Reevaluation of the effect of dietary restriction on different recombinant inbred lines of male and female mice

Abstract Dietary restriction (DR) was reported to either have no effect or reduce the lifespan of the majority of the 41‐recombinant inbred (RI) lines studied by Liao et al. (Aging Cell, 2010, 9, 92). In an appropriately power longevity study (n > 30 mice/group), we measured the lifespan of the four RI lines (115‐RI, 97‐RI, 98‐RI, and 107‐RI) that were reported to have the greatest decrease in lifespan when fed 40% DR. DR increased the median lifespan of female RI‐115, 97‐RI, and 107‐RI mice and male 115‐RI mice. DR had little effect (<4%) on the median lifespan of female and male 98‐RI mice and male 97‐RI mice and reduced the lifespan of male 107‐RI mice over 20%. While our study was unable to replicate the effect of DR on the lifespan of the RI mice (except male 107‐RI mice) reported by Liao et al. (Aging Cell, 2010, 9, 92), we found that the genotype of a mouse had a major impact on the effect of DR on lifespan, with the effect of DR ranging from a 50% increase to a 22% decrease in median lifespan. No correlation was observed between the changes in either body composition or glucose tolerance induced by DR and the changes observed in lifespan of the four RI lines of male and female mice. These four RI lines of mice give the research community a unique resource where investigators for the first time can study the anti‐aging mechanism of DR by comparing mice in which DR increases lifespan to mice where DR has either no effect or reduces lifespan.


| INTRODUC TI ON
The first and the most studied manipulation shown to increase lifespan in mammals is dietary restriction (DR). The classic study by McCay et al. (1935) showed that one could increase the lifespan of rats by dramatically reducing their food consumption early in life.
Since this initial observation, numerous laboratories have confirmed these results and have shown that reducing food consumption 30% to 50% (without malnutrition) consistently increased the mean and maximum lifespan of both laboratory rats and mice (Masoro, 2005;Weindruch & Walford, 1988). The increase in lifespan by DR was found to be similar for laboratory rats and mice used in aging research and similar for females and males, that is, no pronounced sexual dimorphism was observed (Austad, 2017;Turturro et al., 1999), which is different than has been reported for other manipulations that the Intervention Testing Center has shown to increase the lifespan of mice. The exception is DBA2 mice where DR increased the lifespan of female mice twice as much as male mice.
The effect of DR on longevity is not limited to rodents as DR has been reported to increase the lifespan of a large number of diverse animal models in addition to rodents: invertebrates (Kapahi et al., 2017), dogs (Lawler et al., 2005), and non-human primates (Mattison et al., 2017;Pifferi et al., 2018). Because of the broad effect of DR on lifespan, it became accepted that the effect of DR on lifespan was universal, that is, it occurs in all organisms. However, the uni- Surprisingly, less than one-third of the RI lines showed a significant increase in lifespan as was expected. On the contrary, approximately one-third of the RI lines mice showed a decrease in lifespan on the DR diet and one-third showed no effect of DR on lifespan. These data were a surprise to the many in the research community because the data contradicted the prevailing view that DR was a universal, beneficial intervention with respect to lifespan and aging. However, a few previous studies, which had largely gone ignored, also reported that some mouse strains did not show an increase in lifespan when fed a DR diet, for example, male wild-caught mice (Harper et al., 2006) and male DBA/2 mice (Forster et al., 2003), although Turturro et al. (1999) showed DR increased life span of male DBA/2 mice. In addition, Mattison et al. (2012) reported that DR did not significantly increase the lifespan of rhesus monkeys.
One of the major limitations of the study by Liao et al. (2010) was the number of mice used to measure lifespan in each RI line of male and female mice, which was limited to 10 or less mice per group. In addition, a report from the same team was unable to replicate the effect of DR on the lifespan of many of the female RI lines . Therefore, the goal of this study was to determine the replicability of the lifespan data for the RI lines of mice when a larger number of mice (e.g., 30 to 45 mice/group) were used to assess the effect of DR on lifespan. Because we could only study a limited number of strains of mice using larger numbers of mice to measure lifespan, we focused our attention on those RI lines reported to show a decrease in lifespan for the following reasons. First, we felt that the data from the RI lines that showed no significant increase in lifespan could have simply resulted from the small number of mice studied, resulting in the inability to detect a significant difference in lifespan. Therefore, we felt it was more likely that the RI lines showing a decrease in lifespan would give us the best opportunity to identify RI lines that did not respond to DR. Second, we were interested in determining if DR actually resulted in a decrease in the lifespan of the RI lines because such an observation is rare, and in many cases where it has been observed, it has not been replicated. We describe below the effect of DR on the lifespans of male and female mice from four RI lines of mice: 115-RI, 107-RI, 98-RI, and 97-RI. Our data show that four out of the eight groups of mice studied showed a significant increase in lifespan with DR while the other four show either no significant effect of DR on lifespan or reduced lifespan.

| Lifespan analysis
One possible explanation for the contradictory data on the effect of DR on the lifespan of the RI mice could arise because the level of DR required to increase lifespan is genotype-dependent. In other words, it is possible that 40% restriction used by Liao et al. (2010) had a negative effect on lifespan of some of the RI lines. Therefore, a lower level of DR might increase the lifespan of the genotypes that did not respond or responded negatively to DR. This possibility is supported by two studies that showed lower levels of DR are effective at increasing the lifespan of rats (Richardson et al., 2016) and mice (Mitchell et al., 2016). To test this possibility, we first studied the effect of various levels of DR (10%, 20%, and 40%) on the RI line that Liao et al. (2010) reported DR to have the greatest negative effect on the lifespan of male and female mice, 115-RI mice, for example, DR (40%) reduced the mean survival of female and male 115-RI mice ~85% and ~70%, respectively. Figure 1 shows the lifespan curves we obtained from the female and male 115-RI mice fed ad libitum (AL) on the three levels of DR, and Table 1 gives the lifespan data and the statistical analysis of these data. It is apparent that the lifespan of the female 115-RI mice is much shorter than the male mice, for example, median lifespan is ~30% less for female mice compared with male mice. Liao et al. (2010) also reported a similar difference in the lifespan of male and female 115-RI mice. As can be seen from Figure 1a and Table 1, 40% DR significantly increased the lifespan of both female and male mice whether measured by the mixed effects Cox models or the parametric models with Gompertz distribution. However, DR had a much greater effect on the lifespan of the female 115-RI mice than male mice, for example, median survival was increased 50% for female mice compared with only 9% for males. DR (40%) also significantly increased both the median and mean survival of the female 115-RI mice; however, the increase in the median or mean survival of the male 115-RI mice was not statistically significant. The 90th percentile survival was increased by 35% and 12% for female and male mice, respectively, fed 40% DR. We also used the maximum lifespan test developed by Gao et al. (2008) to statistically test for significant differences in maximum lifespan by testing for differences in the upper tail of the distribution in the survival data. The maximum lifespan test did not quite reach statistical significance for the female 115-RI mice but was significant for the male 115-RI mice.
As can be seen from Figure 1 and Table 1, lower levels of DR had a smaller effect on lifespan than 40% DR such that in male 115-RI mice, 10% and 20% DR did not significantly increase any measure of lifespan. In contrast, the survival curves for 10%, 20%, and 40% DR show a graded effect of DR on lifespan of female 115-RI mice, that is, greater the level of DR the greater the increase in survival. A similar trend was observed when the lifespan data were presented as violin plots (Figure 1b). The lifespan of the female 115-RI mice was significantly increased by 20% DR as measured by either the mixed effects Cox models or the parametric models, resulting in a 16%, 27%, and in 24% increase in mean, median, and 90th percentile survival, respectively. However, the increase in mean and median survival was not significant and the maximum lifespan test was not significant. Although 10% DR increased both the median and 90th percentile survival of the female 115-RI mice by 12%, none of the measures of lifespan were significantly increased by 10% DR.
Because 10% and 20% DR did not show any evidence of a greater increase in lifespan of the 115-RI mice compared with 40% DR, we focused our effort on the effect of 40% DR on lifespan, which allowed us to study the effect of DR on three other RI lines: 97-RI, 98-RI, and 107-RI mice. Liao et al. (2010) reported that the mean survival of both female and male 97-RI mice was reduced over 50% by 40% DR. As Figure 2a and Table 1 show, we found that 40% DR significantly increased the lifespan of female 97-RI mice as measured by the parametric hazard analysis, as well as a significant (10%) increase in median survival. DR increased the 90th percentile survival 6%, and the maximum lifespan test was also significant. The violin plots in Figure S1 also show a shift toward the DR mice living longer.
In contrast, 40% DR had no significant effect on any measure of the lifespan of male R97-RI mice. However, as can be readily observed from the survival curves in Figure 2a or the violin plots in Figure S1, DR resulted in an increase in deaths in the first half of life in the male 97-RI mice; however, the survival was similar in the later-half of the lifespan.
We next studied 98-RI mice because Liao et al. (2010) reported that 40% DR reduced the mean survival of both female and male 98-RI mice over 40%. The survival curves and lifespan data for the female and male 98-RI mice in Figure 2b and Table 1 show that 40% DR had no significant effect on the lifespan of the female mice. On the contrary, we observed a statistically significant decrease in the lifespan of male 98-RI mice as measured by both the mixed effects Cox models and the parametric models. The decrease in mean, median lifespan, and 90th percentile of 10%, 2%, and 4% was small and not significant for mean and median lifespan. However, the maximum lifespan test showed a significant difference for the DR mice compared with AL mice. The violin plots in Figure S1 in the supplement also show that the distribution of the lifespan data is similar for AL and DR in both female and male 98-RI mice.
The 107-RI mice were the last RI line we studied. We selected these mice because Liao et al. (2010) reported that this RI line showed one of the greatest sex differences in the effect of 40% DR on lifespan. DR had no effect on the lifespan of female 107-RI mice F I G U R E 1 Lifespan of female and male 115-RI mice fed AL and DR. Panel a shows the Kaplan-Meier survival curves for mice fed AL (blue) and 10% (yellow), 20% (green), and 40% (red) DR. The number of mice in each group and the analysis of the survival data are given in Table 1. Panel b shows the violin plots for the distribution of the lifespans for the age at death for each of the mice in the four groups of 115-RI mice. The solid lines show the quartiles and the dashed line the median but reduced the mean survival of male 107-RI mice by 50%. The lifespan data in Figure 2c and Table 1 show that 40% DR increased the lifespan of female 107-RI mice as measured by either the mixed effects Cox models or the parametric models, resulting in a 24% increase in both mean and median survival and a 10% increase in 90th percentile survival. The increase in median survival was significant; however, either the change in mean survival or the maximum lifespan test was statistically significant. In contrast, 40% DR resulted in a statistically significant decrease in the lifespan of male 107-RI mice as measured by either the mixed effects Cox models or the parametric models, resulting in a 22%-23% decrease in the mean and median survival, which was statistically significant for both. A 13% decrease in the 90th percentile survival was observed, which was a statistically significant difference as measured by the maximum lifespan test. The violin plots in Figure S1 also show that DR shifted the distribution of lifespan of female 107-RI mice to a longer lifespan while DR shifted the distribution to a shorter lifespan in male mice.

| Analysis of body mass/composition and glucose tolerance
In their study with the RI mice, Liao et al. (2010) reported that the effect of 40% DR on lifespan was inversely correlated fat reduction, that is, mice showing the lowest reduction in fat when fed 40% DR were more likely to have extended lifespan. In comparing male and female C57BL/6 and DBA/2 mice fed 20% and 40% DR, Mitchell et al. (2016) also found that the mice that preserved their fat mass in response to DR showed the greatest increase in survival. Therefore, we measured the effect of DR on body and fat mass in the four RI lines of male and female mice to determine if changes in body mass or composition were correlated with the ability of DR to increase the lifespan of the mice. The data in Figure 3 show the body weights of the four RI lines of mice fed AL or 40% AL. As expected, all of the mice showed a decrease in body weight. When measuring body composition, we observed no significant change in the percent of lean body mass with DR in most of the RI lines at any age ( Figure S2). However, the changes in fat mass by DR varied in the four RI lines. Figure S4 in the supplement shows the fat mass in grams, and Figure 4 shows the percent fat for the four lines from ~2 to 18 months of age. DR. The number of mice in each group and the analysis of the survival data are given in Table 1 lifespan, we plotted the percent change in fat mass, body mass, and lean body mass induced by DR at 12 and 18 months of age versus the change in medium lifespan induced by DR. As shown in Figure 5, we found no significant correlation between the changes in fat mass and lifespan. Interestingly, the group (female 98-RI mice) that showed the least change (actually slight increase) in fat mass by DR showed no increase in lifespan by DR. We also observed no correlation between changes in body mass or lean body mass and lifespan.
One of the hallmarks of DR is improved glucose tolerance and insulin sensitivity, and these changes have been proposed to play a role in the life-extending action of DR (Bartke et al., 2001;Barzilai et al., 1998). Therefore, we compared the effect of 40% DR on glucose tolerance in the four RI lines of male and female mice. Because we previously showed that 40% DR can enhance glucose tolerance in C57BL/6 mice within 10 days after implementation of DR (Matyi et al., 2018), we measured glucose tolerance 30 and 90 days after implementing DR (e.g., at 2.5 and 4.5 months of age). Figure S6 shows the curves for the glucose tolerance tests, and Figure 6 shows the data when expressed as the area under the curve. Most of groups showed improved glucose tolerance; however, DR had no significant effect on glucose tolerance in 115-RI females, which showed the greatest increase in lifespan by DR. Glucose tolerance was significantly reduced at 4.5 months of age by DR in female 98-RI mice, which showed no increase in lifespan. Thus, we found no relationship between the impact of 40% DR on glucose tolerance and lifespan.  Williams et al. (2004) to analyze genetic variation in alcohol sensitivity (Bennett et al., 2006) and were derived from an eight-way cross of the inbred strains: A, AKR, BALB/c, C3H/2, C57BL, DBA/2, Is/Bi, and RIII. As described above, Liao et al.

| DISCUSS ION
(2010) observed that less than a third of the RI lines they studied showed a significant increase in lifespan when placed on 40%

DR. Of particular interest was the observation that approximately
one-third of the mice showed a shortened lifespan on 40% DR, which was unexpected. The study by Liao et al. (2010) has been the most extensive study to date on the effect of genotype on the life-extending action of DR because they used a large number of strains of inbred mice, which were genetically diverse because of the RI lines came from an eight-way cross. However, because of the large number of RI lines compared, the study suffered from the small number of mice they used to measure lifespan for each sex and each RI line.
As noted above, the goal of this study was to determine if the inability of certain RI lines of mice to respond to DR could be replicated when a larger number (30 to 45 mice/group) of mice were used, which would allow us to detect a 10% change in lifespan (Liang et al., 2003). Of the eight groups of mice (female and male mice of four RI lines) studied, we were able to replicate the observation of only one of the eight results previously reported by Liao et al. (2010).
This was for the male 107-RI mice, which showed that DR resulted in a 22% decreased in mean survival compared with the 50% decrease in mean survival reported by Liao et al. (2010). We did observe that We also observed that the effect of DR on lifespan was sexually dimorphic in two of the four RI lines studied. In 97-RI mice, DR increased (10%) the lifespan of female mice but had no significant effect on the lifespan of male mice. In the 107-RI line, the sex difference was major. DR increased the median survival of female mice 24% and reduced the median survival of male mice 22%. The sex differences in response to DR that we observed in the RI lines are quite different than Turturro et al. (1999)  Previous studies have suggested that the ability of mice to preserve their fat mass in response to DR was correlated with a greater increase in lifespan Mitchell et al., 2016) and that glucose and insulin sensitivity was important in the anti-aging action of DR (Bartke et al., 2001;Barzilai et al., 1998). Therefore, we measured the changes in body and fat mass and glucose tolerance induced by DR in the eight groups of mice. Table 2 summarizes our findings listing the eight groups of mice in order of the effect of 40% DR on their median survival and the effect of DR on body composition and glucose tolerance. As can be seen from Table 2, we were unable to show any consistent association between the effect of DR on any of these measures and the effect of DR on lifespan.
Thus, we observed a separation of the effects of DR two of the hallmarks of DR, adiposity and insulin sensitivity, and longevity.
In summary, our study was to a large extend unable to replicate the effect of DR on the lifespan of the four RI lines reported by Liao et al. (2010); therefore, the lifespan data in their study should be considered suspect because of the limited number of mice used to measure lifespan. However, our data support the general conclusion of their study that genotype has a significant impact on the response F I G U R E 6 Effect of 40% DR on the glucose tolerance of female and male RI mice. Data from the glucose tolerance curves in Figure S6 in the supplement are expressed as the area under the curve for mice fed AL (blue bars) or DR (red bars) at 3 and 9 months of age. The data are expressed as the mean ± SEM. The number of mice per group is shown above each bar: Panel a 115-RI mice, Panel b 97-RI mice, Panel c, 98-RI mice, and Panel d 107-RI mice. The data for each time point were analyzed as AL vs DR by two-tailed Students t test. Values where the DR mice (red bars) are significantly different from AL mice (blue bars) are shown by *p > 0.05, **p 0.01, and ***p > 0.001 of an animal to DR. While we observed half of the groups of mice we studied showed an increase in lifespan when fed DR, the other half either did not respond to DR or showed a decrease in lifespan.
These RI lines are potentially an important resource for investigators studying the anti-aging mechanism of DR because these strains of mice will allow investigators for the first time to compare mice in which DR increases lifespan to mice where DR has either no effect or reduces lifespan. These comparisons will give investigators a new approach to identifying pathways that are altered only in mice showing an increase in lifespan, which will give us a better understanding of mechanism that is involved in the anti-aging action of DR. Jin et al.
(2020) have shown the power of this approach when they compared the genome and metabolome of strains of Drosophila that varied widely in their lifespan in response to DR. They were able to "pinpoint" cellular pathways and three genes that governed the variation in lifespan by DR. One of these genes was CCHa2R, the Drosophila homolog of the human oxidation resistance 1 (OXR1) gene, which was only identified through this approach. It had not previously been implicated in aging or DR in Drosophila.

| Animals and lifespan analysis
We obtained the following four RI lines from The Jackson The food consumption by the AL group of each RI line and sex was measured every week until 6 months of age and then every month and the amount of NIH-31 diet given to the DR groups each day was adjusted accordingly, that is, the DR groups were fed 90%, 80%, or 60% of the food consumed by the AL mice for 10%, 20%, and 40% DR, respectively. Because food consumption was relatively constant after 12 months of age, we discontinued measuring food consumption in AL mice at 18 months of age and used the food consumption at 18 months of age as the basis of the food given to the DR mice after 18 months of age. We did not do a step-wise reduction in food given the mice; rather the mice were immediately put on 40% (or 10% and 20%) DR at 6 weeks of age to be consistent with the study by Liao et al. (2010). It should be noted that the DR diets were not fortified with vitamins or minerals, which was identical to the DR protocol used by Liao et al. (2010). The mice in the survival studies were allowed to live out their lifespan without any manipulations except for cage changes every other week and the daily feeding of the DR groups. The mice were housed 5 mice/cage initially and were maintained in their respective cages until they died resulting in less than 5 mice/cage as mice in the cage died. The mice were monitored daily, on weekends, and holidays for overall health and morbidity and allowed to die naturally unless they were either unable to move to obtain food/water, experience pain from the presence of large tumors, or exhibit a major loss of weight (20%) indicating they would die within 24 to 48 h.
The statistical analysis of the lifespan data was conducted by

| Body composition and glucose tolerance test
Body composition and glucose tolerance were conducted in a separate cohort of mice for longitudinal analysis that were maintained on AL and DR diet for each line and sex. Body composition of the AL and DR fed live mice was measured using nuclear magnetic resonance spectroscopy (NMR-Bruker minispec) at ~2.5, ~4.5, ~13.5, and ~16.5 months of age (30, 90, 360, and 500 days of DR, respectively). Body fat and lean body mass of the animals in each group were measured.
Glucose tolerance was determined on each strain and sex after an overnight fast of mice at ~2.5 and ~4.5 months of age (30 and 90 days of DR, respectively). Mice were weighed and injected intraperitoneal with 20% glucose (2 g/kg), and blood glucose levels, collected from tail, were measured over a 120-min period using a glucometer (Contour next EZ from Bayer). The area under curve (AUC) for each curve was determined and represented as AUC glucose (mmol × 120 min).

ACK N OWLED G M ENTS
The research was supported by the following NIH grants:

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
AU and AR were involved in the design and implementation of the study, data analysis, interpretation of the results, and manuscript writing. SM, KG, and MRB were involved in the maintenance of mice colonies and longitudinal body composition analysis. DA, KE, XC, and SD conducted the statistical analysis of the lifespan data.

DATA AVA I L A B I L I T Y S TAT E M E N T
The statistical analysis of the lifespan data was conducted by the