Recombinant inbred (RI) strains are formed from an outcross between two well-characterized inbred stains followed by at least 20 generations of inbreeding. RI strains can be utilized for the analysis of many complex phenotypic traits. The LEXF/FXLE RI strain set consists of 34 RI strains derived by reciprocal crossing of LE/Stm and F344/Stm. Here we report on genetic dissections of complex traits using this RI set and their parental strains. We have developed strain distribution patterns for 232 informative simple sequence length polymorphism markers. The framework map covers the rat genome except for chromosome Y. Seventy-six phenotype parameters, which included physiological and behavioral traits, were examined for these RI lines. Quantitative trait locus (QTL) analysis of these parameters revealed 27 significant and 91 suggestive QTLs, illustrating the potential of this RI resource for the detection of underlying gene functions for various phenotypes. Although this RI set was originally developed to study susceptibility to chemical-induced tumors, it has been shown to be equally powerful for a wide spectrum of traits. The LEXF/FXLE RI strains have been deposited at the National Bio Resource Project for the Rat in Japan and are maintained under specific pathogen-free conditions. They are available at http://www.anim.med.kyoto-u.ac.jp/nbr.
- Rattus norvegicus
- recombinant inbred rats
- quantitative trait locus mapping
- physiological traits
the discovery of gene functions related to human diseases is still a major issue in biomedical research. A large number of single genes have already been identified as underlying modifications associated with various monogenic disorders. Moreover, numerous articles on the dysfunctions of single gene defects exist, but the consequences of allelic variations on the complex physiological network as well as the various players of this network remain largely unknown. The NCBI database Online Mendelian Inheritance in Man contains 17,744 entries. of which only 386 are on genes with known sequences and phenotypes (14). Since all complex phenotypes result from interactions between numerous genes, quantitative trait locus (QTL) analysis in rodent models is an important method for unraveling these phenotypes and extrapolating the results to human studies. The number of such QTL experiments that have already been performed is enormous. To date, 3,538 QTLs are described in the Mouse Genome Database (16) and 1,302 QTLs are listed in the Rat Genome Database (19). Even though the majority of the listed QTLs were obtained from F2 or backcross studies, one could be misled to underestimate the role of recombinant inbred (RI) lines since they have been utilized in rodents for more than 40 years (3, 4, 7, 15, 25). However, the majority of the RI lines originated from mouse strains, and only a few rat-derived RI lines are or were available (1, 8, 18, 20, 23).
Successful QTL mapping always depends on diverse phenotypes and genotypes and a statistical method for determining the odds between phenotype and genotype patterns. This diversification of phenotypes combined with numerous recombination events across the rat genome are given requirements for the largest RI rat strain set available, the LEXF/FXLE strains, which were historically generated to study genes involved in tumor genesis. Considering the QTLs that have already been described in other experiments (19) and the theoretical power of the LEXF/FXLE strains, the questions that we wanted to answer in this study cover two aspects: 1) the scientific value of this RI panel as a tool for the dissection of quantitative traits and 2) the number and nature of the detected QTLs themselves. In other words, we asked whether or not these RI strains can be utilized for the determination of QTLs for physiological and other randomly analyzed phenotypic parameters despite the LEXF/FXLE's initial research purpose being only based on their different susceptibility to chemical-induced tumors. Furthermore, if QTLs are detectable, we wanted to know how effective this set of RI strains is for identifying QTLs for randomly examined phenotypic parameters. Finally, we examined whether the QTLs that are obtained are new compared with previously known QTLs or whether they confirm independently computed results from other experiments. For instances where these questions can be definitively answered, the LEXF/FXLE panel could become a universal tool for the detection of virtually every type of physiological QTL.
MATERIALS AND METHODS
The LEXF/FXLE RI strains and their parental strains, F344/Stm and LE/Stm, were originally generated at the Saitama Cancer Center Research Institute by Shisa et al. (20). LE/Stm was derived from a closed Long-Evans colony from the Ben May Laboratory for Cancer Research of the University of Chicago, and F344/Stm originated from F344/DuCrj (Charles River Japan). The strains were inbred at the Saitama Institute for more than 50 and 23 generations, respectively. The RI lines were generated in two phases: first the LEXF strains were established, followed by the FXLE strains. Several RI lines had substrains that branched out at the 7th to 11th generation after an attempt to fix the coat color. These sublines are indicated by the letters B–D following the strain number, e.g., LEXF8D. Further details on the history of these RI strains are described elsewhere (20, 23). The following strains were used for this study, with the inbred generations indicated in parentheses: F344/Stm (F69), LE/Stm (F95), LEXF1A (F51), LEXF1C (F48), LEXF2A (F50), LEXF2B (F54), LEXF2C (F54), LEXF3 (F52), LEXF4 (F50), LEXF5 (F52), LEXF6B (F46), LEXF7A (F51), LEXF7B (F53), LEXF7C (F49), LEXF8A (F51), LEXF8D (F50), LEXF9 (F53), LEXF10A (F54), LEXF10B (F49), LEXF10C (F54), LEXF11 (F53), FXLE12 (F27), FXLE13 (F27), FXLE14 (F26), FXLE15 (F30), FXLE16 (F26), FXLE17 (F25), FXLE18 (F26), FXLE19 (F28), FXLE20 (F27), FXLE21 (F28), FXLE22 (F30), FXLE24 (F24), FXLE25 (F28), and FXLE26 (F26). Since the genotyping performed in this study revealed breeding contamination for the FXLE23 strain, only 33 of 34 RI lines were analyzed. Since the rederivation of FXLE23 from uncontaminated embryos has almost been accomplished, it will be possible for future experiments to be carried out with all 34 RI strains. The rats were maintained at the specific pathogen-free facility of the Institute for Animal Reproduction. At 5 wk of age, six male rats from each strain were shipped to the Environmental Biological Life Science Research Center for phenotype screening. All animals were maintained under a 12:12-h light-dark cycle with lights on at 7:00 AM and ambient conditions of 23 ± 3°C and 55 ± 15% humidity. They were housed in groups of three animals per aluminum cage (dimensions of 240 × 380 × 200 mm) and were given free access to acidified water and chow (CE2, CLEA). Animal care and all experimental procedures were approved by the Animal Research Committee, Graduate School of Medicine, Kyoto University (approved no. MedKyo07001).
Phenotypic profiles for this project consisted of the following 7 categories covering 109 parameters: 1) functional observational battery (FOB, neurobehavioral test), 2) behavior studies, 3) blood pressure, 4) urine parameters, 5) biochemical blood tests, 6) hematology, and 7) anatomy (see Table 2). All measurements were performed on all male rats from each strain from 5 to 10 wk of age. The detailed protocols used for measurements of these parameters are available on our website at http://www.anim.med.kyoto-u.ac.jp/nbr/phenotype and were described previously (13). QTL analysis was performed with a subset of 76 quantitative parameters, which were part of the above-mentioned phenotypic profiles.
The genetic profiles consisted of 357 simple sequence length polymorphism (SSLP) markers with known genomic locations, which are distributed throughout the rat chromosomes except for chromosome Y. Detailed marker information is available at the National Bio Resource Project (NBRP) home page at http://www.anim.med.kyoto-u.ac.jp/NBR/Genotyping.htm. Genomic DNA was extracted from the spleen. The product sizes of the SSLP markers were determined with an ABI3100 DNA sequencer (Applied Biosystems).
The phylogenetic tree of the RI strains was obtained through maximum parsimony analysis implemented in PAUP 4.0b10 (22) on the basis of 259 markers that were polymorphic between the parental strains. An initial heuristic search using Fitch parsimony was carried out with 1,000 random addition sequence replicates, followed by a tree bisection-reconnection (TBR) branch swapping algorithm. Tree stability was estimated by bootstrap analysis on 1,000 replicates where the characteristics were sampled with equal probability. TreeView (http://taxonomy.zoology.gla.ac.uk/rod/treeview.html) was used to display the resulting tree (17).
Two hundred thirty-two markers of 357 tested were informative for the RI strains and were therefore included in the genetic map for subsequent QTL scans. The basis for marker positioning and order, however, was not recombination fractions but their known location on the physical map. Genomewide scans for QTLs were performed with the 76 mean and variance values from 35 strains and the physical map of 232 genetic markers noted above. Calculations were performed with MapManager QTXb20, which is available at http://www.mapmanager.org/ (12). Interval mapping was performed by fitting a regression equation along the genetic map to a hypothetical QTL in 1-cM steps with an additive regression model. Permutation tests were performed to empirically determine the significance thresholds for all QTL mapping results. A minimum of 1,000 permutations for each QTL calculation for the constrained additive regression model were applied to establish individual suggestive, significant, and highly significant thresholds, which correspond to genomewide probabilities for the 37th, 95th, and 99.9th percentiles, respectively, as proposed by Lander and Kruglyak (11).
Two hundred fifty-nine of 357 markers that were tested were polymorphic between the parental strains LE/Stm and F344/Stm. Twenty-seven of these polymorphic markers did not show recombination with neighboring markers among any RI strains and were therefore not included in the physical map; hence, 232 markers were informative and were utilized for QTL calculations (Fig. 1). The markers comprised in total ∼2.4 Gbp on the physical map, which is ∼90% of the rat genome (6). The SSLP markers provided in total 2,821 recombinations in these 33 RI strains and showed an average spacing of ∼12 Mbp (Table 1), with the largest gap being 78 Mbp on chromosome 6.
The genetic relationship among the RI and parental strains on the basis of the utilized SSLP markers is displayed in Fig. 2, which reflects their historical breeding processes and substrains as previously described (20, 23).
QTL mapping efficiency of this RI panel.
Analysis of the phenotypic parameters revealed that 43 of 76 mean values were significantly different between the parental F344/Stm and LE/Stm strains (Table 2). In total, 118 QTLs were detected by interval mapping, of which 27 passed the significant or highly significant criteria and 91 showed suggestive linkages (Table 2). Twenty-two traits (28%) could not be associated with any QTL. Thirty-five QTLs (30%) were associated with 20 traits (26%) that did not show significant phenotypic differences between the parental strains as indicated in Table 2. One hundred two QTLs were to our knowledge new and not described in the RGD QTL database (19). Sixteen were confirmed by this database, of which 11 and 5 were suggestive and significant, respectively. Figure 3 shows six representative highly significant QTLs that were found in this study. Further details on all QTLs detected can be obtained from Table 2 or online at http://www.anim.med.kyoto-u.ac.jp/NBR/RI_SSLP_QTLs/SSLP_QTLs.htm.
The initial screening for QTLs using 232 informative SSLP markers in these 33 LEXF/FXLE RI and 2 parental strains already revealed 118 QTLs for 54 quantitative parameters, which is equivalent to a rate of ∼70% when referring to the 76 parameters examined. These pure numbers indicate that this RI panel is a powerful tool for QTL mapping and shows promise for use in further dissections of quantitative traits. It can be concluded by simple statistics that a QTL can be detected for two of three randomly examined parameters. However, a closer look shows that the strength of the obtained QTL seems to depend on the different natures of the parameters that were examined. All analyzed quantitative parameters are likely to be controlled by more than one gene, and it is thought that the strength of a QTL is higher when fewer genes contribute to it. In other words, the detection of a QTL becomes more difficult if many genes account for the phenotypic variance with a relatively similar, low size. This observation can also be seen in our data. The parameters that were examined can be roughly divided into two groups: simple physiological parameters such as organ weights, enzyme activities, or ion concentrations and more complex behavioral traits like rearing, locomotor activity, or passive avoidance tests. Many high-score likelihood ratio statistics (LRS) values were calculated for physiological parameters, but the LRS levels for all behavioral parameters were always only in the range of the empirically calculated suggestive threshold, confirming the complex characters of these traits. The difficulties in detecting weak QTLs are not only relevant for studies that utilize standard sib-mated RI strains. Valdar et al. (24) simulated a QTL analysis on a basis of 1,000 individuals for several breeding strategies, including normal F2 intercross, backcross, advanced intercross RI lines, heterogeneous stock RI lines, and various forms of collaborative cross approaches. They showed that a simple mapping computation based on a single marker regression model can guarantee the detection only of QTLs with effect sizes of 30% or greater. QTLs with smaller effects can be detected, but they may be overlooked. A more sophisticated mapping calculation such as composite interval mapping may lower this threshold to 10% of the trait variance, provided that 1,000 individuals are utilized in various breeding strategies (24).
To date, QTL studies mostly utilize F2 or backcross animals to map loci related to a specific phenotype for which the parental strains show highly significant differences and therefore highly segregating QTLs. Such crosses are time- and resource intensive but have the advantage that they can be used to produce maps down to the resolution of single genes. This is especially successful in the case of virtual monogenic QTLs (2, 5). In contrast, the benefit of RI strains is the fast experimental approach since the use of RI lines avoids long-term crossing periods as well as genotyping and provides ad hoc a sufficient number of recombination events. This makes it possible to reduce the experimental effort for QTL mapping using RI strains to only the phenotyping. However, this advantage is also a limiting factor in terms of the analytical power of RI strains. Single genes have to our knowledge not yet been mapped in QTL experiments using RI panels. More than 500 sophisticated bred RI lines would be required to detect weak QTLs that account for 5% of the phenotypic variation to within <1 cM (24). This is far more accurate than the resolution of the QTLs obtained in this study. Their confidence intervals are in most cases larger than 20 Mbp, which corresponds to several hundred candidate genes. Logical subtraction can be used to exclude most of them, but too many putative candidate genes remain to allow successful causative gene detection. Nonetheless there remains the potential to increase the accuracy of this RI resource by increasing the density of the markers. This study describes the results of QTL mapping using only 232 SSLP markers, which leaves several huge gaps of >50 Mbp in the rat genome. Currently, the STAR consortium (21) is determining the sequence for up to 100,000 single nucleotide polymorphism (SNP) loci for many rat strains, including those in this resource, and it is expected to generate a SNP map for these RI strains that will consist of ∼30,000 SNPs (Hübner N, personal communication). Not all of these will be informative because of the limited number of recombinations among these 34 RI lines, but it can be assumed that these SNPs will greatly increase the accuracy of QTL mapping using this RI panel. Another way to increase the number and probably also the accuracy for the QTLs for this RI panel is the application of different and more sophisticated calculation methods. Standard interval mapping as used here takes into account only single markers, whereas in contrast composite interval mapping also takes the effect of other loci into account. Such calculations have not been performed yet because the primary goal of this study was the general evaluation of this RI panel for QTL mapping and because the upcoming SNP map will allow for a more detailed dissection of these complex traits. This is also the reason why we are not dissecting every QTL that we obtain and are publishing them without further discussion regarding candidate genes or cross-species comparison. Their value might seem limited because of the relatively rough genomewide 232-marker map, but their correctness—not accuracy—should not be underestimated. The result that in total 102 of 118 QTLs are not contained in the RGD QTL database (19) is due to the majority of these parameters never having been examined in QTL research in the rat before. On the other hand, QTLs for common parameters such as cholesterol, glucose concentration, or heart weight were confirmed by our results, which was also the reason why we decided to publish not only significant but also suggestive QTLs. They confirm the results of other independent experiments and prove the investigative power of the LEXF/FXLE RI strains.
Another interesting finding of this work is the subset of the 20 detected QTLs for which the parameters of the parental strains F344/Stm and LE/Stm are not significantly different. Standard trait dissection in RI strains starts with phenotypic examination of the trait in the parental strains. If the parental strains show distinct values for the parameter it can be assumed that the corresponding genes will segregate among the RI progeny along with the QTLs, which can then easily be detected. This raises the question of how it is possible to find QTLs if the parental strains do not show significant differences for a particular trait. The answer to this lies in the complex regulation of the 20 traits, which show a wide range of phenotypic values among the RI strains and can therefore be dissected by standard statistical methods. The QTLs for these traits impressively show the real interactions between the genes, which regulate the quantitative values of these parameters in the mixed allelic environment of RI strains.
As initially stated, QTL mapping is based on a statistical method that is used to determine the odds between diverse phenotypes and genotypes. If a specific allelic variation is associated with several up- or downregulated measured values, the same genomic location of the allelic variants will always appear as QTL for these regulated values. Hence, there is a bias in the detection of regulatory elements that are responsible for several related parameters as also shown in our data for the QTLs for lipid metabolism parameters such as cholesterol and high-density lipoproteins (Fig. 3). This behavior lies in the nature of the statistical mapping approach and can also be seen in the more recent expression QTL (eQTL) mapping, a variant of QTL mapping in which tissue-specific gene expression data are mapped onto a usually dense genetic map (9). Physiological QTLs combined with eQTLs—not utilized for this RI resource yet—would dramatically increase the power of this resource to the level of candidate gene detection.
Finally, it should be mentioned that these RI strains and all data on them are freely available at http://www.anim.med.kyoto-u.ac.jp/nbr. They have already been used by and can be distributed to interested researchers worldwide. Additional results obtained from this unique and largest available RI rat strain set will be forthcoming in the future. QTLs from this resource will be deposited into proficient QTL databases like the RGD database (19) and will not only improve our knowledge on rat physiology but also support progress in biomedical research across a range of species through comparative research approaches.
This work was also supported in part by Grants-in-aid for Scientific Research from the Japan Society for the Promotion of Science (18300141 to T. Kuramoto and 16200029 to T. Serikawa) and a Grant-in-aid for Cancer Research from the Ministry of Health, Labour and Welfare.
This work was supported by the National Bio Resource Project for the Rat in Japan, which is part of the “Research Revolution 2002” (RR2002) initiative of the Japanese Ministry of Education, Culture, Sports, Science and Technology.
Address for reprint requests and other correspondence: T. Serikawa, Inst. of Laboratory Animals, Graduate School of Medicine, Kyoto Univ., Yoshidakonoe-cho, Sakyo-ku, Kyoto 606-8501, Japan (e-mail:).
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
- Copyright © 2008 the American Physiological Society