RIX mapping: A demonstration using CXB RIX hybrids to map QTLs modulating eye size in mice
Robert W. Williams, Lu Lu, David C. Airey, Anand Kulkarni, Kenneth F. Manly#, Guomin Zhou*, David W. Threadgill**
University of Tennessee, Center for Neuroscience, 855 Monroe Ave, Memphis TN 38163 USA, #Molecular and Cellular Biology, Roswell Park Cancer Institute, Buffalo, NY 14263, *Department of Histology and Embryology, Shanghai Medical University, Shanghai, China, and **Department of Genetics, University of North Carolina, Chapel Hill, NC 27599 USA.
Email questions and comments to email@example.com
Recombinant inbred intercross (RIX) mapping is a powerful new gene mapping method that greatly extends the utility and power of existing sets of recombinant inbred strains. An RIX is a large diallele cross made by systematically crossing one or more sets of RI strains. Here we demonstrate the potential of RIX mapping using the set of 13 CXB lines established by Don Bailey and colleagues ~30 years at the Jackson Laboratory. The CXB RI strains can be intercrossed to generate as many as 156 unique F1 genotypes. Our mapping panel consists of C57BL/6ByJ, BALB/cByJ, all 13 CXB parental strains (~36 cases/strain), and most of the 78 nonreciprocal F1 RIX hybrids (~12 cases/F1). The set of RIX strains has a genotype distribution pattern that resembles an F2 intercross with close to a 1:2:1 ratio of CC, CB, and BB genotypes at most loci. Unlike an RI set, RIXs can therefore be used to estimate dominance deviation. Unlike an F2, large numbers of animals with predetermined genotypes can be phenotyped to gain highly reliable trait statistics. The increase in numbers of genotypes and phenotypes greatly boosts the power to detect QTLs and does so without the expense of genotyping. There is also an appreciable gain in positional precision. Like RI sets, and unlike standard intercrosses, RIX data are cumulative, allowing correlative physiological, anatomical, pharmacological, and behavioral studies over a long period of time. A total of more than 10,000 RIX lines with precisely known genotypes and recombination breakpoints can now be easily generated using over 100 well characterized RI lines (Williams et al., 2001; AXB/BXA, BXD, BXH, and CXB). RIX lines can also be generated selectively with breakpoints in candidate intervals to test and refine positions of existing or putative QTLs.
To study the genetics of myopia and eye growth, we used the CXB RIX set to map QTLs that modulate eye size. Eyes were taken from anesthetized adult mice (50 to 60 days) and weighed immediately. Eyes of the CXB strains weigh 19.8 ± 2.0 (SD) mg; those of RIX progeny weigh 20.6 ± 0.8 mg (SD). Heterosis is therefore mild. We corrected for differences in age, sex, body weight, and brain weight by multiple linear regression. The adjusted data—the weight residuals—range from –1.2 to +1.2 mg. A dense map of the CXB strains (900 MIT markers) was produced and imported into Map Manager QT. Using the 13 parental strains we could only detect weak linkage on Chr 8 and Chr 12, with LODs under 3. Pursuit of these putative QTLs would normally have entailed the arduous construction of reciprocal congenics or the production and genotyping of a large intercross. In marked contrast, the analysis of RIX lines was extremely rapid and involved no additional genotyping. Strong linkage (LOD > 4) was detected on chromosomes 3, 6, 12, and 13. The LOD on Chr 13 peaks at 6.3 near D13Mit20. The approximate confidence band extends from 20 to 40 cM. A new eye weight QTL in this interval accounts for as much as 40% of the genetic variance among RIX F1s, and alleles have a linear effect on eye weight (residuals for the three genotypes: BB = –0.77, CB = –0.14, and CC = 0.33 mg).
An important difference between mapping trait variance with an F2 and an RIX panel is that non-syntenic linkage can produce spurious linkage in the RIX. For example, among the CXB inbred strains, genotypes at D12Mit147 and D13Mit20 are almost perfectly concordant despite their complete physical separation on Chr 12 and Chr 13. Naturally these two markers are also "linked" in the RIX panel. Only the eye weight locus on Chr 13 remains compelling after examining a matrix of correlations of non-syntenic markers and comparing these to group means. To protect against this problem of non-syntenic linkage, we have generated full-genome correlation matrices for four of the major RI sets (AXB/BXA, BXD, BXH, and CXB), in each case by using genotypes of ~900 MIT markers. These matrices can be inspected to reveal and protect against non-syntenic linkage in both RI and RIX sets. In conclusion, this study is a practical verification of the improved statistical power of RIX mapping and highlights a complication introduced by the fixation of particular combinations of genotypes (non-syntenic linkage disequilibrium) in parental RI sets.
Supported by NINDS R01 NS35485, R01EY13070, EY12991, and an NIH/NSF-sponsored Neuroinformatics grant P20 MH62009.
Since 3 August 2000
Neurogenetics at University of Tennessee Health Science Center