Neuronal Sirt1 Regulates Glycolysis and Mediates Resveratrol-Induced Ischemic Tolerance

Abstract

BACKGROUND AND PURPOSE

Resveratrol, at least in part via Sirt1 activation, protects against cerebral ischemia when administered 2 days prior to injury. However, it remains unclear if Sirt1 activation must occur, and in which brain cell types, for the induction of neuroprotection. We hypothesized that neuronal Sirt1 is essential for resveratrol-induced ischemic tolerance and sought to characterize the metabolic pathways regulated by neuronal Sirt1 at the cellular level in the brain.

METHODS

We assessed infarct size and functional outcome following transient 60 minute middle cerebral artery occlusion in control and inducible, neuronal-specific Sirt1 knockout mice. Non-targeted primary metabolomics analysis identified putative Sirt1-regulated pathways in brain. Glycolytic function was evaluated in acute brain slices from adult mice and primary neuronal-enriched cultures under ischemic penumbra-like conditions.

RESULTS

Resveratrol-induced neuroprotection from stroke was lost in neuronal Sirt1 knockout mice. Metabolomics analysis revealed alterations in glucose metabolism upon deletion of neuronal Sirt1, accompanied by transcriptional changes in glucose metabolism machinery. Furthermore, glycolytic ATP production was impaired in acute brain slices from neuronal Sirt1 knockout mice. Conversely, resveratrol increased glycolytic rate in a Sirt1-dependent manner and under ischemic penumbra-like conditions in vitro.

CONCLUSIONS

Our data demonstrate that resveratrol requires neuronal Sirt1 to elicit ischemic tolerance and identify a novel role for Sirt1 in the regulation of glycolytic function in brain. Identification of robust neuroprotective mechanisms that underlie ischemia tolerance and the metabolic adaptations mediated by Sirt1 in brain are crucial for the translation of therapies in cerebral ischemia and other neurological disorders.

INTRODUCTION

Stroke is a leading cause of death and long-term disability¹ with no clinically approved neuroprotective treatments. While translation of therapies that target deleterious mechanisms post-injury has been challenging, the field of preconditioning offers an additional therapeutic window and novel targets for significant neuroprotection. Our lab and others have demonstrated the efficacy of preconditioning with resveratrol (resveratrol preconditioning, RPC), a naturally occurring polyphenolic compound^{2, 3}. RPC-induced ischemic tolerance is observed at both early (24 hours) and late (>1 week) time points post-injury, with validation across species, age, sex and models of cerebral ischemia including neonatal hypoxia-ischemia, ischemic stroke, recurrent stroke and asphyxial cardiac arrest^3–6. Yet, the molecular mechanism(s) of RPC are still unclear.

Resveratrol-induced protection has been attributed to the modulation of systemic processes, such as inflammation⁶ and blood flow⁷, which then manifests in the brain. Still, resveratrol protects isolated neurons in culture⁸ as well as organotypic cultures⁹, which have supporting cell types but lack circulation, demonstrating the existence of a brain-intrinsic response that is sufficient for ischemic tolerance. What’s more, it is postulated that silent information regulator 2 homologue 1 (Sirt1) mediates at least some of these effects⁴. Adding to mechanistic complexity, Sirt1 is expressed in both brain and peripheral tissues and has been implicated in the regulation of inflammation¹⁰, cerebral blood flow¹¹ and neuronal functions¹². It remains to be tested whether Sirt1 is essential for RPC-induced ischemic tolerance and if so, in which cell types. These are important discrepancies to reconcile regarding translation of novel therapies.

Sirt1 is an NAD⁺-dependent deacetylase that acts on histone and non-histone proteins to induce adaptive responses to metabolic stress¹³. These responses are extensively characterized in peripheral tissues, where Sirt1 mediates a switch from anabolic to catabolic processes, promotes the utilization of alternative energy substrates and regulates systemic glucose output¹³. The brain, however, may lack this metabolic plasticity given its tight regulation of energy metabolism. While Sirt1 controls neuroendocrine function at the organismal level from within hypothalamic nuclei¹³, it is still unclear how Sirt1 regulates metabolism at the cellular level in the brain and particularly neurons. This regulation is likely essential for the maintenance of neuronal function, within normal physiology and disease.

The aim of this study was to determine if specifically neuronal Sirt1 is necessary for RPC-induced ischemic tolerance and to identify the potential metabolic pathways through which Sirt1 exerts neuroprotection at the cellular level in the brain. To this end, we used a conditional and inducible genetic approach to delete Sirt1 specifically in adult neurons and subjected mice to transient ischemic stroke. Furthermore, we conducted metabolomics analysis of control and neuronal Sirt1 knockout mice to identify putative metabolic pathways regulated by neuronal Sirt1 in brain. Finally, we assessed the neuroprotective potential of Sirt1-mediated enhancement of neuronal glycolysis induced by RPC.

MATERIALS AND METHODS

Generation of inducible, neuron-specific Sirt1 knockout mice

All animal usage and experimentation was approved by the Institutional Animal Care and Use Committee at the University of Miami and was in accordance with the US Public Health Service’s Policy on Humane Care and Use of Laboratory Animals. SLICK-H mice express the creER^T2 fusion protein and EYFP under back-to-back copies of the Thy1.2 promoter. Sirt1^flox/flox mice harbor loxP sites flaking exon 4 of Sirt1. These strains, on the C57BL/6J genetic background, were purchased from The Jackson Laboratory (Bar Harbor, ME; SLICK-H, stock no: 012708, Sirt1^flox/flox, stock no: 008041) and backcrossed further for 2 generations into the C57BL/6J genetic background. Crossings of these strains generated Thy1.2-creER^T2; Sirt1^flox/flox (Sirt1^neu−/−) mice. Eight to twelve week-old male mice (~20–25 g) were used throughout to eliminate the influence of estrous cycle on ischemic injury. Induction was achieved by IP tamoxifen (0.13 mg/g, Sigma, St. Louis, MO) injections 1×/day for 5 days followed by 5 days of rest.

Transient Middle Cerebral Artery Occlusion (tMCAo)

Resveratrol preconditioning was induced by treatment with trans-resveratrol (3,4’,5-trihydroxy-trans-stilbene, 10 mg/kg (Sigma) in 1.5% DMSO) or vehicle (1.5% DMSO) IP, 2 days prior to tMCAo performed as previously described¹². Mice were anesthetized with 3% isoflurane in 100% O₂ and maintained at 37.0 °C through a rectal monitoring heating pad (Harvard Apparatus, Holliston, MA). tMCAo was induced by insertion of a silicone-coated 8-0 monofilament surgical suture (Doccol, Sharon, MA) into the MCA. After 60 min, the suture was removed. Cerebral blood flow was measured before, during, and immediately following tMCAo using a Laser-Doppler (Perimed System, Stockholm, Sweden) probe. Animals that did not display a ≥ 70% reduction in cerebral blood flow during occlusion were excluded.

Neurological scoring

Twenty-four hours after tMCAo, animals were scored based on a neurobehavioral battery (body symmetry, gait, climbing, circling behavior, forelimb symmetry, compulsory circling, and whisker response) as previously described¹². Focal score ranges from 0 – 28, where 0 is considered normal and 28 indicates extremely severe deficits.

Quantification of infarct volume

Triphenyl tetrazolium chloride (TTC, Sigma) staining for infarct was performed as previously described¹². Infarcted area and total ipsilateral area were measured using ImageJ software and infarct was calculated as a percentage of the total ipsilateral hemisphere.

Non-targeted primary metabolomics

Sample preparation, data acquisition and data processing was performed by the West Coast Metabolomics Center (UC-Davis) as previously described¹⁴. This non-targeted primary metabolism platform was achieved by gas chromatography – time of flight – mass spectrometry (GC-TOF-MS). Please see the Online Supplement.

Metabolomics statistical and biochemical analysis

Peak intensity values normalized to tissue weight were used for statistical analysis in MetaboAnalyst 3.0¹⁵ (n=7–8). One sample was a statistical outlier (Random Forest outlier detection, largest outlying measure >3.0) and was removed from analysis. Values were log transformed and auto scaled. Networks that display structural similarity of metabolites based on PubChem substructure fingerprints were generated by MetaMapR¹⁶ and visualized in Cytoscape.

Acute brain slice preparation

Adult (~20–25g) mice were anesthetized with 2% isoflurane and their brains cubed for sagittal sectioning. 300 µM thick slices were sectioned with a Leica VT1000S microtome (Wetzlar, Germany) in cold artificial cerebrospinal fluid (aCSF) solution: 4.5 mmol/l KCl; 2 mmol/l MgSO₄; 1.25 mmol/l Na₂HPO₄; 126 mmol/l NaCl; 2 mmol/l CaCl₂; 26 mmol/l NaHCO₃; 10 mmol/l Glucose; bubbled with 95% O₂, 5% CO₂; osmolality 302–306. Slices were transferred to aCSF at RT for 1 hr prior to use. Slices were maintained on an interface recording chamber for 30 min prior to experimentation and superfused at 2 ml/min (32°C ± 1°C, aCSF bubbled with 95% O₂, 5% CO₂) and oxygenated (humidified 95% O₂, 5% CO₂).

Extracellular Field Recordings and Anoxic Depolarization (AD)

Field population spikes (fPSs) were recorded in the pyramidal cell body layer of CA1 hippocampus with NaCl-filled (150 mmol/l) micropipettes. Schaffer collateral axons were electrically stimulated by 0.3 ms constant current pulses at half-maximal response (from input-output curve) and excluded if they displayed double fSpike properties, indicative of poor slice health. For AD experiments, extracellular recording monitored the onset of a DC-shift while Schaffer collaterals were stimulated every 15s to monitor fPSs. Baseline was recorded for 5 min in normal aCSF (see above) prior to experimental conditions of aCSF with no glucose or aCSF with glucose (20 mmol/l). Anoxia was induced by bubbling with 95% N, 5% CO₂ and de-oxygenating slices with humidified 95% N, 5% CO₂. Latency until AD was defined as the rate of DC-shift exceeding 10V/s.

Primary neuronal-enriched cultures

Primary neuronal-enriched cultures was prepared as previously described¹⁷, with some modifications. Pregnant female Sprague-Dawley rats (E18–19) were acquired from Charles River (Wilmington, MA). Pups were harvested, their cortical tissue was homogenized, dissociated and washed with neuronal media (MEM, 1% Glutamax, 10% FBS and 20 mmol/l glucose). Cells were plated at a density of 80,000 cells per Seahorse XFp Miniplate well (Aligent, Santa Clara, CA). Cultures underwent half media changes every 3^rd day for 7–10 days prior to experimentation.

Glycolytic rate in vitro

Glycolytic rate was measured using the Seahorse XFp Analyzer (Aligent) according to the manufacturer’s instructions and published standards¹⁸. Neurons were treated with resveratrol (100 µM, 1.5% DMSO), vehicle (1.5% DMSO) and/or the Sirt1 specific-inhibitor EX-527 (Sigma, 1µM, 10µM) 2 days prior to measurements. On the day of measurements, neurons were washed 3× with seahorse media (XF Base Medium-Minimum DMEM, 20 mmol/l glucose, 1 mmol/l sodium pyruvate, 2 mmol/l glutamine) and incubated in a CO₂-free incubator at 37°C for 1 hr prior to plate run. ECAR was normalized to cell count in a reference well.

Western Blot, Immunofluorescence and RT-PCR

For these standard methods, please refer to the Online Supplement.

Statistical analysis and blinding procedures

Mice were randomly assigned groups throughout. Data acquisition and analyses were conducted in a blinded manner. Statistical analysis was performed in Prism6 software (GraphPad, San Diego, CA). Data are presented as mean ± SEM. Two-sample, unpaired student t –test, one- and two-way ANOVA with Bonferroni post-hoc tests and repeated measures two-way ANOVA with Bonferroni post-hoc tests were used as indicated in the text and figure legends. Significance is denoted in each figure. Sample size was determined for each experiment depending upon prior experiments conducted by the lab and power analysis in order to detect an effect size of 0.8 using G*power 3.1 software. Statistical tests meet necessary assumptions as measured in Prism6.

RESULTS

Generation of inducible, neuronal-specific Sirt1 knockout mice

In order to specifically investigate post-developmental neuronal Sirt1 function, we generated tamoxifen-inducible, neuronal-specific Sirt1 knockout mice. SLICK-H mice¹⁹, where creER^T2 and enhanced yellow fluorescent protein (EYFP) are under control of the pan-neuronal promoter Thy1.2, were crossed with Sirt1^flox/flox mice that harbor loxP sites flanking the catalytic domain (exon 4) of Sirt1²⁰. In these mice (Sirt1^neu−/−), tamoxifen treatment results in deletion of exon 4 from Sirt1 exclusively in neurons, rendering a mutant protein that lacks deacetylase activity²⁰. Sirt1^neu−/− exhibit strong EYFP fluorescence throughout major brain regions (Fig. 1A). EYFP did not show robust expression in the hypothalamus and body weight was unchanged one week after induction (Fig. I), suggesting that Sirt1 regulation of systemic metabolism from the hypothalamus is not overtly altered. Immunostaining revealed co-localization of EYFP and the neuronal marker NeuN (Fig. 1B) but not the glial marker GFAP (Fig. 1C), demonstrating neuronal specificity. In non-induced mice, Sirt1 is primarily neuronal, co-localizing with EYFP and NeuN but not GFAP (Fig. 1A and C). Efficiency of Sirt1 deletion was assessed by western blot, where full-length wild-type (WT) Sirt1 is observed as two bands at 110 kDa. In Sirt1^neu−/−, these bands were observed at a lower molecular weight, indicating truncation of Sirt1 protein, which now lacks exon 4 (Fig. 1D). No mutant Sirt1 or EYFP protein was found in other tissues such as heart (Fig. 1D; Fig. II). Background CreER^T2 activity was not observed, as non-induced Sirt1^neu−/− mice do not express mutant Sirt1 protein (Fig. II). Deletion was also confirmed at the RNA level using primers specific for the full-length WT Sirt1 transcript (Fig. II). After induction of Sirt1^neu−/− mice, no overt phenotype was observed. Together, this evidence demonstrates efficient and specific deletion of neuronal Sirt1 in Sirt1^neu−/− mice.

Fig 1. Generation of inducible, neuron-specific Sirt1 knockout mice (Sirt1^neu−/−).

(A–C) Confocal images of immunostaining from tissue sections of non-induced Sirt1^neu−/−. (A) Hipp – hippocampus; Ctx – cortex; scale bar = 400 µm. (B) Str – striatum, scale bar = 70 µm. (C) Ctx – cortex, scale bar = 20 µm. (D) Western blot for Sirt1 reveals efficient deletion upon tamoxifen treatment in major and stroke-affected brain areas but not heart. Notice the smaller size band of the mutant Sirt1 protein (Δex4 – Sirt1 mutant) in Sirt1^neu−/−; Hipp – hippocampus. Tx – tamoxifen.

Neuronal Sirt1 is required for RPC-induced ischemic tolerance

Next we sought to determine if neuronal Sirt1 is required for RPC-induced ischemic tolerance. Tamoxifen treated control and Sirt1^neu−/− mice were treated with RPC (resveratrol 10 mg/kg IP) or Veh (1.5% DMSO) 2 days prior to transient middle cerebral artery occlusion (tMCAo). Mortality from tMCAo did not differ between genotypes or treatments (Table I). Twenty-four hours following tMCAo, functional outcomes were scored from a battery of sensory-motor tasks and infarct size was quantified. RPC significantly reduced infarct size (45.2%, p<0.05) and improved functional outcome (25.9%, p<0.05) in control but not Sirt1^neu−/− mice (Fig. 2A and B). Cerebral blood flow measurements during ischemia or at the onset of reperfusion did not differ between groups (Fig 2C), demonstrating that these results were not due to changes in cerebral perfusion. These data show that RPC-induced ischemic tolerance requires neuronal Sirt1.

Fig 2. Neuronal Sirt1 is required for RPC-induced ischemic tolerance.

(A) Representative TTC-stained brain sections following tMCAo (left). Right – infarct quantified as a percentage of the total ipsilateral hemisphere (Control-Veh [n=8], Control-RPC [n=6], Sirt1^neu−/−-Veh [n=9], Sirt1^neu−/−-RPC [n=5],* = p<0.05, ns = not significant, two-way ANOVA – Bonferroni post-test). (B) Neurological scoring, where a lower value indicates better function (* = p<0.05, ns = not significant, two-way ANOVA – Bonferroni post-test). (C) Cerebral blood flow measured by laser doppler flowmetry (ns = not significant, two-way ANOVA – Bonferroni post-test). BL – baseline; ISC – ischemia; REP – reperfusion; Tx – tamoxifen.

Glucose and purine metabolism are altered in Sirt1^neu−/− mouse brain

Since neuronal Sirt1 is necessary for RPC-induced ischemic tolerance, we sought to identify the cellular metabolic pathways of Sirt1 in neurons, which remain largely unknown. To this end, we generated non-targeted primary metabolite profiles from naïve control and Sirt1^neu−/− hippocampi (most efficient Sirt1 deletion, Fig. III) under basal conditions. In total, 227 peaks were captured, 115 of which were structurally annotated metabolites. Statistical analysis using MetaboAnalyst found 23 metabolites significantly altered in Sirt1^neu−/−(p<0.05, False Discovery Rate (FDR) <0.2; Table 1). Full lists of measured metabolites and metrics are located in Tables II and III. Structural similarity networks revealed that the majority were carbohydrates or purines related to glucose metabolism (Fig. 3A). Metabolites associated with ATP consuming processes of glucose and its derivatives were more abundant and included those in glycolysis (early steps – glucose, glucose 6-phosphate, fructose 6-phosphate, fructose 1,6-bisphosphate), pentose phosphate pathway (PPP) and downstream purines (ribitol, inosine monophosphate, inosine, guanosine, adenine, xanthine) and complex carbohydrates (mannose). These changes were concomitant with increased expression of the glucose transporter GLUT1 (38.1%, p<0.05), two ATP consuming glycolytic enzymes hexokinase 2 (Hk2, neuronal enriched²¹ – 86.2%, p<0.05) and phosphofructokinase (PFK – 36.9%, p<0.05), and the rate-limiting enzyme of the PPP, glucose 6-phosphate dehydrogenase (G6PD, 38.7%, p<0.05)(Fig. 3B–C). These data indicate more flux through the initial steps of glycolysis and the PPP. Interestingly, 3-phosphoglycerate, a metabolite associated with the first ATP producing step in glycolysis, and serine, which can be interconverted to 3-phosphglycerate, were less abundant, suggesting that glucose is diverted to branch point pathways and exits glycolysis prior to its complete oxidation. To determine the involvement of transcription factors that regulate glucose metabolism, we measured expression of hypoxia-inducible factor-1 alpha (HIF-1α), a known target of Sirt1²², and two other sirtuins shown to regulate glycolysis. While HIF-1α (41.7%, p<0.01) expression was increased in Sirt1^neu−/−, Sirt2 and Sirt6 had similar expression levels (Fig. 3D). Together, these data show that neuronal Sirt1 regulates glucose metabolism in the brain.

Table 1.

Significantly Altered Metabolites in Sirt1neu−/−^*

Metabolite	p-value	FDR	Fold Change
serine	7.80E-05	0.008965	0.68474
guanosine	0.0012464	0.042976	4.9566
inosine-5'-monophosphate	0.0013771	0.042976	3.6013
ribitol	0.0014948	0.042976	2.3254
xanthine	0.0018844	0.043342	3.351
inosine	0.0025623	0.049111	1.6934
cyclohexylamine	0.005561	0.085437	2.5741
adenine	0.0068764	0.085437	1.5975
phosphoethanolamine	0.0068935	0.085437	0.68725
glutamine	0.0074293	0.085437	1.793
glucose-6-phosphate	0.009671	0.10111	2.8971
threonine	0.010738	0.1029	0.68093
fructose-6-phosphate	0.012182	0.10496	2.6088
taurine	0.012906	0.10496	3.1329
n-acetyl-d-hexosamine	0.014199	0.10496	1.9403
mannose	0.014749	0.10496	2.5191
cystine	0.015515	0.10496	3.7302
glucose	0.0206	0.13161	2.7584
fructose-1,6-bisphosphate	0.022603	0.13681	1.8943
alanine	0.031272	0.17485	1.3561
glycerol	0.03193	0.17485	0.87199
uridine	0.03837	0.19396	1.2
glucose-1-phosphate	0.038792	0.19396	1.8086

^*Data from non-targeted primary metabolomics analysis of control and Sirt1neu−/− hippocampus. Only structurally annotated metabolites are presented. Fold change is with respect to control

Fig 3. Altered metabolic pathways in Sirt1^neu−/− mouse brain.

(A) Biochemical analysis of non-targeted, primary metabolomics from control and Sirt1^neu−/−. Metabolites are linked based structural similarity, where color represents fold change from control and shape denotes chemical class. (B–D) mRNA levels from control and Sirt1^neu−/− hippocampi shown as fold change of control (* = p<0.05, ** = p<0.01, t-test, n=5–9). (B) Glucose transporters. (C) Glycolysis and branch point enzymes. (D) Transcription factors known to regulate glycolysis. Tx – tamoxifen.

Glycolytic ATP production is impaired in Sirt1^neu−/− mouse brain

Given that the initial steps of glycolysis, purine synthesis and synthesis of complex carbohydrates are all ATP consuming processes, we hypothesized that in Sirt1^neu−/−glucose is being preferentially diverted to these ATP consuming pathways rather than being fully oxidized as an energy substrate. To test glycolytic ATP production, we measured latency until anoxic depolarization (AD) as a readout of glycolytic function. Studies show that when mitochondria are inhibited by anoxia, glucose can extend latency to AD through glycolytic ATP production²³. We measured AD latency by extracellular field recording in the hippocampal CA1 region of acute brain slices. Gross synaptic transmission did not differ between control and Sirt1^neu−/− slices, evidenced by similar evoked field population spike (fPS) amplitudes (Fig. 4A) and expression of Na⁺/K⁺ ATPase subunits (Fig. IV). Latency to AD under anoxia and perfusion of artificial cerebrospinal fluid (aCSF) with no glucose was similar between control and Sirt^neu−/− (control – 3.60 min, Sirt1^neu−/− − 3.36 min, Fig. 4B). In controls, perfusion with 20 mmol/l glucose aCSF under anoxia significantly extended latency to AD (8.31 min, p<0.001), suggesting normal glycolytic ATP production that contributes substantially to AD latency. The glycolysis inhibitor iodoacetate (IAA) abolished the extended latency to AD with 20 mmol/l glucose in control (1.74 min, p<0.001), confirming that this effect is mediated through glycolysis. In Sirt1^neu−/−, however, perfusion with 20 mmol/l glucose aCSF failed to extend latency to AD, indicating impaired glycolytic ATP production compared to control (Fig. 4B). Together, these data suggest that neuronal Sirt1 promotes glycolytic ATP production.

Fig 4. Glycolytic ATP production is impaired in Sirt1^neu−/− mouse brain.

Example field population spikes (fPS) stimulated maximally from control and Sirt1^neu−/− acute hippocampal slices (B) Top – examples traces of the anoxic depolarization (AD) event under different experimental conditions (IAA = iodoacetate, glycolysis inhibitor). Bottom – quantification of latency to AD, used as a readout of glycolytic function when glucose is present under anoxia (*** = p<0.001, two-way ANOVA, Bonferroni post-test, n=3–4; n/m = not measured). Tx – tamoxifen.

RPC increases glycolytic rate in neurons in a Sirt1-dependent manner

Next, we postulated that Sirt1 positively regulates glycolysis as an energy pathway in brain. To this end, we measured glycolytic rate in rat primary neuronal-enriched cultures using the Seahorse XFp and inhibited Sirt1 pharmacologically given the low deletion efficiency of inducible cre/lox systems in vitro and to avoid any confounding effects of tamoxifen. Under basal conditions, inhibition of Sirt1 with the specific inhibitor EX-527 (10 µM) for 2 days prior to measurements modestly reduced (12.1%, p<0.01) glycolytic rate. EX-527 also inhibited maximal glycolytic rate (7.3%, p<0.05) in response to energetic stress, elicited by application of the mitochondrial uncoupler FCCP and complex V inhibitor oligomycin (Fig. V). Activation of Sirt1 with RPC treatment (100 µM resveratrol, a concentration previously shown to protect neurons from ischemia in vitro⁸) 2 days prior to measurements increased basal (20.4%, p<0.001) and maximal (29.2%, p<0.001) glycolytic rate compared to Veh (Fig. 5A). Inhibition of Sirt1 at the time of RPC treatment significantly reduced these increases (basal – 12.1%, p<0.05, maximal – 10.1%, p<0.01), demonstrating the Sirt1-dependence of these effects (Fig. 5A). Enhanced glycolytic function could be important in ischemic penumbra, where energy substrates are significantly decreased but still present. Thus, we tested whether these RPC-induced changes in glycolytic rate occur under conditions of low glucose and mitochondrial dysfunction. Indeed, when cells were exposed to 25% of normal glucose (5 mmol/l glucose instead of 20 mmol/l) and mitochondria were inhibited, glycolytic rate was increased in RPC neurons (16.6%, p<0.001, Fig. 5B). Together these data demonstrate that RPC increases glycolytic rate in neurons through Sirt1, an effect that was evident in ischemic conditions.

Fig 5. RPC increases glycolytic rate in a Sirt1-dependent manner in neurons.

Measurement of glycolytic rate (ECAR – extracellular acidification rate) by Seahorse XFp in primary neuronal-enriched cultures. Energetic stress cocktail: mitochondrial uncoupling – FCCP (2 µM); Complex V inhibition – oligomycin (1 µM); arrow indicates application. (A) Naïve cells treated with RPC, Veh, Sirt1-specific inhibitor EX-527 or a combination. Measurements were taken in normal glucose (20 mmol/l) (* = p<0.05, ** = p<0.01, *** = p<0.001, ns = not significant, one-way ANOVA, Bonferroni post-test, n=5). (D) RPC treatment of naïve cells exposed to 25% of normal glucose (5 mmol/l, *** = p<0.001, ns = not significant, t-test, n=4).

DISCUSSION

Although the efficacy of resveratrol is demonstrated across species, sex, age and models of cerebral ischemia, its mechanism(s) are not well understood spatially and temporally. Studies indicate that modulation of both systemic processes and brain or neuronal functions can afford robust protection. What’s more, the Sirt1-dependence of resveratrol-elicited protection has not been investigated fully. Here, we found that RPC-induced protection is lost in the absence of neuronal Sirt1, demonstrating a robust mechanism that either originates or terminates in neurons. Importantly, cerebral blood flow during ischemia or at the onset of reperfusion was unaffected by deletion of neuronal Sirt1 or RPC. Still, Sirt1promotes angiogenesis through the migration and sprouting of endothelial cells²⁴ and can maintain cerebral blood flow through deacetylation of eNOS¹¹. Mice overexpressing Sirt1 are resistant to global ischemia by retaining cerebral perfusion up to 45–50% of baseline²⁵. Moreover, resveratrol elicits pro-angiogenic effects in brain endothelial cells through MAPK signaling and upregulation of eNOS⁷. Our data indicate that neuronal activities of Sirt1 take precedence RPC-induced ischemic tolerance, most likely orchestrating and acting in concert with systemic and non-neuronal effects. This conclusion is supported by several studies that implicate neuronal function in RPC^{4, 8, 9, 12}. The temporal distribution of resveratrol-elicited effects is also a probable factor in determining their precedence and warrants further investigation.

Interestingly, Sirt1^neu−/− mice displayed similar infarct sizes and neurological scores compared to control mice, indicating a negligible innate role for neuronal Sirt1 in focal ischemic injury. In line with this finding, mice overexpressing Sirt1 specifically in neurons were not protected from tMCAo, evidenced by infarct size and functional outcome comparable to WT mice²⁶. However, Sirt1 null mice are more susceptible to permanent MCAo²⁷ and mice that constitutively overexpress Sirt1 in all tissues are resistant to bilateral common carotid artery occlusion²⁵, a model of global ischemia, suggesting that Sirt1 does contribute to endogenous ischemic neuroprotection. This discrepancy likely stems from several factors that make comparison across these studies difficult: three different ischemia models were employed (transient and permanent MCAo, global ischemia) that have different pathophysiology, injury severity ranged from mild to severe across these models, outcome measures were evaluated at different time points post-ischemia, and constitutive Sirt1 null mice have neurodevelopmental defects that may confound results. A detailed assessment of neuroprotection in Sirt1^neu−/− mice with titrated injury severity and long-term recovery will more accurately evaluate the innate role of neuronal Sirt1.

In the current study we identify a novel role for neuronal Sirt1 in the regulation of brain glycolysis. In liver, under metabolic stress, Sirt1 and PGC-1α work in concert to suppress glycolytic genes (glucokinase, pyruvate kinase), promote gluconeogenic genes and ultimately the production of glucose²⁸. Our results suggest that in the brains of Sirt1^neu−/− mice, PGC-1α activity might not be altered since expression of its target gene pyruvate kinase was unchanged. However, we observed increases in expression of the HIF-1α target genes GLUT1, HK2 and PFK. In line with this result, Sirt1 and HIF-1α are known to interact in brain, where Sirt1-mediated deacetylation of HIF-1α at Lys674 inactivates it and represses its target glycolytic genes²². Alternatively, PGC-1α and another well-characterized target of Sirt1 deacetylation, p53, regulate mitochondrial metabolic pathways. PGC-1α null mouse brains have decreased expression of oxidative genes²⁹ and p53 is involved in various aspects of mitochondrial quality control³⁰. However, Sirt1^neu−/− mice did not display alterations in mitochondria-associated metabolites such as TCA cycle intermediates or fatty acids, suggesting that these Sirt1 targets may be less robust in brain or have compensatory regulation in the absence of neuronal Sirt1. Still, several purine metabolites were significantly increased in Sirt1^neu−/− and could reflect a more subtle defect in mitochondrial metabolism since many of the rate-limiting and ATP-dependent reactions for purines are localized to mitochondria³¹. More specific experiments concerning the effects of neuronal Sirt1 deletion on mitochondrial bioenergetics are needed to address this issue.

In contrast to peripheral tissues, the brain relies heavily on glucose for energy production and thus Sirt1 regulation of glycolysis may differ in this tissue. Whereas Sirt1 deficiency increases glycolytic activity in Th9 cells³², we found that RPC increases glycolytic rate in a Sirt1-dependent manner in neurons, demonstrating a cell type-specific function. While the bulk of brain ATP is generated from the oxidation of pyruvate within mitochondria, recent studies highlight the importance of neuronal glycolysis. Local production of ATP from glycolysis is crucial for fast axonal transport of vesicles³³, the energetic demand of action potential firing³⁴ and the maintenance of synaptic ATP levels under energetic stress³⁵. Our results demonstrate that under mitochondrial stress by anoxia, glycolysis can produce a significant amount of ATP to maintain ion gradients and delay depolarization, an effect that was lost with deletion of neuronal Sirt1. Moreover, studies show that the Na⁺/K⁺ ATPase, which is a main component of ion gradient maintenance in the brain, is preferentially fueled by glycolytic ATP production over mitochondrial-generated ATP³⁶.

More efficient utilization of glucose for energy production could greatly benefit neurons in the ischemic penumbra, where blood flow has been significantly restricted but some energy substrates still reach the affected tissue. We observed that RPC also increases glycolytic rate under low glucose (25% of normal) and mitochondrial stress conditions, supporting a physiological relevance of these effects in ischemia. Consistent with these findings, resveratrol stimulates glycolytic ATP synthesis in yeast³⁷ as well as rodent liver³⁸ and has been shown to inhibit the PPP³⁹. Together, our findings illustrate a neuroprotective mechanism whereby RPC via neuronal Sirt1 promotes glycolytic efficiency to combat energetic stress. These results have far-reaching implications, as dysregulation of glycolysis plays a role in the pathogenesis and/or manifestation of Huntingtin’s disease, Alzheimer’s disease and Parkinson’s disease, in all of which Sirt1 has a neuroprotective role^{40, 41}.

In terms of clinical efficacy, the translatability of resveratrol for the treatment of ischemic stroke is promising. It is safe for use in humans above the equivalent preconditioning doses⁴², desensitization to prolonged treatment does not occur¹² and protection has been validated from oral administration⁵. Moreover, it affords protection and recovery when administered post-injury⁵. A combination of prophylactic treatment, for the large population at risk for stroke, and treatment post-injury will likely prove a robust therapeutic strategy. Sirt1 is also an attractive therapeutic target with translational potential. Aside from resveratrol, recently developed Sirt1-specific activator compounds (SRT1720 and SRT2014) allow for direct, specific and more robust activation of Sirt1⁴³, making them ideal candidates for the induction of Sirt1-mediated neuroprotection.