The evolution and genetic architecture of shape is still poorly understood, but quantitative genetic studies based on morphometric approaches promise to lead a significant step forward . The mouse mandible is a well-established model for studying the quantitative trait genetics of shape [2–11] as well as the analysis of general principles of wild type variation and evolution [12–14]. Given this quantitative genetics focus, the mouse mandible may also be a good model to address general questions of the nature and interactions of quantitative trait loci (QTL). This includes in particular the number and effect sizes of loci, the conservation of QTL across species and the prevalence of epistatic interactions.
In the classic infinitesimal model of quantitative genetic variation with mostly additive effects, one could expect relatively fast evolutionary divergence of the genetic architecture underlying phenotypic variation, since changes in one locus could easily be compensated by changes in another one. Although this is an issue of fundamental interest, only few studies have specifically addressed the question of evolutionary conservation so far .
There is some evidence pointing towards the existence of smaller sets of conserved "trait loci" underlying variation in some traits. For example, a conservation of major effect loci is generally implied in comparisons of medically relevant traits between mice and man. Furthermore, theoretical and empirical studies of metabolic flux control [16–19] suggest that variation at certain enzymes in metabolic pathways will produce most variation in flux, i.e. the propensity of some genes to produce phenotypic variation may be especially high. The same notion is supported by the empirical finding that QTL for univariate traits at certain genomic positions are recurrently identified across studies using different crosses of the same species [19–22], or sometimes even across species [23, 24]. For example, a study on the comparison of atherosclerosis QTLs in humans and mice showed that 17 out of 27 loci mapped in humans correspond to homologous mouse QTLs , although there remains some uncertainty about the validity of this comparison . But little is known so far about the conservation of QTLs for multivariate traits, such as shape. Because QTL studies of mandible shape have so far focused exclusively on mapping populations originating from a LG/J versus SM/J intercross, analysis of genetic variation between another set of mouse strains offers the opportunity to assess the conservation of the genetic basis of shape variation.
A topic closely related to this issue is the background-dependency of QTL effects, or epistasis. If the same loci or genomic regions recurrently exhibit large effect sizes on certain traits, this would mean that the relative effect sizes of QTLs on specific characters are usually background-independent. However, QTL-background interactions could still influence the specific phenotypic manifestation of genetic differences at QTLs, namely with regard to absolute effect sizes and the distribution of pleiotropic effects (multivariate direction of effects). This is likely to be relevant for complex, inherently multidimensional phenotypes such as shape, where the developmental processes affected by QTL variation are likely to affect multiple, albeit closely related "traits" (e.g., landmarks). Epistatic interactions for skull and/or mandible shape have been studied so far using both intercross QTL mapping designs  and a recombinant congenic strain panel . Both studies found significant epistasis. However, this topic deserves further attention. The power to detect physiological epistasis using intercross designs is limited, because the complex patterns of marker and QTL segregation in an intercross mapping population reduce the power to detect epistasis between pairs of loci, since individual genotypes are not replicated and since multiple epistatic interactions in recombinant individuals may cancel each other out. Therefore, the amount of physiological epistasis is likely to be underestimated in such experiments. Congenic strains, which replicate individual genotypes and isolate small fractions of the genome on a uniform genetic background, are potentially a more powerful resource to assess the degree of epistasis. Similarly as congenic strains, chromsosome substitution strain (CSS) panels offer a powerful opportunity to detect physiological epistasis . CSS panels consist of a number of strains derived from a receptor and a donor strain, where in each strain one chromosome of the receptor strain has been replaced by the corresponding chromosome of the donor strain.
According to the differences between a classical QTL experiment and a CSS panel, one could expect that it will be difficult to detect similarities in the genetic basis of quantitative traits in comparisons between such experiments, because the CSS effects are expected to represent unique genetic combinations including the possibility of strong epistatic interactions, whereas a QTL experiment measures additive effects of loci as average effects over many different combinations of genetic background.
We use here a morphometric analysis of mandible shapes in a CSS panel that was derived by introgressing single chromosomes from PWD, representing M. m. musculus, into the genomic background of C57BL/6J, representing mostly M. m. domesticus . We were particularly interested to ask whether relative effect sizes and effect directions of genetic variation are comparable to those that were determined by QTL studies conducted with strains coming from one of the subspecies only (M. m. domesticus). Hence, we compare our results to the QTL mapping study by Leamy et al.  using an almost identical set of landmarks.
We find that there is indeed a correlation between effect sizes between the QTL and the CSS results, although not for all loci. Effect directions were, however, not correlated, suggesting major allelic differences. When we combine the effects of all CSS strains, this leads to an extreme hypothetical phenotype outside the range of known natural variation. This confirms the notion of major epistasis effects uncovered through the analysis of consomic strains . Our results thus provide insights into the genomic architecture and evolution of multivariate quantitative traits.