Numerous physiological and molecular changes accompany dietary restriction (DR), which has been a major impediment to elucidating the causal basis underlying DR's many health benefits. Two major metabolic responses to DR that potentially underlie many of these changes are the body temperature (Tb) and body weight (BW) responses. These responses also represent an especially difficult challenge to uncouple during DR. We demonstrate in this study, using two recombinant inbred (RI) panels of mice (the LXS and LSXSS) that naturally occurring genetic variation serves as a powerful tool for modulating Tb and BW independently during DR. The correlation coefficient between the two responses was essentially zero, with R = −0.04 in the LXS and −0.03 in the LSXSS, the latter averaged across replicate cohorts. This study is also the first to report that there is highly significant (P = 10−10) strain variation in the Tb response to DR in the LXS (51 strains tested), with strain means ranging from 2 to 4°C below normal. The results suggest that the strain variation in the Tb response to DR is largely due to differences in the rate of heat loss rather than heat production (i.e., metabolic rate). This variation can thus be used to assess the long-term effects of lower Tb independent of BW or metabolic rate, as well as independent of food intake and motor activity as previously shown. These results also suggest that murine genetic variation may be useful for uncoupling many more responses to DR.
- food restriction
- caloric restriction
dietary restriction (dr) in rodents produces a vast array of physiological, cellular, and biochemical changes (22). Indeed, it is unusual to find a process that is not affected by DR. This plethora of changes has made it virtually impossible to sort out which responses to DR are relevant to its many health benefits and its ability to extend life span by as much as 50% in rodents (22). Here, we use a comparative and genetic approach that will ultimately function to reveal a set of causally related physiological responses to DR.
Two of the major physiological responses to DR in mammals are lower body temperature (Tb) and reduced body weight (BW). Both result in health benefits, and each could play a role in the extension of life span under DR (1, 14, 15, 17, 18). However, uncoupling these responses to DR represents a major challenge, especially while trying to control for the effects of food intake and activity. In particular, heat production and BW are physiologically related during diet-induced thermogenesis and nonexercise activity thermogenesis, such that greater heat production leads to lower BW (2, 10). Although controversial, it is also widely thought that lowering Tb inherently lowers metabolic rate (5, 19) and would thus mitigate weight loss.
The metabolic trade-off between Tb and BW is especially apparent in mice; their small size and thus large surface area-to-volume cause mice to lose heat quite rapidly. At room temperatures (21–23°C), mice must allocate ∼23% of their energy budget to maintaining a normal mean Tb of 37°C (24). Under the conditions of DR used in this study (60% ad libitum, AL), there is a survival response to conserve energy by temporarily lowering the Tb set point each day to ∼19°C, resulting in daily torpor and often a dramatic drop in Tb (5, 7, 12). This energy savings mitigates weight loss as shown by DR studies conducted at thermoneutrality (3, 8, 9, 24). For example, after 2 wk on a 54% AL diet, weight loss was reduced from 23% at 23°C to just 9% at thermoneutrality (30°C) (24). The BW of mice is thus highly responsive to the energy saved by reducing the differential between Tb and ambient temperature during DR.
There is also a survival response to reduce the rate of heat loss during DR and torpor (5, 19, 20). This decreased heat loss can mitigate a drop in Tb by as much as 5–6°C (19), allowing torpid mice to keep Tb as high as possible without increasing metabolic rate. Although reducing the rate of heat loss also reduces the energy expenditure needed for rewarming from torpor, it does not appreciably reduce the 24-h metabolic rate (20). (For a quantitative analysis, the online version of this article contains supplemental information.). Therefore, in contrast to varying the Tb set point, strain variation in the rate of heat loss would not be expected to appreciably affect weight loss.
Previously, we have demonstrated that mice exhibit a remarkable amount of genetic variation in their response of Tb to DR, with mean Tbs ranging from ∼32 to 35°C (12–14). More recently, we also demonstrated that there is marked genetic variation in the response of BW to DR, with the strain means for BW ranging from about 60 to 85% of the AL controls (15). The variation in these responses is not explained by trivial factors, such as the strain differences in absolute food intake, feces:food ratios, feces calorimetry, gross motor activity, or AL percent body fat (12, 13, 15). Therefore, in this study, we asked to what extent the Tb and BW responses are being coordinately modulated in these strains.
The LXS and LSXSS recombinant inbreds (RIs) were originally derived from a heterogenous stock of eight classical inbred strains, including five of the classical inbreds analyzed in this study (A, BALB/c, C3H, C57BL/6, and DBA) (23). From this stock, the LS line was selected over many generations for “long-sleep” sensitivity to a high dose of ethanol, hence, the designation LS. At the same time the SS line was selected for “short-sleep” sensitivity, hence, the designation SS. The LS and SS lines were used to create the LSXSS RIs by the standard method of inbreeding F2 mice. However, the LS and SS parental strains were not inbred, which makes the LSXSS less useful for mapping on traits not related to ethanol sensitivity, because of the presence of multiple alleles outside the regions of genetic selection. The LXS is a newer, larger, and much more powerful RI panel created from an inbred LS strain (ILS) and inbred SS strain (ISS) (23).
Control of calorie intake.
DR food intake in this study was carefully controlled to be 60% of AL, as determined separately for each strain, with each mouse housed singly (12, 13, 15). The food intake of each AL mouse was measured each week over the entire study, with the food weights being collected at approximately the same time of day each week. Mice were started on DR at ∼2 mo of age and were closely age-matched within cohorts (typically within 2 wk of age). With only a few exceptions, the AL and DR mice were also closely weight matched just before the start of DR (15). The amount of food wasted by both the AL and DR mice was measured approximately every 10 wk and used to adjust the calculation of the DR rations (12, 13, 15). We also measured the feces:food ratio and feces calorimetry of both the AL and DR mice and found that these were not appreciable correlates of either the Tb or the BW response to DR (Refs. 12, 13, 15; and results). The DR mice were fed from the same bag of chow (Harlan Teklad 7012) as the AL mice, ensuring the same calorie content. Every DR ration was weighed to the nearest 0.1 g for each mouse and adjusted each week on the basis of our measures of AL food intake. The mice were fed double rations on Monday and Wednesday and a triple ration on Friday, except LXS Cohort 1, which was initially fed each day for the first 4 mo on DR before switching to the Monday/Wednesday/Friday schedule. We have previously shown that switching to the Monday/Wednesday/Friday schedule produces no appreciable effect on the strain variation in the Tb or BW response to DR (12, 15).
Tb response to DR.
The strain means for Tb in response to DR for the LSXSS, ILS, ISS, and the six classical inbred strains—129S6, A/ibg, BALB/c/ibg, C3H/ibg, C57BL/6/ibg, and DBA/ibg—are the same as reported in our previous studies (12–14). The Tb response to DR is defined as the mean Tb measured after at least 2 mo of DR, adjusted by linear regression to remove a small correlation with the AL strain mean for Tb (Pearson correlation coefficient Rs = 0.1–0.5) (12, 13).
The strain means for the Tb response to DR of the 51 LXS strains are being reported for the first time in this study. As in our previous studies (12, 13), all temperature trials were conducted at room temperature (21–24°C), with Tbs being measured every 4 h (except LSXSS Cohort 2a at 18 wk was measured every 6 h; Table 1) for 1 wk using rice-grain size, transdermally implanted transponders (BioMedic Data Systems, Seaford, DE). For the LXS, all transponders were calibrated against rectal temperature, as previously described (13) (In the online version of this article, see supplemental information, last paragraph of additional discussion online). The LXS cohort characteristics (birth dates, DR start dates, number of mice per strain and diet, early mortalities, and mean daily food intake) have been previously described (15). As in our previous studies, all mice were female. All protocols were approved by the University of Colorado's Institutional Animal Care and Use Committee.
BW response to DR.
The strain means for the BW response to DR used in this study are based only on mice that were also monitored for Tb, typically four DR and two AL mice per strain (only two DRs per classical inbred strain, except only one DR for BALB/c). Therefore, the strain means are very similar, but not identical, to those reported by Rikke et al. (15). Just as before, the BW response to DR is defined as the strain means for diet-restricted BW regressed on the AL means. The residuals were then rescaled to give an adjusted percent AL value by adding the DR grand mean and dividing by the AL grand mean (15). This rescaling was done so that the residuals would be physiologically meaningful; it has no effect on the strain variation. The time periods over which the BW response was assessed are also the same as previously described (15), beginning when the diet-restricted BWs had stabilized in each cohort (typically 6–8 wk after starting DR). The correspondence between the time periods used for the BW measures and the time periods of the Tb trials is shown in Table 1.
With the exception of 5 LXS strains in LXS Cohort 1 after week 14 of DR, all of the BWs are fasted weights measured 0–4 h before feeding (15). We also conducted three trials in LXS Cohort 2, in which we compared the DR grand means for BW on Monday, Wednesday, and Friday just before feeding double, double, and triple rations, respectively, and found that the day of the week had no appreciable effect on BW (trial 1: 17.6 g, 18.0 g, 17.9 g; trial 2: 17.6 g, 17.4 g, 17.7 g; trial 3: 17.7 g, 17.7 g, undetermined, respectively). Therefore, the BWs just before feeding were independent of length of fast, most likely because the length of fasting is compensated for by a proportionately larger ration preceding the fast.
The LXS strain means for absolute food intake, feces:food ratio, feces calorimetry, and home cage activity are the same values used in Rikke et al. (15). The methods for collecting these data and calculating the strain means are also described in that study.
Hair length was measured by collecting a tweezers sample of hair from the mouse's back, ∼1 cm above the base of the tail. The lengths of 11 hairs from each sample were measured using a microscope with an ocular micrometer. To ensure representativeness, the longest and shortest length measurements from each sample were automatically excluded before calculating the mean hair length. These mean lengths were then averaged within strain, typically four mice per strain.
The QTL comparisons are based on our previous genetic mapping conducted using all mice measured for weight loss (15). QTL mapping on the Tb response to DR in the LXS was conducted using MapManager QTL (11), as we have previously used (13, 15).
To test for a correlation between the Tb and BW responses to DR within strain, we mathematically removed the effect of strain variation in these measures using a general linear model univariate analysis, with strain and cohort as fixed factors (SPSS 13.0, Chicago, IL).
We expected that a lower Tb during DR would be reflected in an increased BW relative to the AL controls. We tested for this separately in the LXS and LSXSS panels of RI strains. As previously reported for the LXS (15), the strain means for BW in response to DR stabilized by ∼7 wk after starting DR. Therefore, we used our previous data (15) from weeks 7 to 16 of DR (week 16 is the cutoff because we arbitrarily stopped studying some strains after week 16) but only from the subset of mice that were also measured for Tb. Fifty-one different strains were represented, and the strain variation in the BW response to DR for these mice ranged significantly from 65 to 85% of the AL BW (Fig. 1). The Tb data on these mice were collected every 4 or 6 h for a 1-wk period during weeks 7–16 of DR (see Table 1 for exact time periods for each cohort).
The Tb results during DR for the LXS mice are being reported for the first time in this study. The strain means under DR ranged significantly from 32 to 36°C (P < 10−10, ANOVA; Fig. 1). This strain variation was not correlated with the strain variation in the DR ration sizes (absolute food intake), also indicating no correlation with the absolute differences in calorie, protein, carbohydrate, fat, vitamin, or mineral intake (Fig. 2) (R = 0.20, P= 0.08, n = 51 strains). The strain variation in the Tb response to DR was also not explained by differences in the feces:food ratio (R= −0.18, P = 0.22, n = 20 strains) or feces calorie content per gram (R = −0.45, P = 0.11, n = 9 strains), suggesting no effect of coprophagy or digestive efficiency. There was also no correlation with the differences in home-cage motor activity (R = −0.02, P = 0.89, n = 41 strains).
Surprisingly, we found that the marked variation in the Tb response to DR in the LXS was not significantly correlated with the BW response to DR in these same mice, with P = 0.40 (1-tailed based on an expected negative correlation). Moreover, the Pearson correlation coefficient was essentially zero, with R = −0.04 (Fig. 1).
We also tested for a correlation between the Tb and BW responses to DR in a cohort of 20 LSXSS RIs, plus ILS and ISS (also inbred from the LSXSS parental strains and tested at the same time). Again, we found that there was no correlation (Fig. 3A) (R = −0.01, P = 0.48, 1-tailed), even though these strains also exhibited marked variation in both responses; the strain means for Tb ranged from 32 to 35°C, as previously reported (12), and the means for BW ranged from 61 to 82% of AL BW. There was also no correlation between the BW response to DR and our previously reported measures of torpor duration and the standard deviation of DR Tb (which gives added weight to the magnitude of the Tb drop) (13), with R = 0.05 and 0.19, respectively (Ps ≥ 0.2). Torpor duration and standard deviation of DR Tb are highly correlated with each other and the mean Tb response to DR, with Rs = 0.9. They were also less robust than mean Tb for determining the strain variation in the Tb response to DR (13); therefore, we have not examined these measures further.
The absence of correlation between the Tb and BW responses to DR in the LSXSS was confirmed in a replicate cohort studied 2 yr later (including two additional strains) with R = −0.04 (P = 0.43, 1-tailed) (Fig. 3B). Again, the mean Tbs ranged from 32 to 35°C (13), and the mean BWs in response to DR ranged from 63 to 84% of AL BW. There was also no correlation when we compared the Tb and BW responses to DR of these mice over the next 13 wk of DR (weeks 18–30) with R = 0.05 and P = 0.82 (2-tailed P value because a negative correlation was expected) (Fig. 3C). During this time, there was essentially no change in the means of the Tb and BW responses from the previous time period.
We also tested for a negative correlation between the responses of Tb and BW to DR among the six classical inbreds 129S6, A, BALB/c, C3H, C57BL/6 (B6), and DBA (IBG substrains except 129S6), which also exhibited significant strain variation in their Tb response to DR (P = 0.003) (14). The result was a positive correlation of 0.52 that was not statistically significant (P = 0.29, 2-tailed P value because a negative correlation was expected) (Fig. 3D).
We then asked whether any of the quantitative trait loci (QTLs) that we previously mapped in the LSXSS for the Tb and BW responses to DR were having a pleiotropic effect on both responses (13–15). For example, we previously mapped a statistically significant QTL affecting the Tb response to DR on chromosome 9 in the LSXSS RIs and a provisional QTL on chromosome 17 (13). Neither locus, however, was even suggestive (single-marker P ≤ 0.05) for an effect on the response of BW to DR, with peak logarithm of the odds of linkage (LOD) ≤ 0.35 and single-marker Ps > 0.2 (weeks 7–16 and 17–30 averaged). Likewise, none of the six suggestive QTLs that we mapped in the LSXSS for the BW response (15) was suggestive for the Tb response (peak LODS < 0.3, single-marker Ps > 0.25). The same search for pleiotropic QTLs in the LXS was also negative; none of the eleven suggestive QTLs that we identified for the BW response, including a provisional epistatic interaction (15), was suggestive for an effect on the Tb response to DR.
We also tested for a nongenetic (i.e., within strain) negative correlation between the Tb and BW responses to DR by pooling the data from all mice (n = 387) after removing the strain and cohort effects by univariate analysis (see methods). The correlation was slightly positive (R = 0.07), having a two-tailed P of 0.18 (Fig. 4).
We also tested whether the strain variation in the Tb response to DR might be explained by variables that affect thermal conductivity, such as hair length and absolute BW (a proxy for surface area:volume). Although there are a number of other variables that also affect thermal conductivity; we tested for the effects of these two because we happened to have the data for other reasons (measuring hair growth rates and weight loss). The small amount of strain variation in hair length (range 6.0 to 8.1 mm), based on measures collected from 42 LXS strains, was not a significant correlate of the Tb response to DR (R = 0.04, P = 0.4). However, as previously reported for the LSXSS, the absolute BWs of the DR mice were correlated with the Tb response to DR, even though the correlation only explained ∼3% of the variance (R = 0.18, P = 0.05) and appeared only after the mice have been on DR for several months (10). Nevertheless, the correlation was also significant and considerably higher (R = 0.38, P = 0.003) in the LXS (assessed 2–4 mo after starting DR; Table 1), explaining ∼15% (R2) of the variance in the Tb response to DR.
The results demonstrate that the Tb and BW responses to DR are being modulated independently. A lack of significant correlation was seen in both RI panels, both of which exhibit marked variation in Tb and BW in response to DR. That the correlation was essentially zero in both panels was unexpected, given that the lowering of Tb during DR is known to be a metabolic response to conserve energy expenditure and preserve BW. Our major conclusion, therefore, is that Tb and BW in response to DR are not genetically coupled in these strains, which was further supported by finding that none of the suggestive QTLs for the Tb or BW response appeared to overlap. Uncoupling of DR responses by examining genetic variation is a clear advantage of a comparative approach, as opposed to studying changes entirely within strain or other fixed genotype.
The absence of correlation between the Tb and BW responses to DR also suggests mechanistic information regarding the nature of the strain variation in the Tb response. Long-term differences in the BW response to DR at the same level of energy input (carefully controlled to be 60% of AL for all strains, see methods) cannot occur without differences in energy expenditure. All forms of energy expenditure produce heat as a by-product with 100% efficiency; thus, these differences in energy expenditure also represent differences in heat production (16). It is also well known that there is a direct relationship between heat production (burning calories) and weight loss. Therefore, the marked strain variation in the Tb response to DR is clearly not reflecting these differences in heat production affecting weight loss. Given that the only other determinant of Tb besides heat production is heat loss, it follows that heat loss is probably the major determinant of the strain variation in the Tb response (see also additional discussion online for mathematical analyses in the supplemental data).
One mechanism affecting heat loss is body size. We found that a proxy, absolute diet-restricted BW during the temperature trial was a significant, positive correlate of the Tb response to DR in both RI panels (a result that also argues empirically against a lack of statistical power to detect a correlation between the Tb and BW responses to DR). Interestingly, these strain differences in absolute diet-restricted BW are due in part to the strain variation in the BW response to DR; strains losing less weight tend to be bigger during DR than strains losing more weight, independent of the baseline strain differences in AL BW, as indicated by a significant increase in R2 of 0.20 in the LXS for the BW response to DR for explaining differences in absolute diet-restricted BW beyond the correlation with the AL strain means for BW (P = 8 × 10−6, stepwise multiple regression). Paradoxically, therefore, lower heat production, by virtue of maintaining a larger body size (and perhaps by maintaining more insulating fat as well), decreases the rate of heat loss, thereby having opposing effects on the Tb response to DR (though it still conserves energy). This opposing effect might explain why the strain variation in heat production during DR appeared to contribute little, if anything, to the strain variation in the Tb response to DR.
Besides body size, there are multiple physiological and behavioral factors that can alter the rate of heat loss during DR, including hair length (not a significant correlate in this study as shown), piloerection, posture (curling up), and nesting. However, the major mechanisms cited as minimizing heat loss during torpor are a decrease in pulmonary ventilation and an increase in peripheral vascular constriction (5, 19). Of these two possibilities, it seems more likely that vasoconstriction might continue after rewarming and thus explain why the strain variation in the Tb response persists even after the grand mean for Tb across strains had returned to 37°C (and anecdotally, all strains seemed to have returned to their normal level of motor activity) (13).
The absence of correlation between the Tb and BW responses to DR also argues against a direct effect of temperature differences per se on metabolic rate (5, 19) (see also supplemental data for a quantitative analysis of the expected Q10 effect). This strain variation in the Tb response to DR may thus be especially useful for understanding the life span extension recently reported by Conti et al. (1), in which a mouse engineered to overexpress a mitochondrial uncoupling protein in the region of the hypothalamus controlling Tb had a 0.3°C-lower mean Tb and lived about 15% longer. This mouse was also more resistant to weight loss during short-term food restriction, suggesting that the effects of a lower Tb and a lower rate of energy expenditure are not uncoupled in this system.
In conclusion, our results demonstrate that murine strain variation offers a strategy for determining genetic interactions among physiological responses to DR, making it feasible to narrow down and perhaps pinpoint which physiological, cellular, or biochemical responses to DR are correlated with DR's various health benefits and extension of life span, as well as pinpoint QTLs for these responses. For this reason, genetic mapping on each response is also vitally important, not simply as a prelude to positional cloning, but for establishing that any correlations of interest have a causal molecular basis (i.e., a genetic locus with pleiotropic effects).
Funding was provided by the Ellison Medical Foundation and the National Institutes of Health (R01 AG024354).
We're grateful to Colin Larson, Christine Martin, and Jey Yerg for excellent animal care.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2007 the American Physiological Society