Behavior Genetics, VoL 22, No. 6, 1992

M a p p i n g Quantitative Trait Loci for Behavioral Traits in the Mouse Thomas E. Johnson, 1 John C. DeFries, 1 and Paul D. MarkeP

After many years of studying various behavioral characters in the mouse, it is clear that most are heritable and are specified by complexes of genes or quantitative trait loci (QTLs). In order to attain a more complete understanding of the genetic causes of individual differences in behavior, the mechanism of action of these QTLs must be elucidated. The most straightforward approach to determining the mechanism of action of a particular QTL is to identify and molecularly clone the gene; this can be done by positional cloning, which depends on precise knowledge of the genetic map position. As the genetic data base for the mouse genome continues to develop, such strategies will become increasingly easy to perform. Here we suggest a multistage strategy for QTL mapping using recombinant-inbred strains of mice: (1) characterize genomic DNA from parental strains originally used to generate the R[ strains; (2) characterize the RI strains for a quantitative character and for DNA markers that differ in the parental strains; and (3) assess the quantitative character in Fz mice derived from crosses between the parental strains, then determine the genotypes of extreme F2 mice for markers that account for at least 5% of the additive genetic variance. Data from these F 2 crosses can be used to test hypotheses from the analysis of R l strains, i.e., that a QTL maps to a particular region. Using data from the mouse genome data base, this strategy should allow the molecular identification of the gene based on a candidate-gene approach. We illustrate the approach with examples from our work in mapping QTLs specifying neural sensitivity to the anesthetic effects of ethanol KEY WORDS: recombinant inbred strains; genetic markers; selected lines; Mus mus-

culus; alcoholism.

This work was supported by grants from the National Institutes of Health (R01-AG08332, R01-AG10248, and K04-AG00369), by a gift from the Glenn Foundation for Medical Research, from ADAMHA (P05 AA-03527), by a training grant from the NIMH (MH16880) to support P.D.M., and by BRSG Grant RR-07013 to the University of Colorado. 1 Institute for Behavioral Genetics and Department of Psychology, Box 447, University of Colorado, Boulder, Colorado 80309. 635 0001-8244/92/1100-0635506.50/0 9 1992 Plenum Publishing Corporation

636

Johnson, DeFries, and Markel

INTRODUCTION In 1980 Botstein et al. (1980) proposed the construction of a human genetic map based on restriction fragment length polymorphisms (RFLPs). Only a few years later the first gene to be mapped using this strategy-the gene leading to Huntington's disease (HD)--was localized to chromosome 4 (Gusella et al., 1983). The next step in the unraveling of the causes of this behavioral alteration involves the cloning of the HD gene. This requires positioning the gene precisely to the genetic map based on human pedigree analysis. Such information is then used in conjunction with the physical genetic map to identify the gene in a strategy sometime misnamed "'reverse genetics" but more properly referred to as "positional cloning." The difficulty of this approach is evidenced by the fact that the HD gene has still not been cloned (Bates et al., 1991; Pritchard et al., 1991). In contrast, a recent and spectacular example of the successful use of positional cloning was the isolation of the gene for cystic fibrosis (CF; Rommens et al., 1989). The Human Genome Project (HGP; Watson, 1990) has as one of its immediate goals the construction of a human genetic map with an average spacing between markers of 2 cM and with no gaps greater than 5 cM; this will greatly facilitate positional cloning in humans. The HGP has also funded genome mapping and DNA sequencing in a variety of other organisms including the laboratory mouse. Recent developments in the physical mapping of the mouse genome make positional cloning of genes involved in the specification of various behavioral characteristics increasingly possible. However, in contrast to the situation for HD and CF, most behavioral variations studied in the mouse are not specified by a single genetic locus. Instead, most behaviors are determined by a number of genes (Plomin, 1990) often termed quantitative trait loci or QTLs (Gelderman, 1975; Lander and Botstein, 1989). The development of a high-density genetic map offers the possibility of precise mapping of these QTLs (Lander and Botstein, 1989), thus allowing their eventual cloning by position. Currently, more than 500 RFLPs have been identified and mapped in the mouse (Elliott, 1991). In addition, a new class of molecular markers has recently been developed for use in genetic mapping (Love et al., 1990; Aitman et al., 1991; Cornall et al., 1991; Hearne et al., 1991); these markers are based on polymorphisms in microsatellite sequences. Since microsatellites are very abundant (> 50,000 copies), widely dispersed throughout the mouse genome, and very polymorphic, they serve as excellent DNA markers from which

M a p p i n g Quantitative Trait Loci for Behavior in Mice

637

to map. Moreover, with the advent of the polymerase chain reaction (PCR; Saike et al., 1985), small amounts of DNA (0.1 ~g or less) can be amplified in 3 h, and size differences as short as 2 base pairs (bp) can be readily detected. PCR should facilitate a much more rapid mapping of genes than methods used to detect RFLPs. Plomin et al. (1991), McClearn et al. (1991), and Gora-Maslak et al. (1991) have advocated the use of recombinant inbred strains to map QTLs specifying behavioral traits in mice. Employing the method proposed by this group, QTLs specifying behaviors such as ethanol acceptance (Crabbe et al., 1983; McClearn et al., 1991; Gora-Maslak et al., 1991), audiogenic seizure susceptibility (Seyfried et al., 1980; Neumann and Seyfried, 1990; Neumann and Collins, 1991; Plomin et al., 1991), and drug sensitivity or consumption (Crabbe et al., 1983; Gora-Maslak et al., 1991) have been assigned tentative map positions. Such mapping is based on the association between mean scores of recombinant-inbred (RI) strains on a behavioral test and DNA polymorphisms that have been previously mapped in each RI strain. RI strains of mice were first developed in studies of transplantation immunity by Bailey (1971). In the most widely used series of RI strains, the B x D (C57BL/6J • DBA/2J), strain distribution patterns of several hundred restriction fragment length polymorphisms (RFLPs) and microsatellite STSs have been determined. The major advantage of RI strains is that they are almost fixed (97% homozygosity after 20 generations of sib mating) for one or the other parental allele at every locus for which the parental strains differ. RIs thereby provide a genotype that remains consistent over time, thus allowing repeated sampling of the same genotype. Major disadvantages of RI strains of mice are the length of time and the considerable expense involved in generating and maintaining the collection; thus, relatively few strains are generated from any given pair of parental inbred strains. RI strains have been widely used for genetic mapping of single-gene loci using both phenotypic and molecular markers. These markers are arranged into a series of strain distribution patterns (SDP) and significant similarities in SDPs is indicative of linkage. In genetic terms this similarity results from the decreased probability of recombination events between two genes that are physically close together on the chromosome. However, a fourfold expansion of the genetic map occurs during the generation of the RI series (Haldane and Waddington, 1931; Bailey, 1981), which makes it more difficult to detect linkage between a marker locus and a loosely linked gene that specifies some fraction of the variance of a behavior among the RIs. Many lines of the mouse and rat have been genetically selected to display large differences in sensitivity to, and/or preference for, ethanol.

638

Johnson, DeFries, and Markel

A partial survey of such lines includes long-sleep (LS) and short-sleep (SS) mice (McClearn and Kakihana, 1981), which differ with respect to alcohol-induced loss of the righting response; withdrawal seizure-prone (WSP), withdrawal seizure-resistant (WSR), and control (WSC) mice (Crabbe et al., 1986), which were selected on the basis of seizure susceptibility after ethanol withdrawal; and preferring (P) and nonpreferring (NP) rats (Li et al., 1981), which were selected on the basis of preference for a 10% ethanol solution versus water. Other selected lines have been and are being developed (for a recent review see Crabbe, 1989) and may eventually provide even better animal models for components of the addiction process. The LS/SS mouse lines have been a major focus of alcohol research at the Institute for Behavioral Genetics. The long-sleep (LS) and shortsleep (SS) lines were selectively bred to be differentially sensitive to the anesthetic action of alcohol (McClearn and Kakihana, 1981) and have been maintained by quasi-random mating after selection was suspended. No animal model has been more intensively studied than the LS and SS lines. These lines do not differ markedly in the rate of ethanol metabolism in vivo but display differences in the dose of ethanol required to cause a loss of righting response and in blood ethanol concentration at the time of regaining the righting response (Heston et al., 1974; Erwin et al., 1976). This differential CNS sensitivity has been traced to the level of the Purkinje cell (Sorenson et al., 1980) and is maintained both in vitro (Basile et al., 1983) and in intraocular transplants (Palmer et al., 1982). Thus the LS/SS lines are very suitable as a model for the study of CNS sensitivity to alcohol, even though they may not be an ideal model for human alcoholism. Recently, the generation of LS x SS RI strains (DeFries et al., 1989) provided the basis for estimates that seven loci are involved in specifying the differences in sleep time between the LS and the SS lines. One of these loci may account for over 8% of the additive genetic variance in sleep time and has been tentatively localized near the c locus (tyrosinase) on linkage group 7. Dudek and Abbot (1984), using F1, Fz, and backcross individuals derived from SS x LS reciprocal matings, estimated that there were about eight genes that contributed to the observed differences in sleep time and showed that this differential sensitivity to the anesthetic effect of alcohol fits an additive genetic model. The fit to an additive model suggests that most of the genetic difference in sleep time can be detected in RI strains. Unfortunately, two inherent problems result from the limited number of RI strains available: (1) there is limited precision in the assignment of map position, and (2) there are a large number of false positives due

Mapping Quantitative Trait Loci for Behavior in Mice

639

to statistical artifacts. For example, with only 27 RI lines, a behavior that is completely heritable, with no epistasis, and is specified by a single gene could be mapped to a 95% confidence interval of some 7 cM (Silver and Buckler, 1986)--a distance of some 20 million bp in the mouse. If the behavior results from several QTLs with reduced heritability, a severe decrease in the precision of mapping is likely. Even the optimal level of mapping assignment is unlikely to be useful in directed gene cloning because an interval of this size could contain a thousand genes. More importantly, the assignment of QTLs to a particular genetic region should be viewed as a tentative assignment. Many type I errors can result from this approach when many markers are employed, thereby resulting in false positives in the proposed mapping strategy. These false positives can be eliminated by characterization of F2 or backcross individuals from crosses between the parental strains to confirm QTL assignments. Our goal is to identify the genetic alterations that are causally related to the differences in sleep times between the LS and the SS strains. To do this we are pursuing a strategy to map the QTLs that lead to this differential sensitivity. The precision of our mapping is not limited by the number of RI strains, nor is the statistical analysis prone to type I errors because all correlations detected in the RIs will be confirmed by analysis of F2 mice. Similar strategies could be employed to confirm the tentative associations of behavioral traits with genetic map positions described by Gora-Maslak et al. (1991), Plomin et al. (1991), and McClearn et al. (1991). The strategies we are employing are described below. DESCRIPTION OF MULTISTAGE STRATEGY DeFries et al. (1989) estimated that the heritability of sleep time in LS • SS RIs was 42 __. 7%. They also estimated that there were seven QTLs accounting for this genetic variance. This estimate should be viewed as a minimal estimate because of the assumptions of complete and equal additivity and nonlinkage of each QTL. Thus each QTL, on the average, could explain about 6% of the total phenotypic variance. Because of the problems alluded to above, effects of this size cannot be reliably mapped using only the 27 LS x SS RIs available. We have therefore decided to use a multistage strategy to map sleep-time QTLs. A positive association at one stage indicates that the molecular marker may be linked to a QTL and that marker is then used at the next stage of analysis to confirm or disprove the prior assignment. Stage 1 involves screening with a probe to ensure that a difference between LS and SS lines of mice is detectable. If a difference is detected,

640

Johnson, DeFries, and Markel

a large sample of LS and SS mice will be screened to ensure that the marker is characteristic of all members of the line. Stage 2 involves using that marker to screen LS x SS RI lines. If the mean sleep time of those lines carrying the LS haplotype is at least 14 min longer than the mean sleep time of those mice carrying the SS haplotype (i.e., it explains at least 5% of the genetic variance and has an effect that is significant a t p _< .10), the marker will be used to characterize the 100 longest-sleeping mice and the 100 shortest-sleeping mice from 1000 LS x SS F2 mice (Lander and Botstein, 1989). For those markers that also manifest an association at the .01 level in the F 2 generation, the probability of type I error for linkage to a QTL will be less than .001. This approach is conservative of time and resources. If one uses a 5-cM reference map, then molecular probes that are spaced at this distance must be assayed; thus, some 340 probes will be needed to map the entire mouse genome. With QTLs specifying a small percentage of the total genetic variance, some 1000 F2 mice need to be surveyed (see below) to give reasonable power. Thus 340,000 data points would need to be collected--a formidable undertaking indeed. Using the multistage strategy, many fewer data points are needed because markers that fail to yield significant results in early analyses will not be examined further in F2 mice. Stage 1: Detection of Sequence Variation. In stage 1, LS and SS mice are characterized using probes that map to physical sites distributed approximately every 5 cM throughout the genome--a total of some 340 sites--to determine (a) if the pattern of bands (haplotype) in the LS is different from that in the SS and (b) if that haplotype difference is consistently observed between the LS and the SS mouse lines. Two types of probes have been used: cloned genes that detected restriction fragments that are known to be polymorphic in one or more of the parental strains from which the HS stock was derived (McClearn and Kakihana, 1981) and microsatellite sequences that have been shown to be polymorphic at a known site in the genome (Love et al., 1990). RFLPs are detected using Southern blots, while microsatellite variation is detected by PCR. Since the LS and SS mice were selected from HS mice (derived from an eight-way cross of inbred strains), only those sites displaying RFLPs in two or more of the inbred strains making up the HS stock will be informative for RFLP analysis. (It should be noted that microsatellites evolve rapidly so that new polymorphisms, not in any of the parental inbred strains, could be present, although this has not been documented by examination of the HS or the inbred lines from which the HS was established.) Genomic DNA from two LS and two SS mice are charac-

Mapping Quantitative Trait Loci for Behavior in Mice

641

terized by either Southern or PCR techniques. If a difference is detected, six LS or SS mice from different families having no common parents are examined. Strong selection and substantial inbreeding occurred during the generations of the LS and SS lines. Therefore, any QTL conferring increased sensitivity or decreased senstivity to alcohol will be expected to be isomorphic (display only on haplotype) in the LS and differ from the haplotype in the SS (LS is estimated to be > 80% inbred and SS to be > 65% inbred). If the six LS mice display a consistent haplotype that differs from the haplotype of the six SS mice, a QTL may be nearby and the region is further examined for ethanol sensitivity in Stage 2. If the LS and the SS lines are not consistently different, the locus is not completely fixed; lack of consistent haplotype differences between SS and LS mice thus serves as a strong indicator that the RFLP is unlikely to be linked to a QTL with a major effect on sleep time. Stage 2: Exploratory Data Analysis Using RIs. Polymorphic probes that passed the stage 1 test are next used to characterize haplotypes in each of the 27 LS x SS RI strains. The frequencies of the LS and SS alleles for each QTL are expected to be 0.5 in the F2 population derived from the initial cross of the LS and SS lines. Thus, the additive genetic variance (Falconer, 1989) due to segregation at a QTL is 1//2 (a) 2

where a is the "average effect'" of the LS allele (i.e., half of the difference between the means of the two homozygotes). Given that the additive genetic variance for ethanol-induced sleep time due to segregation at all loci is about 500 (DeFries et al., 1989), we can estimate the average effect of an allele at a QTL that accounts for 5% of the total additive genetic variance by solving the following expression: 1/2 (a)2/500 = 0.05

Thus, a = 7.07, i.e., a QTL accounts for at least 5% of the additive genetic variance in sleep time if the average sleep time of those RI strains with the LS alMe exceeds that of the RI strains with the SS allele by at least 2(7.07) = 14.14 rain. If the marker is not closely linked to the QTL, the percentage variance due to segregation at the QTL could be substantially larger than 5%. A less stringent criterion than 5% could be employed to screen for QTLs in future studies if adequate resources become available. Stage 3: Confirmation of Association by Analysis o f f 2. The final stage involves proving that the putative site is actually linked to a QTL for sleep time. Stage 3 involved the assessment of sleep time in 1000 individual F 2 mice derived from crosses between the LS and the SS selected lines. Similar approaches have been used to map QTLs for traits

642

Johnson, DeFries, and M a r k e l

of agronomic import in tomatoes (Paterson et al., 1988) and are currently being applied to many other species. Genomic DNA and brain RNA from the 10% of the mice that are least sensitive and 10% that are most sensitive will be isolated. Molecular probes that have passed stages 1 and 2 will be used to confirm the QTL maps. In order to detect significant associations between sleep time and RFLP haplotype status, data from F2 mice will be subjected to multiple regression analyses (Klein, 1978). The observed sleep time of individual mice will be regressed on their gene dosage (Falconer, 1989) for each RFLP marker. In order to control for correlations due to linkage among markers, separate multiple regression analyses will be performed for each chromosome because markers on the same chromosome are correlated. The partial regression of sleep time on gene dosage for a given marker will be employed as a test of significance for the effects of any QTLs that are proximal to the marker. Alternative tests, including those developed by Lander and Botstein (1989), can also be used. Statistical Considerations. The multiple regression test for the effects of alleles at QTLs is statistically powerful. Given the estimates of additive genetic and environmental variances for ethanol-induced sleep time obtained from the LS x SS RI strain analysis (DeFries et al., 1989), we predict that the phenotypic variance in the F2 population will be about 1200. By multiplying the expected regression of sleep time on gene dosage by the ratio of the standard deviations for gene dosage and sleep time, we estimate that the standardized partial regression coefficient (analogous to a correlation coefficient) will be at least 0.15 for a QTL that accounts for 5% or more of the variance in sleep time. Thus, the statistical power (Cohen, 1977) to detect an effect of this magnitude (~x = .01, one tailed, because we can predict the direction of the effect from previous tests during Stages 1 and 2) in the total sample of 1000 mice would be approximately 0.90. If the F2 mice are first tested for neurosensitivity to ethanol and only selected samples are genotyped, a considerable savings in cost can be achieved with relatively little loss in statistical power (Lander and Botstein, 1989). With selected samples, we can assess statistical power by comparing the expected frequencies of the LS allele in the two groups. Given the expected regression of sleep time on gene dosage in the unselected Fz population and the variance of both variables, we estimate that the regression of gene dosage on sleep time for a QTL that accounts for 5% of the additive genetic variance in sleep time is about 0.003. If we determine the genotype of only the 10% longest-sleeping and 10% shortest-sleeping of the F2 mice tested, we can use the regression of gene dosage on sleep time to predict the frequency of the LS allele in these

Mapping Quantitative Trait Loci for Behavior in Mice

643

two samples, viz., 0.593 and 0.407, respectively. The power to detect a significant difference (c~ = .01) between proportions of this magnitude is approximately 0.60 (Cohen, 1977). However, if the QTL accounts for 10% of the genetic variance in sleep time, the power to detect a significant difference (c~ = .01) between the expected frequencies of the LS allele in the high and low samples (0.63 and 0.37, respectively) is greater than 0.90. Although the alpha level for statistical significance to be employed in the analysis of the data from F2 mice is .01, the probability of a type I error associated with our multistage test for linkage to each marker will be substantially lower. The criterion employed in the statistically independent LS x SS RI strain analysis (Stage 2) requires an allelic difference that accounts for at least 5% of the additive genetic variance. A mean difference of this magnitude between subgroups of RI strains with either LS or SS alleles approaches statistical significance at the 0.10 level (one tailed). Thus, the probability of a type I error for any RFLP marker that meets both criteria is (0.10)(0.01) = 0.001. Of course, the probability of type I errors for all tests considered simultaneously increases with the number of markers that are employed. However, multistage screening greatly reduces the number of possible chance associations. Moreover, the probability of obtaining at least one significant association due to chance alone is relatively low even when many markers are assessed simultaneously. For example, if 50 RFLP haplotypes are assessed in the F z population, the type I error associated with all 50 tests would only be 1 - (0.999) 5o = 0.05. Thus, this multistage QTL analysis is efficient, statistically powerful, and highly conservative.

DESCRIPTION OF MOLECULAR APPROACHES Detection of RFLPs on Chromosome 7. Since DeFries et al. (1989) suggested that the c locus on chromosome 7 may be linked to a gene specifying sleep time, we have focused some of our initial work on this chromosome. As shown in Table I, we have examined a total of 11 sites on chromosome 7 by probing Southern blots of LS and SS mice after digestion with the appropriate restriction endonuclease. Few differences were detected. We have screened only for RFLP variants that were already reported in the literature; thus, the lack of detected variation between the LS and the SS lines for these probes does not necessarily mean that LS and SS mice are identical by descent for these chromosome 7 sites, because other variants may remain undetected. Moreover, the obvious difference in coat color suggests that some region of the genome

644

Johnson, DeFries, and Markel

Table I. Partial List of Probes Used to Detect LS/SS RFLPs ~

Probe pPS1.25 pMH8 pSM676 p450b-5 11B2 23.3 pHR3.0 pMAK1E p2C1 Int2a pl.8 LMAG pSM479 GABAAal p358 p358 p358 p358

Gene

Chromosome

LS/SS different b

emv comp fact H NGF cyt P450 albino

I I VII VII VII VII VII VII VII VII VII VII VII XI XI ?c ? ? ?

+ + -? ? + + + +

emv-23 ~yr hyd androg reg apoE lnt-2 MLV myln glyprot renin GABAAod polA polB polC polD

Difference in RI means

Minor 9%

Minor Minor 6% Minor Minor

a High molecular weight DNA was prepared for RFLP analysis by extraction from the liver, spleen, kidney, brain, and lungs of male and female LS/Ibg and SS/Ibgt mice; the size of undigested DNA was checked by 0.8% agarose gel electrophoresis. Ten to twenty-five micrograms of genomic DNA from each mouse was digested with the appropriate restriction enzyme (Pharmacia LKB); the digested DNA was electrophoresed and transferred to a nylon or nitrocellulose support and probed with the desired radiolabeled probe. The GABAAed plasmid (obtained from Dr. Jim Sikela) was labeled using random hexamer-primed [~x:2P]dCTP to a specific activity > 109 cpm/ ~g. The blots were prehybridized and hybridized at 65~ using 6 • standard saline citrate (SSC), 5 • Denhardt's solution, 0.1 mg/ml salmon sperm DNA, 0.5% sodium dodecyl sulfate (SDS), and 10% dextran sulfate (for hybridization only). The blots were washed twice, for 2 min and for 15 min, at room temperature in 2 • SSC, 0.1% SDS, followed by two 30-min washes at 65~ in 0.5 • SSC, 0.1% SDS. Autoradiograms were generated by exposing the blots for 1-7 days at - 70~ Methods are essentially those of Sambrook et al. (1989). b A " + " indicates that a consistent difference is found between all LS and all SS mice; i.e., the site passes Stage 1. Cannot be assigned to any chromosome because unique sites are still too sparesly mapped in the LS • SS RI lines.

must be polymorphic. Rather than undertaking extensive searches for new RFLPs on VII, we have used the following alternative methods. Use o f R e p e t i t i v e P r o b e s . A n o t h e r a p p r o a c h is to u s e p r o b e s t h a t d e t e c t m a n y s i t e s t h r o u g h o u t t h e g e n o m e in o n l y o n e b l o t ( e . g . , F r a n k e l et a l . , 1 9 9 0 ) . R i s e et al. ( 1 9 9 1 ) r e c e n t l y d e m o n s t r a t e d t h e u s e o f r e p e t -

Mapping Quantitative Trait Loci for Behavior in Mice

645

itive probes to localize genes predisposing to epilepsy. For example, we have used a 358-bp probe derived from the polymerase gene of an endogenous provirus (Rossomando and Meruelo, 1986) to detect several loci that have differentially segregated between LS and SS lines (Fig. 1). Although this method is somewhat more demanding technically and the gels are complex to interpret, these approaches have the advantage of mapping many sites simultaneously. For example, one locus (tentatively identified as polB) may be linked to a QTL for sleep time (Table

I). Other multipoint mapping probes are also available. Frankel et al. (1990) have established genetic maps of xenotropic and polytropic mufine proviral loci. The advantage of these loci is that they are members of the middle repetitive family of genomic sequences and distributed throughout the genome, so that one probe can detect multiple loci simultaneously. Moreover, different inbred strains have different distributions of proviral integration loci; thus each of these loci can be mapped in crosses between inbred strains. Other probes suggested by Frankel et al. (1990) include two olignonucleotide probes (JS-4, which detects 30 endogenous modified polytropic proviruses, and JS-5, which detects modified polytropic viruses) and JS-6/10, a 116-bp probe that detects about 40 endogenous nonecotropic nonxenotropic proviruses. Another useful probe for mice minisatellites are human VNTR sequences; Juliet et al. (1990) showed that almost 50% of the tested probes produced detectable polymorphisms between inbred strains and that, on average, 240 polymorphic differences could be detected. These multipoint mapping gels maximize information generation, although our experience tells us that the best approach is an appropriate mix of singlesite probes (to be used as anchors) and multisite probes, because individual sites detected by repetitive probes alone are difficult to interpret without known anchors to assist in the linkage determinations. Yet another approach is the use of two-dimensional gels as described by Uitterlinden et al. (1989). Excellent resolution of many loci can be achieved through the use of orthogonal electrophoretic separation techniques: The first dimension is a polyacrylamide sizing gel, and the second dimension is a denaturing gradient polyacrylamide gel which separates primarily on the basis of base composition. After electrophoresis, genomic DNA is transferred electrophoretically and probed with labeled variable-number tandem repeat (VNTR) core sequences (Jeffreys et al., 1985), 33.15, 33.6, or other core sequences. Because the patterns of these two-dimensional gels are complex, they must be interpreted with special care; however, it is clear that the patterns are replicable and can be analyzed as proposed (Uitterlinden and Vijg, 1989).

646

Johnson, DeFries, and Markel

Fig. I . Autoradiogram of genomic DNA from Er digests of LS x SS RI mice probed with 3~p-labeled p358, a probe homologous to an endogenous murine viral gene present in about 25 copies per mouse genome. Individuals LS • SS strains are shown across the top and the position of Hind III-digested lambda DNA is shown at the left for determining the size of each band.

M a p p i n g Q u a n t i t a t i v e T r a i t Loci for B e h a v i o r in Mice

647

Sequence-Tagged Sites. A sequence-tagged site (STS) is a unique genomic region for which there is sequence available such that the entire region can be amplified from genomic DNA by PCR (Olson et al., 1989; Green and Olson, 1989). There are major advantages in using STSs because any laboratory can amplify the desired sequence at will using PCR and no probes need to be transmitted from one lab to another. The use of such sites to anchor mammalian genome mapping has been proposed (Olson et al., 1989); in this case the STS site should have a high information content, i.e., be very polymorphic. A number of oligomer pairs that fulfill these criteria have been developed for mapping the mouse genome (Love et al., 1990; Cornall et. al., 1991; Aitman et al., 1991; Hearne et al., 1991). All of these probes are highly polymorphic among inbred mouse strains and between related species of mice. The variation of these sites results from variable numbers of simple sequence (microsatellite) repeats. Since replication errors occur with a detectable frequency, microsatellite sequences maintain high levels of polymorphisms that can be detected after PCR amplification and analysis on acrylamide gels. We have begun to analyze such sites in the LS and SS lines. Whereas several sites on chromosome 11 are polymorphic, none appear polymorphic on chromosome 7. An example of a polymorphism on chromosome 11 is shown in Fig. 2. Testing of Candidate Genes. Candidate genes for sensitivity to ethanol are genes for which there is independent evidence of direct involvement ['u acida (GABAA) loci (Wafford et al., 1990)]. Candidate genes can be assessed for involvement in the specification of alcoholinduced sleep time. At least three different strategies can be used to choose "'candidate loci" (loci that may be directly involved in the specification of alcohol sensitivity). (1) Clones genes such as the GABAAoL1 receptor subunit (Fig. 3) have been tested, even if such probes are not yet mapped or map to unexplored genomic regions or regions lacking significant associations with QTLs for sleep time. For example, considerable evidence suggests that the GABAA~xl receptor may be involved in alcohol neurosensitivity (Martz et al., 1983; Allan et al., 1988). Recently this gene was cloned (Keir et al., 1991) and provided to us by J. Sikela; this probe detects a RFLP in the LS and SS lines, but this difference does not consistently distinguish LS and SS mice, nor does it explain much of the variation in sleep time among the RI strains (Greenlee and Johnson, unpublished observations). More recent results implicate the GABAA~/2L subunit-an alternative RNA splicing product of the GABAA',/2 primary trans c r i p t - i n sensitivity to ethanol in vitro (Burnett et al., submitted); this probe has been provided to us by Dr. Sikela and in collaboration with

648

Johnson, DeFries, and Markel

Csfgm LS/SS Variants chrom. 11 (30) c~,o~ 000

~

1 7w~ 6 r~ ~i,,,, -~00

to 0

m Fig. 2. Polyacrylamide gel showing use of sequence-tagged site (STS) polymorphism. Oligomers were mixed in a PCR reaction buffer and reactions performed essentially as described in Love et al. (1990). The sequences of the oligomers in this case were 5 ' C T G T G C A A C A G A C T A A G C C T 3 ' and 5 ' C T G T A A C A C A A T A A C C A G G C A 3 ' [oligomer pair No. 3 (Love et al., 1990)], which visualize a CA repeat in the Csfgrn locus at 30 cM on linkage group 11. DNA is from individual LSP and SSP mice as indicated. The polymorphism is indicated at the right. DNA oligomers were synthesized by Oligos Etc., Inc. (Guilford, CT).

Mapping Quantitative Trait Loci for Behavior in Mice

649

Fig. 3. Autoradiogram of a blot of HindlII-digested genomic DNA from several LS (LSP) or SS (SSP) mice probed with 32p-labeled GABA.~xl. Each lane is from an individual mouse; lowercase letters refer to different siblings in the same family.

Dr. Sikela w e are currently seeking (1) RFLP variation that can be detected with this probe, (2) microsatellites that are near to or in noncoding regions o f this gene, and (3) single-strand conformational polymorphisms (SSCPs) that may be used to determine alternate forms of this gene that consistently distinguish the LS and SS lines. (2) Given a genetic map for QTLs for ethanol neurosensitivity in the mouse, it is possible to choose nearby genes as candidates. In addition to those genes mapped in the mouse, it is possible to select candidate genes based on human gene mapping data resulting from the H u m a n Genome Project. This is facilitated by the existence of synteny maps that

650

J o h n s o n , DeFries, a n d M a r k e l

describe chromosome rearrangements that have occurred since the phylogenetic separation of mouse and human; such maps can be used to predict those regions of the human genome where genes for alcoholism might be found. (3) Several brain-specific cDNA libraries have been constructed and are being analyzed under the auspices of the Human Genome Project; such libraries may be of use in identifying candidate loci. For example, a cDNA library is being generated from brains of human alcoholics and will contain genetically mapped transcripts that are specifically expressed in brains of former alcoholics (Sikela, personl communication). Such candidate genes may be ultimately analyzable by means of standard molecular and transgenic analyses using gene replacement strategies.

CONCLUSIONS Using recombinant inbred strains of mice to map QTLs is a valuable initial strategy. However, the power of RIs alone for QTL mapping is relatively weak. Here we have outlined an alternative approach wherein RIs ate screened during the second stage of a multistage strategy to map QTLs to precise positions in the mouse genome. The basis of the increased precision is subsequent analysis of individual Fz mice. Such studies may be useful in attempts to identify human genes predisposing to alcoholism. The identification of QTLs in the mouse that predispose to alcohol sensitivity can serve as a basis for orderly searches of the human genome by taking into account the evolutionary relationships of chromosomal linkage between human and mouse (Nadeau and Reiner, 1990). Although results of quantitative genetic and pedigree analyses of human data have indicated that a genetic vulnerability to alcoholism exists, it has been very difficult to determine its neurobiological and molecular bases. The molecular cloning of loci that influence various components of the addiction process (viz., avidity, initial sensitivity, tolerance, and physical dependence) in mice may allow the generation of new pharmacological agents, "designer drugs" that are tailor-made to ameliorate the effects of these gene products and establish a rational basis for the pharmacological treatment of alcoholism and drug abuse. ACKNOWLEDGMENTS We thank J. Greenlee and J. Johnson for collecting the data and J. Sikela for helpful discussion.

M a p p i n g Quantitative Trait Loci for Behavior in Mice

651

REFERENCES Aitman, T. J., Hearne, C. M., McAleer, M. A., and Todd, J. A. (1991). Mononucleotide repeats are an abundant source of length variants in mouse genomic DNA. Mammal Genome 1:206-210. Allan, A. M., Spuhler, K. P., and Harris, R. A. (1988). Gamma-aminobutyric acidactivated chloride channels: relationship to genetic differences in ethanol sensitivity. J. PharmacoL Exp. Ther. 244:866-870. Bailey, D. W. (1971). Recombinant-inbred strains. An aid to finding identity, linkage, and function of histocompatibility and other genes. Transplantation 11:325-327. Bailey, D. W. (1981). Recombinant-inbred strains and bilineal congcnic strains. In Foster, H. L., Small, J. D., and Fox, F. G. (eds.), The Mouse in Biomedical Research, Academic Press, New York, pp. 223-239. Basile, A., Hoffer, B., and Dunwiddie, T. (1983). Differential sensitivity of cerebellar Purkinje neurons to ethanol in selectively outbred lines of mice: Maintenance in vitro independent of synaptic transmission. Brain Res. 264:69-78. Bates, G. P., MacDonald, M. E., Baxendale, S., Youngman, S., Lin, C., Whaley, W. L., Wasmuth, J. J., Gusella, J. F., and Lehrach, H. (1991). Defined physical limits of the Huntington disease gene candidate region. Am J. Hum. Genet. 49:7-16. Botstein, D., White, R. L., Skolnick, M., and Davis, R. W. (1980). Construction of a genetic linkage map in man using restriction fragment length polymorphisms. Am. J. Hum. Genet. 32:314--331. Cohen, J. (1977). Statistical Power Analysis for the Behavioral Sciences, Academic Press, New York. Cornall, R. J., Aitman, T. J., Hearne, C. M,, and Todd, J. A. (1991). The generation of a library of PCR-analyzed microsatellite variants for genetic mapping of the mouse genome. Genomics 10:874-881. Crabbe, J. C. (1989). Genetic animal models in the study of alcoholism. AlcohoL Clin. Exp. Res. 13:120-127. Crabbe, J. C., Kosobud, A., Young, E. R., and Janowsky, J. S. (1983). Polygenic and single-gene determination of responses to ethanol in BXD/Ty recombinant inbred mouse strains. Neurobehav. ToxicoL TeratoL 5:181-187. Crabbe, J. C., Young, E. R., Tam, B., Kosobud, A., Belknap, J. K., and Laursen, S. E. (1986). Genetic differences in anticonvulsant sensitivity in mouse lines selectively bred for ethanol withdrawal severity. J. PharrnacoL Exp. Ther. 239:154-159. DeFries, J. C., Wilson, J. R., Erwin, V. G., and Petersen, D. R. (1989). LSxSS recombinant inbred strains of mice: Initial characterization. Alcohol Clin. Exp. Res. 13:196-200. Dudek, B. C., and Abbot, M. E. (1984). A biometrical genetic analysis of ethanol response in selectively bred long-sleep and short-sleep mice. Behav. Genet. 14:120. Elliott, R. (1991), DNA restriction fragment variants. Mouse Genome 89:365-409. Erwin, V. G., Heston, W. D. W., McClearn, G. E., and Deitrich, R. A. (1976). Effects of hypnotics on mice genetically selected for sensitivity to ethanol. Pharm. Biochem. Behav. 4:679-683. Falconer, D. S. (1989). Introduction to Quantitative Genetics, Longman House, London. Frankel, W. N., Stoye, J. P., Taylor, B. A., and Coffin, J. M. (1990). A linkage map of endogenous murine leukemia proviruses. Genetics 124:221-236. Gelderman, H. (1975). Investigations on inheritance of quantitative characters in animals by gene markers. I. Methods. Theoret. AppL Genet. 46:319-330. Goldman, D., Lister, R. G., and Crabbe, J. C, (1987). Mapping a putative genetic locus determining ethanol intake in the mouse. Brain Res. 420:220-226. Gora-Maslak, G., McClearn, G. E., Crabbe, J. C., Phillips, T. J., Belknap, J. K., and Plomin, R. (1991). Use of recombinant inbred strains to identify quantitative trait loci in psychopharmacology. Psychopharmacology 104:413--424. Green, E. D., and Olson, M. V. (1989). Chromosomal region of the cystic fibrosis gene

652

Johnson, DeFries, and Markel

in yeast artificial chromosomes: A model for human genome mapping. Science 250:94-98. Gusella, J. F., Wexler, N. S., Conneally, P. M., Naylor, S. L., Anderson, M. A., Tanzi, R. E., Watkins, P. C., Ottina, K., Wallace, M. R., Sakaguchi, A. Y., Young, A. B., Shoulson, I., Bonilla, E., and Martin, J. B. (1983). A potymorphic DNA marker genetically linked in Huntington's disease. Nature 306:234-238. Haldane, J. B. S., and Waddington, C. H. (1931). Inbreeding and linkage. Genetics 16:357-374. I-Iearne, C. M., McAleer, M. A., Love, J. A., Aitman, T. J., Cornall, R. J., Ghosh, S., Knight, A. M., Prins, J.-B., and Todd, J. A. (1991). Additional microsatellite markers for mouse genome mapping. Mammal Genome 1:273-282. Heston, W. D., Erwin, V. G., Anderson, S. M., and Robbins, H. (1974). A comparison of the effects of alcohol on mice selectively bred for differences in ethanol sleeptime. Life Sci. 14:365-370. Jeffreys, A. J., Wilson, V., and Thein, S. L. (1985). Hypervariable "minisatellite" regions in human DNA. Nature 314:67-73. Julier, C., de Gouyne, B., Georges, M., Guenet, J.-L., Nakamura, Y., Avner, P., and Lathrop, G. M. (1990). Minisatellite linkage maps in the mouse by cross-hybridization with human probes containing tandem repeats. Proc. Natl. Acad. Sci. USA 87:4584---4589. Keir, W. J., Kozak, C. A., Chakraborti, A., Deitrich, R. A., and Sikela, J. M. (1991). The cDNA sequence and chromosomal location of the murine GABAA ~1 receptor gene. Genomics 1:390-395. Klein, T. W. (1978). Analysis of major gene effects using recombinant inbred strains and related congenic lines. Behav. Genet. 8:261-268. Lander, E. S., and Botstein, D. (1989). Mapping Mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121:185-199. Li, T.-K., Lumeng, L., McBride, W. J., and Waller, M. B. (1981). Indiana selection studies on alcohol-related behaviors. In McClearn, G. E., Deitrich, R. A., and Erwin, V. G. (eds.), Development of Animal Models as Pharmacogenetic Tools, DHHS Publication No. [ADM] 81-113, U.S. Government Printing Office, Washington, DC, pp. 171-191. Love, J. M., Knight, A. M., McAleer, M. A., and Todd, J. (1990). Towards construction of a high resolution map of the mouse genome using PCR-analysed microsatelites. Nucleic Acids Res. 18:4123--4130. Martz, A., Deitrich, R. A., and Harris, R. A. (1983). Behavioral evidence for the involvement of gamma-aminobutyrie acid in the actions of ethanol. Eur. J. PharmacoL 89:53-62. McClearn, G. E., and Kakihana, R. (1981). Selective breeding for ethanol sensitivity: short-sleep and long-sleep mice. In McClearn, G. E., Deitrich, R. A., and Erwin, V. G. (eds.), Development of Animal Models as Pharmacogenetic Tools, DHHS Publication No. [ADM] 81-113, U.S. Goverment Printing Office, Washington, DC, pp. 147-159. McClearn, G. E., Plomin, R., Gora-Maslak, G., and Crabbe, J. C. (1991). The gene chase in alcohol research. PsychoL Sci. 2:222-229. Nadeau, J. H., and Reiner, A. H. (1990). Linkage and synteny homologies in mouse and man. In Lyon, M. F., and Searle, A. G. (eds.), Genetic Variants and Strains of the Laboratory Mouse, 2nd ed., Oxford University Press, Oxford, pp. 506-536. Neumann, P. E., and Collins, R. L. (1991). Genetic dissection of susceptibility to audiogenic seizures in inbred mice. Proc. Natl. Acad. ScA USA 88:5408-5412. Neumann, P. E., and Seyfried, T. N. (1990). Mapping of two genes that influence susceptibility to audiogenic seizures in crosses of C57BL/6J and DBA/.2J mice. Behav. Genet. 20:307-323. Olson, M., Hood, L., Cantor, C., and Botstein, D. (1989). A common language for physical mapping of the human genome. Science 245:1434-1435. Palmer, M. R., Sorenson, S. M., Freedman, R., Olson, L., Hoffer, B., and Seiger, A.

Mapping Quantitative Trait Loci for Behavior in Mice

653

(1982). Differential ethanol sensitivity of intraocular cerebellar grafts in long-sleep and short-sleep mice. 3". PharmacoL Exp. Ther. 222:480-487. Paterson, A. H., Lander, E. S., Hewitt, J. D., Peterson, S., Lincoln, S. E., and Tanksley, S. D. (1988). Resolution of quantitative traits into Mendelian factors by using a complete linkage map of restriction fragment length polymorphisms. Nature 335:721726. Plomin, t?. (1990). The role of inheritance in behavior. Science 248:183-188. Plomin, R., McClearn, G. E., Gora-Maslak, G., and Neiderhiser, J. M. (1991). Use of recombinant inbred strains to detect quantitative trait loci associated with behavior. Behav. Genet. 21:99-116. Pritchard, C., Cox, D. R., and Myers, R. M. (1991). Invited Editorial: The end in sight for Huntington disease. Am. J. Hum. Genet. 49:1-6. Rise, M. L., Frankel, W. N., Coffin, J. M., and Seyfried, T. N. (1991). Genes for epilepsy mapped in the mouse. Science 253:669-673. Rommens, J. M., Ianuzzi, M. C., Kerem, B.-S., Drumm, M. L., Melmer, G., Dean, M., Rozmahel, R., Cole, J. L., Kennedy, D., Hidaka, N., Zsiga, M., Buchwald, M., Riordan, J. R., Tsui, L.-C., and Collins, F. S. (1989). Identification of the cystic fibrosis gene: Chromosome walking and jumping. Science 245:1059-1065. Rossomando, A., and Meruelo, D. (1986). Viral sequences are associated with many histocompatability genes. Immunogentics 23:233-245. Saiki, R. K., Scharf, S., Faloona, F., Mullis, K. B., Horn, G. T., Erlich, H. A., and Arnheim, N. (1985). Enzymatic amplification of B-globin genomic sequences and restriction site analysis for diagnosis of sickle cell anemia. Science 230:1350-1354. Sambrook, J., Fritsch, E. F., and Maniatis, T. (1989). Molecular Cloning, a Laboratory Manual, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Seyfried, T. N., Yu, R., and Glaser, G. H. (1980). Genetic analysis of audiogenic seizure susceptibility in C57BL/6J x DBA/2J recombinant inbred strains of mice. Genetics 94:701-718. Silver, J., and Buckler, C. E. (1986). Statistical considerations for linkage analysis using recombinant inbred strains and backcrosses. Proc. Natl. Acad. ScL USA 83:14231427. Sorenson, S., Palmer, M., Dunwiddie, T., and Hoffer, B. (1980). Electrophysiological correlates of ethanol-induced sedation in differentially sensitive lines of mice. Science 210:1143-1145. Uitterlinden, G., and Vijg, J. (1989). Two-dimensional DNA typing. T. L BTECH. 7:336-341. Uitterlinden, G., Slagboom, E., Knook, D. L., and Viig, J. (1989). Two-dimensional DNA fingerprinting of human individuals. Proc. Natl. Aead. Sci. USA 86:27422746. Wafford, K. A., Burnett, D. M., Dunwiddie, T. V., and Harris, R. A. (1990). Genetic differences in the ethanol sensitivity of GABAA receptors expressed in Xenopus oocytes. Science 249:291-294. Watson, J. D. (1990). The human genome project: Past, present, and future. Science 248:44-49.