Genetic Analysis of a Quantitative Trait in a Mouse Model of Polycystic Kidney Disease

OLGA A. IAKOUBOVA, HOLLY DUSHKIN, and DAVID R. BEIER

Genetics Division, Brigham and Womens Hospital, Harvard Medical School, Boston, Massachusettes

The development of a variety of powerful tools for genome analysis has facilitated the ability to genetically map loci which contribute to the variation of a quantitative trait. However, the fact that these traits are often determined as a result of complex genetic interactions has made their analysis considerably more difficult then the molecular characterization of qualitative traits that are monogenic in origin. We have described the use of a novel method of chromosomal exclusion to map the recessive mutation juvenile cystic kidney ( jck) to mouse chromosome 11 using an intercross between (C57BL/6J DBA/2J) F1 jck/ mice. The severity of polycystic kidney disease (PKD) in the intercross progeny, which could be quantitated as a function of kidney size, was significantly more variable than that found in the parental C57BL/6J strain, suggesting that a modifier locus or loci introduced from DBA/2J affects expression of jck. Two regions (one from DBA/2J on chromosome 10 and a second from C57BL/6J on chromosome 1) were found to be associated with inheritance of a more severe PKD phenotype. The finding of a highly significant association of inheritance of a C57BL/6J-related locus with disease severity was unexpected since the PKD phenotype in this inbred background is mild. This result suggests that inheritance in the affected F2 mice of loci from the two different parental backgrounds results in the more severe phenotype, presumably as a consequence of a direct or indirect interaction between their protein products. This type of effect, which is an example of genetic epistasis, will make the molecular characterization of loci that contribute to complex traits markedly more difficult than the analysis of monogenic disorders. Iakoubova OA, Dushkin H, Beier DR. Genetic analysis of a quantitative trait in a mouse model of polycystic kidney disease.
AM J RESPIR CRIT CARE MED 1997;156:S72S77.

The development of easily scored highly polymorphic markers, detailed genetic maps, and powerful computational tools has made it possible to investigate the genetic contribution to quantitative traits in mammalian systems. These tools can be used to analyze genotypic and phenotypic information for specific families, populations, or crosses in order to localize the regions of chromosomes whose inheritance appears to correlate with quantitative variation; these regions are presumed to contain quantitative trait loci (QTLs). This approach has been especially well suited for genetic studies in the mouse, since there exist well diverged, genetically inbred laboratory strains that exhibit differences in quantitative phenotypes that are heritable and reproducible. Studies have identified QTLs for a wide variety of traits, including several with particular relevance to respiratory disease, such as airway hyperresponsiveness (1), humoral immunity (2), and lung tumor development (3). However, while the identification of QTLs in mouse crosses is technically feasible, there are several reasons that their localization to specific genetic intervals amenable to po-

This research was supported by grants from the NIDDK (R01DK45639) and from the Life & Health Insurance Medical Research Fund. Correspondence and requests for reprints should be addressed to David R. Beier, M.D., Ph.D., Brigham and Womens Hospital, 75 Francis Street, Boston, MA 02115. Am J Respir Crit Care Med Vol 156. pp S72S77, 1997

sitional cloning analyses may prove considerably more problematic. First, the nongenetic variation in these traits, as well as the logistic limits on the number of animals studied, usually results in the localization of a QTL with only modest resolution; that is, the peaks obtained in the computational analysis of quantitative traits are often broad. Secondly, the positional cloning strategies for mutations that have unambiguously scored phenotypes may not be readily applicable for the cloning of a modifying locus. These strategies generally rely on the ability to narrow the interval containing the gene of interest by identifying closely linked recombination break points. Because a quantitative trait is a continuum, one cannot assume with any degree of certainty that a putative recombinant is a member of a severely or mildly affected class, and not an outlier from the other class. Thus, the ability to localize the modifying trait to a small interval for positional cloning is greatly limited. Finally, quantitative traits are often polygenic in origin, and may exhibit both complex interactions with environmental factors, as well as unexpected intergenic effects as a consequence of epistasis. Our studies of quantitative traits include the analysis of modifying loci that affect the progression of a mouse model of polycystic kidney disease (PKD) (4). There exist a number of recessive murine mutations that predispose to the development of PKD (59). We have described the mutation juvenile cystic kidneys ( jck), in which enlarged kidneys can be palpated between 5 and 7 weeks of age (10). The jck mutation

Iakoubova, Dushkin, and Beier: Genetic Analysis of Quantitative Traits

S73

originally arose in a transgenic line of mice that carries an insertion at the perinatal lethality (ple) locus on mouse chromosome 15 (11). jck is not a transgene insertion mutation, since it segregates independently from the transgene. One of the notable features of both dominant and recessive PKD in humans is the variability of the phenotype, both with respect to disease progression and extrarenal manifestations. While a significant component of this variance is likely due to genetic heterogeneity, marked variability of clinical disease even within kindreds (which presumably share the same mutation predisposing to PKD) is well documented (12). In murine models, effects of genetic background on disease phenotype is also suggested by the reports of increased disease severity when either cpk or pcy mutations were introduced into the DBA/2J genetic background (5, 6). In this report, we describe some of the approaches that were used for the analysis of QTLs that affect disease severity of the jck mutation (4).

GENOTYPE ANALYSIS OF AN INTERCROSS

To map the position of jck, an intercross with DBA/2J mice was done. Male homozygous jck mice, which are descended from mice that had been serially backcrossed with C57BL/6J mice for more than 10 generations and have proved to be C57BL/6J at every locus tested, were mated to DBA/2J female mice. F1 jck/ were mated and F2 progeny were sacrificed at 67 weeks and scored for the presence of PKD. Of 462 F2 progeny tested, 105 (23%) were found to have PKD, consistent with what would be expected for an autosomal recessive disorder. It was noted that the size of the polycystic kidneys found in age-matched F2 mice was quite variable compared to sizes of kidneys in mice with a parental C57BL/6J-like background, with weights of paired fixed kidneys ranging from 0.5 to 3 g (Figure 1). In general, the increase in kidney size reflected more extensive disease (as opposed to simply larger cysts) as judged by histopathologic analysis. The observation that progeny from the intercross had more variable disease than the parental line suggests that one or more modifying loci affecting severity had been introduced from the DBA/2J background. To map the modifying locus kidney weights from affected mice were recorded for quantitive trait locus (QTL) analysis (13). The presence of PKD in smaller kidneys was confirmed by histopathology. The mean size of the affected kidneys in the C57BL/6J parental and F2 progeny mice was not significantly different (0.82 g versus 0.90 g, respectively); however, the variance of the F2 progeny exceeded that of the C57BL/6J parental mice by greater than three-fold (0.28 g2 versus 0.08 gm2, respectively). An F-test of these variances is consistent with their being significantly different (p 0.002). Genotype determination was done by analysis of microsatellite markers that were amplified by PCR using recommended protocols (14) and electrophoresed on a 6% polyacrylamide denaturing gel. Microsatellite markers polymorphic between C57BL/6J and DBA/2J were chosen using information provided by the Genome Center at the Whitehead Institute (Cambridge, MA). Genotype data were recorded using the MapManager computer program (15). To expedite the genetic analysis a strategy of chromosomal exclusion was employed; this method is described in detail in the article by Iakoubova and colleagues (1995). This analysis exploits the fact that in unselected progeny of an intra- or interspecific cross, a significant number of mice will inherit from either parent chromosomes that are apparently nonrecombinant; i.e., they carry markers corresponding to a single strain. If, however, in the analysis of a recessive mutation one exam-

ines only affected mice, then chromosomes that are inherited as nonrecombinant from the unaffected parent cannot carry the mutation. That is, the mutant chromosome in affected mice must be homozygous for the mutant parental strain somewhere along its length. One can approximate the number of nonrecombinant chromosomes by simply testing markers from the proximal and distal ends of the individual chromosomes. This can then serve to exclude chromosomes that are unlikely to carry the mutation, and the remainder candidate chromosomes can be analyzed in more detail. This strategy can also be used as a screen for loci that contribute to a quantitative trait, by comparing cohorts of mice whose phenotypes are grouped at the tails of the distribution of the trait being tested. This strategy differs from simply genotyping across the genome to assess allele distribution in that it is effectively a haplotype analysis of the mutant mice. A computer program to facilitate this analysis is available from the authors. This strategy of haplotype analysis was initially performed on 20 intercross progeny (corresponding to 40 meioses) using a cohort of mice with severely affected kidneys (combined kidney weight 1.4 g), since this would allow the identification of both the jck mutation and the modifying loci introduced from DBA/2J that resulted in increased disease severity. For each chromosome, both C57BL/6J and DBA/2J apparent nonrecombinants were scored. Five chromosomes had markedly less than the number of nonrecombinants that would be predicted according to the formula described in (4). These were analyzed in more detail, with at least 3 additional markers per chromosome. No significant linkage was found for loci on two of the five chromosomes. Significant linkage (p 0.01)

Figure 1. The distribution of fixed kidney weights in wild-type (wt) mice at 67 weeks of age is compared to that found for affected jck mice in the parental background (B6 jck), the C57BL/6J DBA/2J) F1 jck/ intercross progeny (F2 jck), and the N3 DBA/2J backcross mice (D2 jck). Data are from Reference 4.

S74
TABLE 1

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

VOL 156

1997

SEGREGATION ANALYSIS OF MICROSATELLITE MARKERS IN JCK MICE

Chr 1 2 3 4 5 6 7 8 9 10 No. Markers 12 3 4 3 3 6 3 6 5 9 Max. 2 28.1 1.1 1.1 2.7 4.8 3.3 1.2 5.9 3.6 10.0 Chr 11 12 13 14 15 16 17 18 19 No. Markers 17 4 4 5 4 4 3 4 3 Max. 2 60 0.3 1.7 2.7 4.5 3.6 1.2 1.4 2.7

chromosomes were analyzed in detail to more precisely localize the linked loci (Table 2). To ensure the validity of the chromosomal exclusion strategy, additional markers on all other chromosomes were also tested; no linkage at any other site was detected (Table 1).

GENETIC LOCALIZATION OF QUANTITATIVE TRAIT LOCI (QTLS)

To further analyze these results, an additional 20 affected jck mice were tested for linkage to loci on chromosomes 1, 10, and 11, but in this case mice with mild disease (combined kidney weight 0.5 g) were used. For these mice, nonrandom cosegregation was observed only for markers on chromosome 11, confirming that this is the position of the jck mutation (Table 2). The random segregation of markers on chromosome 1 and chromosome 10 in the least severely affected mice implicates these as modifying loci associated with disease severity (Table 1). To assess this in more detail, quantitative trait locus (QTL) analysis was employed to test for nonrandom segregation with a quantitative trait, which in this case is kidney weight. Of the 105 affected mice originally identified, a subset of 55 (including the 40 previously characterized) with a more extreme 0.6 or 1.35 g) phenotype (i.e., combined kidney weight were analyzed at 6 loci on chromosome 1 covering a region of

The randomness of segregation in a cohort of severely affected jck mice of microsatellite markers distributed across the autosomes of the mouse was tested by 2 analysis. The number of markers tested and the maximal 2 value obtained for loci on a chromosome is shown. A 2 value of 6.0 corresponds to p 0.05. Nonrandom distributions with p 0.05 were found for loci on chromosomes 1, 10, and 11; these analyses are shown in more detail in Table 2. Data are from Reference 4.

was found for markers on chromosome 1 and 11 (which had one and no apparent non-recombinant chromosomes of DBA/ 2J origin, respectively) and chromosome 10 (which had two apparent nonrecombinants of C57BL/6J origin) (Table 1). These

TABLE 2 LINKAGE ANALYSIS OF CHROMOSOMES 1, 10, AND 11 TO POLYCYSTIC KIDNEY DISEASE IN THE JCK MOUSE
Map position (cM) 6.1 43.7 55.5 61.7 76.6 82.2 86.1 86.7 99.4 110.4 110.7 116.6 2.3 12.4 21.3 29.9 47.7 52.4 65.6 70.1 17.9 23.5 34.8 35.9 35.9 37 39.2 41.5 42.6 Most Severely Affected Mice* Locus D1Mit1 D1Mit7 D1Mit48 D1Mit11 D1Mit30 D1Mit42 D1Mit16 D1Mit57 D1Mit150 D1Mit17 D1Mit56 D1Mit155 D10Mit28 D10Mit3 D10Mit20 D10Mit32 D10Mit10 D10Mit2 D10Mit14 D10Mit102 D11Mit20 D11Mit24 D11Mit28 D11Mit15 Atp1b2 D11Mit90 D11Nds1 D11Mit34 D11Mit35 D/D 5 0 0 0 0 1 0 1 1 0 0 0 4 5 6 6 8 8 10 10 0 0 0 0 0 0 0 0 0 D/B 7 7 5 7 4 3 4 4 3 4 4 6 12 11 13 12 12 11 10 10 8 5 1 1 1 1 1 0 8 B/B 5 13 12 12 14 15 15 15 16 14 15 13 3 3 1 1 0 0 0 0 11 15 19 19 19 19 19 20 12
2

Least Severely Affected Mice D/D 1 2 3 5 4 3 5 D/B 8 9 13 18 13 10 18 B/B 4 7 3 5 3 6 5 2 2.2 2.7 3.2 0.26 1.9 1 0.26

Chr 1

18.7 17.3 15.7 25 28.1 24.8 26.8 32.6 27.5 24.8 16.4

10

5 7 6 10 10 12.6 27.5 52.7 52.7 52.7 52.7 52.7 60 15.2 5 3 0 0 0 0 0 0 0 0

13 9 11 10 10 4 2 2 2 2 0 2 3

2 4 3 5 6 16 18 18 18 18 20 18 17

2.7 1 1.1 0.4 1.0 32.8 51.0 51.0 51.0 51.0 62.0 51.0 37.4

11

Definition of abbreviations: Chr chromosome; B C57BL/6J; D DBA/2J. * The cohort (20) of F2 affected mice with kidney weight 1.4 g. The cohort (20) of F2 affected mice with kidney weight 0.6 g. Only 2 values 9.2 (p 0.01) are shown. All 2 values are shown. Map positions of markers (with respect to the centromere) are according to MIT Genome Center (Dietrich et al., 1994). In cases of markers that are not resolved in their mapping analysis, data for only one marker is shown. Data are from Reference 4.

Iakoubova, Dushkin, and Beier: Genetic Analysis of Quantitative Traits

S75

88 cM and 5 loci on chromosome 10 covering a region of 55 cM, respectively, with a mean interval size of 16 cM. Genotype data and kidney weights were analyzed using the Mapmaker/ QTL program (16). Inheritance of a C57BL/6J locus on chromosome 1 correlates significantly with increased kidney size, with a maximal LOD score of 16.8 for the interval between D1Mit7 and D1Mit30 (Figure 2); this locus accounts for 74% of the variance observed for kidney size in affected F2 progeny. The QTL/Mapmaker analysis can be constrained to test the likelihood that observed associations are consistent with dominant, recessive, or additive modes of inheritance. When markers on chromosome 1 were tested for their association with disease severity, the maximum LOD score was found for recessive inheritance of the C57BL/6J alleles. The interval between D1Mit7 and D1Mit30 that includes this region is still large; however, since the QTL peak is broad, localization of the modifier gene within this interval can most reliably be accomplished by constructing and testing congenic strains (see DISCUSSION). Consistent with the 2 analysis shown in Table 2, inheritance of a DBA/2J region on distal chromosome 10 between D10Mit14 and D10Mit102 shows an association when analyzed for a QTL for severity of kidney disease (Figure 3). This

accounts for 12% of the observed variance, with a maximal LOD score of 2.1. Modeling of the inheritance using QTL/ Mapmaker suggests that the effect of the DBA/2J allele at this locus is additive, with a maximum LOD score of 2.0 in this interval for the constrained analysis.

QUANTITATIVE TRAIT VARIATION OCCURS AS A RESULT OF A GENE INTERACTION

To further analyze the heritable aspect of PKD severity, an additional test was performed after crossing jck into the DBA/ 2J background for three generations. N3 jck / mice were intercrossed and then genotyped to ensure that loci spanning chromosome 1 and chromosome 10 carried only DBA/2J alleles. Kidney sizes of affected progeny were recorded and are shown in Figure 1. Mean affected kidney weight in these backcross progeny was 0.93 g, compared to 0.82 g for the parental strain and 0.90 g for the F2 intercross progeny. Of note is that the variance in the DBA/2J background was less than half that found for the F2 progeny (0.12 g2 versus 0.28 g2). This variance is not significantly different than that found for the C57BL/6J parental mice. Thus, the most severe PKD phenotype is found in (C57BL/6J DBA/2J)F2 hybrid mice that retain C57BL/6J

Figure 2. LOD score values for intervals on chromosome 1 associated with disease severity in affected jck mice, calculated according to QTL/Mapmaker. The microsatellite markers used for analysis are shown at their map positions (in cM from the centromere). Data are from Reference 4.

S76

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

VOL 156

1997

Figure 3. LOD score values for intervals on chromosome 10 associated with disease severity in affected jck mice, calculated according to QTL/Mapmaker. The microsatellite markers used for analysis are shown at their map positions (in cM from the centromere). Data are from Reference 4.

alleles on chromosome 1. That is, it appears that the severe phenotype occurs only in mice that have some contribution from both parental genetic backgrounds.

DISCUSSION
A novel strategy of chromosomal exclusion was used to map jck, a recessive mutation predisposing to PKD, to mouse chromosome 11 (4). Despite the localization of jck to a small well defined interval, its localization in the human genome cannot be reliably predicted based on conservation of linkage relationships. This is because the human homologues of most genes proven to map immediately proximal to this region lie on chromosome 17p near the telomere (e.g., Trp53, Atp1b2), while those that lie immediately distal map to human 17q (e.g., Nf1). In addition, the observation that F2 progeny of (C57BL/6J DBA/2J) F1 jck/ intercross showed a much more variable phenotype with respect to progression of PKD than the parental C57BL/6J-related strain suggested that one or more modifying loci were introduced from the DBA/2J background. By analyzing separately populations of severely affected and mildly affected mice, presumptive modifying loci on chromosomes 1 and 10 were identified (4). A completely unexpected finding was that the allele associated with increased disease severity on chromosome 1 was derived from the C57BL/6J background. Since the PKD phenotype in this background is not severe, the association of a C57BL/6J-related locus with increased disease severity was not anticipated. We have proposed that it is, in fact, the inheritance of both the homozygous C57BL/6J locus on chromosome 1 and a DBA/2J gene that results in the severe phenotype, presumably as a consequence of an interaction (either direct or indirect) between their protein products. The fact that the variation in kidney size is reduced (compared to the F2 intercross progeny) as the jck mutation is serially backcrossed into a DBA/2J background (and specifically tested for the absence of C57BL/6J alleles on chromosome 1) supports this hypothesis. This observation has significant implications for the understanding of the biology of modifying interactions, since it suggests that it is not only the strain backgrounds that are of import, but also the particular strain combinations. This result is an example of genetic epistasis, in which the observed phenotype does not reflect the simple additive effects of loci that contribute to the

trait, but rather occurs in an unpredictable fashion, most likely because these genes do not function completely independently (17). A DBA/2J allele on distal chromosome 10 was also significantly associated with PKD in severely affected mice, but not in mildly affected mice. QTL analysis of a larger sample of affected mice was done with markers on chromosome 10, and the LOD score for association with increased kidney size was found to be 2.1, which is consistent with linkage, although it does not conclusively prove it. It is possible that there exists another DBA/2J locus that contributes to the variance in disease severity we observed. This is particularly true for a dominantly inherited modifier, which would be identified less efficiently using the mapping strategy we have described. To reliably ascertain a dominantly inherited QTL, a comprehensive screen of the genome is needed. Towards this end a total of 102 markers in the cohort of 20 severely affected mice (representing 40 meioses) were analyzed; no additional DBA/2J loci segregating with disease severity were detected (Table 1). A significant problem is that the gene that interacts with the modifying locus we have identified on chromosome 1 need only be nonrandomly distributed in the 1/16 of intercross progeny that are both homozygous for the C57BL/6J allele of the modifier and for jck. If this gene acts in an additive or dominant fashion, it is possible that a skewing from a random distribution sufficient to generate a LOD score of greater than 3 will not be achieved in our cross. Therefore, an even larger cross will be necessary to confirm the provisional association of the chromosome 10 locus we have found with disease severity, as well as to test for additional DBA loci that might contribute to the PKD phenotype, since this can best be done by performing a genome-wide screen of the distribution of alleles in only those affected mice which are homozygous for the C57BL/6J allele of the modifier. The unambiguous phenotype of the jck mutation and its localization to a small genetic interval suggests that positional cloning of this locus should be feasible. The more precise delineation of the regions containing the putative modifying genes can be best accomplished by generating congenic strains that divide the chromosomes carrying these loci into defined intervals and testing for their effect on PKD severity in an appropriate cross. These can then be examined for candidate genes that may play a role in kidney tubule growth or function. One useful constraint on considering candidate loci is that, given the evidence of background-specific effects, the genes should be different in structure or expression between C57BL/6J and DBA/2J. In a preliminary analysis of congenic mice in which intervals of C57BL/6J-derived alleles were introduced into a DBA/ 2J genetic background using serial backcrosses, we obtained the unexpected result that no mice demonstrated the very severe phenotype found for some affected animals in the C57BL/6J-DBA/2J F2 progeny. One possible explanation for this observation is that there are two B6-derived modifying loci on chromosome 1, and these were separated in the congenic mice, which mitigated their effect. In fact, the QTL analysis shown in Figure 2 does appear to have two peaks; more importantly, the region of significant linkage is distributed over a much larger interval than that which might be expected for a single QTL. The hypothesis that there are two QTLs in this region can be tested by generating a larger congenic that should include them both. This result suggests that a more expedient strategic approach to the analysis of QTLs in a mouse system would be the initial generation of consomic mice (in which an entire chromosome is introduced to a different genetic background),

Iakoubova, Dushkin, and Beier: Genetic Analysis of Quantitative Traits

S77
munity. J. Immunol. 157:24982505. 3. Fijneman, R. J. A., S. S. Devries, R. C. Jansen, and P. Demant. 1996. Complex interactions of new quantitative trait loci, sluc1, sluc2, sluc3, and sluc4, that influence the susceptibility to lung cancer in the mouse. Nat. Genet. 14:465467. 4. Iakoubova, O. A., H. Dushkin, and D. R. Beier. 1995. Localization of a murine recessive polycystic kidney disease mutation and modifying loci which affect disease severity. Genomics 26:107114. 5. Fry, J. L., W. E. Koch, J. C. Jennette, E. McFarland, F. A. Fried, and J. Mandell. 1985. A genetically determined murine model of infantile polycystic kidney disease. J. Urol. 134:828833. 6. Takahashi, H., J. P. Calvet, D. Dittemore-Hoover, K. Yoshida, J. J. Grantham, and V. H. Gattone II. 1991. A hereditary model of slowly progressive polycystic kidney disease in the mouse. J. Am. Soc. Nephrol. 1:980989. 7. Moyer, J. H., M. J. Lee-Tischler, H.-Y. Kwon, J. J. Schrick, E. D. Avner, W. E. Sweeney, V. L. Godfrey, N. L. A. Cacheiro, J. E. Wilkinson, and R. P. Woychik. 1994. Candidate gene associated with a mutation causing recessive polycystic kidney disease in mice. Science 264:1329 1333. 8. Flaherty, L., E. C. Bryda, D. Collins, U. Rudofsky, and J. C. Montgomery. 1995. New mouse model for polycystic kidney disease with both recessive and dominant gene effects. Kidney Int. 47:552558. 9. Nauta, J., Y. Ozawa, W. E. Sweeny, J. C. Rutledge, and E. D. Avner. 1993. Renal and biliary abnormalities in a new murine model of autosomal recessive polycystic kidney disease. Pediatr. Nephrol. 7:163 172. 10. Atala, A., M. R. Freeman, J. Mandell, and D. R. Beier. 1993. Juvenile cystic kidneys (jck): a new mouse mutation which causes polycystic kidneys. Kidney Int. 43:10811085. 11. Beier, D. R., C. C. Morton, A. Leder, R. Wallace, and P. Leder. 1989. Perinatal lethality (ple): a mutation caused by integration of a transgene into distal mouse chromosome 15. Genomics 4:498504. 12. Kaplan, B. S., J. Fay, V. Shah, M. J. Dillon, and T. M. Barratt. 1989. Autosomal recessive polycystic kidney disease. Pediatr. Nephrol. 3:4349. 13. Lander, E. S., and D. Botstein. 1989. Mapping mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121:185 199. 14. Dietrich, W., H. Katz, S. E. Lincoln, H.-S. Shin, J. Friedman, N. Dracopoli, and E. S. Lander. 1992. A genetic map of the mouse suitable for typing intraspecific crosses. Genetics 131:423447. 15. Manly, K. F. 1993. A Macintosh program for storage and analysis of experimental genetic mapping data. Mamm. Genome 4:303313. 16. Paterson, A., E. Lander, S. Lincoln, J. Hewitt, S. Peterson, and S. Tanksley. 1988. Resolution of quantitative traits into Mendelian factors using a complete RFLP Linkage map. Nature 335:721726. 17. Frankel, W. N., and N. J. Schork. 1996. Whos afraid of epistasis? Nat. Genet. 14:371373.

rather than multiple congenics. If the consomic strain demonstrates the phenotype predicted by the original QTL analysis, it can be easily employed to generate sublines of congenic strains that can be used for more specific localization. If the consomic fails to demonstrate the predicted phenotype, it may suggest the quantitative effects being examined occur as a result of a complex interaction, in which single loci of major effect may not be readily isolated. In this case the investigator will have been spared the logistic difficulties and expense required for the generation and maintenance of multiple congenic lines. Using these combined approaches, we hope to identify genes that affect polycystic kidney disease in a murine model. The importance of characterizing the modifying loci should not be underestimated, since these genes may provide useful targets for therapeutic intervention that could significantly ameliorate disease-related morbidity. How can the results described here be applied to the analysis of mouse models of airways disease? It should be apparent that the phenotype analyzed in these studies is not polycystic kidney disease per se, but rather a numerical value (kidney weight) that was utilized as an indirect measure of disease severity. Thus, the same analytic strategy can be applied to any phenotype that can be reported as a continuous variable (for example, airway responsiveness). However, as this report illustrates, the analysis of quantitative traits may prove unexpectedly complicated; care should be used with respect to both experimental design and the interpretation of results.
Acknowledgment : This research was supported by grants from the NIDDK (RO1DK4563902) and from the Life & Health Insurance Medical Research Fund.

References
1. DeSanctis, G. T., M. Merchant, D. R. Beier, R. D. Dredge, J. K. Grobholz, T. R. Martin, E. S. Lander, and J. M. Drazen. 1995. Quantitative locus analysis of airway hyperresponsiveness in A/J and C57BL/6J mice. Nature Genetics 11:150154. 2. Wu, J. M., J. A. Longmate, G. Adamus, P. A. Hargrave, and E. K. Wakeland. 1996. Interval mapping of quantitative trait loci controlling humoral immunity to exogenous antigensevidence that non-MHC immune response genes may also influence susceptibility to autoim-