Qua n t i t a t i v e Trait Analysis

Judith E. Grisel, Ph.D.

Alcoholism is a quantitative disorder that is caused by the combined influences of numerous (i.e., quantitative trait loci [QTLs]) and environmental factors. To identify QTLs for alcoholism, researchers compare subject groups (e.g., inbred mouse strains) that differ in both their genetic makeup (i.e., ) and alcohol-related trait (e.g., sensitivity to certain alcohol effects). Using statistical tests one can then determine whether a specific or DNA region contributes to the trait of interest. This strategy requires that the relevant gene exists in several variants (i.e., is polymorphic). To conduct such QTL analyses, researchers study either a large population of mice that all differ in their or compare several strains, each of which has a fixed genotype. However, QTL analyses still have several limitations. Nevertheless, such studies already have identified several DNA regions and genes that may affect the response to alcohol and thus may contribute to the risk for alcoholism. KEY WO R D S : ; genetic mapping; genetic polymorphism; DNA; animal strains; animal model; genotype;

lc o h o l ’s effects are dependent height or intelligence, var y continuously that may contribute to a quantitative upon both genetic and environ - ac r oss a population and derive from a trait, res e a r chers can employ sever a l Amental factors as well as the co n s t e l lation of both genetic and envi- strategies to determine the role of an interaction of these broad sources of ronmental influences. Vir tually all behav- individual DNA sequence among all influence (e.g., Shuckit 1998). In orde r io r a l characteristics (i.e., ) the other genetic and nongenetic influ- to better understand, treat, and preve n t ar e quantitative traits. The genes (or ences on the trait. For instance, in ve s t i - pr oblems associated with alcoholism, DNA regions that contain the genes) gators can analyze genetic differen c e s res e a r ch has focused on pinpointing th e influencing a quantitative trait are called be t w een a human population with a specific genetic and environmental ele- qu a n t i t a t i v e trait loci (QTLs). This specific trait or prone to that trait (e.g., ments that contribute to a liability for, ar ticle describes some considerations alcoholics) and a control population or protect against, alcohol abuse. Add i n g for the genetic analysis of quantitative (e.g., nonalcoholics), using either link- to the complicated nature of the disease, traits, the general steps invol v ed in QTL age or association studies.1 Such analy- all cases of alcoholism are not caused analysis, the primary strengths and fail- ses could allow res e a r chers to identify by the same set of factors. ings of such approaches, and a few exam- genes whose struc t u r e or ex p r ession dif- As with other , alco- pl e s of how QTL analysis has helped holism is a quantitative or polygenic res e a r chers identify genes contributing di s o r der in which the influences of to differential sensitivity to alcohol. 1For a definition of this and other technical terms used throughout this issue, please see the central glossary, pp. multiple genes combine to contribute 193Ð195. to the pathological state (e.g., Eno c h and Goldman 2000). Unlike qualitative Strategies for Analyzing (i.e., Men d e l i a n ) traits, which have an Quantitative Traits JUD I T H E. GRI S E L , PH.D . , is an assistant “ei t h e r - o r ” expression pattern in a pop- pr ofessor in the Dep a r tment of Psyc h o l o g y, ulation and are generally mediated by a In the absence of evidence implicating Fur man Uni ve r s i t y , Greenville, South single gene, quantitative traits, such as a specific gene (i.e., a ) Caro l i n a .

Vol. 24, No. 3, 2000 169 fers between the two populations. The Fin a l l y , a high degree of genetic con- ing the response. As a general rule, the ul t i m a t e success of such an endeavor , se r vation exists betwee n the human gr eater the number of polymorphic ho weve r , depends largely on a clear and the genome of commonly DNA markers, the greater is the statis- pr ognosis of a person’s risk for the dis- used animal models, such as mice, tical likelihood of finding significant ease and on a set of distinct causes (i.e., indicating that large portions of the co r r elations. Howeve r , this strategy ha s etiology). For alcoholism, this proc e s s ch r omosomes have remained rel a t i ve l y practical limitations, the primary one is made more difficult by the fact that intact across species during evol u t i o n . being that an estimated 100,000 genes the disease derives from a complex In fact, res e a r chers have estimated that in the mouse are contained within interplay of many genetic and environ - ap p r oximately 80 percent of the mouse ap p r oximately 3 billion DNA building mental influences that are not uniform genome is organized similarly to the blocks (i.e., base pairs) (O ’ B rien et al. ac r oss alcoholics. human genome. This finding makes it 1999). In spite of the fact that grea t Studies in laboratory animals, most likely that the location of a mouse pro g r ess has been made in unravel i n g commonly in rodents, provide some DNA segment that influences a trait of the mouse genome, res e a r chers are still distinct advantages over human studies in t e r est suggests the location of a corre- a long way from a physical map of the in the quest for elucidating the genetic sponding gene in the human genome DNA that would make gene identifica- causes of a quantitative trait, such as (Copeland et al. 1993). tion truly facile. alcoholism, even in the absence of spe- cific hypotheses reg a r ding its etiology. Strategies for QTL Mapping First, one can isolate more easily the Mapping QTLs genetic and environmental influences Mice are the species of choice for most in laboratory animals than in humans, In general, QTLs are identified by corre- QTL mapping studies, because thou- because to a large extent, one can con- lating genetic var i a t i o n with trait var i a - sands of polymorphic markers are cur- tr ol the animals’ experimental environ - tion; a significant correlation betwee n rently known in this species, many of ment. Consequently, one can attribute ge n o t y p e and phenotype suggests that which have been genotyped across strains vi r tually all the variance in phenotype DNA status helps determine trait expres - (B elknap et al. 1997). Res e a r chers gen- ac r oss different animal strains to genetic si o n . QTL mapping can be conducted erally use one of two mapping strate- factors. by comparing various inbred strains gies: (1) an analysis of a large popula- Second, many animal strains used (i.e., strains in which each animal has tion of mice with different genotypes in these studies already have been wel l the same genotype) or by comparing or (2) the testing of several strains that ch a r a c t e r i z ed with respect to their genetic animals from a population with diver s e each have a fixed genotype. makeup (i.e., genotype). For instance, a genotypes (i.e., a genetically segreg a t i n g For the first strategy, so-called F2 list of strain-specific polymorphic mark- population). Fol l o wing the initial scree n mice are a popular choice. These mice er s and genes—that is, DNA reg i o n s in which a DNA region has been iden- ar e derived from the cross of two that exist in different forms within a tified that contributes to a specific phe- in b r ed strains (see figure 1). In inbred test population—is readily accessible notype, subsequent testing can be used strains, all animals have the id e n t i c a l fr om database source s . 2 Thus, two strains to further res o l v e the exact location and genotype, and each animal has tw o that differ both in behavior and geno- mechanism of the gene effect. identical copies of each gene3—that is, type may provide a good starting point To identify QTLs, res e a r chers eval u - the animals are homozygous at each for genetic analysis. ate groups of subjects that have poly- locus. Strains used in often Th i r d, one can design animal mod- morphic genes (i.e., genes that exist in ha v e known genetic diversity or large els to isolate a specific aspect of the mo r e than one form) for a differen t i a l phenotypic differences for a trait of inter- alcoholic phenotype (e.g., alcohol pref - sensitivity to a parti c u l a r alcohol effect es t . The offspring of a cross of two such er ence, tolerance, or dependence) in (e.g., sedation). They conduct statistical strains (called F1 animals) will have two or der to clarify this aspect’s specific et i o l - tests to determine whether specific DNA di f f e r ent of each gene that differs og y . As a result, res e a r chers can parce l regions contribute to the variance in be t w een the strains. The F1 animals all the heterogeneous di s o r der into units alcohol sensitivity. If the response depends ar e identical, howeve r , because at each that are more amenable to genetic analy- on the sequence of a certain DNA locus where the parent strains differed , si s . Thus, the mechanisms underlying region, then the region is a QTL an d each F1 animal has rec e i v ed one alcohol pref e r ence, tolerance, or depen- contains either a gene that influences the fr om each parent. Mating animals from dence can be studied independently. qu a n t i t a t i v e trait or a DNA sequence the F1 generation produces the F2 ge n - Factors identified (e.g., genes) as con- that modifies the expression of such a gene. tributing to one of these phenotypes It is crucial for such analyses that 3Many genes exist in different variants, or alleles. Each then can be evaluated for a more gen- the test subjects are polymorphic for the carries two copies of each gene, one inherited eral role in other aspects of alcoholism. rel e v ant genes; one cannot identify QTL s from the mother and one inherited from the father. If both in groups that do not have genetic dif- gene copies carry the same allele, the organism is said to be homozygous for that gene. If the two gene copies carry 2For example, the Jackson Laboratory maintains a list of fe r ences in the critical regions, even if different alleles, the organism is said to be heterozygous rodent markers at its Web site . genes in those regions may be influenc- for that gene.

170 Alcohol Research & Health Quantitative Trait Locus Analysis

eration. In contrast to the F1 an i m a l s , the trait for each animal and then attempt sufficiently large population (i.e., suffi- each F2 animal possesses a unique combi- to correlate the individual test scores ciently large number of genotypes), one na t i o n of progenitor alleles. This diver - with the polymorphic markers. For can detect and map QTLs with rel a - sity results from the recombination of example, investigators might test a ti v ely small effects in a single study. DNA segments between the two mem- large group of F2 animals for their vol - The second major strategy for map- bers of a pair that occurs un t a r y alcohol consumption and then ping QTLs—called a two-step or mul- during the specialized cell division proc e s s look for systematic genetic differen c e s tistep approa c h — re q u i r es sequential (i.e., meiosis) in the production of egg be t w een the highest and lowest con- testing of more than one population. and sperm cells and which results in a suming animals. Although limiting Most often, this strategy invol ve s a set recombination of the progenitor DNA the analysis to the extremes of the trait of recombinant inbred (RI) strains. RI (see figure 2 for an illustration of gene of interest may decrease its ability to strains ar e generated by mating two recombination during meiosis). Thus, detect QTLs with only small effects, it in b r ed strains and then inbree d i n g th e the F2 animals are a genetically diver s e will markedly reduce the costs associ- F2 generation of this cross. This strategy gr oup of subjects. ated with analyzing the animals’ geno- results in the recombination of the alle- When analyzing such a grou p , types. Overall, the greatest advantage les present in the progenitors and in th e res e a r chers determine the test score for of analyzing F2 animals is that given a stable transmission of the rec o m b i n e d

1. Two inbred strains (differentiated by colors) differ in numerous polymorphic regions and Progenitors serve as progenitors for RI strains. (Inbred strains)

2. Mating two inbred strains produces an F1 gener- ation, in which all animals are identical. For each F1 polymorphic region, the F1 animals have received one copy from each parent.

3. F1 animals are mated to produce the F2 genera- tion. As a result of the recombination of genetic material that occurs during the production of

sperm and egg cells in the F1 generation (see figure 2), each animal of the F2 generation shows F2 a unique pattern of the progenitors’ genes.

4. Sister-brother pairs of the F2 generation are mated. This inbreeding is repeated over several generations, eventually resulting in RI strains— that is, animals that are homozygous at each allele but possess unique recombinations of the genes that were polymorphic in the progenitors. Recombinant These recombinations can be maintained by Inbred Strains continued inbreeding.

Figure 1 Generation of recombinant inbred (RI) strains. The progenitors of RI strains are animals of two inbred strains that carry different forms (i.e., alleles) of polymorphic genes (indicated by the colors blue and red). For each animal, the figure shows one chromosome pair in which one chromosome is inherited from each parent. In an inbred strain, each animal has two copies of the same allele of each gene, meaning they are homozygous at each locus. Many alleles do not differ between strains, however, and the white areas on the represent those regions that are not polymorphic.

Vol. 24, No. 3, 2000 171 alleles in a set of related RI strains that ar e bred to homozygosity (see figure 1). The advan t a g e of this strategy is that many RI strains whose genotypes have been well characterized are commer- cially avai l a b l e . Maternal chromosome Paternal chromosome Two sets of RI strains have been fre- quently employed in alcohol res e a rc h . The BXD RIs, which wer e derived fr om C57BL/6J and DBA/2J prog e n i - tors, comprise 26 strains with more than 1,300 genotyped polymorphic alleles distributed ac r oss the genome. The LSXSS RIs, which wer e derived from mice selected for high and low sensitiv- Duplication of chromosome ity to alcohol’s sleep-inducing (i.e., hyp- notic) effects, include 27 strains wi t h mo r e than 180 known polymorphisms. This latter set has been pa r ticularly use- ful for identifying QTLs underlying the se n s i t i v i t y to the hypnotic effects of de p r essant drugs, including alcohol. With either the two-step or multi- step mapping approach, res e a r chers test Crossing over/recombination mice from these or other inbred strains for the phenotype of interest (e.g., alco- hol consumption level). The mean val u e s obtained for each strain are then corre- lated with the strains’ genotypes (see fi g u r e 3). Because the number of “su b - je c t s ” to be included in such an analysis is really the number of avai l a b l e ge n o - types or strains, much fewer genotypes exist than with the F2 ap p r oach. As a result, this approach has less statistical po wer than the F2 ap p r oach and, con- se q u e n t l y , a rel a t i v ely high likelihood of missing some significant associations (i.e., type II errors, which are described in the following paragraph).

Limitations of QTL Mapping QTL mapping has been likened to cast- ing a wide-mesh net across the entire Gametes have new combinations of genetic material sea of genetic information. As such, this method is a power ful tool for con- Figure 2 Crossing over and during the specialized cell ducting initial screens of the whole division (i.e., meiosis) that results in the production of sperm or egg cells genome; howeve r , this method is also (i.e., the gametes). Each member of each chromosome pair (i.e., the one subject to some statistical and practical inherited from the mother and the one inherited from the father) duplicates pitfalls. For example, a single ex p e r i m e n t itself, and genetic material may cross over. In inbred strains, in which can detect only QTLs with a rel a t i ve l y each chromosome pair is identical, crossing over has no observable large influence on the trait of interes t . effect. In animals in which the maternal and paternal chromosomes are Most QTLs, howeve r , probably have not identical (e.g., F1 animals), however, crossing over will result in the only a small influence, and thus recombination of parental genes. Because of this recombination, all the res e a r chers may miss some rel e v ant loci genes on a chromosomes are not always inherited together. using this technique. Such errors are

172 Alcohol Research & Health Quantitative Trait Locus Analysis

ca l l e d “false negatives ” or type II error s . ea r l i e r , it does not mean that they have ping parts of the original QTL. By iden- Furt h e r m o r e, because res e a r chers ana- found a gene contributing to the trait tifying strains with similar phenotypes ly z e a large number of markers for their of interest. The DNA regions identi- and determining which QTL reg i o n s li n k a g e with the trait, some correl a t i o n s fied as QTLs are still so large that they they contain, one can discriminate will appear positive purely by chance likely contain hundreds of genes. Seve r a l DNA reg i o n s most likely to contain the and thus are called “false positives ” or methods exist, howeve r , that can help rel e v ant gene(s). The greater the num- type I errors. na r r ow the DNA region enough to ber of such congenic strains (and thus Type I errors can be minimized by conduct a practical search for a specific of specific QTL regions), the finer or establishing stringent criteria for what gene (for an overv i e w of some of these mo r e precise the mapping will be. is considered a significant correl a t i o n . techniques, see Dar vasi 1998). The more stringent these criteria are, One such strategy uses congenic ho weve r , the more likely it is that some strains—mice in which a parti c u l a r QTL Mapping in Alcohol QTLs with small influences will be region of DNA (e.g., a QTL) from one Research missed, increasing the chance of type II strain is tr a n s f e r r ed (i.e., introg re s s e d ) er r ors. To achieve a balance betwee n into the genome of another strain, fo l - Despite its limitations, QTL mapping these two concerns, the scientific com- lo wed again by inbreeding (for a more is a power ful method for widesprea d munity has rel a x ed the initial criterion detailed description of this process, see analyses of the genetic influences under- for significant correlations in QTL Crabbe et al. 1999). Using this strategy, ly i n g qu a n t i t a t i v e traits, such as those sc r eening while insisting on confirma- one can produce several congenic contributing to alcoholism. Res e a rc h e r s tion of potential QTLs in followu p strains that carry different or over l a p - ha v e provisionally identified numerou s ex p e r i m e n t s . 4 QTLs as affectors of alcohol res p o n s e s ; Another limitation of QTL mapping of those QTLs, approximately 15 to 20 4Lander and Kruglyak (1995) have proposed a widely is that when res e a r chers locate a QTL accepted standard for a significance criterion using ha v e resulted in confirmed loci (for a using one of the strategies described logarithm-of-the-odds-of-linkage (LOD) scores. recent rev i e w, see Crabbe et al. 1999).

1 2 3 4 5 6 7

Trait A Trait B

Trait C

Trait A high low high low high low low

Trait B low high high low low high low

Trait C high low low high high low low

Trait D low low high low low high low

Trait E high high low low high high low

Figure 3 Quantitative trait loci (QTL) analysis of recombinant inbred strains. The example shows seven different strains. For each strain, a single chromosome pair is depicted that is associated with three traits (i.e., A, B, and C). Chromosome regions inherited from each progenitor (differentiated by color) confer either high or low responses for the correspond- ing trait. High correlations between the trait and the genetic status indicate a QTL. For example, for trait A, the gene form (i.e., allele) inherited from the blue progenitor confers a high response, whereas the allele inherited from the red progenitor confers a low response. Conversely, for traits B and C, the red alleles confer a high response and the blue alleles confer a low response. More numerous recombinations increase the ability to resolve regions of influence. Two other traits (i.e., D and E), however, are not associated with QTLs on this chromosome, because no correlation exists between the levels of those traits and any of the chromosome regions analyzed.

Vol. 24, No. 3, 2000 173 These loci influence such phenotypes as an approach leading to a putative References alcohol self-administration, alcohol- mechanism for alcohol sensitivity and induced sedation, and withdrawal. For to a potential difference in risk for alco- BEL K N A P , J.K.; DUB A Y , C.; CRA B B E , J.C.; AN D BUC K , K.J. Mapping quantitative trait loci for example, the LSXSS RI strains wer e used ho l i s m is obvi o u s . behavioral traits in the mouse. In: Blum, K., and to sc r een for QTLs influencing sedation Noble, E.P., eds. Handbook of Psychiatric . induced by alcohol and other depres s a n t Boca Raton: CRC Press, 1997. pp. 435–453. dr ugs. Fol l o wup analyses in over 1,000 Conclusions F mice rev ealed four significant QTL s BUC K , K.J., AN D HOO D H. M . 2 of a GABA(A) gamma2 subunit variant contributing to this phenotype. QTL mapping in animals provides a with severity of acute physiological dependence on In some cases, the identification of valuable, though prel i m i n a r y, contribu- alcohol. Mammalian Genome 9:975–978, 1998. a QTL will readily suggest a candidate tion to the search for genes contribut- BUC K , K.J.; MET T E N , P.; BEL K N A P , J.K.; AN D gene. For example, using a two-step ing to human alcoholism. The identifi- st r a t e g y with the BXD RIs, res e a rc h e r s CRA B B E , J.C. Quantitative trait loci involved in identified three loci that accounted for cation of QTL s — e v en though th e s e genetic predisposition to acute alcohol withdrawal regions typically contain about 15 mil- in mice. Journal of Neuroscience 17 : 3 9 4 6 – 3 9 5 5 , nearly 70 percent of the gene-mediated 19 9 7 . di f f e r ences in acute alcohol withdrawal lion DNA base pairs and an estimated (B uck et al. 1997). One of these QTL s 500 to 1,000 genes—provides a signifi- COP E L A N D , N.G.; JEN K I N S , N.A.; GIL B E R T , D.J.; mapped to a region of mouse chrom o - ca n t ad v ance in the daunting task of E TA L. A map of the mouse: Current elucidating the genetic bases of alcohol applications and future prospects. Sc i e n c e 26 2 : 5 7 – so m e 11 that contains the genes for 66, 1993. se v eral components of a specific rec e p - effects. The subsequent identification tor for the brain chemical (i.e., neuro- of novel candidate genes and the clarifi- CRA B B E , J.C.; PHI L L I P S , T.J.; BUC K , K.J.; CUN N I N G H A M , transmitter) gamma-aminobutyric acid cation of their mechanisms of influence C.L.; AN D BEL K N A P , J.K. Identifying genes for alco- (GABA). Alcohol is known to influ- hol and drug sensitivity: Recent progress and future contribute to the eventual understand- di r e c t i o n s . Trends in Neurosciences 22:173–179, 1999. ence neurotransmitter activity at this ing of functional variability in humans DAR V A S I , A. Experimental strategies for the genetic GA B A A receptor; fu rt h e r m o r e, GABA that confers differential vulnerability to mediates many effects of sedative- h y p n o t i c alcoholism. ■ dissection of complex traits in animal models. dr ugs such as alcohol. Consequently, the Nature Genetics 18:19–24, 1998. genes encoding the GABAA rec e p t o r ENO C H , M.-A., AN D GOL D M A N , D. Genetics of components appeared to be good can- Acknowledgments alcoholism. In: Pfaff, D.W.; Berrettini, W.H.; Joh, didate genes for contributing to alcohol T.H.; and Maxson, S.C., eds. Genetic Influences on withdrawal. Fur ther investigation su p - Neural and Behavioral Functions. Boca Raton: CRC The author wishes to acknowledge the Press, 2000. pp. 147–157. po r ted this hypothesis, suggesting that su p p o r t of Dr. John C. Crabbe and a polymorphism involving a single build- AN D E R AN D RU G L Y A K individual National Res e a r ch Serv i c e L , E., K , L. Genetic dissection in g block (i.e., amino acid) in one of of complex traits: Guidelines for interpreting and Awa r d grant AA–05486 from the reporting results. Nature Genetics 11:241–246, 1995. the components of the GABAA rec e p - tor complex may be responsible for National Institute on Alcohol Abu s e an d Alcoholism during postdoctoral train- O’ B RI E N , S.J.; MEN O T T I -R AY M O N D , M.; MUR P H Y , di f f e r ences in alcohol withdrawal sever - W. J . ; E TA L. The promise of comparative ity (Buck and Hood 1998). Whether ing that included experience with the in mammals. Sc i e n c e 286:458–481, 1999. this polymorphism also plays a role in techniques described in this manuscript. Dr. Pamela Met t e n ’s suggestions on SCH U C K I T , M.A. Biological, psychological, and hu m a n alcohol withdrawal or in the environmental predictors of the alcoholism risk: A vulnerability to alcoholism remains to the manuscript are also thankfully longitudinal study. Journal of Studies on Alcohol 59 : be seen. Nonetheless, the value of such ac k n ow l e d g e d . 485–494, 1998.

174 Alcohol Research & Health