Do epistatic modules exist in the genetic control of blood pressure in Dahl rats? A critical perspective

IN THE INITIAL DISSECTIONS of blood pressure (BP) into its genetic components using hypertensive inbred Dahl salt-sensitive rats (SS/Jr, referred to hereafter as S rats), gene-gene interactions and genetic background effects were evident. The effects of alleles at different genetic loci on a quantitative trait such as BP can be additive (plus or minus), or the combined effects can be higher or lower than the sum of the individual alleles alone (genetic interaction known as epistasis). As the techniques for the analysis of quantitative trait loci (QTL) became available, followed by the use of congenic strains, the importance of powerful genetic interactions was recognized in the Dahl S rat model, as early as 1998 and 2002 (5, 7). More recently Moreno et al. (6) showed, using a series of overlapping contiguous congenic strains involving S and Brown Norway rats, that chromosome 13 contains four highly interacting regions influencing BP. Another recent publication by Chauvet et al. (2) classifies BP QTLs into two main sets (“epistatic modules”) based on their interactions or lack thereof. If this classification were valid it certainly would add a previously unknown dimension to physiological and genomic complexity. The study by the Deng group (2) is timely because recent work by others has also emphasized the importance of epistasis. Using chromosome substitution strains in rats and mice, Shao et al. (9) showed powerful epistatic effects among individual chromosomes for many quantitative traits. In human genetic studies it was recently shown that the failure to explain a large fraction of heritability of common quantitative traits can in large part be due to not taking genetic interactions into account (10). Although the article by Chauvet et al. (2) presents an original and interesting hypothesis regarding epistatic modules it does not provide any theoretical basis for the concept. Moreover, Chauvet et al. (2) do not communicate the structure of the data very well, and consequently the strengths and weakness of the data are obscure. In these comments we propose a visual way to understand and analyze their data that leads immediately to a method to test the hypothesis that such epistatic modules actually exist. The most effective way to evaluate BP QTL is to create congenic strains on the S genetic background, which is highly permissive for expressing genetic differences in BP. In general, alleles at a BP QTL are introgressed into the S rat from another (usually normotensive) strain. To study the interaction between two QTL (call them QTL1 and QTL2) four strains are required: the S strain, a congenic strain homozygous for the normotensive-strain allele at QTL1 on the S background, a congenic strain homozygous for the normotensive-strain allele at QTL2 on the S background, and a double congenic strain homozygous for both the QTL1 and QTL2 normotensive-strain alleles on the S background. The BP data are analyzed in a 2 2 factorial analysis of variance (ANOVA) as given in detail in Ref. 7. This analysis yields a probability for interaction. If the interaction is significant, QTL1 and QTL2 are acting epistatically; if it is not significant, QTL1 and QTL2 are acting additively. The data to be analyzed are from Table 1 of Chauvet et al. (2), which consists of 22 congenic strains each containing a BP QTL where the LEW allele (or in the case of chromosome 2 the Milan normotensive strain allele) is on the S background. Genetic interactions were evaluated by 27 pair-wise tests using the 2 2 ANOVA, and the results were classified as either additive or epistatic. We need to introduce some nomenclature modified from Chauvet et al. (2). Note that many chromosomes contain more than one QTL, which are numbered on each chromosome. For example, C10Q1 designates a congenic strain containing QTL1 on chromosome 10; C10Q2 designates a congenic strain containing QTL2 on chromosome 10; C2Q2 designates a congenic strain containing QTL2 on chromosome 2; C16Q designates the only QTL on chromosome 16; and so forth. The results from Table 1 of Chauvet et al. (2) are arranged in a rectangular grid (Fig. 1) created by listing the 22 strains on the left and upper sides. The blocks in the grid represent the pair-wise comparisons of strains. Comparing a strain to itself is not informative, and the blocks along the diagonal are colored black to indicate this. The blocks above the diagonal are redundant to those below it, so only those below the diagonal are considered. Chauvet et al. (2) categorize the QTL into two “epistatic modules” EM1 and EM2. This was done by starting with two congenic strains (index strains) C10Q1 and C10Q2, the effects of which on BP were additive. C10Q1 was tested against other strains and those that were found to act epistatically with it were classified as being in EM1 (coded gray in Fig. 1). C10Q2 was tested against other strains, and those that were found to act epistatically with it were classified as being in EM2 (coded tan in Fig. 1). The hypothesis is that any two QTL within a module act epistatically (shown as purple blocks in Fig. 1), and any two QTL from different modules act additively (shown as green blocks in Fig. 1). In Fig. 1 the five QTL that could not be classified into modules EM1 and EM2 are listed below the red line at the bottom and those that were classified are listed above that line. We added one data point from the literature (4) shown Address for reprint requests and other correspondence: B. Joe, Prog. in Physiological Genomics, Center for Hypertension and Personalized Medicine, Dept. of Physiology and Pharmacology, Univ. of Toledo College of Medicine and Life Sciences, 3000 Arlington Ave., Toledo, OH 43614-2598 (e-mail: [email protected]). Physiol Genomics 45: 1193–1195, 2013; doi:10.1152/physiolgenomics.00159.2013. Editorial