Human rates of obesity in the developed world have now reached epidemic levels, and the global number of overweight individuals outstrips the number of underweight cases (WHO 2003). This reflects the effects of a high-fat food culture, combined with a genome evolved to efficiently conserve, partition, and store energy. The human gene map has so far catalogued more than 300 genes, markers, and chromosomal regions that are associated with human obesity (Chagnon et al. 2003). The majority of these are quantitative trait loci (QTL) for obesity or body weight (Bw), mapped primarily in the mouse, as reviewed by Brockmann and Bevova (2002), but also including pigs (Rankinen et al. 2002) and chickens (Van Kaam et al. 1999; Ikeobi et al. 2002; Sewalem et al. 2002). However, very few of these QTL have been mapped beyond the original scan, and given that the intervals of many of these QTL regions are relatively large, the possibility that there may be multiple QTL within those regions has been rarely assessed. It is also unclear how many of the Bw QTL reported are related to obesity and adiposity, since many Bw studies have not involved the further dissection of the ‘composite trait’ Bw into its components.

A polygenic obesity mouse model would greatly facilitate the studies of the genetics of any complex trait such as obesity. Such a model has been developed in Edinburgh by divergent selective breeding that has resulted in strains differing substantially in fat content (Bünger et al. 2003). The mice were selected for high fat (Fat, F-lines) or low fat (Lean,L-lines) content from a genetically highly variable base population for more than 60 generations (Sharp et al. 1984). Fat content was determined as the ratio of dry matter/body weight after freeze-drying the carcass; the correlation with true fat percentage was 0.98 (Hastings and Hill 1989). From these populations, two inbred lines were produced by eight generations of full-sib matings (Bünger and Hill 1999); the resulting lines differ more than fivefold in fat content and have 22% (F-line) and 4% (L-line) of body fat (Bünger and Hill 1999). Recent research (Bünger et al. 2003) has indicated that polymorphisms in the genes Lep (Chr 4) and Lepr (Chr 6) have not contributed to the selection response.

A genome-wide QTL analysis using the outbred F and L lines identified four QTL regions for fat content that were significant at a genome-wide level (Horvat et al. 2000): Fob1 (F-line obesity QTL1), Fob2, Fob3, and Fob4 on Chrs 2, 12, 15, and X, respectively. Fob1, Fob2, Fob3, and Fob4 explained 4.9%, 19.5%, 14.4%, and 7.3% of the F2 phenotypic variance for fat%, respectively. In the present study, we chose the Chr 15 QTL region (Fob3) for further detailed analysis because of its large effect on fat content (it explained 14.4% of the total phenotypic variation) and its strong statistical support (LOD score of 11.3; Horvat et al. 2000). Here we demonstrate successful introgression of the Fob3 QTL region from the L-line into the F-line. Moreover, through a multiple-QTL analysis, we identify two regions within the original Fob3 interval, which have significant independent effects, operating in the same direction. In addition, we further characterize the effects of these QTL by analysis of Bw through development, concluding that the two QTL affect the phenotype at different developmental stages.

Materials and methods

Production of congenic lines and progeny test population

Fat and Lean mouse lines (F-line andL-lines respectively) were selected from a base population, which is based on a cross of two inbred lines (JU and CBA) and one outbred line (CFLP) (Sharp et al. 1984). For a recent review on the F and L-lines, see Bünger and Hill (1999) and Bünger et al. (2003). To develop congenic lines for Fob3 QTL region, the L-line was crossed to the F-line; then mice heterozygous for the Fob3 QTL region were selected in each generation and backcrossed five times to the F-line. This produced three congenic sublines termed Fchrl5LA, Fchrl5LB, and Fchrl5LD, collectively termed Fchrl5L (F -line with an L -line segment on chromosome 15). F chrl5L A congenic subline, which was later used in a cross to generate recombinants (Fig. 1), has L-line alleles between markers D15Mit150 and D15Mit107; Fchrl5L B has L-line alleles between D15Mit184 and D15Mit39; and Fchrl5LD has L line alleles between D15Mit184 and D15Mit107 (see Fig. 2 for the map). Fchrl5LD-line mice were used as the preferred congenic line for all further crosses, while Fchrl5L A and B-line mice were used where Fchrl5LD-line mice were in short supply.

Fig. 1
figure 1

Mating scheme for generating the progeny from the F2 mice recombinant within the Fob3 QTL region. Thin and thick lines indicate chromosomal segments homozygous for the Lean-line and Fat-line alleles, respectively.

Full size image
Fig. 2
figure 2

F-value plots for the single-QTL interval mapping analysis of Bw21, Bw42, Bw70, Bw98, Bw115, fat% and rGFPwt traits. Vertical dashed line indicates significance at the F0.01 level. Thick vertical lines with intersections indicate the 95% CI and positions of the peaks for each of the QTL based on the fat% trait multiple QTL analysis.

Full size image

At the fifth backcross, L-line alleles at three other QTL found between the F and L-lines on Chr 2, 12, and X [Fobl, 2, and 4 respectively (Horvat et al. 2000)] were selected against. This was done in order to remove any effects associated with these QTL that might have been retained by chance. After the sixth backcross, the estimated F-line background of the congenic line was over 97%.

To fine map the Fob3 QTL region, recombinants within the Fob3 region were generated. This was achieved by crossing a single homozygote congenic Fchrl5L male with nine F-line (inbred) females (Fig. 1). The resulting 42 heterozygous progeny were inter-se mated in order to produce 41 recombinants from 220 progeny. Of these, 23 were male recombinants and were backcrossed to the F-line and congenic Fchrl5L-line females, to generate 789 progeny from 133 families. The 18 female recombinants were backcrossed first to the F-line males to generate 208 progeny from 35 families. Then selected progeny of the female recombinants possessing the desired recombinant genotype were backcrossed to both F-line and congenic FchrI5L-line mice, generating an additional 182 progeny from 29 families. A general mating scheme is illustrated in Fig. 1.

Microsatellite genotyping

In total, 60 microsatellite markers (Research Genetics) were screened with purified (DNAzol, Life Technologies) spleen DNA from a selected set of F and L-line parents. In total, 16 microsatellite mouse markers, polymorphic between the F and L-line alleles on Chr 15, were identified and used in the analysis (Fig. 2).

For general screening of mice, genomic DNA was extracted from mouse ear clips by a modification of the protocol described by Laird et al. (1991). Following incubation of ear clips in 60 μL of lysis buffer, the lysate was diluted 1:40 in distilled water and the resulting solution used as template for PCR. PCR reaction conditions were optimized for each marker’s primer set pair in 12 μL final reaction volumes [6 μL diluted lysate (1:40) DNA template, 6 μL 2× PCR premix], with 96-well polypropylene PCR plates (Abgene) on Hybaid JMBS thermocyclers. PCR products were separated on 4% agarose gels (2% MetaPhor agarose, FMC BioProducts, and 2% Agarose MP, Boehringer), with 0.5× TBE buffer containing ethidium bromide (0.5 μg/mL) and loaded with 4 μL of Cresol red loading buffer.

Production of linkage map

Genotypic data for 16 microsatellite markers on Chr 15, spanning four generations, were uploaded into the ResSpecies database [Roslin Institute; http://www.resspecies.org (Law & Archibald 2000)], which checked for pedigree errors based on the genotype of the parents in the previous generation. This data set was analyzed with CriMap (Lander and Green, 1987) to produce a marker map on the basis of observed recombination frequencies. The orders of the markers (Fig. 2) were confirmed by the base pair positions as mapped by the Mouse Genome Sequencing Consortium (MGSC) v 13.30.1 (http://www.ensembl.org/Mus_ musculus/).

Phenotyping

Progeny of recombinant males and grandprogeny of recombinant females were weighed at 21, 42, 70, 98, and 115 days (traits = Bw21, Bw42, Bw70, Bw98, and Bw115 respectively), at which point the mice were euthanized. Male gonadal fat pads (GFP) were removed, weighed, and then replaced in the abdominal cavity. Relative GFP weight (rGFPwt) was expressed as the ratio of the GFP to Bw115. All mice were freeze-dried, and a dry carcass (DM) weight was taken. Percentage body fat (fat%) was calculated as a ratio of Bw115 and DM fitted to a regression equation (below), as described by Hastings and Hill (1989):

EquationSource% MathType!Translator!2!1!AMS LaTeX.tdl!TeX -- AMS-LaTeX! % MathType!MTEF!2!1!+- % feaafiart1ev1aaatCvAUfKttLearuavP1wzZbItLDhis9wBH5garm % Wu51MyVXgaruWqVvNCPvMCaebbnrfifHhDYfgasaacH8srps0lbbf9 % q8WrFfeuY-Hhbbf9v8qqaqFr0xc9pk0xbba9q8WqFfea0-yr0RYxir % -Jbba9q8aq0-yq-He9q8qqQ8frFve9Fve9Ff0dmeaabaqaciGacaGa % aeqabaWaaeaaeaaakeaacqqGqbaucqqGYbGCcqqGLbqzcqqGKbazcq % qGPbqAcqqGJbWycqqG0baDcqqGLbqzcqqGKbazcqqGGaaicqqGMbGz % cqqGHbqycqqG0baDcqqGLaqjcaaMc8Uaeeypa0JaaGPaVlabbIcaOi % abbseaejabb2eanjabb+caViabbkeacjabbEha3jabbgdaXiabbgda % Xiabbwda1iabbMcaPiaaykW7cqGHxdaTcaaMc8UaeGymaeJaeGymae % JaeG4mamJaeyOeI0IaeG4mamJaeGimaaJaeiOla4IaeGOmaidaaa!5BB7!
(1)

[1] Fat% and Bw are closely correlated in our phenotyped population. The correlation is significant from Bw21 (r = 0.32; P < 0.001) with an increasing phenotypic correlation over time up to Bw115 (r = 0.78; P < 0.001).

Single marker analysis

As a first step in the analysis, QTL detection was performed by single marker analysis of variance (ANOVA). This was carried out on 989 phenotyped individuals, using the GLM procedure of the SAS System for Windows Release 6.08 (SAS Institute Inc., Cary, N.C., USA). Each marker was fitted as a fixed effect. For each trait, all 16 markers were analyzed with the model,

EquationSource% MathType!Translator!2!1!AMS LaTeX.tdl!TeX -- AMS-LaTeX! % MathType!MTEF!2!1!+- % feaafiart1ev1aaatCvAUfKttLearuavP1wzZbItLDhis9wBH5garm % Wu51MyVXgaruWqVvNCPvMCaebbnrfifHhDYfgasaacH8srps0lbbf9 % q8WrFfeuY-Hhbbf9v8qqaqFr0xc9pk0xbba9q8WqFfea0-yr0RYxir % -Jbba9q8aq0-yq-He9q8qqQ8frFve9Fve9Ff0dmeaabaqaciGacaGa % aeqabaWaaeaaeaaakeaacqqGqbaucqqGYbGCcqqGLbqzcqqGKbazcq % qGPbqAcqqGJbWycqqG0baDcqqGLbqzcqqGKbazcqqGGaaicqqGMbGz % cqqGHbqycqqG0baDcqqGLaqjcaaMc8Uaeeypa0JaaGPaVlabbIcaOi % abbseaejabb2eanjabb+caViabbkeacjabbEha3jabbgdaXiabbgda % Xiabbwda1iabbMcaPiaaykW7cqGHxdaTcaaMc8UaeGymaeJaeGymae % JaeG4mamJaeyOeI0IaeG4mamJaeGimaaJaeiOla4IaeGOmaidaaa!5BB7!
(2)

[2] where M is the trait mean, G is the effect of the genotype (FF, FL, LL), L is the litter effect, S is the sex effect, P is the parity effect (1–2), S*G is the sex-by-genotype interaction, and e is the residual error. For rGFPwt measurements, which were carried out only on males, there was no parity effect detected, so genotype and litter effects only were fitted for this data set.

Single-QTL and Multiple-OTL interval mapping analysis

All traits (Bw21, 42, 70, 98, 115, fat%, and rGFPwt) were examined for normality. This was carried out by using the SAS macro NORMTEST. Additive and dominance coefficients (Haley and Knott 1992) were calculated at 1-cM intervals. These calculations incorporated the progeny’s genotypes as well as those of its parents and accounted for the possibility of new recombination within the recombinant parent, but assumed complete interference (i.e., no double recombination within a meiosis). Interval mapping was carried out with a general linear model (proc GLM, SAS Institute, 1989) with a phenotypic trait as the dependent variable and the additive and dominance coefficients as independent variables (Haley and Knott 1992). This model also included parity and sex as fixed effects, and litter size as a covariate. The F-statistics for the additive and dominance coefficients were used as a measure of support for these effects. Significance thresholds were determined empirically by permuting the marker data, as described by Churchill and Doerge (1994). Thresholds were obtained from 1000 permutations and are presented as F0.05 and F0.01 for significance at α = 0.05 and α = 0.01, respectively.

All traits were first analyzed by single QTL interval mapping. The strategy for the multiple-QTL analyses was to add QTL to the model until the additional QTL was not significant. Two- and three- QTL analyses were performed by estimating the effects of two or three QTL at separate positions simultaneously, examining all possible pairs or triads of locations (on a 1-cM grid), and selecting the pair, or triad of locations, for which the model explained the most variation. Epistasis was also included in the multiple-QTL analyses. Initially, our “full” linear model for the two-QTL analysis included all epistatic interactions, i.e., additive-by-additive, additive-by-dominance, dominance-by-additive, and dominance-by-dominance interaction terms. We later excluded additive-by-dominance, dominance-by-additive, and dominance-by-dominance terms from the model (see Cheverud et al. 2001) because our preliminary analysis revealed that our experimental design was generally not suitable for estimating interactions involving dominance—our data set contained few or no cases where a family was heterozygous at one position and segregating at another, and hence the P-values for the interaction terms involving dominance were undefined. However, the additive-by-additive interaction term was retained in the model if it was found to be significant. If the additive-by-additive term was not statistically significant for a particular trait, it was removed, and the analysis was repeated assuming additivity.

The significance of the additive and dominance effects in the multiple-QTL analyses was assessed by using the threshold F-values determined for a single-QTL analysis. Confidence intervals (CI) for the location of QTL were calculated from 1000 bootstrapped samples (Visscher et al. 1996) assuming there are two QTL. Because bootstrapping of two- and three-QTL analyses would be extremely computationally intensive, we bootstrapped the position of each QTL separately. For example, if a two-QTL analysis identified two QTL, A and B, at locations X and Y respectively, the CI for the position of QTL A would be estimated by performing general linear models with one QTL at position Y and finding the position of a second QTL that maximized the F-value for the entire model; such an analysis would be run for each of the 1000 bootstrap pseudo-samples. The CI for the position of QTL B would then be estimated in a similar way, but by holding the position of one QTL fixed at X. All permutation, bootstrapping and multiple-QTL analyses were performed with SAS.

Results

Single marker (ANOVA) analysis

QTL detection was carried out first using single marker analysis. This was performed on a total of 989 phenotyped progeny and grandprogeny of recombinants. A significant difference (P < 0.05) in fat% between the three possible genotypic classes was found at all markers tested, with the exception of D15Mit175, D15Mit150 and D15Mit39. This shows that the Fob3 QTL region was successfully introgressed by marker-assisted backcrossing to the F-line. For the trait rGFPwt in males, a significant difference was observed only at the markers between D15Mit138 and D15Mit63. Single marker analysis of the Bw traits indicates that differences between the genotypic groups become more significant with age and that the differences between the genotypic groups are apparent by 7 weeks of age (data not shown).

Single-QTL interval mapping analysis

All traits except for rGFPwt were approximately normally distributed. A reciprocal transformation was applied to the rGFPwt trait, to achieve greater normality, although a comparison between the interval mapping results of the transformed and untransformed data showed little difference (data not shown).

A summary of the single-QTL analysis for each trait is given in Table 1 and Fig. 2. As originally described by Horvat et al. (2000), the genetic effect of the Fob3 region (fat% trait), is approximately additive in nature, i.e., the additive effect is significant (P < 0.001), but the dominance effect is not (P = 0.1). For Bw115, both the additive and dominance effects are significant at α = 0.01. For all other Bw traits, in addition to significant additive effects, there are also significant dominance effects in males, but not in females.

Table 1 Single-QTL interval mapping analysis within the Fob3 regiona
Full size table

We have used Bw, which is highly correlated to the fat% trait, as an indicator of the development of obesity in our model over time. The results indicate that at weaning age (Bw21), the Fob3 QTL region is already significant (F-value = 18; P < 0.001), indicating that the effect of the Fob3 QTL region seems to be present early in development. The effect of Fob3 rises with age from Bw21 and stabilizes at 10 weeks (Bw70), as illustrated by the raw additive effects (Table l) and the standardized additive effects (i.e., divided by the standard deviation for each trait). Beyond 10 weeks, the effect remains approximately at the same level 2.0 g raw additive effect; ∼0.4 g standardized additive effect). Figure 2 shows the single QTL analysis plots, for each trait examined, giving the F-value for the additive and dominance terms combined. It can be seen that the peak becomes broader with age, indicating that there may be more than one QTL in the region.

Multiple-QTL Analysis

Multiple QTL analysis was performed on all traits, fitting epistatic interactions (i.e., additive-by-additive) when detected. In the two-QTL analyses the additive-by-additive interaction term for the best pair of locations is significant only for the traits Bw42 and Bw70. In all other cases, no significant interactions were detected, even without correction for multiple tests.

Two significant QTL, designated Fob3a and Fob3b, were found for most traits except for Bw21 (Fob3b only) and rGFPwt (Fob3a only). The locations identified in the two-QTL analysis for each trait are summarized in Table 2. For Bw115, Fob3a and Fob3b mapped to (29cM and 67cM, respectively, and the additive and dominance effects are roughly similar (between 1.2 and 1.5 g) for both QTL (Table 2), and significant in all cases (Fob3a additive F = 45.6 vs. F00.1 = 10.9, dominance F = 13.7 vs. F0.01 = 11.0; Fob3b additive F = 21.5 vs. F0.01 = 10.9, dominance F = 11.6 vs. F0.01 = 11.0. The two-QTL analysis of fat% yielded a similar combination of positions, i.e., 27 cM and 68 cM for Fob3a and Fob3b, respectively (Fig. 3 and Table 2). The estimated additive effect of fat% Fob3a is approximately twice (1.62%) that of fat% Fob3b (0.71%) and both are significant, though Fob3b is significant only at the α = 0.05 level (Fob3a additive F = 70.7 vs. F0.01 = 11.4; Fob3b additive F = 7.7 vs. F0.05 = 7.4). The dominance effects are not significant for either of the fat% QTL (Fob3a dominance F = 5.1 vs. F0.05 = 7.6; Fob3b dominance F = 5.6 vs. F0.05 = 7.6). The 95% confidence intervals (CI) for the positions of the two QTL for all the traits, determined by bootstrapping, are presented in Table 2 and shown schematically in Fig. 2 for fat%.

Table 2 Multiple-QTL interval mapping analysis within the Fob3 region
Full size table
Fig. 3
figure 3

Contour plot of the two-QTL analysis for fat%. Contour lines indicate the F-value from a general linear model fitting additive and dominance effects for QTL at two positions simultaneously. The “peak” indicates the combination of positions for which the model explains the most variation, and these positions are indicated by the arrows. The contour plot would be symmetrical about the 1:1 line, and therefore only one half of the plot has been shown for clarity.

Full size image

Three-QTL analysis was performed on all traits for which we found two significant QTL. As in the two-QTL analyses, the additive-by-additive interaction terms are not significant in the three-QTL analysis of Bw115 (P > 0.05 using threshold F value from single QTL analysis). Without epistasis, the triad of locations that explained the most variation (TMV) for Bw115 is 24–47–67 cM, but the additive and dominance effects of the third QTL (at 47 cM) are not significant (F < 2.0).

In the three-OTL analysis of fat%, however, there is evidence of epistasis. The TMV for fat% is 27–54–63 cM and, as in the two-QTL analysis, the additive effect of the QTL at 27 cM is highly significant (F = 36.3 vs. F0.01=11.4), but its dominance effect is not (F = 5.3 vs. F0.05 = 7.6). The additive(54)-by-additive(63) interaction term is significant for either of these position (F = 10.6 vs. F0.05 = 7.4), although the additive and dominance effects themselves are not significant for either of these positions (F < 4.0 in all cases). The lack of significance of the two QTL involved in the epistatic interaction is difficult to interpret, although it could be that the two QTL identified in the two-QTL analysis are real, and there is a third locus mapping between the Fob3a and Fob3b, which is interacting with Fob3b.

We found no evidence of three-QTL for Bw98; however, there are three significant QTL for the traits Bw42 and Bw70 with TMVs of 23–28–54 cM and 24–43–66 cM respectively. The TMV for Bw42 had significant additive(23)-by-additive(28) interaction; however, no epistasis was seen in Bw70 in contradiction to our results from the two-QTL analysis in this trait, where we saw significant epistasis.

Discussion

Originally, the Fob3 QTL region was detected in an F2 cross as a large effect QTL region with a LOD score of 11.3, and having a two-LOD support interval spanning 10–45 cM. Using an increased number of markers and a congenic line with the Fob3 region from the L-line introgressed into the F-line background, we have been able to dissect the target Fob3 region (fat% trait) to two smaller QTL, Fob3a (27 cM; 22–32 cM;) and Fob3b (68 cM; 44–70 cM;). The effects on fat% and Bw115 traits, contributed by each of the QTL separately, are reduced and demonstrate the effect that smaller linked QTL can have in overestimating a single large QTL effect. This has been shown previously in silico (Visscher and Haley 1996), and in in vivo mouse models (Legare et al. 2000). The importance of carrying out multiple QTL analyses is demonstrated here, as neglecting this may well lead to incorrect conclusions especially in regards to location and effect of the QTL.

We also show through single and multiple QTL analyses that these QTL are closely correlated to Bw and that the effects of the two QTL appear at different developmental stages. QTL for Bw with changing effects over age have been previously identified, for instance by Cheverud et al. (1996), in an F2 intercross between LG/J and SM/J inbred lines, although in that case there was no indication that the effects were correlated to obesity. Similarly, in an F2 cross between C57BL/6J × DBA/2J, Morris et al. (1999) and Christians et al. (2003), found a Chr 15 Bw QTL with evidence of increasing effect with age.

We detect Fob3a as a late-onset obesity QTL. Although it has a significant effect by 42 days, there is significant epistatic interaction between the two QTL detected, which is still present at Bw70. It is not until the later Bw stages and the fat% trait that it is possible to define this QTL to a 10-cM region (22–32 cM). A significant additive effect was also detected for a correlated trait rGFPwt with a similar 95% CI (21–38 cM), confirming the significant effect Fob3a on fat deposition.

Fob3b can be regarded as an early onset obesity QTL, detectable at 21 days, having a 95% CI from 38 to 68 cM, which continues to affect the Bw and fat% traits through development (fat% from 44 to 70 cM) The peak and size of CI for Fob3b seem to fluctuate slightly, especially between the early Bw traits, compared with the later traits (Table 2). This is probably due either to the Fob3a OTL influencing the analysis or Fob3b may itself be a constituent of two QTL with complex epistatic interactions, as indicated by our three-QTL analysis. Interestingly, the related trait rGFPwt is not significant at the Fob3b position, indicating that Fob3b does not affect the GFP, or that the Fob3b QTL also affects fat-free Bw, in addition to obesity. This is reflected in theF-values, which for Bw traits are all significant at the α = 0.01 level, whereas fat% reaches significance only at the α = 0.05 level.

As mentioned above, Fob3a and Fob3b had significant additive and dominance effects in Bw traits, but in obesity traits (fat% and rGFPwt), only the additive effects were significant. A possible interpretation could be pleiotropy of Fob3a and Fob3b—both QTL affect fat tissue mass in an additive manner, but may also affect non-fat tissue mass (e.g., muscle, bone mass), which would be reflected in a composite trait Bw. Although Fob3a and Fob3b may affect fat tissue in an additive manner, they may also increase non-fat tissue in a dominant manner. In this case no significant effect would be seen in heterozygotes on fat%, but there may be a significant accretion of non-fat tissue mass (i.e., lean/bone) in heterozygotes, resulting in a significant dominance effect for Bw traits. Clearly, a more detailed phenotypic characterization during development coupled with future genetic analyses of Fob3a and Fob3b would be needed to determine the specificity of Fob3a and Fob3b on fat tissue mass growth and/or growth of non-fat tissue mass.

Including the Fob3 region, a total of eight obesity/Bw QTL mapping to mouse Chr 15 have been reported to date. Several of these map proximally to Fob3a; the multigenic obesity QTL Mob4 (Warden et al. 1995) maps at marker D15Mit13 (3.2 Mb) and hence is likely to be an independent QTL. However, the diet-dependent QTL Dob3 and Dob4 (West et al. 1994; York et al. 1996) peak at marker D15Nds2 (21.4 Mb), and Bl/Bw, a QTL for body length and suggestive for Bw (Mehrabian et al. 1998), have peaks mapping close to the Fob3a QTL. Two additional QTL, Pfat4 and Pfat5, have been found on Chr 15; however, markers used were not reported (Pomp 1997). Finally, Pfat15, reported by Keightley et al. (1998), was detected as a fat% trait with a peak at D15Mit29 (74.4 Mb). Further analysis of this QTL (Morris et al. 1999) revealed a two-LOD support interval spanning 44.7 Mb, flanked by markers D15Mit5 (43.4Mb) D15Mit159 (8.8 Mb) and thus overlaped with both Fob3a and Fob3a, potentially being influenced by the same genes.

We carried out a search for candidate genes based on the markers flanking the 95% CIs for the fat% Fob3a and Fob3b QTL—therefore, a slightly broader interval than the 95% CI was examined. The data given here are taken from the MGSC v.15.30.1 database (http://www.ensembl.org/Mus_musculus), which was searched for genes having their function assigned to metabolism, obesity, or adipogenesis. In the Fob3a region spanning interval from D15Mit150 to D15Mit184, there were two such candidates: the cytochrome c oxidase–subunit Vic (Cox6c), which is the terminal enzyme in the respiratory chain (Hofmann et al. 1998), and the thyrotropin-releasing hormone receptor (Trhr), which affects obesity via the thyroid hormone pathway (Duthie et al. 1993; Taylor and Rowe 1987). The Fob3b QTL interval (between D15Mit115 and D15Mit1 contains Ndufb9 (NADH dehydrogenase (ubiquinone) 1 beta subcomplex, 9), which is involved in the oxidative phosphorylation of NADH to ubiquinone (Gu et al. 1996); Squalene monooxygenase (Sqle) involved in the sterol biosynthesis pathway (Kosuga et al. 1995); the myelocytomatosis oncogene (Myc, shown to inhibit adipocyte differentiation (Freytag and Geddes 1992); the otoconin 90 precursor (Oc90), which was suggested to be involved in lipid catabolism (Wang et al. 1998); thyrotroph embryonic factor (Tef), which acts in the thyroid hormone pathway (Drolet et al. 1991); Rho-GTPase-binding protein-1 (Rhpn1), found to bind to the glucose response element and hence triggers genes for lipogenesis (Hasegawa et al. 1999); and finally, diacylglycerol O-acyltransferase-1 (Dgat1), which has been shown to be an enzyme in triacylglycerol synthesis (Cases et al. 1998). These candidate genes could be examined in future experiments for differential expression between Fob3a/Fob3b congenic strains and F line to verify whether any of them could be causal for the QTL effects.

The dissection of the Fob3 region into at least two QTL reflects the complexity one would expect to see in a polygenically obese population as seen in humans, i.e., many loci having small effects. The challenge to positionally clone such loci is one that must be taken up, if true understanding of the human obesity epidemic is to be achieved. Such future cloning attempts of the two QTL will inevitably be more formidable, since the smaller effects will make the segregation of individual QTL (Fob3a and Fob3b) harder to track down. Both new QTL intervals are still relatively large, and while candidate genes can be identified, a large proportion of the genomic information pertains to unknown genes. The task to isolate the specific genes responsible for these effects from these intervals is challenging. Ideally, the Fob3a and Fob3b intervals should be fine mapped further with the “high resolution genetic mapping” approach such as using interval-specific congenic strains (ISCS) (Darvasi, 1997). A microarray-based gene expression analysis could be coupled to the fine mapping approach, for instance to examine differences between the F-line and the new ISCS. This could provide a more definitive list of candidates for further investigation and potentially could help elucidate the causal mutations underlying the Fob3a and Fob3b phenotypic effects and the metabolic pathway(s) involved.