Alpha-Ketoglutarate, an Endogenous Metabolite, Extends Lifespan and Compresses Morbidity in Aging Mice

SUMMARY

Graphical Abstract

In Brief

INTRODUCTION

The decline in early life mortality since the 1950s has resulted in a dramatic demographic shift toward an aged population. Aging manifests as a decline in health, multiple organ dysfunction, increased vulnerability to chronic disease, and degraded quality of life. A variety of genetic and pharmacological interventions, identified mainly in non-vertebrate model organisms, extend lifespan (Friedman and Johnson, 1988; Harrison et al., 2009; Kenyon et al., 1993; Wood et al., 2004), but the relationship to healthspan, the disease-free and functional period of life, is unclear (Bansal et al., 2015; Hahm et al., 2015; Hansen and Kennedy, 2016).

The frailty concept was initially developed to quantify human vulnerability to adverse health outcomes (Fried et al., 2001). However, corresponding comprehensive indices in mice were seldom applied to aging studies (Whitehead et al., 2014). Recently, there is increasing awareness that frailty might be a critical tool to measure healthspan in longevity studies (Kane et al., 2016; Rockwood et al., 2017). To study health during an aging intervention, we performed a series of longitudinal, clinically relevant healthspan measurements utilizing the frailty index developed by Howlett and colleagues (Whitehead et al., 2014).

Here, we show that alpha-ketoglutarate (delivered in the form of a calcium salt, CaAKG), a key metabolite in the tricarboxylic acid (TCA) cycle that is reported to extend lifespan in both C. elegans (Chin et al., 2014) and Drosophila (Su et al., 2019), can significantly extend lifespan and healthspan in mice. AKG is involved in various fundamental processes, including central metabolism, collagen synthesis (Myllyharju, 2003), epigenetic regulation (Tsukada et al., 2006), and stem cell proliferation (TeSlaa et al., 2016). Due to its broad biological roles, AKG has been a subject of interest for researchers in various fields (Zdzisińska et al., 2017). We find that CaAKG supplementation promotes longer, healthier life associated with decreased levels of inflammatory cytokines in mice.

RESULTS

C57BL/6 mice were fed regular chow and then switched to a diet containing CaAKG at 540 days of age (Figure S1). We assessed two independent cohorts of mice, each consisting of 45 ± 2 females and 45 ± 2 males (total of 182 animals), allowing us to assess reproducibility for longevity effects of CaAKG. Dietary-supplemented 2% CaAKG (w/w) increases survival in both cohorts of female mice (Figures 1A and ​and1B).1B). In the first cohort of females, median lifespan and survival (age at 90^th percentile mortality) were significantly extended by 16.6% and 19.7% from inception of CaAKG feeding. In the second cohort of mice, lifespan and survival were significantly extended by 10.5% and 8% (Figures 1A and ​and1B;1B; Table S1). Although improved survival for males was not significant in either cohort (Figures 1D and ​and1E),1E), median lifespan was extended for 9.6% and 12.8% from inception of treatment (Table S1).

Open in a separate window

Figure 1.

AKG Extends Lifespan and Decreases Mortality

Post-treatment survival plots graphed for cohort 1 (n = 90) and cohort 2 (n = 93). Comparing control mice to those fed AKG in the diet starting at 18 months of age. Arrows indicate the start of the treatment.

(A–C) Female survival curves for cohort 1 (A) control (n = 24), AKG (n = 20) and cohort 2 (B) control (n = 23), AKG (n = 23), and (C) pooled females.

(D–F) Male survival curves for cohort 1 (D) control (n = 24), AKG (n = 22) and cohort 2 (E) control (n = 24), AKG (n = 23), and (F) pooled males. Survival curve comparisons were performed using Log-rank test; *p < 0.05. Maximum lifespan extensions were calculated using Fisher’s exact test statistics; **p = 0.0064 for female cohort 1 and *p < 0.021 for female cohort 2.

(G and I) Body weight trajectories from cohort 2 mice; n = all live animals in the study at each time point; data are mean ± SEM. Two-way ANOVA tests were applied to check if independent variables including time and treatment (AKG) affect male and female body weight. The comparison provided evidence of significant effects of both treatment and time on male body weight; ***p < 0.0001, ***p < 0.001, and effect of only time on female body weight ***p < 0.001.

(H and J) Male (H) and female (J) body composition; n = all live animals in the study; data are mean ± SEM; no significance changes detected (two-tailed t test).

(K) Food consumption was measured for 3 consecutive days in different times during the lifespan of female control (n = 6) and female AKG-fed (n = 5) mice; the arrows at 19, 23, and 28 months show the age at which food consumption was measured. Data are mean ± SEM; no significance change (two-tailed t test).

To assess healthspan, we applied measurements based on a clinically relevant frailty index (Parks et al., 2012; Whitehead et al., 2014), which consists of 31 phenotypes that are indicators of age-associated health deterioration with each phenotype scored in a blinded manner on a 0, 0.5, or 1 scale, based on severity (Figure S1). Measurements were repeated approximately every 8 weeks, providing eight and seven sets of data, respectively, for male and female groups. A baseline health assessment was assessed at the start of treatment at 18 months. The total frailty score, which is the average of 31 frailty phenotypes, was calculated for each animal at every time point (Figures 2A–2F) and indicates the state of increased vulnerability to adverse health outcomes, comparable to morbidity.

Open in a separate window

Figure 2.

Compression of Morbidity by AKG Treatment

(A and B) Separately graphed (A) female and (B) male total frailty index (FI) scores during lifespan, comparing control mice (blue) to those fed AKG in the diet (pink) starting at the 18th month. Each dot is the total score of one animal at specific age as indicated. Data are mean ± SEM of each group. n = all animals alive at each measurement time. Mixed model was used to analyze the data longitudinally. In the current mixed model, every mouse with a unique ID was treated as a random effect. Treatment and time were treated as fixed effects. The low p value for chi-square for significant matching effect indicates that the pairing was effective; comparing the fit of the current mixed model to a simpler ANOVA, ****p < 0.0001. As the animal ages and gets closer to death (higher percentage of lifespan), it manifests several aging phenotypes and will be at its highest multi-morbidity risk. We considered the total scores of 31 phenotypes as a morbidity score.

(C and D) Separately graphed (C) female and (D) male mice total FI scores as their percentage of lifespan. AKG treatment postpones the occurrence of aging phenotypes during lifespan and compresses the morbidity risk into fewer days of life in both sexes. Each dot is the total score of one animal. Lines are mean ± SEM of the group. n = all animals alive at each time. Two-way ANOVA was used for comparison. The area under the frailty curves (AUC) were calculated to measure the percent comparison of morbidity.

(E and F) The total scores for each mouse were plotted against remaining life (life expectancy). Repeated-measures correlation (rmcorr) was applied for AKG and control groups separately and rmcorr coefficients (r_rm), and associated p values have been indicated. Dots and specific regression lines for each mouse with unique ID are color coded. There are negative correlations between total score and life expectancy in both control and AKG for both sexes. We also reported the slopes of the regression lines, which computed by rmcorr for AKG and control separately (black lines). The equation of each regression line is shown (y = ax + b). Where x is the explanatory variable, a is the slope of the line, and b is the intercept. The results suggest a decrease in magnitude of slope for AKG treatment compared to control. To further prove that this difference is statically meaningful, we applied a mixed model to the combined dataset of AKG and control. In the current mixed model, we treated remaining life and treatment as fixed effects and every mouse as a random effect. The reported p value for regression lines is the result of analysis of variance (ANOVA) for the slopes. AKG treatment can significantly decrease the rate of linear relationship between life expectancy and total score in both males and females.

Since our experimental design takes repeated-measurements on the same mice over time, our analysis took into account the probability that measurements for a given mouse will be correlated. We applied mixed models to analyze our longitudinal data (Figures 2A and ​and2B).2B). In the current mixed model, every mouse with a unique ID was treated as a random effect. Treatment and time were treated as fixed effects (Wu, 2010). Further, the chi-square statistic was calculated for comparing the fit of the current mixed model to a simpler ANOVA, which does not account for repeated measures. The very small p value (p < 0.0001) for chi-square indicates that our current analysis model is very efficient. Our results show that, in both female and male animals, CaAKG decreases incidence and severity of aging phenotypes and postpones morbidity (Figures 2A and ​and2B2B).

Since the age for onset of age-related phenotypes can be quite heterogeneous in mammals, we plotted frailty not only as a function of chronological time, but also in proportion to the lifespan of each mouse by binning scores within ten percentiles (Figures 2C and ​and2D).2D). This allowed us to align assessments with respect to biologic age. Our findings show that CaAKG treatment decreases the proportion of life in which the animal is frail and vulnerable to adverse health incomes (determined as the area under the frailty curve and calculated at a 46% reduction for females and 41% for males; Figures 2C and ​and2D).2D). This improvement in healthy days of life is larger than the increase in lifespan. Through this extensive frailty data analysis, we have demonstrated that dietary CaAKG supplementation results in morbidity compression.

To determine whether there is a correlation between the frailty index and life expectancy, we plotted the total score of each animal in a given time point against its remaining days of life. We applied repeated-measures correlation (rmcorr) to analyze our longitudinal data (Bland and Altman, 1995). rmcorr accounts for dependence of the dataset and evaluates the overall common intra-individual association between measurements (Bakdash and Marusich, 2017). Our data show that higher scores for mice correlates negatively with life expectancy and can be considered as a risk factor (Figures 2E and ​and2F).2F). These are analogous to studies in humans whereby frailty is a property of age, but not all individuals acquire frailty to the same extent. Thus, frailty score predicts mortality significantly but imperfectly.

The result of rmcorr indicates that absolute value of slope corresponding to the AKG treatment cohort is smaller than the one for the control cohort. To further prove that there is a statistically meaningful difference between these two slopes, we applied a mixed model to the combined dataset of AKG and control cohorts. In the current model, we treated remaining life and treatment as fixed factors and each mouse as a random factor. We found that AKG treatment significantly decreases the magnitude of the regression slopes in both males and females (Figures 2E and ​and2F2F).

Since mice in our aging study were older than the previously published paper on frailty (Whitehead et al., 2014), we tested whether individual frailty indicators show significant changes during this stage of aging. We applied the Mann-Kendall trend test to statistically assess if there is a monotonic upward or downward trend of frailty phenotype over time (Lento et al., 2012). Our results show that aging significantly increased the incidence of 11 and 14 frailty phenotypes in female and male, respectively (Figures 3A and ​and3B;3B; Table S2). Among these phenotypes, the nasal discharge showed no changes upon aging. Also, grimace, distended abdomen, and breathing did not show any statistical trend in our analysis. We suggest that these three phenotypes are signs of imminent mortality and simply do not accumulate within the age bracket in this study. When an animal exhibits any of these phenotypes, it will die within a few weeks. Therefore, although these phenotypes are age related, they do not show a trend with age.

Open in a separate window

Figure 3.

AKG Treatment Extends Healthspan and Alleviates Age-Associated Frailty

(A and B) Individually graphed frailty phenotypes that significantly increase with aging, comparing control with AKG treated mice for (A) female and (B) male. The number of animals assessed at each time point are shown as row beneath each graph. Data are mean ± SEM of the group.

(C and D) Trajectories for age-dependent frailty phenotypes for (C) female and (D) male in the study. Mixed model was used to test if each age-dependent phenotype was affected by AKG supplementation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. See also Figure S3.

(E) Simplified anatomy of an anagen hair follicle showing the location of differentiated melanocytes in hair bulb with corresponding immunofluorescence image.

(F and G) Phenotypical characterization of melanocytes in the hair matrix during anagen phase for (F) female AKG-treated and (G) female control mice after 6 months of treatment. Skin sections were stained with antibodies against Dct. nuclear staining (blue) with DAPI.

(H) Dot plot showing the percentage of Dct+ hair follicles in skin sections. Data are mean ± SEM of the group. *p < 0.05 (two-tailed t test).

(I) Aged-matched control (right) and AKG-fed (left) female mice. Animals are 28 months old in the captured picture.

(J and K) Locomotor activity (J) and pedestrian locomotion (K) were measured utilizing metabolic cages at the median life (28 months old) of female control (n = 5) and female AKG (n = 6) mice. Total locomotor activity refers to any movement made by the animal that is picked up by the sensors. The pedestrian locomotion is referred to walking. Data are mean ± SEM; **p < 0.001 (two-tailed t test).

Next, we tested the effects of CaAKG treatment on age-dependent phenotypes utilizing the mixed model (Figures 3C and ​and3D).3D). CaAKG treatment significantly decreased the severity of multiple age-dependent phenotypes in females, including fur color loss, piloerection, poor coat condition, dermatitis, gait disorder, and kyphosis (Figure 3C). In males, severity of body condition, dermatitis, gait disorder, eye discharge, kyphosis, and tumors were all decreased (Figure 3D). Among the frailty measures affected by CaAKG, protection from age-related changes in female coat color was particularly prominent. We chose to characterize this phenotype in more detail by detecting dopachrome tautomerase (Dct), which is expressed in the differentiated melanocytes (MCs) of the hair bulb during anagen phase (Figure 3E).We found that 6 months of AKG supplementation can increase the population of MCs in skin sections (Figures 3F–3I). Gait improvement was also observed in both sexes. We further confirmed gait improvement utilizing metabolic cages. Physical activity and walking were simultaneously recorded over 4 consecutive days (Figures 3J and ​and3K).3K). Interestingly, despite increased locomotion, the levels of oxygen consumption, carbon dioxide production, and energy expenditure were significantly lower in the CaAKG-treated group (Figure S4).

Not all phenotypes were improved by CaAKG. Treated mice failed to perform better in a treadmill exhaustion test and showed no cardiac functional improvement, as determined using echocardiography (Figure S4). Importantly, however, we did not detect any significant adverse changes with CaAKG treatment (Figure S2).

To gain better mechanistic insight, we initiated CaAKG treatment in a new cohort of female mice at 18 months of age. The reported beneficial effect of AKG on lifespan in C. elegans was associated with reduced TOR signaling (Chin et al., 2014). We did not detect any decrease in mTORC1 signaling upon three months of CaAKG treatment in different tissues of mice (Figure S5). However, these experiments were done in absence of any challenge to modify mTOR activity.

Age-associated diseases are accompanied by chronic inflammation, which is generally linked to an age-associated functional decline (Chung et al., 2009). We measured levels of 24 inflammatory cytokines in the serum of aged female mice (28 months old). In untreated mice, the levels of most cytokines increase; however, CaAKG-fed animals were largely refractory to these changes (p < 0.0001; Figure 4A). The reduction in inflammation is consistent with previous findings in the intestine of young pigs receiving an AKG supplementation (He et al., 2017). We repeated the same experiment in both sexes. Interestingly, 6 months of AKG treatment could only affect the cytokine levels in female mice (Figures 4B and S6).

Open in a separate window

Figure 4.

AKG Reduces Inflammation

(A) Heatmap of 24 inflammatory cytokines and chemokines from the plasma of middle-aged (female, age = 18 months, n = 11), aged control, and AKG (female, 28 months old, n = 5) animals. Inflammatory cytokines show a general trend of reductions in AKG group compared to aged control.

(B) Fold changes of 24 inflammatory cytokines for female and male mice after 6 months of treatment (24 months old, n = 5). Levels were calculated using the untreated 18-month-old female or male animals as reference for each treatment grouped. Values were all added together and compared with paired t test. **p < 0.001.

(C and D) The splenocytes were harvested from the spleens of treated and control mice after 6 months of treatment (24 months old, 5 animals for each treatment per sex). Harvested cells were stimulated in culture with PMA, ionomycin, and Brefeldin for 6 h, and IL-10 production was detected using intracellular staining. The gating strategy is shown for CD4/CD8 positive T cells in (C) female and (D) male. Summary of data plotted as bar graphs show positive IL-10 cells as percent of total CD4/CD8 T cells. Data are mean ± SEM, *p < 0.05, **p < 0.01 (two-tailed t test).

(E and F) qRT-PCR analysis of the indicated tissues of female mice.

(G–I) Ionizing radiation (IR) was used to induce senescence in IMR-90 fibroblasts in vitro. Cells were concurrently treated with PBS (control) or 1 mM AKG and were either mock (0 Gy) or irradiated (10 Gy). All assays were performed 10 days post irradiation.

(G) Cells were stained for senescence-associated β-galactosidase activity (left panel) or EdU incorporation (right panel).

(H) IL-6 levels in conditioned media were determined by ELISA, normalized to cell number.

(I) qRT-PCR analysis showing expression of senescence-associated secretory phenotype (SASP) genes, normalized to actin. Each dot is one independent experiment. Data are mean ± SEM, *p < 0.05, **p < 0.01 (two-tailed t test). See also Figure S6.

In order to investigate the cytokine secretion profile of leukocytes, splenocytes were harvested and re-stimulated in culture. Splenic T cells from AKG-treated mice significantly produce higher interleukin (IL)-10; this effect was sex specific with higher levels in AKG-treated female mice (Figures 4C and ​and4D).4D). IL-10 is a potent anti-inflammatory cytokine that plays a central role in limiting host immune response (Ip et al., 2017). Considering the critical role of T cell mediated IL-10 in chronic inflammatory conditions (O’Garra et al., 2004), we believe that induction of IL-10 by AKG treatment can be a potential mechanism for suppression of chronic inflammation.

Studies have shown the accumulation of senescent cells in different tissues of old mice (van Deursen, 2014). These cells contribute to age-associated inflammation by acquiring the senescence-associated secretory phenotype (SASP), and their removal extends lifespan (Baker et al., 2011). We looked into senescent markers in different tissues of mice and did not detect any significant changes (Figures 4E and ​and4F).4F). We further performed cell culture studies to explore possible effects of NaAKG in primary fibroblasts, a well-established model for cellular senescence (Campisi, 2013). While no changes in the senescence formation response to ionizing radiation were observed (Figure 4G), consistent with in vivo findings, AKG altered the SASP (Figures 4H and ​and4I).4I). Specifically, we found a reduction of IL-1b, IL-6, CCL2, and MMP3 without any changes in β-gal and p21.

DISCUSSION

Maintaining function and health are primary outcomes of interest in longevity trials (Newman et al., 2016). A number of different frailty indices have been developed for humans. The most widely used scales are the “Fried frailty phenotype” (Fried et al., 2001) and the “Rockwood frailty index” (Rockwood et al., 2005). Howlett and colleagues have developed a simplified, noninvasive method to quantify frailty through health assessment of C57BL/6J mice. This index exhibits key features of the Rockwood frailty index, including similar exponential increases (Whitehead et al., 2014).

We repeatedly collected longitudinal health measurements and applied statistics to take into account the dependency of each dataset and the random nature of our study units (mice). Our findings demonstrate that AKG, a key metabolite in the TCA cycle, has longevity effects consistent with compressed morbidity in a long-term study of mouse aging. Notably, CaAKG administration started at 18 months of age had robust effects. This is valuable, as human clinical studies are likely to be initiated at a similar relative age. If translated to humans, this effect would be highly desirable, extending lifespan, but more importantly, reducing the debilitating period of functional decline and disease management.

Recent findings have shown that the SASP from senescent cells can drive age-related pathologies through paracrine and likely endocrine effects (Tanaka et al., 2018; Wiley et al., 2019). Our in vitro data indicate that AKG can alter the SASP and reduce inflammation without affecting the formation of senescent cells. Although this is consistent with our in vivo observation of reduced inflammation, there are additional sources of the observed cytokines such as tissue resident macrophages or other immune cells. Inhibition of SASP is a possible mechanism for AKG action in aging mice, and more confirmatory studies need to be performed. Also, senescent cells can activate inflammatory pathways in immune cells (Hall et al., 2016; Vicente et al., 2016), thus it is possible that immune cell activation, the SASP, or likely both are inhibited by AKG.

We show that female T cells significantly produce higher IL-10 upon AKG treatment. This effect, as well as suppression of inflammation and decreased mortality, were sex specific. Interestingly, IL-10 knockout (KO) mice (IL-10^tm/tm) show frailty phenotypes, such as an increase in several pro-inflammatory cytokines including IL-6, IL-1β, TNF-α, CXCL-1, and IL-12, as well as reduced muscle strength and higher mortality rates (Ko et al., 2012; Walston et al., 2008). IL-10 is produced by a variety of immune cell types; however, in context of chronic inflammation, the source of IL-10 is largely T cells (O’Garra et al., 2004). In addition, mice lacking IL-10 specifically in T cells recapitulate the immune phenotype of the whole-body IL-10 KO (Roers et al., 2004). Considering the critical role of T cell mediated IL-10 in chronic inflammatory conditions (O’Garra et al., 2004), we believe that induction of IL-10 by AKG treatment is a potential mechanism for suppression of chronic inflammation in females.

We observed no significant adverse effects of AKG administration. The only phenotypes that had a higher incidence in AKG-treated animals were cataracts and corneal opacity, although neither reached statistical significance. AKG has been used in human clinical studies linked to diseases without associated adverse effects (Filip et al., 2007; Jeevanandam and Petersen, 1999; Riedel et al., 1996). Interestingly, in humans, plasma AKG levels are reported to decline 10-fold between the ages of 40 and 80 (Harrison and Pierzynowski, 2008). The molecule is not available in the human diet, making direct supplementation the only feasible route to restore levels. Given its GRAS status and human safety record, our findings point to a potential safe human intervention that may impact important elements of aging and improve quality of life in the elderly population.

STAR★METHODS

Supplementary Material

ACKNOWLEDGMENTS

Footnotes

REFERENCES