QTL Mapping to QTL Cloning: Mice to the Rescue
With the availability of dense, highly informative marker maps, it has recently become feasible to map genes (Quantitative Trait Loci or QTL) accounting for part of the heritability of continuously distributed traits in experimental crosses as well as outbred populations. QTL mapping efforts have almost invariably revealed a limited number of loci with effects of a magnitude clearly departing from the predictions of the infinitesimal model (a model introduced to facilitate mathematical treatment of quantitative traits rather than to truly reflect their underlying biology). As most experimental designs would have limited detection power, which could lead to an overestimation of the identified gene effects, interpretation of results from QTL mapping studies must be viewed with caution. However, numerous independent confirmation studies leave little doubt that most quantitative traits indeed involve a limited suite of loci with major effect. This assertion seems to hold not only for QTL mapped in crosses between divergent lines, but—more importantly—for QTL segregating in outbred populations as well (for review, see Paterson 1995).
Despite the sometimes unexpected magnitude of the identified QTL effects, the lack of simple correspondence between genotype and phenotype in complex trait analysis precludes the unambiguous identification of recombinant individuals. This may limit the achievable mapping resolution of QTL, posing a serious threat to the efficacy of positional (candidate) cloning for QTL considerably. QTL mapping efforts, whether performed in pedigrees or by exploiting linkage disequilibrium, are likely to leave geneticists with a portion of the genome that contains tens if not hundreds of genes and many DNA sequence polymorphisms to examine to identify the causal variant. Mutations causing monogenic inherited diseases are often destructive enough to leave little doubt about their causality. Even if the functional consequences of such mutations were less transparent, demonstration of a perfect correspondence between genotype and phenotype strongly implicates the corresponding gene, if not mutation. In the case of quantitative inheritance, the ambiguous genotype–phenotype relationship, as well as the possibly more subtle nature of the causal mutations, may complicate considerably the distinction between neutral and causative polymorphisms in the candidate region.
Imaginative geneticists will certainly find a number of strategies to untangle this Gordian knot (e.g., Risch and Merikangas 1996). Increasingly, however, rodent models appear potentially to be one of the most valuable assets in QTL quests. This is illustrated vividly in an article by Hu and colleagues, dealing with comparative QTL mapping for Salmonella resistance in species as distantly related as mice and chicken (Hu et al., this issue)
It is well established that genes contribute to individual differences in resistance to a variety of viral, bacterial, and parasitic pathogens (for review, see Malo and Skamene 1994; Hill 1996). A striking example of a monogenic disease resistance in man is the virtual immunity to HIV infection of individuals homozygous for a deletion in the CCR5 receptor (Liu et al. 1996; Samson et al. 1996). Genetic control of host resistance to infection is often more complex, however, and thought to involve multiple genes.
This is, for instance, the case when studying resistance toSalmonella in poultry. In backcrosses obtained from susceptible and resistant chicken lines, survival time after inoculation behaves very much as a quantitative trait, suggesting the contribution of several QTL (Hu et al., this issue). The importance ofSalmonella contamination of poultry products as a cause of food-borne disease in humans amply justifies efforts to identify the underlying QTL, which could lead to more efficient control strategies.
Genetic resistance to Salmonella infection is also well documented in mice and it is known to be a multifactorial entity as well. However, judicious strain choice and refinement of the phenotype has allowed the dissection of murine resistance/susceptibility toSalmonella in a series of distinct entities, each segregating as simple Mendelian traits in specific matings (for review, see Malo and Skamene 1994). Two of these in particular have been the subject of considerable attention: Lps and Bcg.
Measuring LPS (lipopolysaccharide, a major component of the outer membrane of Gram-negative bacteria)-induced spleen cell proliferation or liver CFU counts after inoculation withSalmonella in (resistant × C3H/HeJ)F1 × C3H/HeJ backcrosses, reveals two nonoverlapping populations. This was interpreted as evidence for the segregation of a locus, Lps, with major effect on responsiveness to LPS and, concomitantly, regulation of preimmune susceptibility to infection with Gram-negative bacteria, including Salmonella. Using >1000 backcross individuals, the proposed Lps locus has been mapped recently to sub-centimorgan resolution on mouse chromosome 4 (Qureshi et al. 1996). Although the actual Lps gene has not been identified yet, it seems reasonable to anticipate its succesful positional cloning in the not-too-distant future.
In a similar approach, using spleen CFU counts after infection withMycobacterium bovis (BCG) as a discrete phenotype in 1000 (C57L/J × C57BL/6J)F1 × C57BL/6J backcross progeny, the Bcg locus was mapped by linkage analysis to a 0.3-cM interval on mouse chromosome 1 (Malo et al. 1993). The Bcglocus (also known as Ity or Lsh) modulates the capacity of macrophages to restrict the growth of ingested pathogens as diverse as Mycobacteria, Salmonella, and Leishmania,and, therefore, the proliferation of these agents in the reticuloendothelial organs during the preimmune phase. In a perfect illustration of the positional cloning strategy, Vidal et al. (1993)isolated the Nramp1 gene as a strong candidate from the region based on its macrophage-restricted expression pattern and a nonconservative Gly–Asp substitution in susceptible mice, which lead to complete absence of the protein in their macrophages. The precise mode of action of Nramp1, an integral membrane phosphoglycoprotein resembling known prokaryotic and eukaryotic transport proteins, remains unknown. Subsequent demonstration thatNramp1 disruption using knockout technology abrogates natural resistance to infection with intracellular parasites provided the formal proof of the causality of the corresponding gene (Vidal et al. 1995).
Pursuing the prediction of Malo and Skamene (1994) that loci contributing to multifactorial resistance to infectious agents identified in mice would be prime candidates for QTL affecting resistance/susceptibility in other species, Hu et al. (this issue)tested the contribution of the Bcg and Lps orthologs with Salmonella resistance variation observed in chicken lines. The entire coding region of the Nramp1 gene was sequenced from three resistant (including W1) and three susceptible chicken lines. Apart from 10 apparently neutral nucleotide substitutions, Hu et al. found a nonconservative Arg–Gln substitution in a highly conserved portion of the fifth transmembrane domain, confined to one of the susceptible lines: C. They found a clear association (P = 0.0004) between the corresponding mutation and survival rate at day 7 post infection in 425 (W1 × C)F1 × C backcross progeny. Nramp1CC homozygous chicken had a mortality rate twice as high as their CW1 contemporaries (27% vs. 13%). Altogether, these data suggest very strongly a direct role of Nramp1 in susceptibility toSalmonella infection in chicken.
To test the effect on Salmonella resistance of theLps ortholog in chicken in the absence of the actual gene,Malo and Skamene (1994) generated a restriction fragment-length polymorphism (RFLP) marker using a human tenascin C (TNC) cDNA probe. In mice, the TNC gene is located at 0.7 cM from theLps locus. The TNC probe revealed alternate alleles in the resistant W1 and susceptible C lines, which allowed them to trace the segregation of the corresponding chromosome segment—shown to be independent of that of the NRAMP1 locus—in the same 452 backcross progeny. Again, there was a significant (P = 0.005) association with survival rate: The mortality of CC homozygotes was nearly twice that of CW1 individuals at day 7 post infection. Although the conservation of synteny between mice and chicken around the Lps locus still needs verification, these data are indicative of a causal role of the Lps locus in these chicken lines.
With this work, Hu et al. (this issue) demonstrate how judicious use of comparative mapping data can, at least in specific instances, accelerate identification of QTL even in species where the available genomic tools are modest compared to those in humans and mice (Georges and Anderson 1996). The generic value of this approach, however, remains to be demonstrated and will depend on the degree of overlap between the suites of QTL underlying the genetic variation for a given trait in different species. This degree of QTL convergence will reflect how far different sets of genes may lead to the same phenotype. Comparative QTL mapping is therefore a potentially powerful approach to address this fundamental issue in developmental biology.
The most extensive comparative data set available at this point probably comes from QTL mapping efforts in plants. Interestingly, an unexpectedly high proportion of QTL affecting seed size, height, flowering, and other complex traits do correspond among different taxa (for review, see Paterson 1995).
Significant interspecies overlap between QTL suites has the potential to play a crucial role in efforts aimed at cloning QTL, particularly in domestic animal species. Indeed, whereas QTL cloning still faces major conceptual hurdles in most outbred species for the reasons mentioned above, mouse as a model organism offers the necessary tools to make QTL cloning an achievable objective in many circumstances: Large numbers of offspring can be generated from selected matings at moderate cost; animals can be reared in well-controlled environmental conditions; tools for genomic analysis are particularly well developed and will continue to improve at a much faster rate than in any other mammal except man; and—last but not least—a panoply of powerful transgenic and gene targeting methods is available to perform the ultimate functional tests.
QTL mapping could be performed in parallel in mice and livestock for traits of interest. Numerous selection lines are already available in mice for traits that are of relevance to agriculture. Not only would this allow the rigorous evaluation of the degree of interspecies QTL overlap, but positional cloning could readily be initiated in mice for QTL yielding evidence for interspecies conservation based on primary QTL mapping data. Once a QTL affecting a trait of interest is cloned in mice, extensive mutation screening could be performed for the ortholog in the domestic species of interest and its contribution to genetic variation examined using adequate association tests, including the transmission/disequilibrium test (Spielman et al. 1993). One could easily justify pursuing QTL mapped in mice alone, not only because of the fundamental interest in identifying a gene underlying a complex trait of interest, but because association studies based on the knowledge of the underlying gene have the potential to be more powerful for identifying segregating QTL than conventional linkage mapping (Risch and Merikangas 1996).
The most difficult part of such an exercise may be to convince agencies funding agricultural research of the value of a detour into mouse genomics.
