DNA-PK promotes the mitochondrial, metabolic and physical decline that occurs during aging

Aging increases DNA DSBs and DNA-PK activity

A naturally occurring condition that may increase DNA DSBs and potentially activate DNA-PK is aging. In order to determine if aging is associated with increased DNA DSBs in skeletal muscle, we isolated skeletal muscle from young (3 mo) and old (25 mo) mice and visualized γ-H2AX (Rogakou et al., 1998), a phosphorylated form of histone H2AX that is a marker for DNA DSBs (Figure 1A). We found that skeletal muscle from old mice has more γ-H2AX, suggesting that aging increases DNA DSBs in skeletal muscle.

We then examined the association of aging and DNA-PK activity in two different organisms: rhesus macaques and mice. We isolated skeletal muscle (gastrocnemius) from healthy rhesus macaques ranging in age from 1 to 15 yr, which is equivalent to 3 to 45 yr in human age. The tissue was isolated from fasted animals to exclude food intake as a variable. We detected DNA-PK activity by visualizing autophosphorylation of S2056 (p-DNA-PK) in DNA-PKcs (Chen et al., 2005). In all cases, the level phosphorylation was normalized to the total protein level. The levels of p-DNA-PK were similar at 1, 3 and 10 yr (image not shown), but rose significantly at 15 yr (Figure 1B). The antibody specific for p-DNA-PK, which was raised against human DNA-PKcs, was too weakly cross reacting for detection of mouse p-DNA-PK (data not shown).

Consistent with the dramatic increase in p-DNA-PK at middle aged, T5,7-HSP90α (p-HSP90α) was significantly higher in middle aged macacque skeletal muscle (Figure 1C). Similarly, p-HSP90α was significantly higher in skeletal muscle from old mice compared to that in young mice (Figure 1D). These findings indicate that both DNA-PK activity and p-HSP90α are elevated in skeletal muscle of older rhesus monkeys and mice. Interestingly, p-HSP90α did not increase with aging in the lung, indicating that aging does not increase p-HSP90α in all tissues (Figure S2).

DNA-PK-mediated phosphorylation of HSP90α triggers release of clients

Although HSP90 is more abundant in the cytoplasm, there is some HSP90 in the nucleus (van Bergen en Henegouwen et al., 1987). In order to determine the cellular compartment to which phosphorylated p-HSP90α is localized, we purified the cytoplasmic and nuclear fractions before and after ionizing radiation (IR), which generates DNA DSBs and activates DNA-PK (Figure 2A). We found a detectable basal level of p-HSP90α in the cytoplasm and nucleus, but after IR, p-HSP90α increased dramatically in both compartments. Moreover, IR increased nuclear accumulation of HSP90α but not HSP90β.

Based on a previous report showing that the ENT region may regulate HSP90 chaperone function by affecting the HSP90-client complex formation (Pullen and Bolon, 2011), we examined whether T5,7 phosphorylation affects HSP90α-client interaction. To test this, we transiently-transfected C2C12 myotubes with vectors encoding Flag-tagged WT, T5,7A (T→A) mutant and phosphomimetic T5,7D (T→D) mutant HSP90α, treated them with IR and examined the interaction of the HSP90α mutants with the clients liver kinase B1 (LKB1) and glucocorticoid receptor (GR). Co-immunoprecipitation studies showed that IR released LKB1 from WT HSP90α, but not from T5,7A HSP90α (Figure 2B). T5,7D HSP90α bound poorly to LKB1 even in the absence of IR. Similarly, IR released GR from WT HSP90α, but not from T5,7A HSP90α (Figure 2C); T5,7D HSP90α also bound poorly to GR even in the absence of IR. Unlike the HSP90α-GR interaction, the HSP90β-GR interaction was not affected by IR (Figure 2D). To demonstrate that the IR-induced release of GR from HSP90α was mediated by DNA-PK-dependent phosphorylation, we performed co-immunoprecipitation between GR and HSP90α after knocking-down DNA-PKcs with siRNA. As shown in Figure 2E, DNA-PKcs siRNA significantly reduced p-HSP90α and prevented IR from disrupting the GR-HSP90α interaction (Figure 2F). The T5,7A mutations also blunted IR-induced release of other HSP90 clients such as endothelial nitric oxide synthase (eNOS) (Figure 2G) and the androgen receptor (AR) (Figure 2H) from HSP90α. Taken together, these findings indicate that DNA-PK-mediated phosphorylation of HSP90α can disrupt HSP90α-client complexes.

Inhibition of DNA-PK activates AMP-activated protein kinase

Since AMPK (Taipale et al., 2012) and its upstream activator kinase LKB1 (Nony et al., 2003) are HSP90 clients, we investigated whether DNA-PK and phosphorylation of HSP90α mediate the aging-related decline in AMPK activity. Although AMPKα2 (the dominant AMPK in skeletal muscle) is an HSP90 client, its interaction with HSP90 is weak to begin with (background intensity) (Taipale et al., 2012), making it difficult to see further decrease in interaction after IR. However, to confirm that AMPK is a client of HSP90α, we knocked-down HSP90α with siRNA and visualized the level of AMPK in C2C12 myotubes. Decreased chaperone function of HSP90 often results in decreased levels of client proteins due to aggregation and/or destabilization. Consistent with the notion that AMPK is a client of HSP90α, HSP90α siRNA decreased AMPK levels (Figure 3A top and bottom left panels) without affecting its mRNA levels (Figure 3A bottom right panel), and treatment with the HSP90 inhibitor 17-AAG also decreased AMPK levels (Figure 3B). We also investigated whether the T5,7 phosphorylation affects the AMPK-HSP90α interaction by transiently expressing WT and T5,7A HSP90α. As shown in Figure 3C, AMPK interaction with T5,7A HSP90α was stronger than with WT HSP90α and consistent with this, p-AMPK (T172), a marker of AMPK activity, was higher with T5,7A HSP90α compared with WT HSP90α (Figure 3D). These findings suggest that phosphorylation of T5,7 decreases HSP90α-AMPK complex formation and AMPK activity.

We then investigated whether DNA-PK affects AMPK activity in vivo. For this purpose, we used severe combined immunodeficiency (SCID) mice, which carry a leaky mutation in DNA-PKcs (Blunt et al., 1996) but have a normal lifespan (Bailey et al., 1999). We visualized p-AMPK in skeletal muscle of young (3–4 mo) and middle aged (12–14 mo) WT and SCID mice (Figures 3E and F). In young SCID mice, p-AMPK was similar to that in WT mice but in middle aged SCID mice, p-AMPK was higher than that in WT mice. In line with this observation, p-HSP90α was also decreased in middle aged SCID muscle relative to WT muscle (Figures 3G and H). To confirm that the DNA-PK-AMPK link was cell autonomous in vivo, we generated a conditional knockout of DNA-PKcs in skeletal muscle (MDPKO) by breeding DNA-PKcs^fl/fl mice (Mishra et al., 2015) with mice expressing the Cre recombinase driven by muscle creatine kinase (MCK) (Figure 3I). As expected, in skeletal muscle, p-HSP90α was lower and p-AMPK and phosphorylation of AMPK substrate acetyl-CoA carboxylase (p-ACC) were higher in 17 mo old MDPKO mice compared to DNA-PKcs^fl/fl (fl/fl) mice; there was less difference in 5 mo old mice. Taken together, these findings support the notion that the aging-related increase in DNA-PK-mediated phosphorylation of HSP90α leads to the decline of AMPK activity with age.

Since DNA-PK is involved in NHEJ, we asked whether DNA-PK-deficiency or treatment with DNA-PK inhibitor NU7441 may affect the DNA DSB burden in skeletal muscle. As shown in Figure S3, γ-H2AX in skeletal muscle of SCID and MDPKO mice were similar or slightly decreased (in old SCID) compared to WT mice; treatment with NU7441 also did not affect γ-H2AX levels.

DNA-PK inhibition increases skeletal muscle mitochondrial content

These findings led us to consider the possibility that DNA-PK may promote aging-associated mitochondrial decline in skeletal muscle. We quantified mitochondrial content in skeletal muscle of 14–16 year old rhesus monkeys and its correlation with DNA-PKcs activity. We found a strong inverse correlation between mitochondrial content and p-DNA-PK in middle aged skeletal muscle (r=−0.93, p=0.008) (Figure 4A). One possible interpretation of this finding is that DNA-PK suppresses mitochondrial biogenesis in middle aged skeletal muscle.

Peroxisome proliferator-activated receptor-gamma coactivator 1α (PGC-1α) (Puigserver et al., 1998) is the master regulator of mitochondrial function. We measured mRNA levels of PGC-1α as well as related PGC-1β and additional genes involved in mitochondrial function in skeletal muscle isolated from WT and SCID young mice (Y, 3–4 mo old), obese mice generated by feeding high-fat diet (HFD) for 4–5 months (Ob, 5–6 mo old) and middle aged mice (MA, 12–14 mo old) (Figure 4B). We found that mRNA levels of PGC-1α and PPARδ, were decreased in muscle from obese WT mice but not from SCID mice fed a HFD. Also, the expression of all the genes tested was decreased in middle aged WT muscle, but not in middle aged SCID muscle. Consistent with this, the protein levels of PGC-1α in skeletal muscle of 17 mo old MDPKO mice were elevated compared to the controls (Figure 4C).

We then calculated the mitochondrial density in skeletal muscle of young and middle aged WT and SCID mice. Morphometric analyses of electron micrographs indicated that middle aged WT muscle had 36% less mitochondrial density than young WT muscle (Figure 4D). However, middle aged SCID muscle had same mitochondrial density as young SCID and WT muscle. In agreement with this, the mtDNA copy number in middle aged muscle was higher in SCID mice (Figure 4E) and tended to be higher in MDPKO mice (Figure 4F) compared to their respective controls.

These findings imply that inhibition of DNA-PK activity with a small molecule will induce mitochondrial biogenesis. We treated WT mice with highly-specific DNA-PK inhibitor NU7441 for 3 mo by incorporating it in the HFD pellet, starting from age 1 mo. After treatment, we measured serum IgG levels to determine if the inhibitor caused any immune dysfunction. Unlike SCID mice, which do not produce any antibody because of the lymphocyte defect, WT mice treated with NU7441 had normal levels of serum antibodies (Figure 4G), indicating that in adult mice, the DNA-PK inhibitor did not significantly affect B-cell function. To determine the role of AMPK in mitochondrial biogenesis, we also treated AMPKα2 KO mice (Viollet et al., 2003) on HFD with the inhibitor. As shown in Figure 4H, NU7441 increased mitochondrial content by approximately 50% in WT mice but did not increase it in AMPKα2 KO mice. Taken together, these findings are consistent with the notion that inhibiting DNA-PK increases mitochondrial biogenesis in an AMPK-dependent manner.

DNA-PK inhibition increases physical fitness in obese and middle aged animals

Skeletal muscle is composed of oxidative and glycolytic fiber types. Type 1 and type 2A fibers are high in mitochondrial content and are oxidative whereas type 2X and type 2B fibers are lower in mitochondrial content and are glycolytic. Previous studies have shown that increased expression of PGC-1α causes fiber switch from glycolytic toward more oxidative fibers (Lin et al., 2002). We analyzed the fiber type composition of gastrocnemius muscle, which is largely glycolytic, from middle aged SCID and WT mice and WT mice treated with the DNA-PK inhibitor for 10–12 weeks. As shown in Figure 5A, SCID muscle had more Type 1 fiber and less Type 2B fiber than WT muscle as did the muscle from the inhibitor treated mice (Figure 5B). Consistent with increased mitochondrial content and oxidative fiber in middle aged SCID muscle, they had higher mitochondrial oxygen consumption rate (VO₂) in permeabilized muscle (Figure S4A).

Treatments that activate AMPK and/or reprogram glycolytic fibers, which are prone to fatigue, to oxidative fibers are known to increase fat oxidation and fatigue-resistance (Narkar et al., 2008; Suwa et al., 2003; Wang et al., 2004). Mice with increased mitochondrial content in skeletal muscle generally have lower serum lactate, the by-product of glycolysis and a major source of muscle fatigue (Hanson and Hakimi, 2008; Mason et al., 2004). Since SCID muscle has more oxidative fibers, we suspected that SCID mice might have lower serum lactate levels. Indeed, serum lactate levels of middle aged SCID mice were 17% lower than those of WT mice (Figure 5C) in the resting state.

These findings suggest that DNA-PK may promote aging-associated decline in physical fitness. To test this, we measured the treadmill running capacity of young and middle aged WT and SCID mice. Young WT and SCID mice ran similar distances, but middle aged SCID mice ran nearly twice the distance of middle aged WT mice, showing no aging-associated decline (Figure 5D). With some high-endurance mice, repeated exercise damages muscle and decreases endurance (Mason et al., 2004). This was a potential concern because increased mitochondrial content could lead to increased oxidative damage, including DNA damage. To test whether this is the case with SCID mice, we exercised obese and middle aged WT and SCID mice for three consecutive days (Figure S4B). By the third day of exercise, the superior endurance of SCID mice compared to WT mice became even more pronounced, indicating that SCID muscle is not prone to damage by repeated exercise.

Results shown above suggest that the DNA-PK inhibitor may be an effective fitness enhancer. Indeed, obese and middle aged WT mice treated with NU7441 ran approximately 60% and 40% greater distance, respectively (Figure 5E). Consistent with our earlier finding that DNA-PK inhibition does not increase mitochondrial biogenesis in skeletal muscle of obese AMPKα2 mice (Figure 4H), the inhibitor increased the running distance by more than 60% in obese WT mice but had no effect on the running distance in obese AMPKα2 KO mice (Figure 5F). Taken together, these findings indicate that DNA-PK inhibition can prevent aging-associated loss of mitochondria and its function and dramatically improve fitness in an AMPK-dependent manner.

To determine if the immune status of SCID plays any role in their physical performance, we exercised mice deficient in Rag1, a nuclease that is essential for VDJ recombination and lymphocyte development (Oettinger et al., 1990). Since Rag1 and DNA-PKcs participate sequentially in the same VDJ recombination pathway in lymphocytes, the immune status of Rag1 KO and SCID mice are very similar. Unlike SCID muscle, Rag1 KO muscle did not express higher PGC-1α or PPARδ expression (Figure S5A) and had normal physical endurance (Figure S5B).

Inhibition of DNA-PK protects against obesity and type-2 diabetes

To investigate the metabolic function of DNA-PKcs in vivo, we measured the body weight of young WT and SCID mice fed either regular chow diet (RCD) or HFD after weaning. After 20 wk on RCD, SCID mice weighed 1–2 gm less than WT mice, but the difference was statistically not significant (Figure S6A). On HFD, SCID mice were relatively resistant to weight gain (Figure S6B) compared to WT mice even though their food intake (Figure S6C) and fat absorption (Figure S6D) were similar to those of WT mice. NMR spectroscopy analysis indicated that on HFD, SCID mice have approximately 25% lower fat mass index than WT mice (Figure S6E). Consistent with this, fat in the abdominal region was noticeably decreased in SCID mice (Figure S6F and G). On the other hand, WT and SCID mice had similar lean mass index (Figure S6H).

Although the body weight difference between WT and SCID mice was insignificant at young age (see Figure S6A), it became much more pronounced at middle aged (14 mo) (Figure S6I). Consistent with this, middle aged SCID mice had lower plasma triglyceride levels than WT mice (Figure S6J). Middle aged SCID mice had similar fasting plasma glucose levels as WT mice (92 ± 8 mg/dl vs 94 ± 14 mg/dl) despite having significantly lower plasma insulin levels than WT mice (0.28 ± 0.01 ng/ml vs 0.64 ± 0.1 ng/ml, p<0.05). This suggested that middle aged SCID mice are less insulin resistant than WT mice. Indeed, the insulin resistance index (HOMA-IR) increased with age in WT mice as expected but did not change in SCID mice (Figure S6K). Taken together, these results indicate that DNA-PK promotes aging-associated increase in insulin resistance.

The results shown above suggested that a DNA-PK inhibitor might have an anti-diabetic effect in obese WT mice. Indeed, we found that including NU7441 in the pellet decreased weight gain on HFD (Figure 6A) and increased glucose tolerance (Figure 6B) and insulin sensitivity (Figure 6C). Activation of AMPK can increase the secretion of glucagon-like peptide-1 (GLP-1), a peptide hormone produced by the L-cells in the gut that helps to maintain normoglycemia (Baggio and Drucker, 2007). Consistent with the DNA-PK-AMPK link, the inhibitor increased the plasma level of GLP-1 (Figure 6D). To determine the site of improved insulin action in these mice, we performed hyperinsulinemic-euglycemic clamps. The glucose infusion rate (GIR) required to maintain euglycemia was approximately 27% higher in inhibitor-treated mice (Figure 6E, Figure S7A, B) and the total glucose disposal rate (GDR) was increased by almost 20% (Figure 6F). The inhibitor increased insulin-stimulated glucose uptake (GU) into skeletal muscle, the primary site of insulin-stimulated glucose disposal, by almost 40% (Figure 6G) and insulin-stimulated glucose uptake into white adipose tissue (WAT) by more than 120% (Figure 6H). However, endogenous glucose production (EGP) by the liver was not improved by NU7441 (Figure 6I). Middle aged MDPKO mice also tended to have lower body weight, fasting glucose and insulin resistance compared to fl/fl mice although the differences were not statistically significant (Figure S7C). These findings indicate that inhibition of DNA-PK activity with a small molecule inhibitor protects against obesity and has an anti-diabetic effect.

Calorie restriction and aerobic fitness are associated with decreased DNA-PK activity

Calorie-restriction (CR), like DNA-PK inhibition, increases mitochondrial biogenesis (Lopez-Lluch et al., 2006) and protects against insulin resistance and metabolic syndrome (Omodei and Fontana, 2011). We examined how CR affected DNA-PK activity and phosphorylation of T5,7 in HSP90α in middle aged rhesus macaques. CR, 30% caloric reduction compared to ad libitum feeding for 3.4 yr, decreased both DNA-PKcs autophosphorylation and HSP90α phosphorylation in middle aged skeletal muscle (Figure 7A).

A number of clinical studies during the past two decades have shown that low exercise capacity is a strong predictor of morbidity and mortality (Kokkinos et al., 2008). As part of an effort to understand the link between exercise capacity and health, rats with low and high intrinsic treadmill running capacity have been produced by two-way selective breeding of genetically heterogeneous founder rats (Koch et al., 2012). The lines, termed the Low Capacity Runners (LCR) and the High Capacity Runners (HCR), differed by approximately 7-fold in running capacity after 27 generations of selective breeding. The sedentary HCR rats have increased mitochondrial content and function, are protected from the metabolic syndrome, hypertension, type 2 diabetes and cancer and have longer lifespan compared to the sedentary LCR rats (Garton et al., 2016). Since the HCR rats have metabolic phenotypes that resemble CR and DNA-PK-deficiency, we hypothesized that breeding for the HCR rats may have led to the generation of rats with decreased DNA-PK activity. We measured total DNA-PKcs level and p-HSP90α in skeletal muscle of 7–9 mo sedentary the HCR and LCR rats. As shown in Figure 7B, DNA-PKcs level was three-fold lower and p-HSP90α was approximately 60% lower in the HCR rats compared to LCR rats. Taken together, these findings suggest that CR and aerobic fitness, which delay aging and protect against chronic diseases, are associated with decreased DNA-PK activity.

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to, and will be fulfilled by the corresponding author Jay H. Chung (chungj@nhlbi.nih.gov)

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHODS DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical comparisons between groups were analyzed by Two-way ANOVA and Two-way repeated measures ANOVA for a parametric test or Mann-Whitney test for a nonparametric test using Prism7. Results are expressed as the mean±s.e.m. Significance was accepted at p<0.05. The Excel software (Microsoft) was used to determine average values, standard errors and standard deviations. For each figure, number of experimental replicates and other information relevant for assessing the accuracy and precision of the analysis are included in the accompanying legend.