A freely eaten diet that extends lifespan like calorie restriction but works through opposite energy regulation

The composition of the diets is shown in Table S1 and the experimental design in Figure 1A. The Con and CR diets contained 18% protein, 67% carbohydrates, 15% fat (as a percentage of dietary energy content) and ~4% fibre (as a percentage of food weight). The mice on the CR diets received 20% less food than consumed by the mice on the Con diets with ad libitum access to food. The LPHC diet contained 6% protein, 79% carbohydrates and 15% fat and was diluted 30% with non‐digestible fibre (final dilution of ~25% from the Con diet). The protein in all diets was exome‐matched and supplemented with AIN93 trace minerals and vitamins. The diets were commenced at 12 weeks of age.

Survival curves (Figure 1B–D) show that the LPHC and CR diets had similar effects on lifespan compared with the ad libitum control diet. LPHC and CR diets significantly increased median lifespan by 17% and 11%, respectively, compared to the control diet. There was no statistically significant difference between median lifespans of the CR versus LPHC. Maximum lifespans were 1008 days for controls, 1179 days for CR, and 1115 days for LPHC diets. Sex was not a significant effect modifier.

Per protocol, food and energy intakes were reduced by 20% in the CR mice compared to the ad libitum‐fed control mice (Figure 1E,F). Although food intake was increased by 16%–17% with the LPHC diet compared to the control diet, this did not fully offset the dilution by cellulose (i.e., low energy density), and consequently, energy intake was decreased by 15%. This value compares with reported values ranging from 20%–50% reductions in energy intake using conventional CR regimes (Speakman et al. 2016; Swindell 2012). The increase in lifespan in mice on the LPHC diets was not secondary to periods of fasting (Pak et al. 2021) because the mice had ad libitum access to food.

After 9 months on diet (12 months of age), body weight, percentage body fat and lean mass were highest with the ad libitum‐fed Con diets (Figure 2A–C) and, overall, were higher in males than in females. In males, body fat was lower with LPHC than with CR diets. Blood glucose (Figure 2D) and insulin (Figure 2E) were reduced similarly by both the CR and LPHC diets compared to controls (Figure 2D). Consequently, the insulin sensitivity index was lower (i.e., improved) in mice fed LPHC diets and CR diets (Figure 2F). Although the oral GTT curves were lower in the CR and LPHC diets (Figure 2G,H), the AUC was only significantly different in the CR group (Figure 2I).

RER in the LPHC and Con groups did not show the same robust diurnal range as the CR (i.e., between 1, where carbohydrates are dominating metabolism, and 0.7, where fat is the predominant nutrient source; Figure 2J,K), likely because of the fasting associated with CR. However, the average 24 h RER was not different (Figure 2L). The reduction in energy intake in mice on the CR diet (M 23%, F 19%) approximated the reduction in energy expenditure (M 21%, F 14%) compared to mice on the Con diet. Likewise, the reduction in energy intake in the mice on the LPHC diet (M 15%, F 15%) matched the reduction in energy expenditure (M 13%, F 8%) compared to mice on the Con diet (Figure 2M–O). Of note, the reductions in energy intake and energy expenditure were greater with CR than with the LPHC diet, suggesting that the LPHC is not only acting via reduced energy intake.

To determine the age‐related effects on the metabolic responses to diet, a second cohort of mice was culled after 1, 6, 12, or 18 months on the diets (4, 9, 15 and 21 months of age, respectively). Body weight, percentage fat mass and lean mass (Figure 3A–C) increased in control mice with age. Percentage fat mass also increased with age in male CR mice, which has been reported with CR previously (Pak et al. 2021), although the mechanisms remain unexplained. Compared to the control and CR diets, weight and fat mass were relatively stable across age for animals on the LPHC diet.

Glucose and insulin homeostasis deteriorated with age in mice on the Con diet. These outcomes were stable in mice with CR and the LPHC diet (Figure 3D–G), so that at the older ages these were improved compared to the mice on the Con diet. Liver weight and liver triglyceride levels increased with age in the control mice and in males on the CR diet (Figure 3H,I). Histology of the livers was consistent with these results, showing hepatosteatosis in the control mice and male CR mice after 18 months on diets (Figure 3J). At no point for these metabolic, body composition, and liver outcomes did CR outperform the LPHC diets. Indeed, in males, the LPHC diet did better than CR for some parameters.

To determine whether the LPHC and CR dietary interventions act via similar or different cellular mechanisms, the liver proteome was studied in the cohorts of mice culled longitudinally (Figure 4). Multidimensional scaling plots (Figure 4A) indicated that sex, age and diet all influenced the proteome. Therefore, subsequent analyses determining the effects of diet accounted for interactions with sex and age. The Venn diagram (Figure 4B) showed that there were more differentially expressed proteins comparing LPHC to controls (1885 of 3233 proteins detected) than CR versus controls (1351 proteins). There were 863 proteins where the changes in abundances were common between LPHC and CR (Figure 4B; heatmap of common proteins Figure 4C; Top 20 proteins Table S2). There were 1022 proteins unique to LPHC (Top 20 proteins Table S3) and 488 proteins unique to CR (Top 20 proteins Table S4).

The main pathways common to both diets were mitochondrial and related to energy metabolism (Figure 4D). Further analyses of the common proteins shared by CR and LPHC accounted for the possibility that the direction of regulation of protein abundances may be opposite between the two diets. Indeed, these mitochondrial pathways were generally upregulated with the CR diets and down‐regulated with the LPHC diets. Examples of some proteins of interest include mitochondrial proteins (e.g., CPT2, AASS, NSUD4 and TMEM135), where the abundance with LPHC is reduced compared with CR and/or controls (Figure 4E).

Proteins with changes in abundance that were unique to CR included those associated with mitochondrial (upregulated), proteolysis, transport and catabolic (down‐regulated) pathways (Figure 5B,D). Proteins with changes in abundance that were unique to LPHC and upregulated included those involved with RNA metabolism and the spliceosome pathways (Figure 5C,E). This included SRSF1, a key signalling protein for the inverse relationship between cellular energy and the spliceosome (Ferrucci et al. 2022).

All experiments performed were approved by the University of Sydney Animal Ethics Committee (protocols 2018/1453 and 2019/1513).

Male and Female C57BL/6J_Ozarc mice were obtained from Ozgene (previously named Animal Resources Centre WA, Australia) at 4 weeks of age and housed at the Charles Perkins Centre's Biological Services Unit at the University of Sydney, Australia. All mice were housed three mice per cage, held at room temperature (21°C–23°C) under a 12 h diurnal light/dark cycle, and were not exercised. Animals were maintained on standard chow ad libitum until commencing experimental diets. Cages used cellulose bedding and were changed fortnightly to minimise negative impacts of high frequency changes on cage microenvironments and animal health (Reeb‐Whitaker et al. 2001).

Animal experiments were split into two parts: a Timepoint (n = 9 sex/diet/timepoint) and a Longevity study (n = 24 sex/diet/timepoint). At 12 weeks of age, mice from both studies were randomly assigned to either a control (Con), a low‐protein, high‐carbohydrate, low‐energy density diet (LPHC), or a 20% calorie‐restricted (20% CR) diet. Dietary compositions are shown in Table S1 and were purchased from Specialty Feeds Ltd. (Perth, Australia). All diets were provided ad libitum, except for the CR group, where animals were fed daily (~3 pm) with a pre‐weighed portion of the control diet, corresponding to 80% of the measured food intake of the control mice.

Body weights were measured fortnightly, and food intake (by cage) was measured monthly. Animals in the timepoint study that reached the ethical endpoint before their scheduled experimental endpoint (8 out of 216 animals in the timepoint study) were euthanised and excluded from the study. Ethical endpoints reached within the longevity study were maintained as data points for lifespan analysis, the exception being mice with a rectal prolapse occurring without a secondary ailment (as per Pak et al. 2021); 10 from 143 animals in the longevity study, as mice are prone to developing this ailment (Pettan‐Brewer and Treuting 2011). These animals were censored as the date of euthanasia and are indicated as vertical dashes in survival curves in Figure 1B–D.

One week prior to the experimental endpoint (1, 6, 12 or 18 months on diet), mice in the Timepoint study were measured for lean tissue, fat mass and fluid by magnetic resonance imaging using an EchoMRI 900 (EchoMRI, TX, USA). Longevity Study animals were assessed for body composition at 9 months on diet only.

Mice (all Timepoint mice and subset of Longevity study) were fasted for 5 h, and basal blood glucose measurements were taken using a handheld glucometer (Accu‐Chek Performa, Roche Diagnostics Australia Pty Ltd) from the tail tip. Animals were gavaged with 25% glucose solution (2 g/kg lean mass), and blood glucose measurements were taken at 15, 30, 45, 60 and 90 min thereafter.

An additional 10 μL blood was taken from the tail at baseline to determine insulin levels using an enzyme‐linked immunosorbent assay (Crystal Chem, USA). The insulin sensitivity index was estimated using the product of fasting glucose and insulin measures.

At 9 months on diet, a subset of Longevity Study animals underwent indirect calorimetry using the Promethion high‐definition respirometry system (Sable Systems, Las Vegas, NV, USA). Mice were singly housed in Promethion cages for 48 h (including 24 h acclimation), and calorimetric data and physical activity were measured and calculated using the software (MetaScreen). Oxygen consumption (VO₂) and carbon dioxide production (VCO₂) were measured to determine the respiratory exchange ratio (RER = VCO₂/VO₂).

Animals designated for the Timepoint Study were culled at their assigned experimental endpoints of 1, 6, 12 or 18 months on diet (corresponding to 4, 9, 15 and 21 months of age) for tissue harvesting and analysis. Mice were anaesthetised with Lethabarb (60 mg/mL pentobarbitone sodium) by IP injection and killed by exsanguination via cardiac puncture while anaesthetised. Tissue collection was performed for all groups between 09:00 and 13:00. Caloric‐restricted mice were not fed on the day of sacrifice. The mouse cull order was randomised to prevent time bias.

Blood from the cardiac puncture was placed into a 1.5 mL microcentrifuge tube with EDTA, spun at 10,000 rpm, and plasma transferred to another tube before being stored at −80°C. After the liver was dissected out and weighed, a small portion was fixed in 10% neutral buffered formalin for 24 h and transferred to 70% ethanol before being paraffin‐embedded (Garvan Institute Histopathology Facility). The rest of the liver was snap frozen and stored at −80°C.

Liver samples from the Timepoint Study were analysed for triglyceride content following a simple lipid extraction protocol (Folch et al. 1957). 20–30 mg of powdered tissue was submerged in 4 mL methanol:chloroform (1:2) mixture and vortexed well. Lipids were extracted by incubating overnight on a rotary tube spinner at 4°C. The next day, 2 mL of 0.6% NaCl was added, samples vortexed and then centrifuged at room temperature (2000 rpm, 10 min). The lower phase was transferred to a glass tube and dried down with a nitrogen evaporator until completely dry. The dried sample was then resuspended in 500 μL ethyl alcohol (Sigma Aldrich).

Triglyceride content was quantified against a glycerol standard (Precimat). Samples and standards (10 μL) were added to a 96‐well plate, dried down at 37°C, and 300 μL Triglyceride reagent (Thermo Fisher Scientific) was added to each well. Samples were incubated at 37°C (30 min), and absorbance was read at 490 nm.

30–50 mg of frozen powdered tissue, from the liver of all Timepoint Study mice, was weighed out into a 2 mL microfuge tube and lysed in 1 mL of SDC buffer (4% Sodium Deoxycholate, 100 mM TrisHCL, pH 8.0) at 95°C for 10 min on a thermomixer at 1000 rpm, sonicated in a water bath sonicator for 30 min at room temperature and then centrifuged at 18,000 g for 10 min at room temperature. The clarified lysate was transferred to a new tube and stored at −30°C. 5 μL of reduction/alkylation mix (50 mM TCEP, 200 mM chloroacetamide in SDC buffer) was added to 20 μL of clarified tissue lysate in a 1 mL 96‐deep well plate and incubated on a thermomixer at 95°C at 1000 rpm for 10 min, and then 75 μL of water was added. Following the addition of 1 μg trypsin, proteins were digested for 16 h at 1000 rpm and 37°C.

Peptides were cleaned using SDB‐RPS stage tips mounted in a custom 96‐well plate centrifuge rack as described (Harney et al. 2019). Peptides were eluted in 100 μL of 12.5 mM NH₄OH, 80% acetonitrile, dried down in a vacuum centrifuge and resuspended in 40–60 μL of 5% formic acid. Peptides in 5% (v/v) formic acid were directly injected (injection volume 50 μL) onto a 2.1 × 50 mm, 1.9 μm, InfinityLab Poroshell 120 EC‐C18 analytical column (Agilent). Peptides were resolved over a gradient from 9% acetonitrile to 29% acetonitrile over 13.8 min and then from 29% to 42% over 0.6 min with a flow rate of 0.8 mL/min and then 80% acetonitrile for 2 min at 40°C. The Sciex TripleTOF 6600 system was operated using the TurboSpray ion source in scanning SWATH mode with positive polarity. The scanning SWATH DIA method consisted of 51 variable‐width SWATH DIA windows that spanned the Q1 mass range 390–900 Da.

Raw data were processed in DIANN (v 1.81) (Demichev et al. 2020; Messner et al. 2021) with an in silico generated spectral library, standard settings with "high accuracy", "robust LC" and "MBR" activated. Peptides were searched against a reference mouse proteome (UP000000589) downloaded from www.uniprot.org↗ on 01/11/2022. The resulting unique gene table was read into R (version 4.3.1), and samples with high numbers of missing values were removed (> 40%). Intensity values were log transformed (base 2). To deal with missing values, proteins were sorted into missing at random (MAR) and missing not at random (MNAR) on the basis of whether there was a difference in the percentage (> 30%) of missing values within the main contrasts of sex (M and F) and age (1 month and 18 months). Proteins were filtered out that were present in fewer than 70% of samples unless determined to be MNAR for sex and/or age. Imputation of missing values from proteins determined as MNAR was performed using a left‐censored imputation method (QRILC) from the QFeatures package (https://www.bioconductor.org/packages/release/bioc/html/QFeatures.html↗). Data were analysed using limma (Ritchie et al. 2015) with the following model (model.matrix(~diet + timepoint + sex + batch)) and settings (eBayes: trend = T, robust = T). p values were corrected for multiple comparisons using the standard Benjamini‐Hochberg correction in limma (FDR). MDS plots were generated using the plotMDS() function in limma. Protein overrepresentation analysis was performed using the clusterProfiler package (Yu et al. 2012) on a background of all detected proteins.

The number indicates individual mice, or individual cages, for food intake measurements. Data are expressed as mean ± SEM unless otherwise stated in the figure legend. Apart from proteomic data, as described above, results were analysed by a two‐way or three‐way ANOVA. If the ANOVA reached statistical significance (p < 0.05), a Tukey's post hoc test was used. Statistical analysis was performed using RStudio (Version 2024.04.2 + 764). Statistical significance was set at p < 0.05.