- Email: [email protected]
Prospects for quantitative trait locus methodology in gerontology
ExperimentalGerontology,Vol. 32, Nos. 1/2,pp. 49-54, 1997 Copyright© 1997Elsevier ScienceInc. Printed in the USA.All fightsreserved 0531-5565/97 $t 7.00 + .00 ELSEVIER
PII S0531-5565(96)00066-6
PROSPECTS FOR QUANTITATIVE TRAIT LOCUS METHODOLOGY IN GERONTOLOGY
GERALD E. MCCLEARN Center for Developmental and Health Genetics, Pennsylvania State University, UniversityPark, Pennsylvania USA Abstract--The remarkable advances in mapping of genomes have made it feasible to search among the hitherto anonymous polygenes of complex systems for loci of intermediate effect size (quantitative trait loci--QTL). Pursuing the strategy of identifying these QTL in complex systems will open the possibility of exploring the genetic architecture and the anatomical, physiological, biochemical, and molecular mechanisms underlying the phenotypes under investigation. One study (Gelman et al., 1988) has pioneered the exploration of QTL in the study of aging. Their engaging results, dealing with age at death as the endpoint measure, portend the power of the search for quantitative trait loci in agirlg processes. Copyright © 1997 Elsevier Science Inc.
Key Words: longevity, quantitative trait loci, recombinant inbred strains
INTRODUCTION DATA FROM inbred strains and derived generations constitute a very large part of the corpus of knowledge about the genetics of aging. The remarkable recent advances in mapping the mouse genome have made possible an extension of inbred strain methodology that has promise for bridging the gap between complex gerontological phenotypes and molecular mechanisms. Until very recently, there have been two general approaches to the genetic basis of individuality in complex traits. Though it is something of a simplification, these two generic classes of research on heredity may be called the Mendelian and the biometric or quantitative genetic approaches. The former approach is applicable to those situations in which alternative genotypes at a single genetic locus are associated with such large phenotypic differences that they generate a dichotomous outcome (e.g., affected or unaffected) regardless of the environmental circumstances of the individuals or their genotypes at other loci. The advantages of identifying a single locus are clear. Mendelian rules of transmission provide for clear, thongh stochastic, prediction,
Correspondence to: Gerald E. McCleam, Center for Developmental and Health Genetics, Pennsylvania State University, 101 Amy Gardner House, University Park, PA 16802 49
50
G.E. McCLEARN
and the study of physiological and molecular mechanisms is enormously facilitated by the capability of identifying genotypes of individuals. However, Mendelian analyses are often simply inapplicable to some of the most important phenotypes in a domain--those that are continuously distributed. The quantitative method, on the other hand, was designed explicitly to deal with the joint effects of numerous loci. In these cases, evidence is sought for the role of genetics in locating individuals along a continuum rather than in assigning them to a category. The quantitative genetic theory underlying analysis of this type of situation basically posits that numerous gene loci, each with only a small effect, influence the same phenotype. Furthermore, the role of environmental influences is explicitly included, and a major analytical outcome is the decomposition of the measured phenotypic variance into various genetic and environmental components. With elegant elaboration over the decades (see Falconer and Mackay, 1996), the theory and the accumulated empirical data pertinent to the quantitative model have contributed greatly to basic understanding of the workings of the genome. QUANTITATIVE TRAIT LOCI It has long been conjectured that the loci influencing many complex, continuously distributed phenotypes, instead of possessing traditionally defined small and equal effects, might vary in effect size, with some being sufficiently robust to be detectable with special effort (for example, Robertson, 1968). In the past, there were some successes in the search for these loci of intermediate magnitude of effect, but progress was slow until the recent avalanche of information concerning molecular markers of genomes of a variety of organisms. With much impetus from plant genetics, the identification of individual loci or chromosomal regions having detectable influence on variability of continuously distributed phenotypes has become a major research emphasis (see McClearn et al., 1991; Plomin et al., 1991a). The phrase, "quantitative trait locus" (QTL; used here both as singular and plural), has emerged as standard terminology to describe such loci. QTL research promises to be a major tactic in merging quantitative and Mendelian genetics. Such a merger will provide a potent methodology with access to the theoretical and empirical bases of both parent areas. There are two particularly important advantages to be expected from this merger. First is the route provided for the study of mechanism. Identification of a QTL provides the potential for the application of the full panoply of Mendelian and molecular genetics methods. Even if only one or a few QTL are identified, and even if they account for only a relatively small percentage of the variance of the phenotype, their study could illuminate significant features of the complex causal field that mediates the combined effects of all genetic and environmental influences on the trait (see McClearn, 1993). A second major advantage of the merger will be the clear guidelines for moving from the animal model research arena to that of human research. For example, the percentage of the mouse genome that is known to be syntenic to the human genome is substantial and new research increases the known correspondence almost daily. RI QTL methodology In essence, QTL research involves assessment of the relationship of some quantitatively distributed phenotype to marker genotypes. Several different methods of making this assessment have been utilized. One with some special advantages is the recombinant inbred (RI) strain method.
PROSPECTS FOR QTL IN GERONTOLOGY
51
RI strains are derived from the F2 between two progenitor inbred strains. The genotype of each RI strain thus constitutes a "reshuffling" or recombination of the alleles at loci for which the two progenitor strains differ. RIs were developed for the principal purpose of identifying and mapping single genes (Bailey, 1971). The basic logic is as follows: over the years, the RI strains have been characterized in respect to many genetic markers with known location on different chromosomes. If a new trait is identified that displays categorical distribution (i.e., the RIs fall into two distinct groups), then the pattern of this phenotypic distribution (the so-called strain distribution pattern or SDP) can be compared to the distribution of the alleles across the RIs separately for each already-known locus. If an SDP is found for which all the RI strains with one allele are in one phenotypic group and all those with the other allele are in the other group, the conclusion is warranted that a major Mendelian gene is involved, and that that gene is located on the chromosome in the general region of the known marker. In recent years, the explosion of information on molecular markers has provided chromosome maps of ever-increasing detail. It was early recognized that RI strains "should be especially useful for the analysis of complex characters" (Taylor, 1976). Of the many studies employing RIs in the last couple of decades, however, few have addressed continuously distributed characteristics. The logic extending the usefulness of the RIs for detecting QTL is straightforward (McClearn et al., 1991, Plomin et al., 1991a). If a locus has some effect on the phenotype, it might be possible to detect it by examining each known marker locus to determine if a significant mean difference exists between those RIs homozygous for one allele and those homozygous for the other allele. Whether such a locus will be detected in this manner will depend upon the effect size of the locus relative to the amount of variance contributed by other loci and by environmental sources, and also by the nearness of a marker of known location. Since its introduction, the RI QTL methodology has been rapidly deployed by a number of laboratories. A recent review (Crabbe et al., 1994) illustrates both the popularity and the success of the method in application to pharmacogenetics. Power and restrictions of the RI QTL approach There are two aspects of the RI approach which require comment. First, because of the large number of statistical tests performed (there are now nearly 1500 markers of the mouse genome) there will inevitably be a number of false positives. One response to this problem is to set a very high alpha level (Lander and Kruglyak, 1995). The risk of doing so is that the number of false negatives will be high. A useful compromise is to use a moderately stringent alpha level, but undertake independent replication in a different population of animals, such as Fls or F2s. A disadvantage to the RI QTL method is that its power is limited by the number of RI strains, not the total number of animals, because the RI means are the units of analysis. Thus, only QTL above a certain effect size will be discernible. For many research purposes this limitation is only a modest one, because the objective is often to discover some QTL, those of largest effect size, but not necessarily to obtain an exhaustive identification of all QTL in the polygenic system. On the other hand, the RI procedure has a sterling merit--the fixed nature of the genotypes of each of the RI strains (with the reservation, relevant to all inbred strain discussions, of the possible influence of mutational events). By using serial samples from the RI strains, it is possible to characterize the same genotypes on any number of phenotypes for which the animals must be sacrificed, or for which one measurement would prejudice later measurements of the same or of different phenotypes. The genotypic "repeatability" of the RIs has the effect that all newly generated data are immediately assimilable to the same pool of information. A new
52
G.E. McCLEARN
hypothesis about a biochemical mechanism underlying some phenomenon, for example, can be assessed by making only observations on the new variable, and relating the outcome to the data base already established. Thus, the value of an RI series continues to increase as data, both about phenotypes and about genotypes, flow together from all of the laboratories utilizing them. With methods using groups in which each animal has a unique genotype, such as F2s or backcrosses, such cumulation of information is impossible, of course.
A pioneering application to gerontology The recency of development of the QTL methodologies has not permitted extensive application to aging. A pioneering study (for mammalian QTL work generally and in application to gerontology specifically) was the work of Gelman et al. (1988), who studied longevity of the BXD recombinant inbred strains of mice derived from the progenitor C57BL/6 and DBA/2 strains. These two strains are widely utilized in studies on aging because of a well-established difference in lifespans. Six chromosomal regions were identified as putatively containing QTL influencing age at death. These QTL are "dispersed," in that for three of the six loci, the allele from the shorter-lived DBA/2 progenitor strain is associated with longer life in the RIs. The type of evidence that justifies nomination of a QTL is shown in Fig. i for the D12Nyul marker locus on chromosome 12. It is clear that a mean difference in longevity exists between those RI strains with the C57BL/6 allele and those with the DBA/2 allele at that locus. In addition to strain mean differences, differences in within-strain variability were commented upon by the authors. It is naturally of considerable interest to determine if this sensitivity
@
900
>- 800
%
700
u.I
60O
z
500
I
C57BL/6 Allele
DBA/2 Allele D12Nyu1 Locus Chromosome 12
FIG. 1, Relationship between mouse longevity and allelic status at the marker locus D 12Nyul. Drawn from data published by Gelman et al., 1988.
PROSPECTSFORQTLIN GERONTOLOGY
53
is itself under the detectable influence of a QTL. Upon screening in the present author's laboratory of the 974 currently available marker loci, two were found (one on chromosome 9 and one on chromosome 11) to be associated with RI standard deviations, but not with means for longevity. This "finding" must, of course, be viewed with considerable caution, because one such observation would be expected by chance. Nonetheless, the possibility is raised of identifying loci that influence sensitivity to environmental factors affecting longevity; assuming all strains to be equally genetically uniform, and assuming them to have been exposed to the same range of environmental circumstances, the differences in intrastrain variance must reflect differences in sensitivity to those environmental effects. The existence of such genetic influences would certainly enrich our conceptualization of aging processes (see Berg, 1987, for a discussion of "variability genes"). The RI advantage in making it possible to revisit the same genotypes for different phenotypes can also be illustrated by an example comparing the longevity data to other information in the literature concerning the BXD series. Of some 137 different phenotypes recorded in the Penn State BXD RI-QTL Cooperative Data Registry (see Plomin et al., 1991b) none were significantly (p < 0.01) correlated with longevity; two, however, were correlated with within-strain variability. These were brain weight (p < 0.004) and cerebellar weight (p < 0.003) from the data of Lassalle et al. (1994). Reservations about this outcome must be even stronger than in the previous example, because, in addition to the false-positive problem, there is a part-whole relationship between these two variables. If this outcome could be confirmed, however, it would offer an interesting supplement to the well known data on brain weight-longevity correlations across species (see Finch, 1990), and might suggest that the mechanism includes features relating to sensitivity to environmental agencies. From the density of caveats in this presentation, it should be clear that the search for QTL is, at this stage, largely a hypothesis-generating activity. Nominated loci require subsequent verification; correlated phenotypes similarly must be confirmed in other data sets; power is limited and false-positives are an inevitable feature of the approach. So, it is a risky business. But the payoff will be very substantial, indeed, opening complex, whole-organism, age-related phenotypes to the analytic might of molecular genetics. Acknowledgments--The author's aging research has been supported in part by grants from the National Institute on Aging (AG04948, AG09333, AG04563, AG10175) and the John D. and Catherine T. MacArthur Foundation Research Network on Successful Aging.
REFERENCES BAILEY, D.W. Recombinant inbred strains. Transplantation 11, 325-327, 1971. BERG, K. Genetics of coronary heart disease and its risk factors. In: Molecular Approaches to Human Polygenic Disease, Ciba Foundation Symposium 130, Bock, G. and Collins, G.M. (Editors), pp. 14-33, John Wiley & Sons Ltd., Chichester, UK, 1987. CRABBE, J.C., BELKNAP, J.K., and BUCK, K.J. Genetic animal models of alcohol and drug abuse. Science 264, 1715-1723, 1994. FALCONER, D.S. and MACKAY, T.F.C. Introduction to Quantitative Genetics, 4th ed., Longman Group, Ltd., Essex, UK, 1996. FINCH, C.E. Longevity, Senescence and the Genome, The University of Chicago Press, Chicago, IL, 1990. GELMAN, R., WATSON, A., BRONSON, R., and YUNIS, E. Murine chromosomal regions correlated with longevity. Genetics 118, 693-700, 1988. LANDER, E. and KRUGLYAK, L. Genetic dissection of complex traits: Guidelines for interpreting and reporting linkage results. Nat. Genet. 11, 241-247, 1995.
54
G.E. McCLEARN
LASSALLE, J.M., ROULLET, P., and HALLEY, H. An RI QTL analysis of hippocampal mossy fiber distribution using the BXD recombinant inbred mouse strains. Behav. Genet. 24, 519, 1994. MCCLEARN, G.E. Genetics, systems, and alcohol. Behav. Genet. 23, 223-230, 1993. MCCLEARN, G.E., PLOMIN, R., GORA-MASLAK, G., and CRABBE, J.C. The gene chase in behavioral science. Psychol. Sci. 2, 222-229, 1991. PLOMIN, R., MCCLEARN, G.E., GORA-MASLAK, G., and NEIDERHISER, J.M. Use of recombinant inbred strains to detect quantitative trait loci associated with behavior. Behav. Genet. 21, 99-116, 1991a. PLOMIN, R., MCCLEARN, G.E., GORA-MASLAK, G., and NEIDERHISER, J.M. An RI QTL cooperative data bank for recombinant inbred quantitative trait loci analyses. Behav. Genet. 21, 97-98, 1991b. ROBERTSON, A. The spectrum of genetic variation. In: Population Biology and Evolution, Landon, R.C. (Editor), pp. 5-16, Syracuse University Press, Syracuse, NY, 1968. TAYLOR, B.A. Development of recombinant inbred lines of mice. Behav. Genet. 6, 118, 1976.
PII S0531-5565(96)00066-6
PROSPECTS FOR QUANTITATIVE TRAIT LOCUS METHODOLOGY IN GERONTOLOGY
GERALD E. MCCLEARN Center for Developmental and Health Genetics, Pennsylvania State University, UniversityPark, Pennsylvania USA Abstract--The remarkable advances in mapping of genomes have made it feasible to search among the hitherto anonymous polygenes of complex systems for loci of intermediate effect size (quantitative trait loci--QTL). Pursuing the strategy of identifying these QTL in complex systems will open the possibility of exploring the genetic architecture and the anatomical, physiological, biochemical, and molecular mechanisms underlying the phenotypes under investigation. One study (Gelman et al., 1988) has pioneered the exploration of QTL in the study of aging. Their engaging results, dealing with age at death as the endpoint measure, portend the power of the search for quantitative trait loci in agirlg processes. Copyright © 1997 Elsevier Science Inc.
Key Words: longevity, quantitative trait loci, recombinant inbred strains
INTRODUCTION DATA FROM inbred strains and derived generations constitute a very large part of the corpus of knowledge about the genetics of aging. The remarkable recent advances in mapping the mouse genome have made possible an extension of inbred strain methodology that has promise for bridging the gap between complex gerontological phenotypes and molecular mechanisms. Until very recently, there have been two general approaches to the genetic basis of individuality in complex traits. Though it is something of a simplification, these two generic classes of research on heredity may be called the Mendelian and the biometric or quantitative genetic approaches. The former approach is applicable to those situations in which alternative genotypes at a single genetic locus are associated with such large phenotypic differences that they generate a dichotomous outcome (e.g., affected or unaffected) regardless of the environmental circumstances of the individuals or their genotypes at other loci. The advantages of identifying a single locus are clear. Mendelian rules of transmission provide for clear, thongh stochastic, prediction,
Correspondence to: Gerald E. McCleam, Center for Developmental and Health Genetics, Pennsylvania State University, 101 Amy Gardner House, University Park, PA 16802 49
50
G.E. McCLEARN
and the study of physiological and molecular mechanisms is enormously facilitated by the capability of identifying genotypes of individuals. However, Mendelian analyses are often simply inapplicable to some of the most important phenotypes in a domain--those that are continuously distributed. The quantitative method, on the other hand, was designed explicitly to deal with the joint effects of numerous loci. In these cases, evidence is sought for the role of genetics in locating individuals along a continuum rather than in assigning them to a category. The quantitative genetic theory underlying analysis of this type of situation basically posits that numerous gene loci, each with only a small effect, influence the same phenotype. Furthermore, the role of environmental influences is explicitly included, and a major analytical outcome is the decomposition of the measured phenotypic variance into various genetic and environmental components. With elegant elaboration over the decades (see Falconer and Mackay, 1996), the theory and the accumulated empirical data pertinent to the quantitative model have contributed greatly to basic understanding of the workings of the genome. QUANTITATIVE TRAIT LOCI It has long been conjectured that the loci influencing many complex, continuously distributed phenotypes, instead of possessing traditionally defined small and equal effects, might vary in effect size, with some being sufficiently robust to be detectable with special effort (for example, Robertson, 1968). In the past, there were some successes in the search for these loci of intermediate magnitude of effect, but progress was slow until the recent avalanche of information concerning molecular markers of genomes of a variety of organisms. With much impetus from plant genetics, the identification of individual loci or chromosomal regions having detectable influence on variability of continuously distributed phenotypes has become a major research emphasis (see McClearn et al., 1991; Plomin et al., 1991a). The phrase, "quantitative trait locus" (QTL; used here both as singular and plural), has emerged as standard terminology to describe such loci. QTL research promises to be a major tactic in merging quantitative and Mendelian genetics. Such a merger will provide a potent methodology with access to the theoretical and empirical bases of both parent areas. There are two particularly important advantages to be expected from this merger. First is the route provided for the study of mechanism. Identification of a QTL provides the potential for the application of the full panoply of Mendelian and molecular genetics methods. Even if only one or a few QTL are identified, and even if they account for only a relatively small percentage of the variance of the phenotype, their study could illuminate significant features of the complex causal field that mediates the combined effects of all genetic and environmental influences on the trait (see McClearn, 1993). A second major advantage of the merger will be the clear guidelines for moving from the animal model research arena to that of human research. For example, the percentage of the mouse genome that is known to be syntenic to the human genome is substantial and new research increases the known correspondence almost daily. RI QTL methodology In essence, QTL research involves assessment of the relationship of some quantitatively distributed phenotype to marker genotypes. Several different methods of making this assessment have been utilized. One with some special advantages is the recombinant inbred (RI) strain method.
PROSPECTS FOR QTL IN GERONTOLOGY
51
RI strains are derived from the F2 between two progenitor inbred strains. The genotype of each RI strain thus constitutes a "reshuffling" or recombination of the alleles at loci for which the two progenitor strains differ. RIs were developed for the principal purpose of identifying and mapping single genes (Bailey, 1971). The basic logic is as follows: over the years, the RI strains have been characterized in respect to many genetic markers with known location on different chromosomes. If a new trait is identified that displays categorical distribution (i.e., the RIs fall into two distinct groups), then the pattern of this phenotypic distribution (the so-called strain distribution pattern or SDP) can be compared to the distribution of the alleles across the RIs separately for each already-known locus. If an SDP is found for which all the RI strains with one allele are in one phenotypic group and all those with the other allele are in the other group, the conclusion is warranted that a major Mendelian gene is involved, and that that gene is located on the chromosome in the general region of the known marker. In recent years, the explosion of information on molecular markers has provided chromosome maps of ever-increasing detail. It was early recognized that RI strains "should be especially useful for the analysis of complex characters" (Taylor, 1976). Of the many studies employing RIs in the last couple of decades, however, few have addressed continuously distributed characteristics. The logic extending the usefulness of the RIs for detecting QTL is straightforward (McClearn et al., 1991, Plomin et al., 1991a). If a locus has some effect on the phenotype, it might be possible to detect it by examining each known marker locus to determine if a significant mean difference exists between those RIs homozygous for one allele and those homozygous for the other allele. Whether such a locus will be detected in this manner will depend upon the effect size of the locus relative to the amount of variance contributed by other loci and by environmental sources, and also by the nearness of a marker of known location. Since its introduction, the RI QTL methodology has been rapidly deployed by a number of laboratories. A recent review (Crabbe et al., 1994) illustrates both the popularity and the success of the method in application to pharmacogenetics. Power and restrictions of the RI QTL approach There are two aspects of the RI approach which require comment. First, because of the large number of statistical tests performed (there are now nearly 1500 markers of the mouse genome) there will inevitably be a number of false positives. One response to this problem is to set a very high alpha level (Lander and Kruglyak, 1995). The risk of doing so is that the number of false negatives will be high. A useful compromise is to use a moderately stringent alpha level, but undertake independent replication in a different population of animals, such as Fls or F2s. A disadvantage to the RI QTL method is that its power is limited by the number of RI strains, not the total number of animals, because the RI means are the units of analysis. Thus, only QTL above a certain effect size will be discernible. For many research purposes this limitation is only a modest one, because the objective is often to discover some QTL, those of largest effect size, but not necessarily to obtain an exhaustive identification of all QTL in the polygenic system. On the other hand, the RI procedure has a sterling merit--the fixed nature of the genotypes of each of the RI strains (with the reservation, relevant to all inbred strain discussions, of the possible influence of mutational events). By using serial samples from the RI strains, it is possible to characterize the same genotypes on any number of phenotypes for which the animals must be sacrificed, or for which one measurement would prejudice later measurements of the same or of different phenotypes. The genotypic "repeatability" of the RIs has the effect that all newly generated data are immediately assimilable to the same pool of information. A new
52
G.E. McCLEARN
hypothesis about a biochemical mechanism underlying some phenomenon, for example, can be assessed by making only observations on the new variable, and relating the outcome to the data base already established. Thus, the value of an RI series continues to increase as data, both about phenotypes and about genotypes, flow together from all of the laboratories utilizing them. With methods using groups in which each animal has a unique genotype, such as F2s or backcrosses, such cumulation of information is impossible, of course.
A pioneering application to gerontology The recency of development of the QTL methodologies has not permitted extensive application to aging. A pioneering study (for mammalian QTL work generally and in application to gerontology specifically) was the work of Gelman et al. (1988), who studied longevity of the BXD recombinant inbred strains of mice derived from the progenitor C57BL/6 and DBA/2 strains. These two strains are widely utilized in studies on aging because of a well-established difference in lifespans. Six chromosomal regions were identified as putatively containing QTL influencing age at death. These QTL are "dispersed," in that for three of the six loci, the allele from the shorter-lived DBA/2 progenitor strain is associated with longer life in the RIs. The type of evidence that justifies nomination of a QTL is shown in Fig. i for the D12Nyul marker locus on chromosome 12. It is clear that a mean difference in longevity exists between those RI strains with the C57BL/6 allele and those with the DBA/2 allele at that locus. In addition to strain mean differences, differences in within-strain variability were commented upon by the authors. It is naturally of considerable interest to determine if this sensitivity
@
900
>- 800
%
700
u.I
60O
z
500
I
C57BL/6 Allele
DBA/2 Allele D12Nyu1 Locus Chromosome 12
FIG. 1, Relationship between mouse longevity and allelic status at the marker locus D 12Nyul. Drawn from data published by Gelman et al., 1988.
PROSPECTSFORQTLIN GERONTOLOGY
53
is itself under the detectable influence of a QTL. Upon screening in the present author's laboratory of the 974 currently available marker loci, two were found (one on chromosome 9 and one on chromosome 11) to be associated with RI standard deviations, but not with means for longevity. This "finding" must, of course, be viewed with considerable caution, because one such observation would be expected by chance. Nonetheless, the possibility is raised of identifying loci that influence sensitivity to environmental factors affecting longevity; assuming all strains to be equally genetically uniform, and assuming them to have been exposed to the same range of environmental circumstances, the differences in intrastrain variance must reflect differences in sensitivity to those environmental effects. The existence of such genetic influences would certainly enrich our conceptualization of aging processes (see Berg, 1987, for a discussion of "variability genes"). The RI advantage in making it possible to revisit the same genotypes for different phenotypes can also be illustrated by an example comparing the longevity data to other information in the literature concerning the BXD series. Of some 137 different phenotypes recorded in the Penn State BXD RI-QTL Cooperative Data Registry (see Plomin et al., 1991b) none were significantly (p < 0.01) correlated with longevity; two, however, were correlated with within-strain variability. These were brain weight (p < 0.004) and cerebellar weight (p < 0.003) from the data of Lassalle et al. (1994). Reservations about this outcome must be even stronger than in the previous example, because, in addition to the false-positive problem, there is a part-whole relationship between these two variables. If this outcome could be confirmed, however, it would offer an interesting supplement to the well known data on brain weight-longevity correlations across species (see Finch, 1990), and might suggest that the mechanism includes features relating to sensitivity to environmental agencies. From the density of caveats in this presentation, it should be clear that the search for QTL is, at this stage, largely a hypothesis-generating activity. Nominated loci require subsequent verification; correlated phenotypes similarly must be confirmed in other data sets; power is limited and false-positives are an inevitable feature of the approach. So, it is a risky business. But the payoff will be very substantial, indeed, opening complex, whole-organism, age-related phenotypes to the analytic might of molecular genetics. Acknowledgments--The author's aging research has been supported in part by grants from the National Institute on Aging (AG04948, AG09333, AG04563, AG10175) and the John D. and Catherine T. MacArthur Foundation Research Network on Successful Aging.
REFERENCES BAILEY, D.W. Recombinant inbred strains. Transplantation 11, 325-327, 1971. BERG, K. Genetics of coronary heart disease and its risk factors. In: Molecular Approaches to Human Polygenic Disease, Ciba Foundation Symposium 130, Bock, G. and Collins, G.M. (Editors), pp. 14-33, John Wiley & Sons Ltd., Chichester, UK, 1987. CRABBE, J.C., BELKNAP, J.K., and BUCK, K.J. Genetic animal models of alcohol and drug abuse. Science 264, 1715-1723, 1994. FALCONER, D.S. and MACKAY, T.F.C. Introduction to Quantitative Genetics, 4th ed., Longman Group, Ltd., Essex, UK, 1996. FINCH, C.E. Longevity, Senescence and the Genome, The University of Chicago Press, Chicago, IL, 1990. GELMAN, R., WATSON, A., BRONSON, R., and YUNIS, E. Murine chromosomal regions correlated with longevity. Genetics 118, 693-700, 1988. LANDER, E. and KRUGLYAK, L. Genetic dissection of complex traits: Guidelines for interpreting and reporting linkage results. Nat. Genet. 11, 241-247, 1995.
54
G.E. McCLEARN
LASSALLE, J.M., ROULLET, P., and HALLEY, H. An RI QTL analysis of hippocampal mossy fiber distribution using the BXD recombinant inbred mouse strains. Behav. Genet. 24, 519, 1994. MCCLEARN, G.E. Genetics, systems, and alcohol. Behav. Genet. 23, 223-230, 1993. MCCLEARN, G.E., PLOMIN, R., GORA-MASLAK, G., and CRABBE, J.C. The gene chase in behavioral science. Psychol. Sci. 2, 222-229, 1991. PLOMIN, R., MCCLEARN, G.E., GORA-MASLAK, G., and NEIDERHISER, J.M. Use of recombinant inbred strains to detect quantitative trait loci associated with behavior. Behav. Genet. 21, 99-116, 1991a. PLOMIN, R., MCCLEARN, G.E., GORA-MASLAK, G., and NEIDERHISER, J.M. An RI QTL cooperative data bank for recombinant inbred quantitative trait loci analyses. Behav. Genet. 21, 97-98, 1991b. ROBERTSON, A. The spectrum of genetic variation. In: Population Biology and Evolution, Landon, R.C. (Editor), pp. 5-16, Syracuse University Press, Syracuse, NY, 1968. TAYLOR, B.A. Development of recombinant inbred lines of mice. Behav. Genet. 6, 118, 1976.