INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

THE OBESITY MUTATION diabetes (Lepr^db) was originally assigned to mouse chromosome 4 by demonstrating that the Lepr^db locus is tightly linked with the locus affected by the misty (m) mutation (5). Since then it has become standard practice to propagate m and Lepr^db in the same populations, so that m can be used as a marker for prediction of Lepr^db genotype (2, 3). Populations that segregate m and Lepr^db in two different configurations have been developed for this purpose. When m and Lepr^db segregate in cis configuration, mice that develop misty coat color during the 1st wk of life are likely to bear the m Lepr^db/m Lepr^db genotype and become visibly obese during the 4th wk of life. When m and Lepr^db segregate in trans configuration, misty mice are likely to bear the m Lepr⁺/m Lepr⁺ genotype and nonmisty lean animals are likely to bear the M Lepr^db/m Lepr⁺ genotype; this configuration is generally used to identify Lepr^db heterozygotes for breeding.

An assumption underlying the use of m as a marker for Lepr^db genotype is that m has no independent effects on traits affected by Lepr^db. If m were to influence these traits, the combined effects of m and Lepr^db would be difficult to separate and would confound the interpretation of data collected from populations segregating m and Lepr^db. Despite this risk, no published data address the suitability of m as a marker for Lepr^db. An influential paper by Coleman (2) states that m affects no traits other than coat color, but provides no supporting data. In contrast, our experience with populations segregating m and Lepr^db has been that m Lepr⁺/m Lepr⁺ mice appear smaller and leaner than their M Lepr^db/m Lepr⁺ cohorts. This might reflect either a Lepr^db heterozygote effect or an effect of m. We sought to resolve this issue by estimating m genotype effects in the absence of Lepr^db.

	RESEARCH DESIGN AND METHODS

Top Abstract Introduction Methods Results Discussion References

Animals. The mice used in this study were treated under a protocol approved by the Pennington Center Institutional Animal Care and Use Committee. A colony of C57BL/6J-M Lepr^db/m Lepr⁺ mice was established at the Pennington Biomedical Research Center in 1992 from founders provided by Beverly Paigen of The Jackson Laboratory (Bar Harbor, ME). In 1995, m and Lepr^db were separated into different lines by crossing C57BL/6J-M Lepr^db/m Lepr⁺ mice with C57BL/6J-M Lepr⁺/M Lepr⁺ mice purchased from The Jackson Laboratory. Most progeny of this cross are expected to inherit M Lepr⁺/m Lepr⁺ or M Lepr⁺/M Lepr^db genotypes. These genotypes are not visually distinguishable, but can be reliably identified with a simple sequence-length polymorphism (SSLP) described below. Mice identified by this assay as M Lepr⁺/M Lepr^db were mated to create a C57BL/6-Lepr^db population that does not segregate m; this population has been maintained for use in other experiments. Mice identified as M Lepr⁺/m Lepr⁺ were mated to create a C57BL/6-m population that does not segregate Lepr^db; data presented here derive from this C57BL/6-m population.

Genotype assay. An SSLP for D4Mit255 (assay name MT2795, Research Genetics, Huntsville, AL) maps 0.6 centimorgan from Lepr^db, placing D4Mit255 about halfway between the loci affected by m and Lepr^db (9). The C57BL/6-m Lepr⁺/M Lepr^db population segregates two D4Mit255 alleles, the shorter allele segregates with m and the longer allele segregates with Lepr^db. This assay was used to distinguish between M Lepr⁺/m Lepr⁺ and M Lepr⁺/M Lepr^db progeny of the M Lepr^db/m Lepr⁺ × M Lepr⁺/M Lepr⁺ cross, which helped us select appropriate breeding pairs.

The C57BL/6-m population segregates two D4Mit255 alleles, the shorter allele segregates with m and the longer allele segregates with M. This assay allowed us to genotype members of the C57BL/6-m population as M/M, M/m, or m/m.

Primers for the D4Mit255 assay were 174 nM each in 50 mM KCl; 10 mM Tris · HCl (pH 9 at 25°C); 0.1% Triton X-100; 1.5 mM MgCl₂; 0.2 mM each dATP, dCTP, dGTP, and dTTP; 0.3 U Taq polymerase; and 200 ng genomic DNA in a 25 µl volume. Target DNA was amplified by denaturing at 95°C for 1 min, followed by 30 cycles of 95°C for 30 s, 55°C for 30 s to anneal, and 72°C for 30 s to extend. Ten microliters of PCR product were electrophoresed through a vertical 8% polyacrylamide gel at 13 V/cm for 1.6 h in 1× TBE (in mM: 89 Tris · HCl; 89 H₃BO₃; and 2 EDTA; pH 8.0). Gels were stained in 1× TBE containing 1.0 mg/ml ethidium bromide and fluoresced under ultraviolet light. Images were captured as TIFF (Tagged Image File Format) files via a charge-coupled device camera and AMBIS Image Acquisition & Analysis software (San Diego, CA).

Protocol. When mice in the C57BL/6-m population reached 5 days of age, we marked them by toe clipping, collected a small piece of tail for DNA isolation, and weighed them for the first time. Body weights were measured daily thereafter. Dams were removed when the pups reached 24 days of age, but littermates were housed together through 35 days of age. Mice were killed by CO₂ asphyxiation at 35 days of age; body weight was again measured, along with nose-to-anus length and inguinal adipose pads weight.

Data analysis. The body weight dataset was analyzed as a growth curve under a mixed linear model. Mixed linear models include both fixed-effects factors and random-effects factors (7). Fixed-effects factors have discrete levels, all of which are included in the dataset. Random-effects factors have many possible levels that are randomly distributed, with the dataset including a random sample of all possible levels.

The model for growth curve analysis included three fixed-effects factors: sex, m genotype, and age. Sex had two levels, male and female; m genotype had three levels, M/M, M/m, and m/m; and age had fifteen levels (age = 6, 8, 10,...34). Although we measured body weight every day from 5 to 35 days of age, we analyzed data from even days only, because analysis of all data required more computer memory than was available. Litter size was also considered to be a potentially important fixed effects factor, but preliminary analyses indicated that litter size had negligible effects, so litter size was excluded from the model.

The model for growth curve analysis also included one random-effects factor, within-subject covariance among ages. This factor accounts for the fact that body weights of a given subject are correlated among ages. Because covariance estimates differ considerably among ages and between sexes, separate covariance estimates were made for each age by sex group under a heterogeneous autoregressive covariance structure (7). We also considered within-litter covariance to be a potential random-effects factor. This factor would account for the shared environment among littermates. However, preliminary analyses indicated that accounting for within-litter covariance had negligible influence on the results, so within-litter covariance was excluded from the model.

A second analysis was performed on three traits measured after mice were killed at 35 days of age: body weight, nose-to-anus length, and inguinal adipose pads weight. Because within-subject responses of these dependent variables are expected to be correlated, we analyzed the data under a multivariate mixed linear model. Sex and m genotype were treated as fixed effects nested within trait. An unstructured covariance structure was used to estimate different variances for each sex within trait combination (7).

For each of the three traits, two comparisons were made: heterozygotes means were compared with wild-type means, and homozygous mutant means were compared with wild-type means.

Data were analyzed under PROC MIXED of SAS version 6.11.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Experimental population. The experimental population included 52 mice from six litters (Table 1). Litter sizes ranged from 7 to 10 mice. Five additional animals, three m/m and two M/m, died during the first few days of life before data collection began. The population size is too small to test whether m affects neonatal survival.

m Affects growth. The effects of m on growth are represented in Fig. 1. The overall effect of m is highly significant [F(2,46) = 13.77, P < 0.0001]. There is also a significant interaction between m and age [F(28,644) = 2.17, P < 0.0005], suggesting that the response to m may differ among ages. The "m by sex by age" interaction is small [F(28,644) = 1.42, P = 0.0735], indicating that the growth curves for males and females are essentially parallel.

View larger version (23K): [in this window] [in a new window]   Fig. 1.   A: sex-averaged growth curves for misty (m) genotype groups. Lines connect least-square means within genotype group. B: body weights as a percent of wild-type (M/M) means. To remove the influence of scale changing with age, all genotype groups means were divided by the wild-type mean at that age and multiplied by 100.

We can represent the "m by age" effect by plotting growth curves for each genotype group (Fig. 1A). Because the growth curves for males and females are parallel, sex-averaged growth curves can be used, but remember that males will be shifted higher and females shifted lower. The growth curves all display a period of fast growth for the first 2 wk of life, followed by a period of slow growth during the 3rd wk of life and a period of fast growth during the 4th and 5th wk of life.

The response to m appears to be fairly similar across these periods. It may appear in Fig. 1 that the means diverge with time, but this is partly an illusion caused by the change in scale with age (that is, the range and magnitude of variation differs among ages). Scale effects can be removed by representing body weight as a percentage of the wild-type means (Fig. 1B). The m/m means average 83% of M/M means from 6 to 34 days of age, ranging from 80 to 87% of M/M means during this period. M/m means average 95% of M/M means and range from 93 to 99% of M/M means during this period. Thus the body weight response to m changes little during the first 5 wk of life.

Because the analysis of variance indicates that there is an m by age interaction, we compared m/m means and M/m means to M/M means at every age to determine when the effects of m become detectable. We considered the differences between the means to be statistically significant when the P value was 0.01 or less. Under these conservative criteria, m/m mice were significantly smaller than M/M mice at every age from 10 to 34 days of age. M/m mice were never significantly smaller than M/M mice under these conservative criteria. These results demonstrate that m/m mice are smaller than M/M mice by the 2nd wk of life and remain smaller at least through the 5th wk of life.

m Affects multiple growth traits. We also analyzed the effect of m on three traits: body weight, body length, and inguinal adipose pads weight, measured after the mice were killed at 35 days of age (Fig. 2). All three traits are affected by m. At 35 days of age, m/m mice weigh 15% less than M/M mice, are 8% shorter, and have 21% less inguinal adipose mass. M/m are not significantly different from M/M mice in any of these measures. These results demonstrate that m affects multiple traits associated with growth.

View larger version (15K): [in this window] [in a new window]   Fig. 2.   m Affects body weight (A) [F(2,86) = 15.71, P < 0.0001], nose-anus length at 35 days of age (B) [F(2,86) = 11.32, P < 0.0001], and inguinal adipose pads weight (C) [F(2,86) = 30.26, P < 0.0001]. A-C: no M/m means differ from M/M means. All m/m means differ (P < 0.0011) from M/M means. Superscripts a and b are different for means that differ.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

The original reports of m described its effects on coat color and did not comment on body size or other traits (10, 11). Similar to other coat color mutations, m subsequently became an important marker for linkage analysis. In fact, we know more about the chromosomal location of m than we know about its physiological effects, because a fine scale genetic linkage map of the chromosomal region around m has been prepared (4), whereas the physiological effects of m have been ignored until recently.

It is now clear that misty affects traits other than coat color. Sviderskaya and others (8) recently reported that misty mice have no detectable brown adipose tissue on the 1st day of life. They also found that misty mice have prolonged bleeding time. The defect that links these phenotypes may interfere with adenine nucleotide metabolism (8). We find that body weight, body length, and adipose mass are also affected in misty mice. Their smallness is so easy to see by simple visual comparison that it is surprising that this has not been reported previously as a barrier to interpretation of data. Together these observations indicate that misty has much broader effects than previously reported.

Because m was transferred from another strain of mice into this population, it is also possible that a separate locus within the chromosomal segment that contains m causes the smaller size of m/m mice. Whether m or another locus is responsible, the implications are the same, the DNA segment that contains m influences growth of mice.

These observations call for caution when interpreting data from populations segregating both m and Lepr^db. When m and Lepr^db segregate in the same population, interpretation of data collected from the population is confounded by linkage. Some fraction of the variance in body weight, adiposity, or any other trait may be attributable to m and some may be attributable to Lepr^db. Because there is no way of accounting for the bias that m introduces into statistical analyses, spurious conclusions become inevitable. Furthermore, the relative contributions of m and Lepr^db may vary under different environmental, genetic, or developmental conditions and may vary among different traits. Under some conditions, Lepr^db effects may be exaggerated by m, under others Lepr^db effects may be underestimated. Although m heterozygote effects for the traits we measured were small enough that differences between M/M and M/m means were undetectable, m heterozygote effects may still be large enough to bias test statistics. Because of these considerations, we recommend that investigators who wish to draw accurate inferences about Lepr^db genotype effects obtain their mice from Lepr^db populations that do not segregate m.

Now that a base substitution in a leptin receptor gene has been identified as the molecular basis of the Lepr^db mutation (1, 6), m can be replaced as a marker for Lepr^db genotype by assays that detect the base substitution. Although the mutation affects no restriction sites that would be useful for genotype assays, restriction sites can be created by changing a single base in a PCR primer that anneals near the mutation. This assay is independent of phenotype, so Lepr^db genotype can be measured directly, rather than inferred from lean, obese, and misty phenotypes. Measuring Lepr^db genotype directly also eliminates the possibility of recombination between marker and affected locus. Furthermore, Lepr^db genotype can be identified as early as the first day of life if needed. A disadvantage is that the direct assay requires some time and effort.

When the disadvantages of m as a marker for Lepr^db and the advantages of a direct assay are considered, we prefer to use mice segregating Lepr^db in the absence of m. During this study we removed m from our population of C57BL/6J-M Lepr^db/m Lepr⁺ mice, creating a C57BL/6-Lepr^db population that allows Lepr^db genotype effects to be accurately estimated. In particular, Lepr^db/Lepr^db mutants can be directly compared with their wild-type littermates to accurately estimate the effects of a dysfunctional leptin receptor in comparison to a fully functional leptin receptor. This population also enables investigators to accurately estimate Lepr^db heterozygote effects, which are seriously distorted in populations that also segregate m. Furthermore, it enables accurate estimates of Lepr^db genotype effects to be obtained at any age, whereas populations that segregate both m and Lepr^db are especially prone to give poor estimates of Lepr^db effects early in life, when the influence of m on growth is greater than the influence of Lepr^db (our unpublished data).

Perspectives