NAD⁺ precursor modulates post-ischemic mitochondrial fragmentation and reactive oxygen species generation via SIRT3 dependent mechanisms

Abstract

Global cerebral ischemia depletes brain tissue NAD⁺, an essential cofactor for mitochondrial and cellular metabolism, leading to bioenergetics failure and cell death. The post-ischemic NAD⁺ levels can be replenished by the administration of nicotinamide mononucleotide (NMN), which serves as a precursor for NAD⁺ synthesis. We have shown that NMN administration shows dramatic protection against ischemic brain damage and inhibits post-ischemic hippocampal mitochondrial fragmentation. To understand the mechanism of NMN-induced modulation of mitochondrial dynamics and neuroprotection we used our transgenic mouse models that express mitochondria targeted yellow fluorescent protein in neurons (mito-eYFP) and mice that carry knockout of mitochondrial NAD⁺-dependent deacetylase sirt3 gene (SIRT3KO). Following ischemic insult, the mitochondrial NAD⁺ levels were depleted leading to an increase in mitochondrial protein acetylation, high reactive oxygen species (ROS) production, and excessive mitochondrial fragmentation. Administration of a single dose of NMN normalized hippocampal mitochondria NAD⁺ pools, protein acetylation, and ROS levels. These changes were dependent on SIRT3 activity, which was confirmed using SIRT3KO mice. Ischemia induced increase in acetylation of the key mitochondrial antioxidant enzyme, superoxide dismutase 2 (SOD2) that resulted in inhibition of its activity. This was reversed after NMN treatment followed by reduction of ROS generation and suppression of mitochondrial fragmentation. Specifically, we found that the interaction of mitochondrial fission protein, pDrp1(S616), with neuronal mitochondria was inhibited in NMN treated ischemic mice. Our data thus provide a novel link between mitochondrial NAD⁺ metabolism, ROS production, and mitochondrial fragmentation. Using NMN to target these mechanisms could represent a new therapeutic approach for treatment of acute brain injury and neurodegenerative diseases.

1. Introduction

Several lines of evidence suggest that NAD⁺ degradation and associated bioenergetics failure of cellular metabolism is one of the major factors leading to post-ischemic brain damage. NAD⁺ is an essential cofactor in most enzymatic reactions supporting fundamental mitochondrial functions including oxidative phosphorylation and enzymatic reactions of the tricarboxylic acid cycle (TCA cycle) (Belenky et al., 2007; Kristian et al., 2011; Owens et al., 2013). When NAD⁺ is degraded, mitochondria become incapable of ATP synthesis. Furthermore, there are several NAD⁺-dependent enzymes that utilize NAD⁺ as a substrate for their functions, including, poly(ADP-ribose) polymerase 1 (PARP1), the ADP-ribosyl cyclase (CD38), and class III histone deacetylases (sirtuins) (Belenky et al., 2007; Kristian et al., 2011; Owens et al., 2013). It has been proposed that post-ischemic, uncontrolled PARP1 activation might deplete intracellular NAD⁺ and consequently ATP, leading to mitochondrial depolarization and cell death (Chiarugi and Moskowitz, 2003). We also reported that CD38, a major NAD⁺ glycohydrolase, is activated following ischemic insult and contributes to significant degradation of cellular NAD⁺ pools (Long et al., 2017). Finally, sirtuins require cytosolic and mitochondrial NAD⁺ for their activity (Michan and Sinclair, 2007; Owens et al., 2013; Imai and Guarente, 2014; Yoon and Eom, 2016).

Ischemic insult also leads to excessive mitochondrial fragmentation that can further compromise cellular bioenergetic metabolism and contribute to cell death mechanisms (Owens et al., 2015; Kumar et al., 2016). Excessive fission results in mitochondrial depolarization, cyto-chrome c release, increased free radical production and cell death (Knott et al., 2008; Tanaka and Youle, 2008; Stetler et al., 2013). It is commonly accepted that ischemia/reperfusion is associated with increased reactive oxygen species (ROS) production that mediates, at least in part, the post-ischemic brain injury (Sugawara and Chan, 2003). Mitochondria is considered one of the main sources of ROS during the reperfusion period (Chouchani et al., 2014; Abramov et al., 2007).

To preserve post-ischemic NAD⁺ levels one can directly stimulate NAD⁺ synthesis by administration of nicotinamide mononucleotide (NMN) (Belenky et al., 2007, Yoshino et al., 2011, Klimova et al., 2019). NMN is adenylated to form NAD⁺ by nicotinamide nucleotide adenylyltransferase (NMNAT) (Belenky et al., 2007). We have recently shown that by using NMN as a direct substrate for NAD⁺ synthesis one can dramatically ameliorate ischemic brain damage induced by global cerebral ischemia (Park et al., 2016). Interestingly, we found that in addition to the dramatic neuroprotection effect against ischemic insult, NMN also inhibits PARP1 and CD38 NAD⁺ glycohydrolase activity (Park et al., 2016; Long et al., 2017). Thus, NMN-induced preservation of post-ischemic NAD⁺ pools is probably due to multitargeted effect of this metabolite (Klimova and Kristian, 2019). Furthermore, we demonstrated that when naïve mice were treated with NMN neuronal mitochondrial dynamics were shifted towards fusion (Long et al., 2015), and that NMN induced increase in mitochondrial NAD⁺ levels, reduced acetylation of mitochondrial proteins, and ROS generation in hippocampal tissue (Klimova et al., 2019). Therefore, we decided to determine whether NMN treatment will also reverse the ischemia-induced mitochondrial fragmentation and reduce the post-ischemic ROS generation.

2. Methods

2.1. Animals

We used adult, 3 month old male, C57Bl6 wild type (WT), SIRT3KO mice (B6.129S6(Cg)-Sirt^tm1.1Fwa/j) that were obtained from Jackson Laboratories and our transgenic animals that expresses mitochondria-targeted yellow fluorescence protein (mito-eYFP) in neurons (Chandrasekaran et al., 2006). We also used mito-eYFP-SIRT3KO mice generated by crossbreeding the mito-eYFP animals with SIRT3KO mice (Klimova et al., 2019). All animal experiments were performed in accordance with the Guide for the Care and use of Laboratory Animals of the National Institute of Health and were approved by the Animal Care and use Committee of the University of Maryland Baltimore. The animals were maintained in a 12-h light/dark cycle and were housed in temperature-controlled room at 23 °C. All cages contained bedding and nesting material for environmental enrichment. The mice were free of all viral, bacterial, and parasitic pathogens. All mice were allowed free access to water and a maintenance diet. Mice were divided at random into several groups for all experiments.

2.2. Transient forebrain ischemia

Forebrain ischemia was induced as described in (Onken et al., 2012), see also (Park et al., 2016). Briefly, before the surgery, the animals were fasted with free access to tap water. Under isoflurane anesthesia common carotid arteries (CCA) were isolated and the ischemia was induced by clamping the CCA with micro-vessel clamps with concomitant reduction of mean arterial blood pressure (MABP) to 30 mmHg. Both, core and head temperature were monitored and maintained at 37.0 ± 0.5 °C. Following 10 min ischemic period the clamps were removed, anesthesia discontinued, and the animals were placed into a temperature-controlled incubator for 3 h before they were moved to pre-heated cages. Sham operated animals underwent the same surgical procedure without the occlusion of common carotid arteries. The ischemia surgery schedule was adjusted to accommodate for NAD⁺ content determination at the desired time of the day.

2.3. NMN administration

Nicotinamide mononucleotide (NMN) (Sigma N3501) was prepared in 0.2 mL sterile phosphate buffered saline (PBS) and was administered to mice at a dose of 62.5 mg/kg (Park et al., 2016). The drug or vehicle solution (PBS) was injected intraperitoneally (i.p.) using 29G needle 30 min after the start of reperfusion. NMN and vehicle solution were prepared the same day as administered. To avoid the effect of circadian rhythm on brain tissue NAD⁺ levels (Peek et al., 2013), all samples were collected at the same time of the day between 9 and 10 AM.

2.4. Mitoquinol and Apocynin administration

Vehicle treated post-ischemic animals were i.p. administered with mitoquinol (MitoQ) (Cayman Chemical #89950) (4 mg/kg) at 22 h of reperfusion. MitoQ was prepared in a solution of 1.8% DMSO in filtered PBS (200 μL). Apocynin, an NADPH oxidase (NOX) inhibitor was prepared as a stock solution in DMSO (30 mg/mL) and then diluted with filtered PBS prior to administration. Animals received a dose of 5 mg/kg (200 μL volume) as i.p. injection at 30 min of recovery. All drugs administration, sample collection and processing was performed by investigator that was blinded to the identity of the experimental groups.

In vivo reactive oxygen species (ROS) detection by dihydroethidium (DHE).

For ROS detection studies in vivo we administered to animals dihydroethidium that is oxidized to fluorescent hydroxyethidium by reactive oxygen species (Zielonka et al., 2008). DHE (Axxora) stock solution was prepared by dissolving 10 mg in 200 μL DMSO (50 mg/mL) and stored at −20 °C. The stock was diluted in filtered PBS for administration. DHE (50 mg/kg) was injected i.p. (Hall et al., 2012). Sham operated, vehicle or NMN treated post-ischemic animals received DHE at 1 h or at 23 h of recovery. In another group of animals, DHE was injected 10 min before the onset of ischemia at 25 mg/kg dose to measure ROS generation during the early first hour of reperfusion.

In all experimental groups, 1 h after the DHE administration mice were either perfusion fixed with 4% paraformaldehyde and their brain was removed from skull and processed for histology or the animals were decapitated, and the hippocampi were isolated on ice and homogenized in isolation medium (see below) with protease inhibitors. Perfusion-fixed brains were further processed by cutting coronal sections of 40 μm on a freezing microtome. Brain sections were mounted on slides and hippocampal images were collected using Zeiss LSM 510 laser scanning confocal microscope. Z-stack images were obtained from the CA1, CA3, and DG neuronal cell body layer of the hippocampal tissue from sham operated and vehicle or NMN treated post-ischemic animals. Four images were taken from each section. Images were analyzed with Volocity software (PerkinElmer, Waltham, MA). Mitochondrial localization of ROS was determining by the Pearson’s colocalization coefficient between mito-eYFP and hydroxyethidium signal and the total intensity of all objects within the recorded volume was counted and normalized to the unit volume (μm³).

ROS generated fluorescence signal (red fluorescent hydroxyethidium) from homogenized hippocampal tissue was also measured using Fluostar Optima plate reader. Excitation wavelength was set at 460 nm and emission wavelength set at 590 nm. Samples were run in triplicates. Data are reported as fluorescence units/μg protein. Protein concentration of samples was determined by Lowry method using bovine serum albumin (BSA) as standard.

2.5. Hippocampal mitochondria isolation

Non-synaptic hippocampal mitochondria were isolated from sham, vehicle or NMN treated ischemic animals at 2, 4, and 24 h of reperfusion as described in (Kristian et al., 2000; Kristian, 2010). Briefly, following decapitation, hippocampi were isolated on ice and then homogenized in ice-cold isolation medium (225 mM sucrose, 75 mM mannitol, 1 mM EGTA, 5 mM Hepes) with protease inhibitors (AG Scientific T-2495), acetylation inhibitors (10 μM TSA, 10 mM nicotinamide, 10 mM sodium butyrate), and PARP1 inhibitors (2 mM Naorthovanadate, 5 mM NaF, 5 mM glycerol-2-phosphate, 1 μM ADP-HPD, 40 μM PJ-34). Homogenate was mixed 1:1 with 30% percoll, loaded on a 24% – 40% percoll gradient, and centrifuged at 30,700g for 8 min at 4 °C. The layer at the 24% and 40% percoll interface was collected as purified non-synaptic mitochondria. The collected fraction was resuspended in isolation medium and centrifuged at 16,700g for 10 min at 4 °C. Supernatant was removed, isolation medium was added to the pellet, and centrifuged at 6900g for 10 min at 4 °C. Finally, isolation medium lacking EGTA was added to the obtained pellet and used for metabolite measurement or western blots. Hippocampi from one mouse yielded to about 0.25 mg of mitochondrial protein. The whole sample was used for NAD⁺ extraction procedure.

2.6. Blood sample collection for metabolite determination

Arterial blood was collected at 2, 4, and 24 h after the start of reperfusion from sham, vehicle, and NMN treated ischemic animals. Plasma was isolated by centrifugation of blood at 10,000g for 10 min at 4 °C. The supernatant was then stored at −80 °C till further processing.

Extraction of metabolites.

Metabolites from isolated mitochondria and plasma were extracted by 7% ice-cold perchloric acid (PCA) and were incubated on ice for 15 min. Samples were then centrifuged at 10,000g for 10 min at 4 °C. The pellet was used for Lowry protein measurements while the supernatant was neutralized with 1 M Trizma and 9 M KOH. The supernatant was then filtered by 0.22 μm PVDF filter and used for NAD⁺ determination by enzymatic cyclic assay (Kristian and Fiskum, 2004).

2.7. Measurement of mitochondrial NAD⁺ levels

NAD⁺ content in plasma and mitochondrial samples were measured by a cycling enzymatic assay that generates a fluorescent product (Kristian and Fiskum, 2004, Klimova et al., 2019). Levels were normalized to the respective protein amount.

2.8. Western blots

After hippocampal dissection and mitochondrial isolation was performed twenty-five μg of mitochondrial protein were heated at 75 °C, loaded into Bio-Rad 10 well mini-protean TGX precast gels and separated through sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). trans-Blot Turbo system (Bio-Rad) was used to transfer proteins from gels to immobilon PVDF-FL membranes. Membranes were incubated in Odyssey blocking buffer (Li-cor Biosciences) for 1 h and then incubated in primary antibody overnight at 4 °C (acetylated lysine (1:1000, Cell Signaling #9814); SIRT3 (1:1000, Cell Signaling #5490); VDAC (1:5000, Abcam #14734); VDAC (1:5000, Cell Signaling #4661); MFN1 (1:1000, ProteinTech #13798–1-AP); MFN2 (1:1000, Abcam #56889); OPA1 (1:1000, BD Biosciences #612606); SOD2 (1:1000, ProteinTech #24127–1-AP); pDrp1 (S616) (1:1000, Cell Signaling #4494); Drp1 (1:1000, Abcam #56788)). Membranes were washed in phosphate buffered saline with 0.1% tween-20 (PBS-T) and then incubated in infrared (IR) fluorophore conjugated secondary antibody (Li-Cor) for 30 min in the dark at room temperature (800CW Goat anti-Mouse IgG (1:50,000, Li-cor #925–32,210); 680RD Goat anti-Rabbit IgG (1:10,000, Li-cor #925–68,071)). Membranes were washed with PBST, PBS, and scanned using Odyssey infrared imaging system (Li-Cor). Bands were quantified with Li-cor imaging system and VDAC signal was used as normalization control in samples. All antibodies were titrated to appropriate working concentrations before use.

2.9. Immunoprecipitation

Hippocampal tissue was dissected from sham operated or vehicle/NMN treated ischemic animals 24 h after reperfusion and homogenized in 500 μL RIPA buffer containing protease and acetylation inhibitors (10 μM TSA, 10 mM nicotinamide, 10 mM sodium butyrate). The volume was adjusted to 1.5 mL with PBS and concentration was measured by Lowry assay. Four mg of hippocampal tissue was incubated with protein A sepharose bead slurry (Sigma #P9424) for 30 min on a rotator at 4 °C, then centrifuged at 1000g for 3 min at 4 °C, and supernatant transferred to a new tube. Four μg of SOD2 antibody (Protein Tech #24127–1-AP) was added to pre-cleared supernatant and incubated rocking for 12 h at 4 °C. Next, 50 μL of the protein A sepharose bead slurry was added and incubated rocking for 12 h at 4 °C. The immunocomplex-beads were washed 3 times with 1 mL 0.2% TBSTriton X-100. SOD2 protein was eluted using elution buffer (0.1 M Glycine, 0.05 M Tris-HCl, 0.5 M NaCl, pH 1.5–2.5) and neutralized by 10× PBS. 4× Laemmli Sample Buffer was added to the eluted protein samples and heated at 95 °C for 5 min. Western blot analysis was performed using primary antibodies for SOD2 (1:1000, ProteinTech #24127–1-AP) and acetylated lysine (1:1000, Cell Signaling #9814) and normalized to the amount of SOD2 protein.

2.10. Immunohistochemistry

Sham, vehicle or NMN treated ischemic mito-eYFP mice were perfusion fixed under deep anesthesia 24 h after the start of reperfusion. Mice were intubated and ventilated and then transcardialy perfused for 1 min with oxygenated cold PBS. This was followed by perfusion with warm (37 °C) 4% paraformaldehyde in PBS for 7 min (Owens et al., 2015). Afterwards brains were post-fixed overnight in paraformaldehyde at 4 °C and the next day transferred into 30% sucrose for 3 days. Coronal sections of 40 μm were cut on a freezing microtome and collected. Sections were washed with KPBS and incubated with pDrp1 (S616) antibody (1:500, Cell Signaling #4494) at 4 °C overnight in 0.3% KPBS-T (Triton X-100). Afterwards, sections were washed with KPBS and incubated in goat anti-rabbit (Alexa Fluor 594) secondary antibody (1:600, ThermoFisher #A-11012) for 1 h shaking at room temperature. After subsequent washes with KPBS sections were mounted on slides for analysis using laser scanning confocal microscopy. Four Z-stack images were obtained from the CA1 stratum oriens, CA3 stratum oriens, and dorsal molecular layer of the DG for each experimental group. Volocity software was used to measure colocalization by determining the Pearson’s coefficient between mito-eYFP and pDrp1 (S616) signal.

2.11. Mitochondrial dynamics quantification

Mitochondrial length in sham operated, vehicle or NMN treated ischemic mito-eYFP mice was measured 24 h after the start of reperfusion. Z-stack images of mitochondria from hippocampal sub-regions, CA1 stratum oriens, CA3 stratum oriens, and dorsal molecular layer of the DG, (4 images/region) were collected using a laser scanning confocal microscope. Volocity software was used to analyze and quantify images. While defining individual mitochondria the program measures morphometric parameters of identified objects including skeletal length and shape factor (measure of circularity, 1 being spherical) of each individual mitochondrion. After measurements were completed, mitochondria were sorted based on length and shape factor and data exported to excel for final calculations and graph constructions.

2.12. Laser scanning confocal microscopy

A Ziess LSM 510 laser scanning confocal microscope using a Plan-Apochromat 63×/1.4 oil lens was used to acquire images from brain sections. Single planes of 1024 × 1024 pixels were recorded at 1.0–1.5 Airy unit pinhole every 0.5 μm z-spacing throughout the whole tissue section. For mitochondrial length analysis 0.2 μm z-interval was used.

2.13. Statistics

Statistical analysis was performed using Prism (GraphPad) version 6.0c. All data are expressed as averages ± standard error of mean (SEM). Data was assumed normally distributed based on the skewness or kurtosis parameter. Statistical significance was assessed by Student t-test when two groups were compared, and one-way or two-way ANOVA test followed by the Tukey HSD test for multiple comparisons. Figure legend indicates which test was used for each corresponding experiment. The p values < .05 were considered to be statistically significant. However, due to small sample sizes we confirmed any significant differences between experimental groups also using non-parametric Kruskal Wallis test, or Mann-Whitney U test.

3. Results

3.1. NMN administration prevents post-ischemic depletion of mitochondrial NAD⁺ and consequent increase in mitochondrial protein acetylation

We previously showed that NMN administration significantly reduced post-ischemic NAD⁺ degradation in hippocampal tissue (Park et al., 2016). To determine whether the mitochondrial NAD⁺ pools are also affected by NMN treatment we isolated mitochondria from mouse hippocampus after ischemic insult and measured the mitochondrial NAD⁺ content. As Fig. 1A shows, 4 h after ischemia mitochondrial NAD⁺ levels were significantly reduced from 5.9 nmol/mg protein to 3.5 nmol/mg protein and this decrease in NAD⁺ content was even more pronounced at 24 h of recovery (2.5 nmol/mg protein) (Fig. 1A). NMN administration prevented the NAD⁺ degradation, resulting in no significant change in mitochondrial NAD⁺ content following ischemia.

Fig. 1.

NMN administration reverses post-ischemic decline in hippocampal mitochondria NAD⁺ levels and induces SIRT3 mediated decrease in protein acetylation. (A) NAD⁺ levels were measured in isolated hippocampal mitochondria in sham operated and post-ischemic vehicle (PBS) or NMN (62.5 mg/kg) treated animals at 2, 4, and 24 h of recovery. NMN administration resulted in no significant change in mitochondrial NAD⁺ content following ischemia when compared to sham levels. *** p < .001 when compared to sham, ## p < .01 compared to 4 hR vehicle, ### p < .001 compared to 24 hR vehicle, two-way ANOVA followed by Tukey’s HSD test, n = 4–5 animals/group. (B) Mitochondrial protein acetylation changes were analyzed in post-ischemic vehicle and NMN treated animals. NMN normalized post-ischemic mitochondrial protein acetylation levels. *** p < .001 when compared to sham, # p < .05 when compared to corresponding vehicle treated animals, ### p < .001 when compared to corresponding vehicle treated animals, two-way ANOVA followed by Tukey’s HSD test, n = 6 animals/group. (C) Mitochondrial protein acetylation in SIRT3KO animals following ischemia in vehicle and NMN treated animals. Neither Ischemic insult nor NMN administration affected mitochondrial protein acetylation in SIRT3KO mitochondrial samples, one-way ANOVA followed by Tukey’s HSD test, n = 4 animals/group. (D) SIRT3KO mice show elevated levels of mitochondrial NAD⁺ when compared to wild-type (WT) mice. * p < .05, Student t-test, n = 4 animals/group. (E) NAD⁺ levels in plasma of post-ischemic animals injected with NMN. NMN was administered 30 min after the start of reperfusion. Blood was collected at 2 and 4 h of recovery. Plasma of NMN treated animals show an increase in NAD⁺ levels at 2 h of reperfusion that returned to normal physiologic levels 4 h after the start of recovery. ## p < .01 when compared to 2 hR vehicle. ** p < .01 when compared to sham, two-way ANOVA followed by Tukey’s HSD test, n = 4–5 animals/group.

Sirtuins’ deacetylase activity depends on NAD⁺, therefore we decided to examine the changes in acetylation of mitochondrial proteins in post-ischemic samples. As expected, there was a dramatic increase of protein acetylation at 4 and 24 h after ischemia (Fig. 1B). This increase in acetylation was partially inhibited by NMN at 4 h of recovery and then fully reversed at 24 h of recovery, when acetylation returned to normal pre-ischemic levels (Fig. 1B). To confirm that the changes in mitochondrial protein acetylation are controlled by mitochondrial deacetylase SIRT3 (Anderson and Hirschey, 2012, Ahn et al., 2008, Klimova et al., 2019), we subjected SIRT3KO animals to ischemic insult and treated them with vehicle or NMN. As expected, neither ischemic insult nor NMN administration affected the mitochondrial protein acetylation in SIRT3KO samples (Fig. 1C). However, although the mitochondrial protein acetylation is significantly increased in SIRT3KO samples obtained from naïve animals (Klimova and Kristian, 2019), the mitochondrial NAD⁺ content was about 4 nmol/mg protein, thus significantly higher when compared to wild type (WT) naïve mice (about 2.8 nmol/mg protein, Fig. 1D).

To determine the effect of NMN i.p. injection on the plasma NAD⁺ levels post-ischemia we collected blood samples from NMN injected mice and measured the plasma NAD⁺ content. As Fig. 1E shows there was about 30% increase in plasma NAD⁺ at 2 h of reperfusion that returned to normal physiologic levels 4 h after the start of recovery.

3.2. NMN reverses post-ischemic mitochondrial fragmentation

Using our transgenic animals models we demonstrated that transient global ischemia leads to permanent extensive fission of neuronal mitochondria in CA1 neurons (Owens et al., 2015). Furthermore, we showed that when naïve animals received NMN there was a shift in mitochondria length distribution towards longer organelles (Long et al., 2015, Klimova et al., 2019). Therefore, we decided to examine the effect of NMN treatment on ischemia-induced mitochondrial dynamics. As Fig. 2A demonstrates, ischemia resulted in massive fragmentation of mitochondria at 24 h of recovery. The relative number of small spherical organelles (0.2–1 μm length) increased from 30 to 50%, and the number of rod-shaped (1–5 μm length) mitochondria was reduced from 70 to 50% (Fig. 2B). However, the post-ischemic mitochondrial fragmentation was inhibited in animals treated with NMN (Fig. 2B). The relative distribution of mitochondrial length in animals that received NMN was not significantly different form the control sham operated mice (Fig. 3B). To quantify the structural changes of mitochondrial organelles we also calculated the shape factor of individual mitochondria in different hippocampal subregions. The shape factor has a numerical value between 1 and 0, where number 1 represents an object with perfect spherical shape and the value closer to 0 indicates that the object is a more rod-shape. Fig. 2C shows the relative number of CA1 mitochondria with spherical-like shape (shape factor 0.7–1.0), and rod-shape (shape factor 0.35–0.65) in control, sham operated animals and post-ischemic vehicle or NMN treated mice. At 24 h of recovery the number of rod-shape mitochondria decreased and the number of spherical increased in vehicle treated animals. This change in shape factor was not observed after NMN administration (Fig. 2C).

Fig. 2.

NMN administration leads to inhibition of ischemia-induced mitochondrial fragmentation in CA1 neurons. (A) Transgenic mice with neuron-specific expression of mitochondria targeted enhanced yellow fluorescence protein (mito-eYFP) were subjected to 10 min of transient global cerebral ischemia. Thirty minutes after ischemia an intraperitoneal (i.p.) injection of vehicle (PBS) (24 hR) or NMN (24 hR NMN) was administered and animals were perfusion-fixed at 24 h of recovery. Mitochondria were visualized and their morphometric parameters measured in the hippocampal CA1 oriens. Mitochondria in NMN treated ischemic brain appeared less fragmented when compared to mitochondria in vehicle treated animals. Scale bar represents 100 μm. (B) Quantification of mitochondrial fragmentation. Following recording of z-stack images by confocal microscope the data were processed and analyzed by Volocity software. Mitochondria were divided into 3 populations based on their length (0.2–1 μm, 1–5 μm, and 5–15 μm). The relative distribution of individual mitochondria populations show that ischemic insult increases the number of short spherical mitochondria (0.2–1 μm) while decreasing the number of longer rod-shaped mitochondria (1–5 μm and 5–15 μm) when compared to sham. NMN treated ischemic animals did not display this shift in mitochondrial length population and retained physiological morphology. *** p < .001 compared to sham (0.2–1 μm); ### p < .001 compared to 24 hR (0.2–1 μm); ^^^ p < .001 compared to sham (1–5 μm); $ $ $ p < .001 compared to 24 hR (1–5 μm); @ p < .05 compared to sham (5–15 μm); & p < .05 compared to 24 hR (5–15 μm), one-way ANOVA followed by Tukey’s HSD test, n = 16 images/group. (C) Relative counts of mitochondria with shape factor (number one represents spherical shape and numbers close to 0 a rod-like elongated shape). Ischemic insult increased the population of spherical mitochondria compared to sham. NMN treated ischemic animals retained physiological levels of the mitochondrial shape factor. * p < .05 compared to sham (0.35–0.65); ## p < .01 compared to 24 hR (0.35–0.65); ^ p < .05 compared to sham (0.7–1); $ $ p < .01 compared to 24 hR (0.7–1), one-way ANOVA followed by Tukey’s HSD test, n = 16 images/group.

Fig. 3.

Ischemia-induced mitochondrial fission in CA3 neurons. CA3 neuronal mitochondria exhibit increased fragmentation after the ischemic insult that was reduced by NMN administration. (A) Mitochondria were visualized from the CA3 oriens in perfusion-fixed ischemic mito-eYFP mice treated with vehicle (PBS) (24 hR) or NMN (24 hR NMN). Mitochondria fragmented into spherical organelles at 24 h of recovery that was reversed after NMN treatment. Scale bar represents 100 μm. (B) The population of short mitochondria (0.2–1 μm) increased after ischemic insult while the longer population of mitochondria decreased (1–5 μm and 5–15 μm). NMN administration after ischemia inhibited the fragmentation process. *** p < .001 compared to sham (0.2–1 μm); ### p < .001 compared to 24 hR (0.2–1 μm); @@@ p < .001 compared to sham (0.2–1 μm); ^ p < .05 compared to sham (1–5 μm); $ $ $ p < .001 compared to 24 hR (1–5 μm); && p < .01 compared to 24 hR (5–15 μm), one-way ANOVA followed by Tukey’s HSD test, n = 16 images/group. (C) Mitochondrial shape factor changes. The population of spherical mitochondria (0.7–1) was elevated post-ischemia while rod-shaped mitochondria (0.35–0.65) decreased. NMN administration reversed these changes. ** p < .05 compared to sham (0.35–0.65); ## p < .01 compared to 24 hR (0.35–0.65); ^^p < .05 compared to sham (0.7–1); $$ p < .01 compared to 24 hR (0.7–1), one-way ANOVA followed by Tukey’s HSD test, n = 16 images/group.

Similarly, we also compared the effect of NMN on ischemia-induced mitochondrial fragmentation in CA3 and DG neurons (Fig. 3 and Fig. 4). In CA3 neurons there was a moderate increase in the number of small organelles at 24 h of recovery that was reversed after NMN treatment. Mitochondria did not fragment after ischemic insult in DG neurons, thus, the relative distribution of spherical, rod-shape and tubular organelles (5–15 μm length) was not significantly different in ischemic samples when compared to sham (Fig. 5). Correspondingly, there was no change in the shape factor of mitochondria in DG neurons after ischemia.

Fig. 4.

Dentate gyrus (DG) neuronal mitochondria are resistant to ischemic-induced fragmentation. (A) Mitochondria are visualized from the dorsal molecular layer of the DG in perfusion-fixed ischemic mito-eYFP mice treated with NMN (24 hR NMN) or vehicle (PBS) (24 hR). Scale bar represents 100 μm. Mitochondria did not show fragmentation after ischemic insult at 24 h of recovery. (B) Quantification of mitochondrial length using Volocity software. There were no changes in the population of 0.2–1, 1–5, 5–15 μm mitochondria after ischemic insult at 24 h of recovery with vehicle or NMN treatment compared to sham. One-way ANOVA followed by Tukey’s HSD test, n = 16 images/group. (C) The population of rod-shaped and spherical mitochondria measured by shape factor remained unchanged between experimental groups, one-way ANOVA followed by Tukey’s HSD test, n = 16 images/group.

Fig. 5.

NMN prevents post-ischemic increase in hippocampal pDrp1 (S616) protein levels. Ischemic groups were treated with vehicle (PBS) (Veh) or NMN and hippocampal tissue and mitochondria was isolated at 2, 4, and 24 h of reperfusion. Mitochondrial fusion and fission protein levels were measured by western blot and normalized using VDAC. (A) Mitofusin 1 (MFN1) levels increased after ischemic insult up to 24 h of recovery when compared to sham (2 hR - 2 h of recovery; 4 hR −4 h of recovery). At 24 h of recovery (24 hR) in both vehicle and NMN treated group the MFN1 levels we slightly lower when compared to 4 hR levels. (B) Mitofusin 2 (MFN2) levels were reduced after injury up to 24 h of recovery. (C) The ratio of OPA1 long (pro-fusion) to OPA1 short (pro-fission) levels was reduced after ischemia compared to sham. NMN treatment had no effect on ischemic-induced changes in mitochondrial fusion protein levels. (D) The phosphorylated form of mitochondrial fission protein, dynamin-1-like at serine 616 (pDrp1)(S616), increased at 24 h after injury while NMN treatment prevented this ischemia-induced change. (A) * p < .05, ** p < .01, *** p < .001 compared to sham; ^^ p < .01 compared to 4 hR NMN (B and C) *** p < .001 compared to sham (D) * p < .05 compared to sham; ^^ p < .01 compared to 24 hR vehicle, two-way ANOVA followed by Tukey’s HSD test, n = 6 animals/group.

3.3. NMN prevents ischemia-induced phosphorylation of Drp1

Since mitochondrial dynamics is controlled by proteins that regulate the fission and fusion process, we thought to examine whether NMN treatment is affecting the expression of these proteins following ischemic insult. As Fig. 5A–C demonstrates ischemia-induced changes in the levels of mitochondrial outer membrane fusion proteins, MFN1 and MFN2, and inner membrane fusion protein OPA1 were not affected by NMN administration. However, there was an increase in the active, phosphorylated form of fission protein Drp1 (pDrp1 (S616)) 24 h after ischemia. Interestingly, in NMN treated animals the pDrp1 (S616) levels did not increase following ischemic insult (Fig. 5D).

After Dpr1 is phosphorylated, it can translocate to the outer mitochondrial membrane and initiate the fission process. To determine the changes in pDrp1 (S616) colocalization with mitochondria we determined the Pearson colocalization coefficient between mitochondrial eYFP (mito-eYFP) and pDrp1 (S616), which was visualized by immunostaining (Fig. 6). There was a significant increase in pDrp1 (S616) colocalization with mitochondria in CA1 neurons at 24 h of reperfusion, and this change in pDrp1 (S616) colocalization with mitochondria was prevented by NMN administration (Fig. 6A, B). In CA3 and DG neurons the pDrp1 (S616) colocalization with mitochondria 24 h after the start of reperfusion was not different from sham animals.

Fig. 6.

NMN administration prevents ischemic-induced trans-location of pDrp1 (S616) to the mitochondrial outer membrane in CA1 neurons. (A) Hippocampal brain sections from mito-eYFP transgenic animals were immunostained with pDrp1 (S616) antibody (red). Z-stack images were collected and the degree of colocalization of mitochondria (green) with pDrp1 (S616) was determine by the Pearson coefficient. (B) The Pearson coefficient was calculated by Volocity software in the CA1 oriens, CA3 oriens, and dorsal molecular layer of the DG. A significant increase in pDrp1 (S616) colocalization with mitochondria in the CA1 oriens was observed at 24 h of reperfusion when compared to sham. This change was prevented by NMN administration after ischemic insult. No changes were observed in pDrp1 (S616) colocalization with mitochondria in CA3 and DG neurons. Scale bar represents 100 μm. *** p < .001 compared to all other groups; ### p < .001 compared to CA1 24 hR vehicle, two-way ANOVA followed by Tukey’s HSD test, n = 16 images/group.

3.4. NMN treatment inhibits post-ischemic increases in SOD2 acetylation and in reactive oxygen species (ROS) production

One of the well-characterized mitochondrial protein that is modulated by acetylation is the mitochondrial superoxide dismutase (SOD2) (Cheng et al., 2016, Klimova et al., 2019). To confirm the acetylation of SOD2 is altered following ischemia we immunoprecipitated acetylated SOD2 at 24 h of recovery from sham, vehicle and NMN treated post-ischemic animals. There was about a 30% increase in SOD2 acetylation following ischemic insult that was prevented by NMN administration (Fig. 7A). Acetylation inhibits SOD2 activity (Liu et al., 2017; Zheng et al., 2018) and since SOD2 is a key enzyme of the mitochondrial mechanism that detoxifies superoxide, we next decided to determine the effect of NMN on ischemia-induced reactive oxygen species (ROS) generation. To this end we injected animals with dihydroethidium (DHE) and 1 h after the DHE administration we measured the fluorescent product of the DHE reaction with ROS, hydroxyethidium in hippocampal tissue homogenate (Suh et al., 2008, McCann et al., 2008, Zielonka et al., 2008, Hall et al., 2012). We performed three sets of experiments. First, we collected the hippocampal samples from sham operated animals and animals treated with vehicle and NMN at 2 h and 24 h of recovery. As Fig. 7C shows at 2 h after the start of reperfusion there was no difference in ROS levels between sham, vehicle and NMN treated samples. However, at 24 h of recovery the post-ischemic animals showed an increased ROS levels. As expected NMN administration prevented this post-ischemic rise in ROS (Fig. 7C). Second, to determine the changes in ROS generation during early recovery period, the first minutes of reperfusion, DHE was administered 10 min before the onset of ischemia and hippocampal homogenate was collected at 40 min of recovery. The hippocampal ROS levels during this early reperfusion period increased by about 120% when compared to sham samples (Fig. 7B).

Fig. 7.

NMN administration inhibits the increase in hippocampal post-ischemic reactive oxygen species (ROS) generation. (A) SIRT3 target, SOD2, was immunoprecipitated from sham operated, 24 hR, and 24 hR NMN hippocampal tissue samples and acetylation levels analyzed by western blot. There is a 30% increase in SOD2 acetylation post-ischemia, which was prevented by NMN administration. * p < .05 compared to sham, # p < .05 compared to 24 hR, one-way ANOVA followed by Tukey’s HSD test, n = 4 animals/group. (B) ROS indicator, dihydroethidium (DHE), was injected i.p. 10 min before start of ischemia and its’ fluorescent product, hydroxyethidium, was measured in hippocampal homogenate by plate reader. Fluorescence was normalized to total protein (represented here as fluorescence units/μg protein). ROS levels at early reperfusion (1 hR) were significantly increased compared to sham. *** p < .001 compared to sham, Student t-test, n = 9 measurements/group. (C) DHE was administered 1 h before the hippocampal tissue collection. At 2 h of reperfusion there was no difference in ROS production compared to sham. However, at 24 h of recovery there is a significant increase in ROS levels that was reduced by NMN administration. *** p < .001 compared to all groups; ### p < .001 compared to 24 hR vehicle, two-way ANOVA followed by Tukey’s HSD test, n = 6–8 animals/group. (D) The specific contribution of mitochondria and NADPH oxidase to ROS generation was examined by using mitochondria targeted anti-oxidant, Mitoquinone (MitoQ), and NADPH oxidase inhibitor, Apocynin (Apo). ROS generation was measured via DHE administration. MitoQ and Apo suppressed ischemic-induced ROS production. *** p < .001 compared to sham; ^^^ p < .001 compared to 24 hR vehicle; # p < .05 compared to 24 hR vehicle, one-way ANOVA followed by Tukey’s HSD test, n = 5–6 animals/group.

To confirm the specific contribution of mitochondria to post-ischemic ROS generation we treated the animals with mitochondria targeted antioxidant Mitoquinone (MitoQ) (Ahmed et al., 2015, Skulachev et al., 2009). As expected MitoQ suppressed the post-ischemic ROS generation leading to similar ROS levels as in sham samples (Fig. 7D). Finally, we examine the contribution of NADPH oxidase (NOX) to the increased free radical levels by administering NOX inhibitor, Apocynin (Apo) (Suh et al., 2008, Yenari et al., 2006, Abramov et al., 2007). Predictably, the effect of Apo was less effective, the post-ischemic ROS levels remained slightly higher than in sham animals (Fig. 7D).

We also studied the effect of NMN on post-ischemic changes in ROS generation by quantifying the oxidized DHE signal intensity within the different hippocampal subregions (Fig. 8A, B). Interestingly, the most dramatic increase was detected in CA3 pyramidal layer, about 100% increase when compared to sham (Fig. 8B). In CA1 pyramidal and DG molecular layer there was only about 40% increase when compared to sham (Fig. 8B). NMN treatment prevented the ischemia-induced ROS increase in all hippocampal subregions (Fig. 9B).

Fig. 8.

NMN treatment prevents ischemia-induced increase in mitochondrial ROS generation. (A) Hydroxyethidium fluorescence (red) following DHE administration in hippocampal CA1, CA3, DG neurons in sham operated, and post-ischemic animals treated with vehicle or NMN. Scale bar represents 100 μm. (B) NMN’s effects on post-ischemic changes in ROS generation were analyzed in various hippocampal subregions via quantification of the hydroxyethidium signal. ROS production significantly increases at 24 h of reperfusion in CA1, CA3, and DG. NMN treatment prevented the post-ischemic ROS increase in all areas. Fluorescence intensity was normalized to unit volume (μm³). (C) To confirm NMN effects were specifically affected mitochondrial ROS generation the Pearson colocalization coefficient between hydroxiethidium (red) and mito-eYFP (green) was determined by Volocity software. In all hippocampal subregions 24 h after the ischemic insult there is a significant increase in the colocalization between the oxidized DHE signal and mito-eYFP when compared to sham group. Administration of NMN prevented this increase in colocalization. *** p < .001 compared to sham; ### p < .001 compared to vehicle, one-way ANOVA followed by Tukey’s HSD test, n = 16 images/group.

Fig. 9.

SIRT3 KO mice exhibit elevated ROS levels and are resistant to NMN-induced effects on post-ischemic free radicals generation. Mito-eYFP and mito-eYFP-SIRT3KO mice were injected with DHE 1 h before perfusion-fixation. (A) Hippocampal CA1 hydroxiethidium florescence (red) at 24 hours post-ischemia in WT and SIRT3 KO mice. Scale bar represents 100 μm. (B) ROS levels in sham operated SIRT3 KO mice are significantly higher compared to WT sham animals. Ischemia results in the rise in ROS in both WT and SIRT3 KO mice while NMN inhibits the ischemia-induced increase. Fluorescence intensity was normalized to unit volume. ** p < .01, *** p < .001 compared to WT Sham; ## p < .01, ### p < .001 compared to WT vehicle; ^^^ p < .001 compared to WT NMN; •• p < .01 compared to SIRT3KO Sham; ■■■ p < .001 compared to SIRT3 KO vehicle, two-way ANOVA followed by Tukey’s HSD test, n = 16 images/group. (C) To identify whether the source of ROS was mitochondrial we determined the Pearson colocalization coefficient between hydroxiethidium (red) signal and mito-eYFP (green) using Volocity software. Colocalization between oxidized DHE and mito-eYFP in SIRT3 KOs are significantly higher when compared to WT sham group. Ischemic-induced increase in the colocalization between hydroxiethidium and mito-eYFP is prevented by NMN in WT mice. However, NMN did not have any significant effect in SIRT3 KO animals. *** p < .001 compared to WT Sham; ### p < .001 compared to WT Vehicle; ^^^ p < .001 compared to WT NMN; •• p < .01 compared to SIRT3 KO Sham; ■ p < .05 compared to SIRT3 KO Vehicle, two-way ANOVA followed by Tukey’s HSD test, n = 16 images/group.

To confirm that NMN is specifically affecting mitochondrial ROS generation we determined the Pearson colocalization coefficient between hydroxyethidium and mito-eYFP in our transgenic mice (Fig. 8C). The colocalization data in Fig. 8C shows that there is a significant increase in the hydroxyethidium signal colocalization with mito-eYFP following ischemia, and NMN administration prevented this increase in all hippocampal subregions (Fig. 8C).

To determine the role of SIRT3 in post-ischemic oxidative stress we subjected the SIRT3KO animals expressing also mito-eYFP to ischemic insult and recorded the ROS levels during recovery. Sham operated SIRT3KO animals showed higher ROS levels when compared to WT sham mice (Fig. 9 A, B). Ischemia further increased the ROS generation in vehicle treated SIRT3KO animals which was reduced to sham levels in animals that received NMN (Fig. 9A, B). However, there was no effect of ischemia or NMN treatment on the colocalization of the oxidized ethidium with mitochondria (Fig. 9C).

Finally, we used cell-type specific markers to determine in which cells there was a higher ROS generation. Most free radicals were generated by mitochondria in pyramidal neurons in all hippocampal subregions and also interneurons showed a dramatic increase in the hydroxyethidium signal. Interestingly, there was no colocalization of oxidized DHE with astrocytic or microglial marker (Fig. 10).

Fig. 10.

Post-ischemic ROS is generated in hippocampal pyramidal neurons and interneurons. Sham operated and ischemic animals were treated with DHE at 23 h of recovery and 1 h later they were perfusion fixed and their brain was processed for immunohistology. Neurons were identified by NeuN antibody, microglia were stained with Iba1 antibody, astrocyte were visualized by GFAP and S100β antibody, and finally interneurons were marked by GAD 65/67 or Parvalbumin antibody (all green). The hydroxiethidium signal (red) colocalized only with NeuN or interneuron markers.

4. Discussion

We recently reported that treatment of naïve animals with a single dose of NMN resulted in an increase in brain mitochondrial NAD⁺ levels, significant reduction of mitochondrial protein acetylation, decrease in mitochondrial ROS production and suppression of mitochondrial fission (Klimova et al., 2019). Since we used an NMN dose that has shown a dramatic protective effect against ischemic brain damage (Park et al., 2016), we were interested in studying the effect of this NAD⁺ precursor on post-ischemic oxidative stress and mitochondrial dynamics (Owens et al., 2015; Kumar et al., 2016).

Several laboratories have previously reported NAD⁺ catabolism following ischemic insult, which was associated with the increase in poly-ADP-ribose polymerase 1 (PARP1) and in ADP-ribose-cyclase (CD38) activity (Endres et al., 1997; Strosznajder et al., 2003; Cozzi et al., 2006; Park et al., 2016; Long et al., 2017). The post-ischemic NAD⁺ consumption was prevented either by pharmacological or genetic inhibition of PARP1 or CD38 (Endres et al., 1997; Eliasson et al., 1997; Kauppinen and Swanson, 2007; Long et al., 2017). In another approach the post-ischemic NAD⁺ levels were maintained via stimulation of NAD⁺ synthesis after administration of NAD⁺ precursors, nicotinamide (Nam) or NMN (Ayoub et al., 1999; Yang et al., 2002; Liu et al., 2009; Park et al., 2016). NMN showed more effective neuroprotection when compared to Nam since only about a 60 mg/kg dose was sufficient to completely ameliorate ischemic brain damage comparing to the most effective dose of Nam (500 mg/kg) (Yang et al., 2002), (Park et al., 2016). NMN is metabolized extracellularly to nicotinamide ribose (NR) that is transported into the cell and converted to NAD⁺ (Ratajczak et al., 2016). It was reported that NR elevates tissue NAD⁺ with distinct and superior pharmacokinetics to Nam (Trammell et al., 2016). In naïve animals, already 15 min after NMN injection there is a significant increase in plasma, hippocampal tissue and mitochondrial NAD⁺ levels (Klimova et al., 2019, Yoshino et al., 2011). Similarly, we detected an increase in plasma NAD⁺ at 2 h after ischemia (1.5 h following NMN injection) in NMN treated animals. After NMN administration the post-ischemic mitochondrial NAD⁺ levels were increased at 2 h of recovery and reached physiological levels 24 h after the ischemic insult. The fast uptake of NMN and its conversion to NAD⁺ is supported by a study showing that there is an NMN transporter in the small intestine (Grozio et al., 2019), suggesting the ability to rapidly take up NMN after i.p. injection from the gut into the blood. However, so far, the presence of this transporter was not confirmed in brain tissue. Thus, NMN is rapidly absorbed from the gut into the blood where it is partially converted to NAD⁺ or nicotinamide riboside (NR) (Klimova et al., 2019) and also transported to the brain cells in either form of NMN or NR (Klimova et al. 2019). In neurons NR can be converted back to NMN by nicotinamide riboside kinase (Belenky et al., 2009). NMN is then used by NMNAT enzyme to generate NAD⁺ (Belenky et al., 2007; Owens et al., 2013).

It is important to note that NAD⁺ levels in sham operated hippocampal mitochondria were significantly increased when compared to NAD⁺ levels in mitochondria of naïve mice. To avoid the variability in blood/tissue glucose levels and its effect on post-ischemic brain damage, animals are fasted for several hours before they are subjected to global cerebral ischemia (Lin et al., 1998; Siesjo et al., 1996; Kristian and Siesjo, 1996). The sham-operated animals undergo the same pre-surgery fasting procedure as animals undergoing ischemic surgery. Since fasting increases the NAD⁺/NADH ratio, the increased mitochondrial NAD⁺ pools detected in sham samples are in agreement with the reported increase of tissue and mitochondrial NAD⁺ levels following caloric restriction (Rodgers et al., 2005; Yang et al., 2007). The pre-surgery fasting procedure thus prevents the effect of variable glucose levels on ischemic outcome, however it also increases the brain mitochondrial NAD⁺ levels and consequently mitochondrial protein acetylation that might also have impact on post-ischemic brain injury. Therefore, further studies are needed to resolve this issue.

Loss of mitochondrial NAD⁺ after ischemic insult could be a result of two mechanisms. First, intracellular post-ischemic conditions, including high levels of cytosolic calcium, high inorganic phosphate, low pH and elevated levels of free radicals favor opening of the high conductance pore in the inner mitochondrial membrane, called the mitochondrial permeability transition (MPT) pore (Bernardi et al., 1992; Kristian et al., 2001; Kristian, 2004; Kristal et al., 2004; Kristian et al., 2007). The MPT pore allows solutes of molecular weight up to 1500 Da diffuse across the inner mitochondrial membrane. Consequently, MPT activation can lead to the depletion of matrix NAD⁺ (Di Lisa et al., 2001; Kristian, 2004). Second, intramitochondrial NAD⁺ can be degraded by activation of mitochondrial PARP following acute brain injury (Du et al., 2003; Lai et al., 2008; Owens et al., 2013).

NAD⁺ in mitochondria serves as substrate for mitochondrial deacetylase SIRT3. Therefore, the dramatic post-ischemic consumption of mitochondrial NAD⁺ pools results in inhibition of SIRT3 activity and excessive increase in mitochondrial protein acetylation. Since NMN prevented the mitochondrial NAD⁺ depletion, the post-ischemic increase in acetylation was inhibited. The SIRT3 activity-dependent effect of NMN on changes in mitochondrial protein acetylation was confirmed in naïve SIRT3KO animals (Klimova et al., 2019). Furthermore, ischemia or NMN administration did not affect the mitochondrial protein acetylation in mice with the disrupted SIRT3 gene. Additionally, data showing significantly higher mitochondrial NAD⁺ levels and also increased protein acetylation levels in SIRT3KO mice when compared to naïve WT animals suggests that NAD⁺-dependent SIRT3 activity is required to reduce the acetylation levels of mitochondrial proteins (Hirschey et al., 2011, Klimova et al., 2019).

Acute brain injury or neurodegenerative diseases are associated with disturbance in mitochondrial fission/fusion homeostasis that can result in profound fragmentation of mitochondria and deterioration of mitochondrial bioenergetics (Westermann, 2010; Nakamura et al., 2012; Owens et al., 2015; Kumar et al., 2016; Klimova et al., 2018). Recently, we have shown that transient global cerebral ischemia has a cell-type and region-specific effect on mitochondrial dynamics (Owens et al., 2015). After initial fragmentation, mitochondria in ischemia resistant CA3 and dentate gyrus neurons underwent fusion and regained their pre-ischemic morphology at 3 days of recovery (Owens et al., 2015). However, in the more vulnerable CA1 neurons the mitochondrial fission-induced changes were irreversible (Owens et al., 2015). Here we demonstrate that NMN administration shifts the mitochondrial dynamics towards fusion in post-ischemic tissue and our data showed that when animals are treated with NMN 30 min after the ischemic insult the mitochondria in CA1 neurons become less fragmented. Thus, 24 h after the ischemic insult neuronal mitochondria in NMN treated animals regained their pre-insult length distribution. The increased fragmentation of mitochondria following ischemia can be induced by changes in expression levels of fission or fusion proteins (Owens et al., 2015; Kumar et al., 2016; Kumari et al., 2012) or due to modulation of their activity by post-translational modifications (Kumar et al., 2016; Klimova et al., 2018). Interestingly, NMN administration affected only the post-ischemic levels of the fission active form of Drp1, phosphorylated Drp1 at serine 616 (pDrp1 (S616)). Ischemia-induced changes in expression levels of Mfn1, Mfn2, and the increased proteolytic cleavage of OPA1 was not affected by NMN treatment. However, pDrp1 (S616) was significantly increased only at the later 24 h reperfusion time point. Data showing increased colocalization of pDrp1 (S616) with mitochondria suggest that this fission protein is significantly contributing to the increased mitochondrial fragmentation. Furthermore, the pDrp1 (S616) colocalization was region specific as it was observed only in CA1 neurons, indicating the possible cause for irreversibility of post-ischemic mitochondrial fragmentation in these cells.

The inner mitochondrial membrane fusion is controlled by OPA1 (Baker et al., 2014; Griparic et al., 2007; Ishihara et al., 2006). OPA1 fusogenic ability is modulated by proteolytic processing and depends on the interaction of both the long and short (cleaved) forms (Ishihara et al., 2006; Anand et al., 2014; Klimova et al., 2018). It was suggested that the release of OPA1 from mitochondria following ischemic insult could also contribute to inhibition of the fusion process and increase mitochondrial fragmentation (Kumar et al., 2016). Although, we did not detect changes in the ratio between the levels of long and short OPA1 in hippocampal tissue samples after NMN treatment, it does not exclude the possibility that the short OPA1 was translocated from mitochondria into the cytosol (Kumar et al., 2016).

SIRT3 deacetylates several proteins in mitochondria including the mitochondrial superoxide dismutase 2 (SOD2) that converts superoxide to hydrogen peroxide and plays a key role in the mitochondrial anti-oxidant defense mechanism (Liu et al., 2017; Ren et al., 2017; Zheng et al., 2018). The activity of this enzyme is inhibited by acetylation (Liu et al., 2017). Highly acetylated SOD2 thus leads to increased production of ROS by mitochondria (Cheng et al., 2016, Klimova et al., 2019). The increased SOD2 acetylation following ischemia suggests that mitochondria contribute to the elevated post-ischemic ROS generation. There were two periods during reperfusion when high ROS levels were detected. First, increased ROS generation was observed during early reperfusion, during the first 1 h after the start of recovery. This was followed by period of normalized ROS levels between 2 and 24 h of recovery and then secondary increase in ROS generation was observed later at 24 h of recovery.

Several reports show that during the intra-ischemic period there is an accumulation of succinate in mitochondria (Folbergrova et al., 1974) that results in increased superoxide production by mitochondria during early reperfusion (Chouchani et al., 2014; Niatsetskaya et al., 2012; Murphy, 2009; Kristian, 2004). This is because after the start of reperfusion, the accumulated succinate is re-oxidized by succinate dehydrogenase, which drives the extensive ROS production due to reverse electron transport from mitochondrial complex II to complex I. Consequently, decreasing intra-ischemic succinate accumulation or pharmacologic inhibition of complex I leads to reduced ROS generation by mitochondria and amelioration of ischemia-reperfusion injury (Piantadosi and Zhang, 1996, Niatsetskaya et al., 2012, Chouchani et al., 2014, Mohsin et al., 2019).

The secondary, delayed increase in post-ischemic ROS generation was observed at 24 h of reperfusion in neurons of all hippocampal subregions. Neurons in the pyramidal layer of the hippocampus and also interneurons displayed strongest signal generated by ROS reaction with DHE. Furthermore, there was also increased colocalization of the oxidized DHE product (hydroxyethidium) with mitochondria, suggesting that mitochondria were the main source of the ROS. Interestingly, animals treated with NMN did not show increased ROS levels at 24 h of recovery and also the colocalization of ROS and mitochondria was reduced to control pre-ischemic levels. Data related to ROS levels in SIRT3KO animals confirmed that the increased ROS production is modulated by SIRT3-driven deacetylation since oxidized DHE signal in sham operated SIRT3KO animals was dramatically increased when compared to sham operated WT mice. However, interestingly, ischemic insult also moderately increased the ROS levels in SIRT3KO mice when compared to sham SIRT3KO animals. Furthermore, when NMN treated, the post-ischemic SIRT3KO mice ROS levels were not significantly different from their respective control sham samples. Since the colocalization of the oxidized DHE signal with mitochondria was not affected by ischemia or NMN treatment, these data suggest that the additional increased in ROS after ischemia in SIRT3KO neurons was not due to mechanisms that were related to changes in mitochondrial protein acetylation. Simply, in SIRT3KO animals there is probably a moderate non-mitochondrial source of ROS that is affected by NMN administration.

To confirm that ROS was mainly originated from mitochondria we treated the animals with mitochondria targeted antioxidant MitoQ. As expected MitoQ reversed the delayed increase in ROS generation. It was reported that administration of MitoQ showed neuroprotection in both the traumatic brain injury model and subarachnoid hemorrhage induced brain damage (Zhou et al., 2018, Zhang et al., 2019). Although the effect of MitoQ on ROS generation was not determined in these studies, the Nrf2 activated pathway, which is regulated by free radicals, was identified as major player in neuroprotection.

Mitochondria together with NADPH oxidase (NOX), a superoxide-generating enzyme, are believed to be the major source of free radicals in post-ischemic tissue (Abramov et al., 2007; Suh et al., 2008; McCann et al., 2008). NOX was originally described in neutrophils, and later it has been identified in microglia, astrocytes, vascular cells, and neurons (Bedard and Krause, 2007; Serrano et al., 2003; Sorce and Krause, 2009). It was shown that NOX is activated following 22 min global cerebral ischemia at early (3 to 6 h) and late (72 h) recovery phases (Yoshioka et al., 2011). Inhibitory effect of Apocynin on ROS generation and its neuroprotective effect was also reported in 30 min 2-vessel occlusion model of cerebral ischemia (Suh et al., 2008), 2 h middle cerebral artery occlusion (MCAO) model of stroke (Qin et al., 2017; Tang et al., 2008) and traumatic brain injury (Loane et al., 2013). In our 10 min global cerebral ischemia model we did also detect a significant effect of Apocynin on ROS generation. This suggest that the neuronal and/or microglial NOX is activated 24 h after the start of reperfusion (see Owens et al., 2015).

We present here an ischemia-induced pathologic mechanism that is initiated by degradation of mitochondrial NAD⁺, followed by increase in mitochondrial protein acetylation and high ROS generation, and finally leading to excessive post-ischemic mitochondrial fragmentation. Here, for the first time, we report post-ischemic degradation of hippocampal mitochondria NAD⁺ that leads to increase in mitochondrial protein acetylation and the ability of NMN administration to reverse these changes. Our data provides a novel link between mitochondrial NAD⁺ metabolism, ROS production, and mitochondrial fragmentation. Using NMN to target these mechanisms could represent a new therapeutic approach for neurobiological diseases.