Identification of Major Quantitative Trait Loci Controlling Body Weight Variation in ob/ob Mice

We have previously shown that two background genomes, BTBR (6) and B6, modify the type 2 diabetes syndrome in obese mice (7). Since BTBR-ob/ob mice are severely diabetic and markedly glycosuric, we expected to observe substantial secondary weight loss. Surprisingly, BTBR-ob/ob mice gained significantly more weight than B6-ob/ob controls despite having similar weaning weights (Fig. 1A and B).

At least part of this weight gain difference can be explained by food intake. Over a 20-day period, BTBR-ob/ob mice consumed significantly more food than B6-ob/ob controls (Fig. 1D). This increased food intake can be directly associated with weight gain during this period, as feed efficiency was not significantly different (Fig. 1G).

We assembled 350 F₂-ob/ob mice derived from BTBR and B6. Sewall-Wright analysis (Fig. 1C) indicates that 30% of the F₂ population variance in body mass is heritable. We genotyped the F₂ mice at an autosome-wide panel of microsatellites and detected quantitative trait loci (QTLs) as described previously (7).

We detected two major body mass QTLs (Fig. 2). One locus, on chromosome 2, exhibits highly significant linkage to body mass (logarithm of odds [LOD] = 9.48) (Fig. 2B). At the peak marker, D2Mit9, each BTBR allele semidominantly increases 10-week body mass by 3.4 g (∼7% of the total mass of the B6-ob/ob mouse). This locus, designated modifier of obese 1 (Moo1), explains 14.1% of the total F₂ population variance.

Another highly significant linkage, Moo2, resides on chromosome 13 around D13Mit66 (Fig. 2C). The Moo2^B6 allele is dominant to BTBR, adding ∼2.7 g to 10-week body mass, explaining ∼12.5% of the F₂ population variance. Multivariate regression analysis of Moo1 and Moo2 revealed no significant nonlinear interactions (data not shown). Although Moo2 cannot explain any of the parental strain difference reported here, it supports other studies showing B6 to bear alleles promoting increased feeding efficiency (8), adiposity, and insulin resistance (9,10).

Two other loci were discovered on chromosomes 5 and 17 (Moo3 and Moo4) (Fig. 2A). Both loci exhibited suggestive linkage with interval mapping but showed significant linkage using multiple interval mapping (Moo3 peak LOD = 3.60, D5mit136; Moo4 peak LOD = 2.49, D17Mit109).

Although the F₂ data establishes the existence of QTLs, it does not provide the positional resolution to support cloning without a candidate gene. We therefore refined the position of Moo1 using interval-specific congenic (ISC) mouse lines (11). We introgressed segments of the Moo1 support interval from B6 into BTBR. We defined the support interval as the 10-cM interval surrounding the LOD peak. Three congenic mouse lines are shown (Fig. 3E), named Moo1-A through -C.

Congenic lines Moo1-A and Moo1-B have inherited the centromeric half of the Moo1 support interval from B6. The two distal recombinational break points arose independently in different founder mice. Each line’s proximal recombination lies at least 19 cM (34 Mb) centromeric to the Moo1 LOD peak. Longitudinal analysis of 28 growth curves from line Moo1-A (Fig. 3A) by sigmoidal regression shows no association between weight gain and genotype (difference in mean asymptotic body mass = b₃ = −0.31 ± 1.71 g, P = 0.82). Similarly, 27 growth curves from line Moo1-B showed no association with genotype (b₃ = −0.21 ± 2.7 g, P = 0.94) (Fig. 3B). Thus, ISC lines Moo1-A and -B detect no evidence of Moo1 in the region of chromosome 2 centromeric to (and including) D2Mit61.

Line Moo1-C has retained an ∼8-Mb segment of the support interval from B6. Inheritance of this region significantly alters expected weight gain among mice from this line (Fig. 3C). Analysis of 24 growth curves from line Moo1-C showed strong association between genotype and body mass (b₃ = −8.70 ± 1.72 g, P < 0.0001), but no sex dimorphism (b₂ = 0.4 ± 1.8 g, P = 0.82). Moo1-C thus establishes the existence of an obesity-modifier locus in an ∼8-Mb interval between D2Mit56 and D2Mit159 (Fig. 3E). Of course, it remains a formal possibility that another modifier locus exists somewhere else in the Moo1 QTL. We need to be mindful of the fact that other leptin modifier QTLs may exist elsewhere in our Moo1 support interval not covered by the introgressed segment in the congenic mice.

Body composition analysis was performed on a randomly selected, representative subset of 15-week-old obese mice from Moo1-C. Nearly 90% of the observed difference in total carcass mass can be accounted for by differences in ether-extractable (lipid) mass (−7.1 ± 2.4 g per B6 allele) (Fig. 3E). No significant differences between genotypes were observed in moisture content or fat-free dry mass, indicating that Moo1 regulates body mass by controlling accretion of lipid.

The Moo1 position established by Moo1-C corresponds to a region associated with subcutaneous fat and diet-induced HDL levels in (B6 × CAST/Ei) F₂ mice (12), and with obesity in (AKR × C57BL6/J) F₂ mice (13). Moo1 is distinct from an adiposity QTL in (NZB/B1NJ × SM/J) F₂ mice (14). Other linkages to body weight or adiposity on chromosome 2 are broad enough to potentially include Moo1 (15,16). Together, the number of linkages on chromosome 2 suggests that a common locus may explain some or all findings and thus be of particular importance. Such a locus would be syntenic with human 2q24-32 in a region containing <50 genes.

Since obesity and type 2 diabetes are so closely interrelated, it is tempting to hypothesize that Moo1 and Moo2 influence body weight by altering diabetes susceptibility or, conversely, that toggling the genotype at Moo1 or Moo2 could produce enough weight loss to ameliorate the risk of type 2 diabetes elicited by leptin deficiency. However, neither locus associates with fasting plasma glucose or insulin levels, despite the presence of an insulin locus, t2dm3, 30 cM telomeric to Moo1 (7). Despite their strength, the obesity-regulating effects of Moo1 and Moo2 are insufficient to overcome the effects of ob on diabetes susceptibility.

The obesity-regulating effects of Moo1-2 are silent in Lep^+/+ mice housed in standard conditions (data not shown). How then can our findings be extended to humans, in whom leptin deficiency is rare? Leptin has been shown to play a predominant role in regulating the physiological response to hunger and weight loss, but more recent data show that leptin’s ability to regulate hypothalamic proopiomelanocortin and CART mRNA levels diminishes when plasma leptin concentrations exceed the physiological range (17), even though proopiomelanocortin mRNA levels decrease in overfed individuals (18). These findings suggest that at higher concentration ranges, leptin may not be the primary mediator of the satiety response in obese or overfed individuals or in high-fat fed mice. The majority of obese humans are hyperleptinemic (19), indicating that regulation of energy homeostasis in the obese state occurs downstream, or independent of leptin. We targeted mediators of such pathways by eliminating the leptin arm directly. The genes thus identified may help elucidate these leptin-independent pathways and therefore be of particular relevance to human obesity.

BTBR, B6, and B6-ob/+ mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred at the University of Wisconsin. Mice were weaned at 3 weeks of age onto a 6% fat diet (Purina 5008). Mice had ad libitum access to food and water, except for fasts (0730−1130) before blood draws and killing (by CO₂ asphyxiation). The facilities and research protocols were approved by the University of Wisconsin Institutional Animal Care and Use Committee. The lineage and characteristics of the BTBR strain have been reviewed by Ranheim et al. (6).

Detection and mapping of QTLs was performed as previously described (7). Briefly, genotypes of 350 F₂ mice at 99 markers were assembled using MAPMAKER/EXP (20). Interval mapping with MAPMAKER/QTL and multiple-interval mapping with QTL Cartographer (21) were used to detect segregating loci. Other analyses, including multiple-trait interval mapping (21), composite interval mapping (21), and stepwise regression analyses, gave results consistent with those stated herein.

A support interval for each locus was defined as the 10-cM region surrounding the LOD peak. BTBR × B6 F₂ mice bearing recombinations in a support interval founded each ISC line. We used a marker-assisted backcrossing program (22) to introgress each recombination into BTBR. We referenced the mouse genome from the Draft Sequence Browser (23) and Celera (24) to map the recombinations and search for new microsatellites.

Each ISC line was scored as “retaining” or “not retaining” the B6 allele of the QTL by comparing growth curves of obese backcross mice within each line, by genotype, i.e., comparing mice heterozygous for the congenic insert with their sibs who did not inherit the insert. No historical controls or cohorts were employed. Longitudinal body mass data for each mouse were fitted to a sigmoidal growth model:

where b₁, b₄, and b₅ are curve-fitting parameters, b₂ quantifies the sex dimorphism, b₃ measures the effect of substituting the B6 allele for BTBR, and u₁ measures random effects. The model was fitted using PROC NLMIXED (25).

From ages 40 to 60 days, ob/ob mice were housed individually. Food was weighed and replenished daily. Body weights were recorded at the beginning, end, and on every 3rd day during the study. Data were compared by Student’s t test.

At 15 weeks of age, fasted mice were killed by asphyxiation. The brain was removed for future study, the gastrointestinal tract was emptied, and the carcass was stored at −20°C. Carcasses were autoclaved for 7 h at 250°C. Water lost during autoclaving was replaced to within 0.01 g of the original mass, and the carcass was homogenized in a blender. Triplicate aliquots were quantitatively lyophilized, transferred to filter paper envelopes (Whatman #1), and extracted for 20 h with ethyl ether in a refluxing soxhlet apparatus. Each measured component of the carcass was regressed on sex and genotype using PROC GLM (25).

This work was supported by National Institutes of Health grant DK-58037, Xenon Genetics, and an American Diabetes Association Mentor-based Fellowship. J.E.B was supported by National Institutes of Health Predoctoral Training Grant DK-07665-11.

We thank the NHLBI Mammalian Genotyping Service at the Marshfield Clinic (Marshfield, WI) for their genotyping efforts. We also acknowledge K. Albright, M. Pariza, W. Dove, and M. Rabaglia for their support and assistance.