethanol-induced locomotor activation in the B6 strain has been observed in these studies, activation is small, of a magnitude less than that seen in typically activating strains.
Biometrical genetic analysis of the total distance traveled measure (difference score) revealed predominantly additive inheritance, with some evidence for epistasis (Fig. 3). Model-fitting procedures indicated that a model including the mai parameters (mean (m), additive effect (a), and homozygote by homozygote interaction (i); Mather and Jinks 1982) provided a good fit, while a model including the mad (dominance (d)) did not fit the data. Because epistasis was detected, an estimate of the number of effective factors (QTL) was not feasible, as it would lead to a gross underestimate (Falconer & Mackay 1996). The finding of epistasis from the biometrical analysis is consistent with a polygenic mode of inheritance.
Analysis of the other indices provided additional insight into the nature of the behavioral activation. Results (manova-SPSS) showed that in relation to B6 mice, C3H mice had decreased rest time following ethanol administration and traveled at a higher rate of speed. In addition, C3H mice increased the amount of time spent and distance traveled in the center of the activity chamber and had decreased activity in the margins of the chamber (P < 0.001). Vertical activity was also suppressed less in C3H mice following ethanol administration (P < 0.001). The only measure on which the parental strains did not differ following ethanol administration was the movement length variable; ethanol increased the distance traveled per movement to the same degree in both strains. Visual inspection of the genetic triangles for the other indices (data not shown) indicated a relatively similar pattern of inheritance (largely additive) as that seen in the total distance traveled measure, with some degree of dominance for the margin and center distance measures and the running speed measure.
Data were next analyzed using mapmaker and mapmanager in order to determine peak LOD scores, support intervals and effects sizes. Marker positions were obtained from the Mouse Genome Database (MGD), and intermarker distances were calculated using these cM positions. We searched for QTL mediating both our difference score and regression residual variables. Regression residuals were calculated in SPSS as deviations of the observed postdrug scores from the value predicted by the regression of postdrug score (day 2 ethanol score; cm) on predrug baseline (day 1 saline score). Results from both mapping programs gave similar results for both traits.
The best evidence for QTL mapped to regions of chromosomes 1, 6 and 15. Table 1 shows the three QTL, one suggestive and two significant, mediating ethanol-induced activation as determined by mapmaker. On chromosome 1, our difference score measure showed a QTL with a LOD score of 3.4 (suggestive linkage) at approximately 80 cM, while the regression residual measure showed a QTL with a LOD score of 6.4 (significant linkage) slightly more distal, at approximately 100 cM. This QTL accounts for 3–5% of the phenotypic variance. On chromosome 6 a QTL was identified at approximately 47 cM for both our difference score and regression residual measures, with a LOD of 9.8–10.4, that shows significant linkage. This QTL accounted for 9–10% of the phenotypic variance. Finally, we found good evidence for a QTL for our difference score and regression residual measures on chromosome 15 at approximately 28 cM. This QTL had a LOD score of 7.3–8.8 (significant linkage) and accounted for about 9% of the phenotypic variance.
Table 1. : QTL detected, location and effect sizes for our difference score and regression residual measures of ethanol-induced locomotor activation using mapmaker qtl
Position of maximum LOD score for our difference score measure of ethanol-induced locomotor activation; for example, for the QTL on chromosome 6, the peak LOD score occurred at D6Mit9 (36.5 cM) + 10 cM = 46.5 cM, taken from the interval mapping output from mapmaker. All cM positions are taken from the MGI database. According to the Lander & Kruglyak (1995) criteria the QTL on chromosome 1 showed suggestive linkage while the QTL on chromosomes 6 and 15 showed significant linkage.
† Percent of phenotypic variance accounted for by QTL taken from mapmaker qtl output.
Position of maximum LOD score for our regression residual measure of ethanol-activation. All three QTL showed significant linkage.
D1Mit15 + 0 cM
D1Mit403 + 0
cM (100.0 cM)
D6Mit9 + 10 cM
D6Mit9 + 10 cM
D15Mit111 + 10
+ 10 cM (27.8 cM)
cM (27.8 cM)
One useful feature of mapmaker is that it will determine the probability that the effect of a QTL allele follows a Mendelian model, either dominant, recessive or additive. For the QTL on chromosome 1, comparison of LOD scores calculated with the constraint of a dominant, recessive or additive mode of inheritance against the LOD score calculated under the free model shows the highest likelihood for the free model, which can be seen in Fig. 4(a). The dominant model had a log-likelihood similar to the free model, and provided a better fit than the additive and recessive models. Inspection of marker class means (Table 2) for 3 markers on chromosome 1 flanking the QTL (D1Mit 15, 26 and 403) also supports the observation of dominance at this QTL. T-tests (P < 0.01) showed that the F1 value was significantly different from the midparental value; mice heterozygous for these 3 markers showed dominance toward mice homozygous for the C3H allele. For the QTL on chromosome 6 and 15 (Fig. 4b,c) the free and additive models had the greatest likelihood.
Table 2. : Marker class means for markers on chromosomes 1, 6 and 15 flanking QTL mediating our difference score measure (total distance traveled in cm) of ethanol-induced activation. BB = B6J homozygote, BH = heterozygote, HH = C3H homozygote
For the QTL on chromosomes 1 and 6, mice homozygous for the B6 allele actually had significantly higher ethanol-activation scores than mice homozygous for the C3H allele, while for the QTL on chromosome 15, mice homozygous for the C3H allele had higher ethanol-activation scores (Table 2). anova also showed a significant sex by marker interaction at D1Mit15 (P < 0.02). The homozygote marker class mean difference was apparent for females [mean difference score of 2897 ± 447 (SEM) for B6 homozygotes vs. 297 ± 254 for C3H homozygotes], but not for males (mean difference score of 1974 ± 326 for B6 homozygotes vs. 1745 ± 428 for C3H homozygotes), which implies that this QTL was specific for females.
The results from mapping with mapmanager were quite similar to the results we found using mapmaker. For our difference score measure of ethanol-activation, permutation tests (5000) showed that a LOD score of 2.00 was necessary for suggestive linkage, 3.45 for significant linkage, and 5.27 for highly significant linkage. For our regression residual measure, a LOD score of 2.00 was necessary for suggestive linkage, 3.44 for significant linkage, and 5.36 for highly significant linkage. Table 3 shows all QTL which showed suggestive or significant linkage. As with mapmaker, we found good support for QTL mediating ethanol-activation on chromosomes 1, 6 and 15. In addition, mapmanager identified suggestive QTL on chromosomes 2, 3, 5 and 9.
Table 3. : QTL detected, location and effect sizes for our difference score and regression residual measures of ethanol-activation using mapmanager qtx
Position of maximum LOD score for our difference score measure of ethanol-activation; for example, for the QTL on chromosome, the peak LOD score occurred at D1Mit15 (87.9 cM) + 2 cM = 89.9 cM. Marker positions are based on MGI cM positions. Permutation tests showed that a LOD score of 2.00 was necessary for suggestive linkage, 3.45 for significant linkage and 5.27 for highly significant linkage.
Percent of phenotypic variance accounted for by QTL, dividing by a factor of 1.59, as described in Darvasi and Soller (1992) to account for the effects of selective genotyping (see text).
Position of maximum LOD score for our regression residual measure of ethanol-activation. Permutation tests showed a LOD score of 2.00 was necessary for suggestive linkage, 3.45 for significant linkage and 5.36 for highly significant linkage.
D1Mit15 + 2 cM
D1Mit15 + 10 cM
D2Mit241 + 14 cM
D3Mit40 + 0 cM
D5Mit233 + 0 cM
D6Mit9 + 12 cM
D6Mit9 + 14 cM
D9Mit133 + 12 cM
5.4 × 10−8
14 cM (31.8 cM)
14 cM (31.8 cM)
When performing selective genotyping using maximum likelihood methods, if one enters phenotypic data from the entire population tested and enters genotype data from only the extremes, this should (theoretically) provide unbiased estimates of effect size; maximum likelihood methods can account for missing data (Lander & Botstein 1989). However, when mapping using a regression residual approach, one enters phenotypic and genotypic data from only the extremes, which leads to an upward bias in effect size estimates (Darvasi & Soller 1992). Darvasi and Soller (1992) have derived calculations to take into account this effect of selective genotyping on effect size estimates. Therefore, in regard to the mapmanager estimates of effect size, we have divided these values by 1.59 (difference score) and 1.62 (regression residual), using the calculations from Darvasi and Soller (1992), to take into account the effects of selective genotyping.
Permutation tests (5000) in mapmanager for our interaction analysis (all possible two–marker interactions) showed that a LOD score of 6.06 was necessary for a suggestive interaction, 7.88 for a significant interaction and 10.36 for a highly significant interaction. Using the criterion that the total effect of the two loci must have a P-value < 10−5, and the above LOD thresholds must be met for declaring a significant interaction, we found no suggestive, significant or highly significant interactions for our difference score measure of ethanol–activation. Digenic interaction analysis for our regression residual variable revealed similar threshold LOD scores for declaring a significant interaction (data not shown); similar to our difference score measure, no suggestive, significant or highly significant interactions were detected.
Results from QTL mapping for baseline activity (day 1 saline scores; cm) are presented in Table 4. mapmaker identified 3 QTL mediating baseline activity, on chromosomes 6 (suggestive), 12 (significant) and 15 (suggestive). These QTL each explained approximately 18–24% of the phenotypic variance. Permutation tests (5000) in mapmanager showed that for suggestive linkage a LOD score of 1.97 was necessary, 3.47 for significant linkage, and 5.01 for highly significant linkage. mapmanager identified three QTL showing suggestive linkage, on chromosomes 5, 6 and 15, and a QTL on chromosome 12 showed significant linkage. Using calculations from Darvasi and Soller, the effect sizes reported by mapmanager were divided by 1.77 to take into account the effects of selective genotyping. The 4 QTL identified by mapmanager each accounted for 8–14% of the phenotypic variance.
Table 4. : QTL detected, location and percent of phenotypic variance explained for our baseline activity measure using both mapmaker qtl and mapmanager qtx
Position of maximum LOD score for our baseline activity measure (total distance traveled on day 1) as determined by mapmaker qtl. All marker positions are based on MGI cM positions. According to the guidelines of Lander and Kruglyak (1995) the QTL on chromosomes 6 and 15 showed suggestive linkage while the QTL on chromosome 12 showed significant linkage.
Percent of phenotypic variance accounted for as determined by mapmaker qtl.
Position of maximum LOD score for baseline activity as determined by mapmanager qtx. Permutation tests showed that a LOD score of 1.97 was necessary for suggestive linkage, while a LOD score of 3.47 was necessary for significant linkage.
Percent of phenotypic variance explained as determined by mapmanager qtx, divided by 1.77 using equations from Darvasi and Soller (1992) to account for selective genotyping.
D5Mit233 + 16
cM (45.0 cM)
D6Mit254 + 6
D6Mit9 + 12
cM (66.5 cM)
cM (48.5 cM)
D12Mit124 + 8
D12Mit124 + 4
cM (21.0 cM)
cM (17.0 cM)
D15Mit156 + 6
D15Mit156 + 4
cM (45.1 cM)
cM (43.1 cM)
The psychomotor stimulant response is a measure of drug sensitivity which is believed to reflect the rewarding or reinforcing properties of alcohol. Results from this study support previous work showing the typical activating response to a low dose of ethanol in the C3H strain and little activation in the B6 strain. Crabbe et al. (1994c) have shown that B6 and C3H mice do not differ in brain ethanol levels following a low dose of ethanol (1 or 2 g/kg), which implies that the differences in ethanol-induced activation between these strains are not due to differential ethanol absorption or metabolism, but are more likely due to differences in neural sensitivity.
Previous biometrical genetic analyses have characterized the locomotor stimulant response to ethanol as being under relatively simple genetic control, comprised of a primarily additive system with few complicating factors such as maternal or sex effects, dominance or epistasis (Dudek & Tritto 1994; Dudek et al. 1991). Results from the present study partially support this characterization. In our cross between B6 and C3H mice, a primarily additive mode of inheritance was observed for ethanol-induced locomotor activation (difference score). However, in contrast to earlier work, some evidence for epistasis was detected. Model-fitting indicated that a model including a homozygote by homozygote interaction fit the data, which is also supported visually in the genetic triangle in Fig. 3. This result was also observed in the Caldarone study (1997), which showed that for ethanol-induced activation in a (B6 × C3H) F1 backcross to B6, the average value for the backcross mice was significantly different from the average of the ((F1 + B6)/2), which indicates the presence of epistasis. One reason epistasis was detected in these two studies (Caldarone 1997; present study), and not in earlier biometrical genetic analyses of this phenotype (Dudek & Tritto 1994; Dudek et al. 1991) could very well be the use of mapping populations derived from C3H mice.
While the best fit model did not include the dominance parameter, one could argue that since the F1 value was significantly different from the midparental value for our difference score measure of ethanol-activation (Fig. 4), dominance was present. An alternative explanation, however, is that the F1 difference from the midparental value is due to a homozygote by homozygote interaction. The finding that the constrained model of dominant inheritance closely matched the unconstrained model for the QTL on chromosome 1, and provided a better fit than the additive or recessive models, suggests that the QTL affecting ethanol activation on chromosome 1 may be inherited in a dominant manner. This is further supported by the data in Table 2, which shows that for markers on chromosome 1, heterozygotes exhibited dominance toward the C3H strain. Thus, this phenotype appears to be primarily additive, with some evidence for a homozygote by homozygote interaction; the QTL on chromosome 1 may exhibit dominant inheritance.
Given that epistasis is present, the use of Sewall Wright's formula to determine the number of loci contributing to ethanol-activation will lead to an underestimate; biometrical and QTL analyses from our study support a polygenic model. This is in line with previous biometrical and QTL studies showing a polygenic mode of inheritance for this phenotype (Cunningham 1995; Demarest et al. 1999, 2001; Dudek & Tritto 1994; Dudek et al. 1991; Erwin et al. 1997; Gill et al. 2000; Hitzemann et al. 1998, 2000; Phillips et al. 1995). The finding of no suggestive or significant interactions in the mapmanager program is not in agreement with our biometrical analysis, which indicated a homozygote by homozygote interaction. However, this is perhaps not too surprising given that we only genotyped 120 animals for each measure; extremely large sample sizes are necessary to reliably detect epistasis, particularly homozygote by homozygote interactions. Genotyping additional mice from our dataset should help resolve this issue.
Several previous studies have identified chromosomal regions containing QTL which mediate the psychomotor stimulant response to ethanol. In the present study, good support has been shown for QTL influencing ethanol-induced locomotor activation on chromosomes 1, 6 and 15. The QTL observed on chromosome 1 is in the same region as a QTL mediating ethanol-activation identified by the Hitzemann laboratory (Stonybrook, NY) (Demarest et al. 1999, 80–90 cM, BXD RIs; Hitzemann et al. 2000, 70–100 cM, B6 × D2 F2 intercross), and is near a marker/QTL identified by Erwin et al. (1997, 103 cM, Long-Sleep × Short-Sleep RIs). This region of chromosome 1, from about 83–102 cM, has been identified as a ‘hotspot’, where several QTL have been localized which influence behavioral and physiological responses to alcohol, including alcohol preference, acute alcohol withdrawal, alcohol-induced hypothermia, sensitivity and tolerance to alcohol-induced ataxia and alcohol conditioned taste aversion (Crabbe et al. 1994a; Gallaher et al. 1996; Risinger & Cunningham 1998).
We also found good support for a QTL affecting ethanol-activation on chromosome 6. The one-LOD support interval overlaps with a QTL on chromosome 6 identified by Cunningham (1995, 41 cM, BXD RIs). The Hitzemann laboratory has also mapped what may be two separate QTL to a similar region of chromosome 6, one at 20–55 cM, and one at approximately 62 cM (Demarest et al. 1999, BXD RIs; Hitzemann et al. 2000, BALB/cJ × lP/J F2 intercross). This general region of chromosome 6 (32–45 cM) has also been identified as a hotspot for drug-related processes, including ethanol consumption, tolerance to the ataxic properties of alcohol, ethanol-induced hypothermia and chronic ethanol withdrawal severity (Buck et al. 1997; Crabbe et al. 1994a,b; Erwin et al. 1997).
For the QTL on chromosomes 1 and 6, the B6 allele was associated with greater activation scores. While reversal of genotype based on parental phenotype may seem puzzling, it is in line with results from other studies showing B6 alleles are associated with greater ethanol-activation (Cunningham 1995; Phillips et al. 1995). As noted above, the Hitzemann laboratory has also identified QTL on chromosomes 1 and 6 mediating ethanol-activation (Demarest et al. 1999, 2001; Hitzemann et al. 1998, 2000); when reported, B6 alleles on chromosomes 1 and 6 were associated with greater ethanol-activation (Demarest et al. 2001). One reason for this is likely due to use of inbred strains to create the mapping populations. Unlike selective breeding, fixation of alleles in inbred strains is a chance occurrence; thus, while B6 mice most likely carry alleles conferring resistance to ethanol-stimulated activity, they also carry alleles for sensitivity.
Interestingly, QTL which influence cocaine-induced locomotor activation have also been mapped to this region of chromosome 15. Using BXD RI mice, both the Phillips laboratory (Portland, OR) and the Jones laboratory (University Park, PA) have identified regions on chromosome 15 (34–48 cM, and 39.6 cM, respectively) which affect the psychomotor stimulant response to cocaine (Jones et al. 1999; Phillips et al. 1998). Boyle and Gill (2001), using AXB/BXA RI mice, have also mapped a QTL mediating cocaine activation to approximately 47 cM on chromosome 15. This region of chromosome 15 has also been associated with amphetamine (14.5 cM, 47.5 cM, 58.1 cM) and phencyclidine (40 cM) induced-locomotor activation (Alexander et al. 1996; Grisel et al. 1997). While the QTL identified in the present study affecting ethanol-induced locomotor activity had a peak LOD score slightly more proximal than some of these regions, all markers tested on chromosome 15 showed significant associations with ethanol-induced locomotor activation. It is possible that one or more QTL influencing the psychomotor stimulant response to several drugs of abuse are located in this region of chromosome 15, from approximately 23–50 cM. The Jones et al. study (1999) also mapped QTL influencing dopamine receptor densities (D1, D2, and dopamine reuptake transporter) to this region of chromosome 15. The role of dopamine systems in the actions of ethanol and other psychomotor stimulants has been well documented.
The QTL we identified for baseline activity replicated other studies mapping QTL for open-field activity (OFA) and locomotor activity. While studies which measure OFA are typically conducted in a brightly lit open-field (Flint et al. 1995; Turri et al. 2001), studies which test for locomotor activity are conducted under either light (Koyner et al. 2000; Phillips et al. 1995) or dark (Radcliffe et al. 1998; present study) conditions, typically with a fan providing a masking background noise. Despite these procedural differences, the QTL we identified for baseline activity on chromosomes 6, 12 and 15 have been identified in previous studies measuring OFA and locomotor activity (Flint et al. 1995; Phillips et al. 1995; Radcliffe et al. 1998; Turri et al. 2001). Thus, while OFA and locomotor activity may not be precisely the same constructs, it appears as if a subset of genes may mediate these two phenotypes.
The fact that two of the three QTL regions identified for ethanol-induced locomotor activation also contained QTL influencing baseline activity suggests that they may be the same QTL; alternatively, they may be different QTL mediating different phenotypes, located in the same region. One possibility is that these two QTL, on chromosomes 6 and 15, mediate spontaneous locomotor activity, and are ‘hyper-expressed’ following a low dose of ethanol. However, results from our study suggest that these QTL may mediate different phenotypes. The purpose of using difference scores and regression residuals is to account for differences in baseline activity. The finding that both measures of locomotor activation, difference scores and regression residuals, showed virtually identical QTL positions and LOD scores suggests that mapping for behavioral activation following a low dose of ethanol was relatively unconfounded by baseline activity differences. It further supports the use of difference scores for measuring individual differences in drug sensitivity.
The results of the present study represent a first stage of gene mapping for QTL which influence the locomotor response to ethanol. The three provisional QTL identified need to be confirmed. The next step will involve verification using the same F2 mapping population, by saturating these three regions with additional markers and genotyping additional mice. These provisional QTL also need to be confirmed using a different mapping population, such as our backcross populations and short-term selection mice which we have created from an independent F2 population derived from B6 and C3H mice.
In summary, it has been shown that the genetic architecture underlying ethanol-induced locomotor activation can be characterized as a primarily additive system, with some evidence for dominance and epistasis. Three provisional QTL were identified, on chromosomes 1, 6 and 15, which together account for approximately 22% of the phenotypic variation in ethanol activation. Two of these regions, on chromosomes 1 and 6, have previously been associated with ethanol activation, while to the best of our knowledge, this is the first study to identify a region on chromosome 15 affecting this phenotype. This region of chromosome 15 is also interesting because QTL affecting the stimulant properties of several other drugs have been mapped nearby. Identification of QTL affecting the psychomotor stimulant response to ethanol in mice should help to identify homologous genes in humans.
This research was supported by National Institute on Alcohol Abuse and Alcoholism grants AA09038 and AA00170 (BCD) and National Institute of Mental Health grant MH58599 (LF). The authors wish to thank Barbara Ashe, Corey Bennett and Jason Isabelle for their excellent technical assistance.