Subcellular NAMPT-mediated NAD⁺ salvage pathways and their roles in bioenergetics and neuronal protection after ischemic injury

Abstract

NAD⁺ is a cofactor required for glycolysis, tricarboxylic acid cycle and complex I enzymatic reaction. In mammalian cells, NAD⁺ is predominantly synthesized through the salvage pathway, where nicotinamide phosphoribosyltransferase (NAMPT) is the rate-limiting enzyme. Previously, we demonstrated that NAMPT exerts a neuroprotective effect in ischemia through the suppression of mitochondrial dysfunction. Mammalian cells maintain distinct NAD⁺ pools in the cytosol, mitochondria and nuclei. However, it is unknown whether mitochondria have an intact machinery for NAD⁺ salvage, and if so, whether it plays a dominant role in bioenergetics, mitochondrial function and neuronal protection after ischemia. Here, using mouse primary cortical neuron and cortical tissue preparations, and multiple technologies including cytosolic and mitochondrial subfractionation, viral overexpression of transgenes, molecular biology and confocal microscopy, we provided compelling evidence that neuronal mitochondria possess an intact machinery of NAMPT-mediated NAD⁺ salvage pathway, and that NAMPT and nicotinamide mononucleotide adenylyltransferase 3 (NMNAT3) are localized in the mitochondrial matrix. By knocking down NMNAT1–3 and NAMPT with siRNA, we found that NMANT3 has a larger effect on basal and ATP production-related mitochondrial respiration than NMNAT1–2 in primary cultured neurons, while NMNAT1–2 have a larger effect on glycolytic flux than NMNAT3. Using an oxygen glucose deprivation model, we found that mitochondrial, cytoplasmic and non-subcellular compartmental overexpressions of NAMPT have a comparable effect on neuronal protection and suppression of apoptosis-inducing factor translocation. The current study provides novel insights into the roles of subcellular compartmental NAD⁺ salvage pathways in NAD⁺ homeostasis, bioenergetics and neuronal protection in ischemic conditions.

Graphical Abstract

We investigated whether neuronal mitochondria have an endogenous NAD⁺ salvage pathway and the roles of subcellular compartmental NAD⁺ salvage pathways in bioenergetics and neuronal protection after ischemia. Using multiple approaches, we found that both NAMPT and NMNAT3 are expressed in neuronal mitochondria, and are localized in the matrix. We further found that mitochondrial NAD⁺ salvage pathway has a larger effect on basal and ATP-production related oxygen consumption than nuclear and cytoplasmic NAD⁺ pathways, while nuclear and cytoplasmic NAD⁺ pathways have a larger effect on glycolytic flux. Moreover, mitochondrial, cytoplasmic and non-subcellular compartmental overexpressions of NAMPT have a comparable effect on neuronal protection in ischemia. Our findings provide novel insights into the roles of subcellular NAD⁺ salvage pathways in bioenergetics and neuronal protection in ischemia.

Introduction

In mammalian cells, the nicotinamide phosphoribosyltransferase (NAMPT)-mediated nicotinamide adenine dinucleotide (NAD⁺) salvage pathway is the predominant pathway for NAD⁺ biosynthesis (Luk et al. 2008;Dahl et al. 2012;Revollo et al. 2007). In this pathway, the first step is catalyzed by NAMPT, which converts nicotinamide (NAM) to nicotinamide mononucleotide (NMN), and the second step is catalyzed by nicotinamide mononucleotide adenylyltransferases (NMNATs), which converts NMN to NAD⁺. NAMPT is the rate-limiting enzyme in this pathway. It is reported that three isoforms of NMNAT, namely NMNAT1–3, are expressed in different tissues and subcellular compartments in mammalian cells. NMNAT1 is localized in the nucleus, NMNAT2 at the Golgi-cytoplasmic interface, and NMNAT3 in the mitochondria (Berger et al. 2005). Studies showed that NAD⁺ is highly compartmentalized in the cytosol and mitochondria of mammalian cells including neurons (Alano et al. 2007; Alano et al. 2010; Yang et al. 2007), and cytosolic and mitochondrial NAD⁺ concentrations are largely different (Yang et al. 2007;Cambronne et al. 2016), suggesting that these subcellular compartments have distinct NAD⁺ salvage pathways. However, there is no evidence that NAMPT bears a mitochondrial localization sequence (Pittelli et al. 2010), and there have been contradictory reports regarding the expression of NAMPT in mitochondria in mammalian cells (Pittelli et al. 2010; Yang et al. 2007; Nikiforov et al. 2011). Moreover, mitochondrial occurrence of NMNAT3 in mammalian cells is also controversial (Felici et al. 2013; Zhang et al. 2003; Berger et al. 2005) and no studies indicating that NMNAT3 is expressed in neuronal mitochondria. Therefore, it remains unknown whether neuronal mitochondria have an intact machinery of NAMPT-mediated NAD⁺ salvage pathway, and if so, which submitochondrial compartments NAMPT and NMNAT3 are localized.

Focal ischemic stroke (FIS) causes human disability and even death, and has a huge impact on public health. Ischemia depletes global NAD⁺ (Zhang et al. 2010; Bi et al. 2012; Liu et al. 2009), which stimulates proapoptotic pathways including translocation of apoptosis-inducing factor (AIF) from mitochondria to nuclei (Yang et al. 2007; Alano et al. 2010; Di Lisa and Bernardi 2006; Wang et al. 2016). Previously, we and other groups have demonstrated that NAMPT plays a neuroprotective role in ischemia (Zhang et al. 2010; Bi et al. 2012; Wang et al. 2014; Wang et al. 2016; Wang et al. 2011; Wang et al. 2012). Global heterozygous NAMPT knockout mice exhibit a much larger infarction than wild-type (WT) mice after photothrombosis (PT)-induced FIS (Zhang et al. 2010). Overexpression of WT NAMPT, but not mutant NAMPT lacking enzymatic activity, can reduce neuronal death and inhibit mitochondrial membrane potential (MMP) depolarization after glutamate excitotoxicity and oxygen-glucose deprivation (OGD) (Bi et al. 2012; Wang et al. 2016). Moreover, both exogenous NAD⁺ and NAM repletion, and NAMPT overexpression suppress ischemia-induced mitochondrial fragmentation, AIF translocation, and the impairment of mitochondria biogenesis (Wang et al. 2014; Wang et al. 2016; Bi et al. 2012). These studies indicate that NAMPT-mediated NAD⁺ biosynthesis can protect neurons after ischemia through maintaining mitochondrial function.

Given the importance of NAD⁺ in cellular bioenergetics under physiological and pathological (such as ischemic) conditions, in current study, we sought to determine whether neuronal mitochondria possess a distinct NAMPT-mediated NAD⁺ salvage pathway, and to investigate submitochondrial localization of NAMPT and NMNAT3 using mouse primary cortical neuron and mouse cortical tissue preparations. Using molecular and pharmacological methods including siRNA to knock down the enzymes involved in the NAD⁺ salvage pathway, we subsequently investigated the effects of NMNAT1–3 and NAMPT on NAD⁺ and NADH levels, mitochondrial respiration and glycolytic function by measuring oxygen consumption rate (OCR) and extracellular acidification rate (ECAR). Finally, we investigated whether mitochondria- and cytoplasm-targeted NAMPT overexpression have different effects on neuronal death and AIF translocation after OGD. The current study provides novel insights into the roles of subcellular compartmental NAD⁺ salvage pathways in NAD⁺ homeostasis, bioenergetics and neuronal protection in ischemic conditions.

Material and methods

Animals

Adult male or female C57BL/6J mice (Jackson Laboratory, Stock No. 000664, RRID:MGI:5657312) aged 8–10 weeks were used for mitochondrial and cytosolic subfractionation and embryo isolation for preparation of primary cultured cortical neurons in this study. Adult mice (~25 g) were housed in ventilated cages (4 animals per cage) and maintained on a 12 h light:12 h dark cycle in the AAALAC-accredited animal facility at the University of Missouri. Animal husbandry technician conducted cage check to make sure mice were accessible to food and water on the top of the cage. Mice were sacrificed with cervical dislocation without using anesthetic for mitochondrial and cytosolic subfractionation and embryo isolation. The study was not preregistered. The animals were not randomized and no sample calculation was performed. All experimental procedures were performed according to the NIH Guide for the Care and Use of Laboratory Animals and approved by the University of Missouri Animal Care Quality Assurance Committee (Protocol No. 9444).

Primary neuronal cultures and OGD model

Primary mouse cortical neuronal cultures were prepared from embryonic day 15/16 (E15/16) C57BL/6J mice as previously described (Bi et al. 2012;Wang et al. 2014;Wang et al. 2016). Briefly, pooled cortical tissues of multiple embryos isolated from a pregnant mouse were dissociated by mild mechanical triturating after digestion with 0.05 % trypsin (Cat. No. 25300, Life Technologies, CA). The isolated cells were plated onto poly-L-lysine-coated (50–100 μg/ml) (Cat. No. P9155, Sigma-Aldrich, MO) tissue culture plates or glass coverslips (12 mm in diameter) in culture plate with Dulbecco’s modified Eagle medium (DMEM)/F12 (Cat. No. 25030, Life Technologies) supplemented with 10% heated-inactivated fetal bovine serum (FBS) (Cat. No. S11150, Atlanta Biological., GA). Five hours later, the medium was then changed to Neurobasal Media (NBM) (Cat. No. 21103, Life Technologies) containing 2% B27 serum free supplements (Cat. No. 17504, Life Technologies), 0.5 mM L-glutamine (Cat. No. 25030, Life Technologies) and 50 U/ml penicillin/streptomycin (Cat. No. 15140, Life Technologies). Cytosine β-D-arabino-furanoside (5 μM, Cat. No. C1768, Sigma-Aldrich) was added into medium 24 h after plating to suppress the growth of glial cell and increase the purity of neuronal cultures. The cultures were maintained at 37 °C in a humidified atmosphere of 5% CO₂ and 95% air with 50% medium changes every 3 days. Experiments were conducted between 9–11 days in vitro (DIV) as the neurons are matured at this age.

To mimic ischemic conditions in vitro, neuronal cultures were subject to OGD (Wang et al. 2014, Wang et al. 2016). Briefly, the medium in cultured neurons was removed in culture plates and replaced with serum- and glucose-free balanced salt solution (BSS). The culture plates were then placed within an anaerobic chamber at 37 °C. The sealed chamber was flushed with 99% N₂ and 1% air for 60 min controlled by a HypoxyDial (STARR Life Sciences, PA). After OGD, the BSS was replaced by normal feeding medium and neurons were then returned to the culture incubator for 24 h. Control neurons were replaced with BSS containing 5.5 mM glucose and incubated under normoxic conditions.

DNA constructs, DNA and siRNA transfection, and adeno-associated virus (AAV) infection in cultured neurons

For the construction of plasmids for subcellular compartment expressions of transgene, mitochondrial matrix-targeting sequence (mito) (Rizzuto et al. 1995; Li et al. 2014) and nuclear exporting sequence (NES)(Rodrigues et al. 2007) were inserted in the backbone of plasmid pZac2.1 under the control of neuronal CaMKII promoter (Dittgen et al. 2004; Li et al. 2014). The cistrons of interest, mRFP and mRFP-NAMPT were subcloned into the cloning sites to form new expressing plasmids, i.e., pZac2.1-CaMKII-mito-mRFP, pZac2.1-CaMKII-NES-mRFP, pZac2.1-CaMKII-mito-mRFP-NAMPT and pZac2.1-CaMKII-NES-mRFP-NAMPT to target transgene expression in mitochondria or cytoplasm of neurons, while pZac2.1-CaMKII-mRFP and pZac2.1-CaMKII-mRFP-NAMPT plasmids were constructed for non-subcellular compartmental expression of transgenes. For detailed information about construction of DNA plasmids and DNA sequences, see Fig. S1 and Supplemental information. The plasmids and other custom-made materials are available up a reasonable request.

For overexpressions of transgene using DNA plasmids, neurons were transfected using lipofectamine 2000 reagent (Cat. No. 1168019, Fisher Scientific Inc.) as in our previous studies (Bi et al. 2012; Wang et al. 2014; Wang et al. 2016). Before transfected at 8–9 DIV, neurons were replenished in regular growth medium without antibiotics. For transfection of neurons in each well in 24-well plates, 0.8 μg DNA plasmid was diluted in 50 μl NBM without serum and gently mixed with 2 μl lipofectamine 2000 in 50 μl NBM. After incubation for 20 min at room temperature, DNA and lipofectamine 2000 complexes were added drop by drop to each well. The medium was changed back to growth medium containing penicillin/streptomycin 6 h later. The neuronal cultures were ready for experiments 48 h later.

To study the effect of knockdown of NMNAT1–3 and NAMPT on protein, NAD⁺ and NADH levels, neurons were cultured in 6-well plates and transfected with 150 pmole/well of scrambled, NMNAT1–3 and NAMPT targeting siRNA (Cat. No. D-001210-01-05, D-051136-02-0005, D059190-01-0005, D-051688-01-0005, D-040272-01-0005, GE Healthcare Dharmacon, Inc., IL) using lipofectamine 2000. Neurons were then harvested 2 days later for NAD⁺ and NADH assays using commercially available kits (Cat. No. E2ND-100, Bioassay systems, CA) (Wang et al. 2017; Bi et al. 2012; Zhang et al. 2010; Wang et al. 2016) and Western blotting analysis of NMNAT1–3 and NAMPT.

For overexpression of NAMPT by viral transduction, primary neuronal cultures were infected with serotype 2/9 AAV vectors (AAV2/9) encoding NAMPT with CMV promoter, i.e., AAV2/9-CMV-NAMPT vectors (Wang et al. 2016). The vectors with a genome copy (GC)/cell ratio of 3×10⁵ were added to neuronal cultures of 3–5 DIV for 6 h in NBM medium without serum. The medium was then replaced by normal growth medium. The neurons were cultured for one week before experiment. The overexpression of NAMPT in neurons was confirmed by immunocytochemistry and/or Western blotting analysis using the anti-His-tag and anti-NAMPT antibodies, respectively.

Immunostaining

The procedures were described as in our previous studies (Bi et al. 2012; Wang et al. 2014; Wang et al. 2016). Neurons cultured on the coated-glass coverslips were washed with 1×PBS three times and fixed with ice-cold 4% paraformaldehyde (PFA) in 1×PBS for 15 min and then rinsed three times with 1×PBS carefully. Neurons were then permeabilized with ice-cold 0.1% Triton X-100 in 1× PBS for 10 min and blocked with 5% serum in 1×PBS at room temperature for 1 h. Neurons were then immunostained with primary antibodies including rabbit anti-NAMPT polyclonal antibody (1:300; A300–372A, Bethyl Laboratories, TX. RRID: AB_2251232), rabbit anti-AIF polyclonal antibody (1:200; PRS2301, Sigma-Aldrich. RRID:AB_2081981) in 1× PBS containing 1% serum, 0.1% Triton X-100 at 4°C overnight. The neurons were washed with 1× PBS for 5 min three times and then incubated with fluorescently labeled secondary antibodies Rhodamine-conjugated donkey anti-rabbit IgG (1:200; AP182R, Millipore. RRID: AB92592); FITC-conjugated goat anti-rabbit IgG (1:200; Millipore), Alexa fluor-488-conjugated donkey anti-mouse IgG (1:400; A21202, Life Technologies. RRID: AB_141607) in 1% serum, 0.1% Triton X-100 in 1× PBS for 1 h in the dark at room temperature. The nuclei were counterstained with DAPI. The samples were subjected to fluorescence detection using a Nikon FN1 epi-fluorescence microscopy equipped with a CoolSNAP-EZ CCD camera (Photometrics, AZ) or an Olympus FV1000 laser scanning confocal fluorescence microscope and analyzed by MetaMorph software (Molecular Devices, CA). Validation data of antibodies for immunostaining are available in the data sheets from the company and were compared with our experimental results. All the antibodies were purchased in the years of 2015–2019.

Subcellular fractionations

Mitochondria and cytosol were isolated from both cortical tissue of adult C57BL/6J mice and cultured cortical neurons using Dounce homogenization method with mitochondrial isolation kit according to a manufacturer’s protocol with slight modification (Cat. No. 89874, Fisher Scientific Inc. The kits were purchased in the years of 2015–2019). Neurons cultured in flasks were harvested by 0.1% Trypsin in PBS (6×10⁶ cells per sample). The cells or cortical tissues were added with Reagent A and then homogenized on ice using a 2-ml glass Dounce homogenizer (grind tube) with a tight-fitting pestle (Cat No. K885300–0002, Fisher Scientific Inc.). Crude lysates were added with Reagent C and centrifuged at 1,000g for 10 min at 4 °C to remove nuclei and unbroken cells in the resulting pellet (P1). The supernatant (S1) was collected and then subjected to centrifugation at 1,000g for 10 min again, yielding the 1,000g pellet (P2). The low-speed supernatant (S2) was collected and then subjected to centrifugation at 3,000g for 15 min again, yielding the 3,000g pellet (P3, containing mitochondria and heavy ER). The supernatant (S3) from the P3 pellet was centrifuged at 15,000g for 10 min to remove any remaining mitochondria. Finally, the 15,000g supernatant (S4) was separated into cytosol (S5) and light membrane fraction by centrifugation at 100, 000g for 1 h. The P3 pellet was washed twice in 500 μl Reagent C plus equal volume of H₂O and was centrifuged at 12,000g for 15 min. After the final wash, the mitochondrial pellet was lysed with 2% CHAPS in Tris buffered saline (TBS) (Cat. No. 28379, Fisher Scientific Inc.). The purities of mitochondrial and cytosolic fractions were assessed by probing mitochondrial and cytosolic markers Cox IV and LDH using Western blotting analysis. To assess sub-mitochondrial compartment localization of NAMPT, the purified mitochondrial pellet was digested by proteinase K (0.1–25.6 μg/ml) (Cat. No. E00492, Thermal Fisher Scientific) in PBS for 45 min on ice (McGee and Baines 2011). The digestion was terminated by the addition of 1 mM phenylmethanesulfonyl fluoride (PMSF) and the mitochondria were then subjected to Western blotting.

Western blotting analysis

Western blotting was used to analyze protein expression in whole cell lysate, mitochondrial and cytosolic subfraction from neuronal cultures and cortical tissues. Protein samples were separated by 10% SDS-polyacrylamide gel electrophoresis and transferred to PVDF membranes. The membranes were blocked by 5% non-fat dry milk and incubated with the following antibodies: rabbit anti-NAMPT polyclonal antibody (1:500, A300–372A, Bethyl Laboratories, TX. RRID:AB_2251232), mouse anti-His-tag monoclonal antibody (1:1000; sc-803, Santa Cruz, CA. RRID:AB631655), goat anti-NMANT3 (1:1000, ab121030, Abcam, MA. RRID: AB_10903780), mouse-anti-NMNAT1 (1:1000, sc-271557, Santa Cruz. RRID: AB_1064722), mouse-anti-NMNAT2 (1:1000, sc-515206, Santa Cruz), rabbit anti-lactate dehydrogenase (LDH) antibody (1:1000, ab52488, Abcam. RRID: AB_2134961), mouse-anti-Complex IV (Cox IV) antibody (1:1000; sc-58348, Santa Cruz. RRID: AB_29779213), rabbit anti-mitofusin 1 (Mfn1) (1:1000, sc-50330, Santa Cruz. RRID: AB_2250540), mouse anti-AIF (1:500, ab110327, Abcam. RRID: AB_1089118), mouse anti-ATP synthase subunit C (AtpS C) (1:1000, ab54880, Abcam. RRID: AB_2243245), rabbit anti- Complement 1q-binding protein (C1qbp) (1:1000, HPA026483, Sigma. RRID: AB_1847108), mouse anti- Optic atrophy 1(OPA1) (1:1000, BD612606, BD Bioscience. RRID: AB_399888), mouse anti-Cytochrome C (Cyto C) (1:1000, ab 110325, Abcam. RRID: AB_10864775), mouse anti-β-actin (1:2000, a3853, Sigma-Aldrich. RRID: AB_262137), mouse anti-GAPDH (1:1000, ab8245, Abcam. RRID: AB_2107448), rabbit anti-mouse IgG-peroxidase (A9044, Sigma-Aldrich. RRID: AB_258431), and goat anti-rabbit IgG-peroxidase (A0545, Sigma-Aldrich. RRID: AB_257896). The protein bands were visualized by ECL Western blotting detection system. β-actin and GAPDH were used as controls for equal loading of total protein of whole cell lysate. Validation data of Western blotting for all the antibodies are available in the data sheets from the company and were compared with our experimental results. The total protein amount of 20–30 μg/well was loaded, which was within the dynamic range (Fig. S2). All the antibodies were purchased in the years of 2015–2019.

OCR and ECAR analyses

OCR and ECAR assays in cultured neurons were conducted using Seahorse XFe96 extracellular flux analyzer as described previously (Agilent Tech., CA) (Wang et al. 2017). Cultured neurons were plated with density of 20,000 cells per well in XFe96 cell culture microplates and transfected with scrambled, and NMNAT1–3 and NAMPT-targeting siRNA (40 pmol siRNA and 0.5 μL lipofectamine 2000 per well) in quadruplicate. The individual wells in XFe96 microplate were washed and replaced by assay medium (Agilent minimal DMEM medium supplemented with 10 mM glucose, 1mM sodium pyruvate, 2 mM L-glutamine, pH 7.4) 48 h after transfection. The plate was then incubated in a 37 °C, non-CO₂ incubator for 1 h. The plate and pre-calibrated (hydrated) sensor cartridge were transferred to the microplate stage of a Seahorse XFe96 flux analyzer for OCR measurement. Three cycles of baseline measurement of OCR were taken followed by 3 cycles of sequential measurements after injection of ATP synthase inhibitor oligomycin (1 μM), FCCP (mitochondrial respiration uncoupler, 0.8 mM) (Cat. No. C2920, Sigma-Aldrich), and antimycin A (Complex III inhibitor, 1 mM) (Cat. No. A8674, Sigma-Aldrich) in conjunction with rotenone (Complex I inhibitor, 1 mM) (Cat. No. R8875, Sigma-Aldrich). For ECAR assay individual wells in XFe96 microplate were washed and replaced by assay medium (minimal DMEM medium supplemented with 1mM sodium pyruvate and 2 mM L-glutamine, pH 7.4) and was then incubated in a 37 °C, non-CO₂ incubator for 1 h. Three cycles of baseline measurement of ECAR were taken followed by 3 cycles of sequential measurements after injection of glucose (20 mM), oligomycin (1 μM) (Cat. No. 75351, Sigma-Aldrich) and 2-deoxyglucose (2-DG) (100 mM) (Cat. No. D8375, Sigma-Aldrich). After OCR and ECAR measurements, the cell numbers of each well were counted using a Countess II cell counter (Life technologies). OCR and ECAR data were normalized by 10⁴ cells and averaged from 4 replicates of representative experiments. For study the effect of NAD⁺ and FK866 on OCR and ECAR, culture neurons were incubated with different concentrations of NAD⁺ (Cat. No. N0632, Sigma-Aldrich) and FK866 (Cat. No. 481908, Sigma-Aldrich) for 3 h, and OCR and ECAR were conducted. All the chemicals were purchased in the years between 2018–2019.

Neuronal death assay

We used terminal dinucleotidyltransferase-mediated UTP end labeling (TUNEL) staining to evaluate apoptotic neuronal death after OGD (Wang et al. 2014; Li et al. 2013; Wang et al. 2016). Cultured neurons were fixed with a freshly prepared 4% PFA in PBS for 30 min, incubated with permeabilization solution (0.1% sodium citrate in 0.1% Triton X-100) for 2 min on ice, and subsequently with TUNEL reaction mixture (In Situ Cell Death Detection Kit, Fluorescein; Cat. No. 11684795910, Roche. Purchased in 2015) in a humidified environment for 1 h at 37 °C in the dark. Samples were imaged by a Nikon FN1 epi-fluorescence microscopy using a 40× magnification objective. The total number of neurons based on Dapi stained nuclei and the number of TUNEL+ apoptotic neurons were blindly counted by other investigators who was not doing the same experiments. The neuronal survival rate was defined as the ratio of the number of TUNEL- cells to the number of DAPI-stained nuclei.

Data analysis and statistics

Data were expressed as means ± SEM. All experiments were performed at least three times and the exact numbers of experiment or replicate are indicated in the figure legends. An assessment of normality of data and a test of outlier were not carried out. Statistical comparisons were made by t-test for two groups (Microsoft Excel software, Microsoft, WA 98052) and a one-way ANOVA test for multiple groups followed by Bonferroni’s post-hoc test (Origin Pro software, OriginLab Corporation Northampton, MA 01060). The statistical significance was set by *p< 0.05, **p<0.01.

Results

Neuronal mitochondria express enzymes involved in the NAD⁺ salvage pathway

We previously demonstrated for the first time that NAMPT was primarily expressed in neurons, but not in glial cells, in the mouse brain under normal conditions (Zhang et al. 2010). Here using immunostaining and confocal imaging, we show that while NAMPT is highly expressed in the neuronal nuclei (arrows), it is also evident that NAMPT is expressed in neuronal processes (arrow heads) in the mouse brain as well as in primary cultured mouse cortical neurons (Fig. 1a–b), suggesting that NAMPT could be expressed in the cytosol and mitochondria. To determine whether mitochondria express enzymes involved in the NAD⁺ salvage pathway, we isolated highly pure mitochondrial and cytosolic subfractions from mouse cortical tissues and cultured neurons. Our results showed that NAMPT was detected in both mitochondrial and cytosolic fractions, while NMNAT3 was detected exclusively in mitochondrial fraction from both preparations (Fig. 1c–d). The detection of both NAMPT and NMNAT3 in the mitochondrial fraction indicates the presence of an endogenous NAD⁺ salvage pathway in this organelle.

Fig. 1. Expression of enzymes involved in the NAD⁺ salvage pathways in mitochondria and cytosol.

(a-b) Immunostaining of NAMPT in mouse cortex and primary cultured mouse cortical neurons. Notice NAMPT is expressed in cell body and neuronal processes. (c-d) Western blot images of NAMPT and NMNAT3 in mitochondrial and cytosolic subfractions isolated from adult mouse cortical tissues (c) and primary cultured mouse cortical neurons (d). Notice NAMPT is detected in both mitochondrial and cytosolic fractions, while NMNAT3 is only detected in the mitochondrial fraction. (e-h) AAV-mediated NAMPT overexpression in the mitochondria and cytosol in the primary cultured mouse cortical neurons. (e-f) Fluorescence images of His-tag and NAMPT immunostaining in cultured neurons without and with AAV-NAMPT transduction. (g) Western blot images of NAMPT and His-tagged NAMPT in cultured neurons without (Control) and with AAV transduction. (h) Western blot images of His-tagged NAMPT in mitochondrial and cytosolic fractions in cultured neurons without (Control) and with AAV transduction. The purities of mitochondrial and cytosolic fractions were indicated by mitochondrial and cytosolic markers Cox IV and LDH. Weak LDH signal was visible in mitochondrial subfraction in cultured neurons. Fluorescent images (a-b, e-f) are represent images from at least three different neuronal cultures and three animals. Western blot images are represent images from three different neuronal cultures and three animals for subfractionation.

To further define the subcellular localization of NAMPT, we infected neurons with AAV vectors encoding human NAMPT (hNAMPT) with a His-tag in C-terminal (Wang et al. 2016). Neurons infected with AAV exhibited His-tag immunofluorescent signal and increased NAMPT immunofluorescent signal, while His-tag signal was absent and NAMPT signal was weak in neurons without AAV infection (Control) (Fig. 1e–f). The overexpression of hNAMPT by AAV infection was also confirmed by Western blotting analyses of NAMPT and His-tagged NAMPT of the whole lysate of cultured neurons (Fig. 1g). Moreover, His-tagged NAMPT was detected in the mitochondrial as well as cytosolic subfractions from infected neurons, but it was absent in the both subfractions from non-infected neurons (Fig. 1h), indicating AAV-mediated overexpression of ectopic NAMPT occurs in both mitochondria and cytosol.

To provide additional evidence of mitochondrial expression of NAMPT in neurons, we constructed DNA plasmids that encode mRFP with mitochondrial targeting sequence (mito) and nuclear export sequence (NES) under the control of neuronal promoter CaMKII, i.e., pZac-CaMKII-mito-mRFP, pZac-CaMKII-NES-mRFP plasmids (see Fig. 2a, Fig. 6a i–ii, Fig. S1i–ii, and Supplemental information of DNA plasmid sequences) and transfected them in cultured neurons. These molecular approaches have been widely used for protein expression in mitochondria and cytoplasm (Cambronne et al. 2016; Rodrigues et al. 2007; Imamura et al. 2009; Wang et al. 2016; Rizzuto et al. 1995; Li et al. 2014). High-resolution confocal images confirmed mitochondrial and cytoplasmic patterns of mRFP expression in transfected neurons (Fig. 2a–b), demonstrating that mito and NES indeed lead to the expression of mRFP in respective subcellular compartments. We then co-transfected neurons with pZac-CaMKII-mito-mRFP plasmid and pCAGGS-hNAMPT plasmid which encodes non-subcellular compartment-targeting hNAMPT. Single optical section confocal images revealed that NAMPT immunofluorescent signal in the dendrite of the transfected neuron is stronger than in the dendrites of the non-transfected neurons, and mRFP fluorescence is colocalized with NAMPT signal (arrows, Fig. 2c–d). Although confocal microscopy cannot completely exclude out-of-focus florescent signal (see explanation in the figure legend), these results provide structural evidence that NAMPT is expressed in mitochondria in addition to cytoplasm, corroborating the results from Western blot analysis of NAMPT in mitochondrial and cytosolic subfractions. Thus, our results from Western blotting analysis and confocal microscopy provide compelling evidence that neuronal mitochondria possess an intact NAD⁺ salvage pathway.

Fig. 2. Localization of ectopic NAMPT in the mitochondria.

(a-b) Confocal images of mRFP expression in mitochondria (a) and cytoplasm (b) in the neurons transfected with pZac-CaMKII-mito-mRFP and pZac-CaMKII-NES-mRFP. c) Single optical confocal images of NAMPT immunofluorescence of Alexa 488 (Left) and mito-RFP (Middle) in the neurons cotransfected with pZac-CaMKII-mito-mRFP and pCAGGS-hNAMPT. (d) High resolution confocal images of the boxed region in (c). Notice a high NAMPT signal in a dendrite is colocalized with mito-mRFP (arrowheads). Images are representative confocal images of n=7–11 transfected neurons for each condition. Here a 60× oil objective (Plan Apo, NA1.42, WD 0.2 mm) was used, and an image size of 1024×1024 pixel was acquired. For imaging mRFP, excitation wavelength was 559 nm from a diode laser and emission wavelength was >575 nm using a long pass emission filter. For imaging NAMPT stained with Alexa 488, excitation wavelength was 488 nm from an argon ion laser and emission wavelength was 525/50 nm using a band pass emission filter. Note: Confocal microscopy can reduce out-of-focus fluorescence (the amount can be adjusted by pinhole size). Based on the parameter of 60x objective (NA=1.42), excitation wavelengths for mRFP (λ_mRFP=559 nm) and Alexa 488 2^nd antibody for NAMPT staining (λ_{Alexa 488}=488 nm), we calculated lateral resolution (r_xy) and axial (or z) resolution (r_z), i.e., r_xy=0.4λ/NA=126 nm for NAMPT and 157 nm for mRFP, and r_z=1.4nλ/NA²=314 nm for NAMPT and 360 nm for mRFP, here n=1.3, the refractive index of oil. Since dendritic mitochondrion has a size of a few micrometer (Lewis et al. 2018;Wang et al. 2016), the resolution of our confocal microscopy should be high enough for mitochondrial imaging.

Fig. 6. The effects of NAMPT overexpression in mitochondria and cytoplasm on neuronal death after OGD.

(a) DNA construct maps. (b-e) Epi-fluorescent images of TUNEL staining and mRFP expression in neurons transfected with pZac-CaMKII-mRFP (b), pZac-CaMKII-mRFP-NAMPT (c), pZac-CaMKII-NES-mRFP-NAMPT (d) and pZac-CaMKII-mito-mRFP-NAMPT (e) 24 h after 1.5 h OGD. Nuclei were counterstained by DAPI. (f) Neuronal survival rates based on TUNEL staining. Data were averaged from n=10–24 images for each condition for three independent preparations. In box overlap plot, box, line, square and whiskers indicate 25–75% interquartile range (IQR), median value, mean value, and 1.5IQR above and below the box, respectively. Here we must point out that because we already demonstrated that NES and mitochondrial matrix targeting sequence indeed lead to mRFP expression in cytoplasm and mitochondria (Fig. 3) and due to the low cell density of neuronal cultures, epi-fluorescent imaging can reliably identify the co-staining of mRFP, TUNEL and DAPI for cell death analysis. **p<0.01 versus mRFP, ANOVA test.

NAMPT and NMNAT3 localize to the mitochondrial matrix

Since Western botting analysis of intact mitochondrial subfraction cannot determine which soluble subcompartment of NAMPT and NMNAT3 reside in the mitochondria, we digested mitochondrial subfractions isolated from primary cultured neurons and cortical tissue with different concentrations of proteinase K (McGee and Baines 2011). The digested mitochondria were then subject to immunoblotting with proteins of known submitochondrial localization. The outer membrane protein Mfn1 was fully digested at relatively low protease concentrations, the inter-membrane protein OPA1 was digested slowly compared with Mfn1, while the inner membrane AtpS C and matrix C1qbp were increasingly resistant to digestion. The digestion profiles of NAMPT and NMNAT3 matched that of matrix protein C1qbp, being only digested at the high proteinase K concentration (Fig. 3a–b). These results indicate that NAMPT and NMNAT3 are localized in the mitochondrial matrix to mediate NAD⁺ salvage (Fig. 3c).

Fig. 3. NAMPT and NMNAT3 are soluble matrix proteins.

(a-b) Mitochondria isolated from mouse primary cultured cortical neurons (a) and mouse cortex (b) were digested by proteinase K with different concentrations for 40 min on ice and then immunoblotted for Mfn1, OPA1, AtpS C, C1qbp, NAMPT and NMANT3. OM-outer membrane, InterM-Inter membrane space, IM-Inner membrane, MM-mitochondrial matrix. Data are representative images from n=3 different mitochondrial subfractionation of neuronal culture and n=4 mouse cortical tissue. (c) Schematic illustration of mitochondrial NAD⁺ salvage pathway in neurons. Based on our results, both NAMPT and NMNAT3 are expressed in the mitochondria and are localized in the mitochondrial matrix. We propose neuronal mitochondria have an intact NAD⁺ salvage pathway and the substrates NAM and PRPP can be transported from cytoplasm or synthesized within the organelle (route 1). For mammalian cells lacking NAMPT expression in the mitochondria, NMN is transported into mitochondria for NAD⁺ salvage (route 2) (Nikiforov et al. 2011).

The effect of subcellular NAMPT-mediated NAD⁺ salvage pathways on cellular bioenergetics

NAD⁺ is highly compartmentalized in the cytosol and mitochondria of mammalian cells including neurons (Alano et al. 2007; Alano et al. 2010; Yang et al. 2007). To determine the effect of mitochondrial NAMPT-mediated NAD⁺ salvage pathway on overall cellular bioenergetics, we knocked down NAMPT and NMNAT1–3 in primary cultured neurons via siRNA transfection. Western blotting analysis with the total protein in dynamic range confirmed significant reductions of NAMPT and NMNAT1–3 levels by respective siRNA, but there was no reduction of non-targeting protein levels (Fig. 4a–b, Fig. S3). Morphologically, neurons transfected with NAMPT and NMNAT1–3 siRNA exhibited dendritic beading compared with neurons without siRNA transfection or transfected with scrambled siRNA (Fig. 4c–h). Transfection of NAMPT and NMNAT1–3 siRNA significantly reduced cellular NAD⁺ and NADH levels as compared with control condition and transfection of scrambled siRNA (Fig. 4i–j), but the transfection of NMNAT3 and NAMPT siRNA did not cause a significant reduction in NAD⁺ and NADH⁺ levels as compared with the transfection of NMNAT1–2 siRNA.

Fig. 4. Knockdowns of NAMPT and NMANT1–3 in cultured neurons reduces cellular bioenergetics.

(a) Western blot images of targeting proteins and other proteins in the salvage pathway. The boxed images (red) indicate the signal reduction of targeting protein. Here, cultured neurons were subject to Western blotting 48 h after the transfection of scrambled and respective targeting siRNAs. (b) Western blotting analysis of targeting proteins. Total amount of 20 μg protein was loaded, which is in the dynamic range of analysis (see Fig. S2). Data were averaged from n=3 different neuronal cultures for each condition. *p<0.05, vs Ctrl, ANOVA test. (c-h) Representative phase contrast images of neurons without siRNA transfection (c) and 24 h after the transfection of indicated siRNA (d-h). (i-j) Summary of normalized NAD⁺ (i) and NADH (j) levels in neurons transfected by different siRNA. Data were averaged from n=3 replicates of each condition from a representative experiment. *p<0.05, vs Ctrl, ANOVA test.

Cellular bioenergetics is largely affected by mitochondrial respiration and glycolysis. We therefore determined the effects of NAMPT and NMANT1–3 knockdowns on OCRs and ECAR of primary cultured neurons using Seahorse extracellular flux analysis. OCR time courses were generated by sequentially injecting electron transport chain (ETC) inhibitors oligomycin (an ATP synthase inhibitor), FCCP (a mitochondrial uncoupler) and rotenone (Rot) and antimycin A (AA) (Complex I and III inhibitors) to probe mitochondrial oxygen consumption under different states (Fig. 5a). Our results show that knockdowns of NAMPT and NMNAT1–3 significantly reduced maximal respiration, but did not affect spare respiratory capacity and proton leak (Fig. 5c, 5e–f). Only knockdown of NMNAT3 significantly reduced basal respiration, while knockdowns of NMNAT3 and NAMPT reduced ATP production-related respiration (Fig. 5b and d). Overall, the results suggest that mitochondrial NAD⁺ salvage pathway has a slightly stronger effect on mitochondrial respiration than cytosolic and nuclear pathways. Pharmacologically, inhibition of NAMPT by FK866 suppressed OCR in a dose-dependent manner (Fig. S4). It was reported that repletion of NAD⁺ has a neuroprotective effect after ischemia (Bi et al. 2012;Wang et al. 2008), and extracellular NAD⁺ can be transported into cell membrane in cell lines (Pittelli et al. 2011;Billington et al. 2008), thus we also tested the effect of extracellular NAD⁺ on OCR. Our results show that repletion of NAD⁺ can largely increase cellular NAD⁺ and NADH levels without affecting NAD⁺/NADH ratio (Fig. S5a–c), and enhance OCR in a dose-dependent manner in neurons (Fig. S5d–i).

Fig. 5. The effects of knockdowns of NMANT1–3 and NAMPT on oxygen consumption and glycolysis.

(a) OCR traces of neurons transfected with scramble, NMNAT1–3- and NAMPT-targeting siRNA. (b-f) Analyses of basal respiratory capacity (b), maximal respiration (c), ATP production-linked respiration (d), spare respiratory capacity (e), and proton leak (f). (g) ECAR traces of neurons transfected with scramble, NMNAT1–3- and NAMPT-targeting siRNA. H-J) Analysis of basal glycolysis (h), glycolytic capacity (i) and glycolytic reserve (h). For OCR and ECAR assays, data were averaged from n=4 replicates of each condition from a representative experiment of three preparations. *p<0.05, **p<0.01, vs Ctrl, ANOVA test.

Next, we assessed glycolytic flux by analyzing ECARs, an indirect measurement of lactate production, after sequential addition of glucose, oligomycin and 2-DG (Fig. 5g). Knockdowns of NMNAT1–3 and NAMPT significantly reduced glycolytic capacity; however, only knockdown of NMNAT1 significantly reduced basal glycolysis, while knockdowns of NMNAT1–2 and NAMPT reduced glycolytic reserve (Fig. 5h–j). FK866 reduced basal glycolysis, glycolytic reserve and glycolytic capacity in a dose-dependent manner (Fig. S6a–d), while NAD⁺ repletion has an opposite effect (Fig. S6e–h). These results indicate that disruption of NAMPT-mediated NAD⁺ biosynthesis causes defective glycolytic function with NMNAT1 and NMNAT2 having a larger effect than NMNAT3.

Overexpression of NAMPT mitochondria or cytoplasm alone is sufficient to protect against OGD induced neuronal death

It has been known that ischemia depletes global NAD⁺ (Zhang et al. 2010; Bi et al. 2012; Liu et al. 2009). Giving that both mitochondria and cytosol have their own NAD⁺ synthesis pathway, we next determined whether mitochondrial and cytoplasmic NAD⁺ salvage pathways have a different effect on neuronal death after ischemia. To test this hypothesis, we constructed DNA plasmids pZac-CaMKII-mito-mRFP-NAMPT and pZac-CaMKII-NES-mRFP-NAMPT for expressions of mRFP-NAMPT in the mitochondria and cytoplasm, while plasmids pZac-CaMKII-mRFP and pZac-CaMKII-mRFP-NAMPT were constructed as internal controls for non-subcompartmental expressions of mRFP and mRFP-NAMPT (Fig. 6a iii–vi, Fig. S1iii–vi). We then transfected these plasmids in neurons which were subjected to 1 h OGD two days after transfection. One day after OGD, we conducted TUNEL staining to assess apoptotic neuronal death to mimic neuronal degeneration in the penumbra in in vivo ischemic stroke. Fluorescent images showed that most of neurons expressing mRFP were TUNEL+ after OGD, while most of neurons overexpressing mRFP-NAMPT, NES-mRFP-NAMPT and mito-mRFP-NAMPT were TUNEL- (Fig. 6b–e). Quantitative analyses found that the overexpression of mRFP-NAMPT, NES-mRFP-NAMPT and mito-mRFP-NAMPT significantly promoted neuronal survival after OGD as compared with the overexpression of mRFP (Fig. 6f).

AIF is a mitochondrial protein that relocates to the nucleus after ischemic stroke to mediate caspase-independent apoptotic cell death (Hong et al. 2004; Chiarugi 2005; Wang et al. 2014; Wang et al. 2016). Consequently, we examined the effects of NAMPT overexpression in different subcellular compartments on OGD-induced AIF translocation. Similar to previous experiments, primary cortical neurons were transfected with DNA plasmids pZac-CaMKII-mRFP, pZac-CaMKII-mRFP-NAMPT, pZac-CaMKII-mito-mRFP-NAMPT and pZac-CaMKII-NES-mRFP-NAMPT. AIF localization under control and OGD conditions was evaluated using AIF immunostaining. Under the control conditions, AIF was largely localized outside the nucleus in neurons (Fig. 7a), while OGD stimulation extensively induced the re-localization of AIF to the nuclei in neurons transfected with mRFP (Fig. 7b, 1^st row). Consistent with the effects of subcellular compartment NAMPT overexpression on apoptosis, AIF remained extra-nuclear after OGD in neurons overexpressing mRFP-NAMPT, mito-mRFP-NAMPT and NES-mRFP-NAMPT (Fig. 7b, 2^nd-4^th rows). Line-scan analysis shows that AIF levels were high in the cytoplasm of neurons transfected with different constructs under normal conditions (Fig. 7c). Neurons overexpressing mRFP exhibited lower AIF signals in the cytoplasm than in the nuclei after OGD, while neurons overexpressing mRFP-NAMPT, mito-mRFP-NAMPT and NES-mRFP-NAMPT exhibited opposite AIF expression pattern (Fig. 7d). Quantitative analyses indicate that neurons overexpressing mRFP, mRFP-NAMPT, mito-mRFP-NAMPT and NES-mRFP-NAMPT exhibited no significant difference in the percentage of AIF translocation under normal conditions, but exhibited significant reductions in the percentage of AIF translocation after OGD as compared with neurons overexpressing mRFP (Fig. 7e). However, there was no difference in AIF translocation in neurons overexpressing mRFP-NAMPT, mito-mRFP-NAMPT and NES-mRFP-NAMPT.

Fig. 7. The effects of NAMPT overexpression in mitochondria and cytoplasm on AIF translocation after OGD.

(a-b) Fluorescent images of AIF staining and mRFP expression in neurons transfected with pZac-CaMKII-mRFP, pZac-CaMKII-mRFP-NAMPT, pZac-CaMKII-mito-mRFP-NAMPT and pZac-CaMKII-NES-mRFP-NAMPT under control (a) and OGD (b) conditions. (c-d) Representative line-scan analysis of AIF and DAPI fluorescence under control and OGD from the cells indicated in (a-b). The fluorescence was normalized to the background. Notice the translocation of AIF after glutamate stimulation and the prevention of AIF translocation by NAMPT overexpression. (e) Summary of AIF translocation under different conditions. Data were averaged from n=8–18 images for each condition for three independent preparations. In box overlap plot, box, line, square and whiskers indicate 25–75% interquartile range (IQR), median value, mean value, and 1.5IQR above and below the box, respectively.**p<0.01 versus mRFP, ANOVA test.

Taken together, these results indicate that mitochondrial-, cytoplasmic- and non-subcellular compartmental- overexpressions of NAMPT have comparable effects on neuronal protection, and cytoplasm- or mitochondria-targeted NAMPT overexpression alone is sufficient to confer a protective effect against OGD-induced neuronal apoptosis and prevent OGD-induced AIF translocation.

Discussion

The expression of NAMPT in mammalian cells can be cell-type specific. Our and other groups showed that NAMPT is primarily expressed in neurons in mouse brain under normal conditions (Zhang et al. 2010; Wang et al. 2011). Our previous studies also highlight the importance of the NAD⁺ salvage pathway in maintaining mitochondrial function to mediate neuronal protection after ischemia (Bi et al. 2012;Wang et al. 2014;Wang et al. 2016). NAD⁺ was detected within distinct subcellular compartments with different concentrations in mammalian cells, and their NAD⁺ pools are likely independently regulated under normal conditions (Cambronne et al. 2016; Alano et al. 2007; Alano et al. 2010; Yang et al. 2007; Canto et al. 2015). Mitochondrial NAD⁺ levels can be preserved when significant depletion of NAD⁺ occurs in the cytosol during genotoxic stimulation (Pittelli et al. 2010). It was reported that NAMPT is involved in the regulation of both cytosolic and mitochondrial NAD⁺ pools in mammalian cells, but results whether NAMPT is expressed in mitochondria in mammal cells were inconsistent (Yang et al. 2007; Pittelli et al. 2010; Morris-Blanco et al. 2014; Nikiforov et al. 2011). On the other hand, mitochondrial occurrence of NMNAT3 in mammalian cells is also controversial (Felici et al. 2013; Zhang et al. 2003; Berger et al. 2005). The current model for mitochondrial NAD⁺ biosynthesis is that NAM is converted to NMN and NMN is then transported into mitochondria for subsequent NAD⁺ synthesis by NMNAT3 (Nikiforov et al. 2011; Formentini et al. 2009; Stein and Imai 2012; Davila et al. 2018) (Fig. 3c). Thus, it is unknown whether there is an intact functional machinery of salvage pathway to synthesize NAD⁺ from NAM within neuronal mitochondria. In the current study, using multiple cellular and molecular approaches including mitochondrial and cytosolic subfractionation, viral transduction, enzymatic digestion of mitochondrial subfraction, Western blotting analyses and confocal imaging, we clarified that mitochondria indeed possess an intact NAD⁺ salvage pathway, and provided compelling evidence for the first time that both NAMPT and NMNAT3 are localized in mitochondrial matrix to mediate NAD⁺ synthesis (Fig. 3c). It is an open question regarding how NAMPT is transported into matrix since NAMPT does not have mitochondrial localization sequence. However, structural features such as positively charged amino acids (Lys, His, Arg) and amphipathic α-helices are known to facilitate matrix import (Schmidt et al. 2010). In this regard, NAMPT has a cluster of 5 Lys and 1 Arg between residues 42 and 53 and an α-helix (residues 98–108) that is predicted to be amphipathic (http://heliquest.ipmc.cnrs.fr). Alternatively, NAMPT could be imported via the chaperone-dependent pathway (Becker et al, 2019). Indeed import of various presequence lacking molecules such as protein kinase Cε, granzyme-B and P450 enzymes into the matrix has been shown to be Hsp70/Hsp90-dependent (Budas et al. 2010; Chiusolo et al. 2017; Anandatheerthavarad et al. 2009).

In NAD⁺ salvage pathway, both NAM and 5-phosphoribosyl pyrophosphate (PRPP) are substrates for NAMPT. PRPP is synthesized in cytoplasm by ribose-phosphate diphosphokinase or PRPP synthetase (PRPS) from ribose-5-phosphate produced in pentose phosphate pathway (PPP). It is also possible that PRPP can be synthesized within or transported into the mitochondria. It was reported that PRPS3 has a mitochondrial localization amino acid sequence and minor PRPP synthase activity was observed in mitochondria in eukaryotic cells (Hove-Jensen et al. 2017), but there was no report regarding PRPP transportation through mitochondrial membrane. For NAM, it can either enter mitochondria from cytoplasm or be synthesized within mitochondria. Sirtuins, PARPs and CD38 consume NAD⁺ as a co-substrate and release NAM. In mammal cells, some members of the three NAD⁺-consuming enzyme families are localized in the mitochondria (Saunders and Verdin 0 AD;Leung 2014;Orciani et al. 2008), therefore, NAMPT could use NAM within the mitochondria to mediate NAD⁺ salvage (Fig. 3c).

Since NAD⁺ is a cofactor for glycolysis, TCA cycle and complex I enzymatic reaction, we studied the effects of NAMPT and NMNAT1–3 on cellular NAD⁺ and NADH levels in neurons. Knockdowns of NAMPT and NMNAT1–3 reduced overall intracellular NAD⁺ and NADH levels compared with control conditions, indicating NAMPT and NMNAT1–3 indeed regulate energy production through the NAD⁺ salvage pathway; but there is no significant difference in NAD⁺ and NADH levels among the knockdowns of NAMPT and NMNAT1–3. There are a couple of possibilities to explain the results. First, since NMNAT1–3 are expressed in distinct compartments, knockdowns of them only reduce NAD⁺ and NADH in respective compartments and their contributions to overall cellular NAD⁺ and NADH levels depend on their pool size and may not be significant. In future, it is feasible to conduct live cell imaging using subcompartment-targeted NAD⁺ biosensor to estimate pool size and NAD⁺ levels in respective subcompartments (Cambronne et al. 2016). Second, the transfection efficiency of siRNA in neurons is not high enough to distinguish their effects on cellular NAD⁺ and NADH.

We also studied the effects of NAMPT and NMNAT1–3 on OCR and ECAR in neurons. The knockdowns of NAMPT and NMNAT1–3 all significantly reduced maximal OCR, but only knockdown of NMNAT3 significantly reduced basal OCR, and NAMPT and NMNAT3 significantly reduced ATP-production related OCR as compared with NMNAT1–2. In ECAR assay, knockdowns of NMNAT1–3 and NAMPT all reduced glycolytic capacity, but only knockdown of NMNAT1 reduced basal ECAR, and knockdowns of NMNAT1–2 and NAMPT, not NMNAT3 affected glycolytic reserve, suggesting NMNAT1–2 have a stronger effect on glycolytic function than NMNAT3. These data underscore the differential effect of subcellular compartmental NAMPT-mediated NAD⁺ biosynthesis pathways on energy metabolism, mitochondrial and glycolytic function.

In the current study, we also found that mitochondria- and cytoplasm-specific and non-subcompartmental overexpressions of NAMPT have a similar effect on neuronal protection and AIF translocation after OGD. Since mitochondria have a higher NAD⁺ concentration than cytoplasm under normal conditions (Cambronne et al. 2016), we expected that the overexpression of NAMPT in mitochondria would have a stronger protective effect against neuronal death after OGD. There are a few possibilities to explain the results. First, the mitochondrial and cytoplasmic NAD⁺ salvage pathways have similar contributions to overall cellular bioenergetics. Second, since there is an endogenous NAMPT-mediated NAD⁺ salvage pathway in different subcellular compartments, overexpression of NAMPT may not be able to differentiate their effect on neuroprotection. Third, although the mitochondrial and cytoplasmic compartments can maintain their distinct NAD⁺ pools under normal conditions, due to the opening of mitochondrial permeability transition pore (PTP) resulting from the breach of mitochondrial membrane under ischemic conditions, mito- and cyto-NAD⁺ pools become exchangeable and equal (Di Lisa et al. 2001). In the future studies, it will be important to test whether overexpression of NAMPT will inhibit PTP opening and whether the PTP inhibitor Cyclosporin A (CsA) (Di Lisa et al. 2001) can ameliorate the depletion of mitochondrial NAD⁺ pool after ischemia. Also, to exclusively test the effect of mitochondrial and cytoplasmic NAMPT on neuronal protection after ischemia, it is required to use neurons lacking endogenous NAMPT. This could be achieved using cultured neurons from conditional and tamoxifen (TAM)-inducible knockout mice (Wang et al. 2017; Rongvaux et al. 2008). Since the deletion of NAMPT causes neurodegeneration in mouse model (Wang et al. 2017), these neurons can be transfected to overexpress NAMPT in mitochondria and cytoplasm prior to deletion of endogenous NAMPT, and subsequently are subject to OGD to test their respective neuroprotective effects.

In the present study, we did not test the effect of nuclear NAMPT on neuronal protection since NAD⁺ is readily exchangeable between the nucleus and cytoplasm due to the large nuclear pores and in fact similar NAD⁺ concentrations were observed in mammalian cells (Berger et al. 2005), thus it is expectable that nucleus NAMPT will also exhibit a similar neuroprotective effect with cytoplasmic NAMPT in ischemia. Ischemia causes the reduction of NAD⁺, but the dynamic changes of NAD⁺ depletion and the relative speeds in different subcompartments are unknown. In the future study, we will conduct live cell NAD⁺ imaging to monitor NAD⁺ levels in the cytoplasm, nucleus and mitochondria using genetically encoded NAD⁺ biosensors after ischemic insult (Cambronne et al. 2016), which should provide direct evidence of dynamic depletion of NAD⁺ in different subcellular compartments. It will be also interesting to conduct combined live cell MMP and NAD⁺ imaging during OGD or glutamate excitotoxicity to establish the relationship between breach of the mitochondrial membrane and NAD⁺ decline during ischemia.

In addition to modulating cellular energy homeostasis and metabolism in glycolysis, TCA cycle and oxidative phosphorylation in the mitochondria, NAD⁺ is also a substrate of NAD⁺-consuming enzymes such as Sirtuins (SIRT1–7) and Poly (ADP-Ribose) Polymerases (PARPs) (Canto et al.2015 ). Thus, SIRTs and PARPs might play important roles in neuronal fate in ischemia. In this regard, SIRT1 can regulate aging, apoptosis, gene expression and cell differentiation in human cells (van der Veer et al. 2007; Dvir-Ginzberg et al. 2008; Saunders and Verdin 2007; Yang et al. 2006). NAMPT has also been implicated in the regulation of SIRT1 (Dvir-Ginzberg et al. 2008; Revollo et al. 2004; Skokowa et al. 2009). Moreover, activation of SIRT1 stimulates mitochondrial biogenesis through regulating PGC-1α, NRF-1, Tfam and Cox IV (Chen et al. 2010; Csiszar et al. 2009; Lagouge et al. 2006; Nisoli et al. 2005). Thus, NAMPT overexpression might mediate a neuroprotective effect through SIRT1 activation. Consistent with this idea, it was reported that NAMPT overexpression can reduce neuronal death and brain damage through the SIRT1-dependent AMPK pathway and autophagic induction (Wang et al. 2011; Wang et al. 2012). Global overexpression of NAMPT in transgenic mice promotes regenerative neurogenesis in vivo through the NAD⁺-dependent deacetylase activity of SIRT1 after middle cerebral artery occlusion (MCAo) and thus improves brain recovery (Zhao et al. 2015). PARP-1 is a key mediator of cell death in excitotoxicity and ischemia. PARP-1 activation leads to cytosolic NAD⁺ depletion and AIF translocation after ischemic stroke. Restoration of cytosolic NAD⁺ prevents mitochondrial failure, AIF translocation, and neuronal death. Thus, overexpression of NAMPT may protect neurons through the suppression of mitochondrial dysfunction that otherwise results in extensive PARP-1 activation (Alano et al. 2010). NAD⁺ depletion is a causal event in PARP-1-mediated cell death in the upstream of mitochondrial AIF release. These studies are consistent with the results in the current study.

In summary, the current study provides a few novel findings. First, neuronal mitochondria have an intact NAD⁺ salvage pathway and NAMPT and NMNAT3 are localized in the mitochondrial matrix. Second, disruption of subcellular NAMPT-mediated NAD⁺ salvage pathways in neurons by siRNA causes reduction in oxygen consumption and defective glycolytic function. Finally, mitochondria- and cytoplasm-targeted and non-subcellular compartmental NAMPT overexpressions have a comparable effect on neuronal protection after ischemia. These results provide insights into the roles of subcellular NAD⁺ salvage pathways in NAD⁺ homeostasis, bioenergetics and neuronal protection in ischemia, and suggest that promoting cellular energy homeostasis through NAMPT-mediated NAD⁺ salvage pathway is a valuable neuroprotective strategy in ischemic stroke therapy.

Supplementary Material

Supp info

Fig. S1 The illustration of DNA construction and restriction enzyme sites of DNA plasmids.

Fig. S2 The dynamic range of total protein amount for Western blotting analysis.

Fig. S3 Quantification of non-targeting proteins in NAD+ salvage pathway after siRNA transfection.

Fig. S4 The effect of inhibition of NAMPT by FK866 on oxygen consumption of cultured neurons.

Fig. S5 The effect of NAD+ repletion on oxygen consumption of cultured neurons.

Fig. S6. The effect of FK866 and NAD+ on glycolytic function of cultured neurons.

Abbreviation list:

2-DG

2-Deoxyglucose

AA

Antimycin A

AIF

Apoptosis-inducing factor

AtpS C

Atp synthase subunit C

C1qbp

Complement 1q-binding protein

ECAR

Extracellular acidification rate

FCCP

Carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone

FIS

Focal ischemic stroke

GAPDH

Glyceraldehyde-3-phosphate dehydrogenase

Mfn1

Mitofusin 1

NAMPT

Nicotinamide phosphoribosyltransferase

NMNAT1–3

Nicotinamide mononucleotide adenyltransferase1–3

NAD⁺

Nicotinamide adenine dinucleotide

NAM

Nicotinamide

NMN

Nicotinamide mononucleotide

OCR

Oxygen consumption rate

OGD

Oxygen glucose deprivation

OPA1

Optic atrophy 1

PRPP

5-Phosphoribosyl pyrophosphate

Rot

Rotenone

RRID

Research Resource Identifier (see scicrunch.org)

WT

Wild-type