Conclusions from a behavioral aging study on male and female F2 hybrid mice on age-related behavior, buoyancy in water-based tests, and an ethical method to assess lifespan

An estimation of C57BL/6N×BALB/c F2 natural lifespan

C57BL/6N×BALB/c F2 female and male hybrid mice were followed in a lifespan study, strictly following Swedish animal ethics regulations; if an animal displayed signs of pain or severe disease the animal was euthanized. The fate of the animals in the study is displayed in Table 1. The three shortest-lived and longest-lived animals are included as an indication of onset of death and maximum lifespan. As depicted, 15% of the males and 25% of the females were found dead in the cage without previous signs of pain or severe disease and thus died from intrinsic causes, generally defined as natural deaths in survival analyses. No autopsies were made but brief descriptions of extraordinary physical features were noted (e.g. enlarged spleen and various tumors; Supplementary Table 1).

For lifespan analysis, two survival curves per sex were made. In Survival Curve A, euthanization was counted as natural death, while in Survival Curve B euthanized animals were considered healthy when removed from the study. Since animals of severe disease are likely to live longer than the onset of disease, but not as long as healthy littermates, Survival Curve A is an underestimation of natural lifespan and Survival Curve B is an overestimation. Taking both curves into account provides an interval of natural lifespan (Figure 1). Data from Survival Curves A and B are presented in Table 2. There were no statistically significant differences when comparing lifespan of male and female F2 hybrids, though trends of female longevity over male were observed on mean and median lifespan, as well as on the age of the three longest-lived animals (Table 1).

Figure 1. Survival of C57BL/6N×BALB/c F2 hybrid male and female mice presented as intervals of natural lifespan. The lower limit of the lifespans, referred to as “Survival Curve A” in Table 2, was obtained by considering euthanization of animals as the same fate as natural death and the upper limit of the lifespans, referred to as “Survival Curve B” in Table 2, by considering euthanized animals as healthy upon removal (censored) and only animals that died of intrinsic causes were counted as deaths. Animals in the cohorts for behavioral assessment were included in the survival analysis until the time of their first test (7, 15, or 22 months of age). Timepoints are indicated by dotted vertical lines. Total number of animals: n_males=48, n_females=51.

Physiological analyses demonstrated that females have less lean mass and more body fat than males

Cohorts of mice were analyzed for behavioral phenotyping at 7, 15, and 22 months of age representing mature adults, middle-aged, and old. The mice were naïve before behavioral testing at all timepoints. As shown in Table 2, no survival loss had yet occurred at the 7-month timepoint and the mice were considered healthy adults. At the 15-month timepoint, survival was around 90% and for the final behavioral analyses at 22 months, 71% of the male and 74% of the female mice remained.

Physiological parameters of the mice in the cohorts were assessed to validate for general health status and to identify factors that may influence behavioral analyses. As shown in Figure 2A, average body weight did not differ significantly between 7, 15, and 22 months of age or between sexes. Despite similar body weights, there were clear changes in body composition. DEXA analyses demonstrated that males had higher percent lean mass (Figure 2B; p₇=0.0078, p₁₅=0.0011) and less percent fat mass (Figure 2C; p₇=0.0023, p₁₅=0.0012), as compared to females. Percent lean mass also increased with age (Figure 2B; p_M7-22=0.040, p_M15-22=0.036, p_F15-22=0.0098). At 15 months of age, the most distinct sex differences observed were an average of 11.8% more fat mass and 10.5% less lean mass in females as compared to males. Females also continuously displayed a higher core body temperature, as compared to males (Supplementary Table 2; p₇=<0.0001, p₁₅=0.0044, p₂₂=0.00013). Neither bone mineral content (BMC) nor density (BMD) declined for either sex at the last timepoint (Supplementary Table 2).

Figure 2. Physiological parameters of mice in the cohorts for behavioral assessment. (A) There was no difference in body weight between the cohorts, but (B) DEXA analysis showed that (p_M7-F7=0.0078, p_M15-F15=0.0011; and lean mass increased with age in males p_M7-22=0.040, p_M15-22=0.036; Mann Whitney), whereas (C) females exhibited increased fat mass as compared to males (p_M7-F7=0.0023, p_M15-F15=0.0012; Welch’s t-test). Values are mean ± SEM; n_M=7, n_M15=10, n_M22=10, n_F7=9, n_F15=12, n_F22=7.

Exploratory behavior, but not necessarily activity, decreased with age

The activity box, an open-field test, was used to analyze horizontal and rearing activities of naïve hybrid female and male mice. A mouse standing on its hind limbs, termed rearing, is considered both exploratory and general vertical locomotor activity [46]. Data were separated into the first 5 minutes, representing the initial exploratory phase of the test, and the full one-hour to enable analysis of general activity and behavior in a more acquainted environment, also known as intrasessional (within session) activity. An age-effect was detected during the exploratory phase of the open-field session, with horizontal and rearing activity clearly declining from 7 to 22 months in both sexes (Figure 3A; horizontal p_M7-22=0.0065, p_F7-22=0.0006; rearing p_M7-22=0.0003, p_F7-22=0.0020). The one-hour test results demonstrated that horizontal activity did not change comparing the 7-month and the 22-month age groups of both males and females (Figure 3B; the activity was maintained through 15 months of age in males, while females exhibit an activity drop that may be caused by different handling of this particular group due to technical reasons, see Methods). Moreover, horizontal activity was found to decline within session as time elapsed in the open-field and the novelty of the environment decreased (Supplementary Figure 1A). Rearing activity over the full hour also declined with age in both males and females (Figure 3B; p_F7-22=0.043, p_M7-22=0.043). In contrast to horizontal activity, no differences in rearing within session were seen, i.e. no intrasessional changes (Supplementary Figure 1B). Thus, activity in the explorative phase was more significant and uniform in exhibiting an age-related decline than the activity during the longer sessions. Taken together, diverse exploratory behaviors decreased consistently with age in both sexes.

Figure 3. Horizontal and rearing activity per minute in an activity box open field test. (A) Activity in the first 5 minutes – the exploratory phase – decreased with aging for both males and females. The decline in horizontal (p_M7-22=0.0065, p_F7-22=0.0006; Student’s t-test) and rearing activities (p_M7-22=0.0003, p_F7-22=0.068; Mann-Whitney) indicated that exploratory behavior was highly affected by age. (B) Activity in 1-hour sessions of the same test. Horizontal activity was not altered from 7 to 22 months of age, whilst rearing activity declined for both female and male hybrids, thus illustrating an aging effect (p_F7-22=0.043, p_M7-22=0.043; Mann-Whitney). The decrease in horizontal activity observed in 15-month-old females was likely caused by different handling of this particular group due to technical reasons (see Methods). Values are mean ± SEM; n_M7=7, n_M15=10, n_M22=10, n_F7=10, n_F15=12, n_F22=7.

Hybrid mice exhibited unaltered learning and memory with aging in the shuttle box passive avoidance test

The shuttle box passive avoidance test is a two-day procedure to assay hippocampal and amygdala-dependent learning and memory through pain conditioning [25, 30, 47, 48]. On training day, mice were allowed to enter a dark compartment, which subsequently resulted in a small electric shock. On the following day, the difference in time to re-enter the dark compartment compared to the previous day was recorded as a measure of recollection of the unpleasant experience. In order to clearly illustrate the memory response, results are presented as time to enter on Day 2 subtracted by time to enter on Day 1. Although 7-month-old males tended to keep away from the avoidance-trained area the longest, indicating the best learning and memory of all groups, large individual variations were observed within the cohorts and no statistical significances were found. Thus, we unexpectedly found no trends of declining learning and memory with age in neither sex (Figure 4).

Figure 4. Learning and memory assessed by shuttle box passive avoidance test. Data are presented as the time it took to enter the conditioned area on the testing day (Day 2) subtracted by the time to enter on the conditioning day (Day 1), calculated for each individual animal, and indicate no major age or sex difference (p_M7-F7=0.055, p_M15-F15=0.97, p_M22-F22=0.74, p_M7-15=0.11, p_M7-22=0.23, p_M15-22=0.19, p_F7-15=0.62, p_F7-22=0.30, p_F15-22=0.60; Mann-Whitney). Values are mean ± SEM; n_M7=7, n_M15=10, n_M22=10, n_F7=9, n_F15=12, n_F22=7.

Increased immobility of females in forced swim test which is partially due to higher fat mass

Immobility in the forced swim test is an indicator of depressive-like behavior. We did not find any age-related differences in forced swim test performance in males or females (Figure 5A; p=0.244; 2-way ANOVA)); however, when comparing males and females we observed a clear effect of sex on immobility (p=0.001; 2-way ANOVA). Since there was little evidence for a sex by age interaction (p=0.267; 2-way ANOVA), the three age groups were combined in subsequent analyses. Comparing pooled age groups demonstrated that females had around 21% increased immobility compared to males (Figure 5B; p=0.0002; Mann-Whitney).

Figure 5. Forced swim test to assay depressive-like behavior. (A) Immobility in the forced swim test is considered an indicator of depressive-like behavior, and F2 hybrid females were found to be more immobile (p=0.0013, 2-way ANOVA), as compared to males. Values are mean ± SEM. (B) Due to no age effect found, all timepoints for each sex were pooled to demonstrate a 20% difference in immobility between females and males (p=0.0002, Mann-Whitney). Values are mean ± SEM. (C) Immobility plotted against fat mass for each animal. Females exhibited increased fat mass as compared to males, and immobility was found to correlate to percent fat mass. The correlation slopes were constrained to be equal since they were not significantly different (p=0.58, ANCOVA). (D) Bayesian mediation analysis separated the total effect of the difference between males and females in the forced swim test into effect mediated by fat mass or by unknown mechanisms. The sex difference in fat mass accounts for 30% of the sex difference in immobility (Bayesian p=0.96) and other mechanisms account for the remaining 70% (Bayesian p=0.99; Bayesian mediation analysis). Values are mean ±50% (thick lines) and 95% confidence interval (thin lines); n_M7=7, n_M15=10, n_M22=10, n_F7=9, n_F15=12, n_F22=7.

We controlled whether body weight affected performance in all our behavioral tests but could not find any correlations. However, with females having a greater proportion of body weight comprised of fat (Figure 2C) they are likely to have increased buoyancy which may affect their performance in the forced swim test. Indeed, for both males and females, animals with a higher percent fat mass were less active (Figure 5C). We therefore used a Bayesian mediation analysis to test if, and to what extent, sex differences in the forced swim test were mediated by fat mass, and/or via other mechanisms [49, 50]. We found that fat mass accounted for about 30% of the sex effect (Figure 5D; Bayesian p=0.96; a Bayesian p-value close to 1.0 indicates that the conclusion is likely), while the remaining 70% was caused by other mechanisms (Figure 5D; Bayesian p=0.99). The differences in fat mass and immobility between males and females were most distinct at 7 and 15 months of age, despite no significant age, or age to sex interaction effect. As forced swim tests are generally not performed on aged animals, we performed the same Bayesian mediation analysis on the data set excluding the 22-month-old cohort, and found that 46% of the sex effect is caused by fat mass in this data set (Supplementary Figure 2; Bayesian p=0.986), while the difference in immobility between males and females was in the same range (24% compared to 21%). Hence, we can conclude that fat mass accounted for 30%-46% of the sex effect in the forced swim test, depending on the age groups examined.

Animals and diets

Subjects were male and female C57BL/6N×BALB/c F2 hybrid mice crossed from C57BL/6N×BALB/c F1 hybrids, with C57BL/6N fathers (Charles River, Lyon, France) and BALB/c mothers (Harlan Laboratories, Horst, the Netherlands), with coat color ranging from white to brown and black. Males were cohoused in groups of three until 6 months of age. In 11 cages with male mice, fighting led to severe injury and all animals in those cages were removed from the study. This was done in order to maintain a representative population, since high-ranked animals would otherwise be selected for. All males were single-housed for the remaining time from 6 months of age. Females were cohoused in groups of four, with the number of animals per cage decreasing as mice deceased in the lifespan study. The mice were not pooled. Cages were equipped with nesting material (paper), cardboard houses, and wooden sticks, and were cleaned every two weeks for single housed males and every week for cohoused mice. Old nesting material was transferred to the new cages upon cleaning. The mice had ad libitum access to regular chow diet (R3; Lactamin, Kimstad Sweden) containing 12% fat, 62% carbohydrates, and 26% protein (energy percentage), with a total energy content of 3 Kcal/g. The mice were maintained on a 12:12 h light/dark cycle (gradual light increase from 5:30-6:00 am and decrease at 5:30–6:00 pm) at 21°C. All experimental protocols of the study were carried out in accordance with the ethical certificate approved by the Animal Ethics Committee in Gothenburg, Sweden (Permit Number: 164-2015) and with EU Directive 2010/63/EU for animal experiments. Causes of euthanization were symptoms indicative of severe ill-health. These include: hunched shoulders, shabby fur, inactivity, failure to eat or drink, enlarged organs and tumors. Upon termination, mice were euthanized by decapitation under 5% isoflurane anesthesia.

Study design of physiological and behavioral phenotypic profiling

The lifespan study consisted of 20 male and 22 female mice, which were not part of the behavioral testing cohorts. Littermates to these mice were tested for physiological and behavioral phenotypic profiling at three ages representing mature adults (7 months; 7 males and 10 females), middle-aged (15 months; 10 males and 12 females) and aged (22 months; 10 males and 7 females) [60]. Specifically, the mice’s ages at behavioral profiling were 6.6-7.8±0.2 (age at test period start - age at test period end ±SD of age variance in the cohort; 7 months), 14.5-15.6±0.1 (15 months), 21.8-22.5±0.2 (22 months). These mice were initially included in the lifespan analysis but taken out (marked as censored) at the testing timepoint, and thus were only included in the calculation of percent survival in the lifespan analysis until their removal. Results from the 7-month cohorts served as control in a previous study [61]. Test periods occurred at different times of the year without any attempt to control for seasonal effects (7-month-olds: September-October; 15-month-olds: May-June; 22-month-olds: late December - January). Prior to the testing period, all mice were handled and acclimated for one week. The sequence of analyses is provided in Supplementary Table 3. Due to technical and practical reasons in our animal facility, the order of the behavioral tests could not be identical for males and females for all timepoints. Behavioral tests that are known to be impacted by handling (such as activity box) were performed early in the testing period (except for the 15-month-old female activity box timepoint; conclusions from which are therefore omitted). No correlation between behavioral test results and coat color could be found. Tests that affect the mice to a greater extent were planned later in the testing periods and/or the animals were given an extended recovery time in order to not impact the subsequent analyses. All animals were acclimated to the behavioral testing room for at least one hour prior to testing, and all experiments started between 10-11 am with the time spent for each experiment noted for each individual test. One female in the 7-month timepoint was removed from testing due to an eye injury.

Body composition and core temperature measurements

Core body temperature, body weight, and body composition were obtained at the same time. Core body temperature was taken with a rectal probe thermometer (ELFA AB, Sweden) followed by dual energy X-ray absorptiometry (DEXA) scanning (Lunar PIXImus Densitometer, GE Medical Systems, Madison, WI, USA) while under 2.5% isoflurane sedation for approximately 4 minutes. The parameters recorded were: body length (cm), body fat (g), lean body mass (g), bone mineral density (BMD; g/cm2), and bone mineral content (BMC; g) [47].

Activity box

The activity box is an open-field activity-like test to assess general activity, exploratory behavior, and signs of anxiety [47]. The activity box records movement of each mouse three dimensionally using infrared sensors built into the walls (8Lx8Bx8H) of a sound-proof opaque box (50x50x50 cm) with a low intensity lamp over the lid (Kungsbacka mät och regler, Fjärrås, Sweden). The mice were placed in the middle of the box and their movements were recorded for a total of 1 hour. The first 5 minutes were used to assess exploratory behavior in this novel environment and the full hour was used to assess general activity and behavior. The parameters that were recorded as events every 5 minutes include: horizontal activity, peripheral activity, rearing activity, peripheral rearing, rearing time, locomotion, and corner time.

Forced swim test

The forced swim test is used to analyze mice for signs of depression [47, 62, 63]. A transparent plexiglas cylinder (25 cm inner-diameter, 60 cm length) was filled with room temperature (22 °C) water level with a grey circular plastic platform hanging from wires on the outside of the cylinder, approximately 20 cm from the top (bespoke construction, AstraZeneca, Gothenburg, Sweden). A single mouse was placed on the water surface and monitored for 6 minutes and 20 seconds by a video camera placed directly above the cylinder, with the last 4 minutes used to analyze behavior and activity (MouseTracker analysis software, Mölndal, Sweden).

Shuttle box passive avoidance test

The shuttle box passive avoidance test is used to study memory performance in mice [47, 64]. The shuttle box system (Accuscan Instruments Inc., Columbus, OH, USA) is made up of a cage centrally divided by a wall with a sliding door creating two compartments, one bright with transparent walls and one dark covered on all sides with opaque walls. Both compartments are equipped with sensors that determine the location of the mouse, and the floor of the cage is made of a stainless-steel grid that can deliver a mild electric shock. On the first day, each mouse was placed into the bright compartment, and after 60 seconds the central door opened allowing the mouse to migrate into the dark compartment, an environment which they should prefer. Upon entry into the dark compartment, immediately after the door closed, the mouse received a mild electric shock (0.3 mA) and remained in the dark compartment for at least 30 seconds thereafter. On the second day, the mouse was again placed into the bright compartment and after 60 seconds when the central door opened the mouse was given the same choice to enter into the dark compartment (300 seconds maximum time). The time of entry into the dark compartment was recorded for both days, and no entry or entry after a longer interval on the second day as compared to the first day was considered a memory response.

Statistical analysis

Comparisons between two groups were analyzed for normal distribution by Shapiro-Wilks and Levene’s test for homogeneity of variance. Two-tailed independent t-test (Student’s t-test) was used for groups that met these criteria, two-tailed independent unequal variance t-test (Welch’s test) for normal distributed groups which failed Levene’s test, and nonparametric Mann-Whitney U test for unevenly distributed groups. Comparisons of survival curves were analyzed with the log-rank Mantel-Cox test. Interactions and comparisons of age and sex effects were calculated with 2-way ANOVA. Covariance was analyzed with ANCOVA. Differences were considered significant at P < 0.05. A Bayesian mediation analysis was used to test whether sex differences on the forced swim test were due to difference in fat mass between sexes, or due to other mechanisms [described in 49, 50]. The Bayesian p-value reported in the results has a direct and intuitive interpretation: it is the probability of the conclusion given the data, so a high value means that the conclusion is highly probable. Statistics were calculated using IBM SPSS Statistics 25, GraphPad Prism 7, or R/Stan.