NAD catabolism, GTP and mitochondrial dynamics

Mitochondrial NUDIX hydrolases: A metabolic link between NAD catabolism, GTP and mitochondrial dynamics

2. Pathology of NAD+ catabolism and downstream metabolic products
NAD+ is an important cofactor involved in multiple metabolic reactions that have a central role in cellular metabolism and energy production (Belenky et al., 2007; Frederick et al., 2016; Mouchiroud et al., 2013). Normally, NAD+ levels decline with age (CamachoPereira et al., 2016; Massudi et al., 2012; Verdin, 2015; Zhu et al., 2015) and NAD+ pools have been shown to decrease during neurodegenerative diseases and after ischemia-reperfusion or TBI (Kauppinen and Swanson, 2007; Martire et al., 2015; Park et al., 2016; Verdin, 2015; Zhou et al., 2015). This decline could be the result of the increased activity of several enzymes that use NAD+ as their substrate; these include: sirtuins, ADP-ribosyl transferases (ARTs), and the cyclic ADP-ribose synthases/ADP-ribosyl cyclases (CD38 and CD157) (Belenky et al., 2007; Feijs et al., 2013; Jesko et al., 2016; Malavasi et al., 2008; Mayo et al., 2008). These NAD+ dependent enzymes hydrolyze NAD+ to ADP-ribose (ADPr) or an ADPr variation (e.g. cyclic ADPr), and nicotinamide (Nam). The resulting Nam can function in a negative feedback fashion by inhibiting the activity of NAD+ dependent enzymes (Avalos et al., 2005; Long et al., 2016; Suzuki et al., 2010), however levels of Nam are usually too low to inhibit sirtuins activity under physiological conditions (Liu et al., 2013). Nam is also the precursor for nicotinamide mononucleotide (NMN), the immediate precursor to NAD+ via the salvage pathway (Imai and Guarente, 2014). The conversion of NMN to NAD+ is ATP dependent and catalyzed by the NMN adenyl transferases (NMNATs): NMNAT1 (nucleus), NMNAT2 (Golgi apparatus/endosomes), and NMNAT3 (mitochondria)

Additionally, commonly in acidic environments, NAD+ levels can also be depleted by NAD+ kinase, by phosphorylating NAD+ to NADPþ (Zhang et al., 2016a).

3. NAD+/NADH
Other then being a substrate, NAD+ plays a key role in cellular bioenergetic metabolism by reversibly being reduced to NADH and ultimately contributing to ATP generation via the glycolytic pathway and in the mitochondria through oxidative phosphorylation (Srivastava, 2016). The reduction of NAD+ is most apparent in the mitochondria, with liver mitochondrial NAD+/NADH ratios being tightly controlled around 7 to 8, while cytoplasmic ratios being much higher, ranging from 60 to 700 in most cells (Stein and Imai, 2012). Techniques to get more accurate measurements are relatively new but suggest that NAD+ pools and NAD+/NADH ratios can vary based on cell type (Cambronne et al., 2016; Christensen et al., 2014). The reduction of NAD+ to NADH is essential for the glyceraldehyde 3-phosphate (GAPDH) step of glycolysis and multiple steps in the tricarboxylic acid cycle (TCA) (Akram, 2014; Sirover, 1999). NADH generated in the mitochondria will be oxidized to NAD+ and donate electrons to complex I of the electron transport chain (Sazanov, 2015).

4. ARTs/PARP1/mtPARP1
There are 22 known human genes that encode proteins with an ADP-ribosyltransferase (ART) catalytic domain. These proteins transfer ADPr from NAD+ onto targeted amino acid residues of proteins. To generate the ADPr chains, ART’s release Nam from NAD+, and then form an a(1e2)O-glycosidic bond between two ADPr molecules. This post-translational modification is named either monoor poly-ADP-ribosylation (MARylation or PARylation), depending if the ART’s are transferring a single ADPr or are generating a chain (Gibson and Kraus, 2012; Hottiger et al., 2010). In vitro, these chains of APD-ribose can be around 200 residues, with the length and branching being dependent on the concentration of available NAD+ (Alvarez-Gonzalez and Jacobson, 1987; Alvarez-Gonzalez and Mendoza-Alvarez, 1995). This process of PARylation is reversible and is constantly being regulated by poly (ADP-ribose) glycohydrolase (PARG) and the mainly mitochondrial, ADP-ribosyl hydrolase-3 (ARH3) (Di Meglio et al., 2003; Mashimo et al., 2013; Niere et al., 2012). PARG and ARH3 associated degradation of PAR can result in releasing intact PAR chains (endoglycohydrolase) or by releasing free ADR-ribose (exoglycohydrolase) (Gibson and Kraus, 2012).
Poly (ADP-ribose) polymerase 1 (PARP1), also known as ADPribosyltransferase diphtheria toxin-like one (ARTD1), is responsible for the majority of PARylation and is involved in the repair of moderate single stranded DNA damage.

Historically, PARP1 has been reported to be exclusively nuclear, although there is growing evidence for PARP1 associated mitochondrial functions. Several studies report PARP1 can modulate mitochondria from the nucleus through: PAR translocation from the nucleus to the mitochondria, depletion of cellular NAD+ pools, and epigenetic regulating of nuclear genes that are involved in mitochondrial DNA transcription and repair (Alano et al., 2010; Fang et al., 2014; Fatokun et al., 2014; Lapucci et al., 2011). Additionally, several studies report that PARP1 can directly interact with mitochondria and that there is an intramitochondrial localized PARP1 (mtPARP1) (Lai et al., 2008; Rossi et al., 2009; Szczesny et al., 2014; Zhang et al., 2016b). Brunyanszki and colleagues, reviewed the growing body of work suggesting there is a mitochondrial PARP1, in which they hypothesize mtPARP1 is activated prior to nuclear PARP1 in the event of oxidative stress (Brunyanszki et al., 2016).

When discussing the possibility of mtPARP1 it is important to note that mitochondrial and cytosolic NAD+ pools are mostly distinct from each other, with studies showing mitochondrial NAD+ pools are maintained for over 24 h after cytoplasmic NAD+ depletion (Alano et al., 2007; Pittelli et al., 2010; Stein and Imai, 2012). Furthermore, the mechanism of mammalian mitochondrial NAD+ regulation is still unclear and it is unknown whether there is a mitochondrial membrane transporter to shuttle NAD+, NMN, or nicotinamide riboside (NR) between the mitochondria and cytosol (Stein and Imai, 2012). Using a novel NAD+ biosensor, Cambronne and colleagues determined that mitochondrial NAD+ pools appear to be regulated differently in various cell types. They found NMNAT3 depletion caused a significant decrease in mitochondrial NAD+ in HEK293T cells but not HeLa cells (Cambronne et al., 2016).

Additionally, Nikiforov and colleagues, reported using stably transfected 293mitoPARP cells that mitochondrial NAD+ pools are regulated by cytosolic NMN being transported to the mitochondria and then being converted to NAD+ by NMNAT3 (Nikiforov et al., 2011). These studies suggest NMNAT3 is essential for mitochondrial NAD+ production in some cells while other cells might use a yet to be discovered mitochondrial NAD+ shuttle. Furthermore, mitochondrial NAD+ pools can vary greatly depending on the cell type: cardiac myocytes (10.02 ± 1.82 nmol/mg protein), neurons (4.66 ± 0.37 nmol/mg protein), and astrocytes (3.20 ± 1.02 nmol/ mg protein). The study also reported cytoplasmic NAD+ pools were most similar to those in neuronal mitochondria. This suggests NAD+ pools are split between the cytoplasm and mitochondria differently depending on the cell type (Alano et al., 2007). Therefore, it is reasonable to suggest that the pathologic outcome of PARP1-dependent depletion of mitochondrial or cytosolic NAD+ could vary greatly depending on the cell type.

For example, Modis and colleagues reported PARP1 depleted mitochondrial NAD+ pools at a much faster rate than cytosolic. They used PARP1 (shPARP1) silencing in A549 cells, which resulted in a four-fold increase in mitochondria NAD+, while showing no change in total cellular NAD+ (Modis et al., 2012).

Several studies have shown substantial amounts of mitochondrial targeted proteins that are subjected to PARylation (Brunyanszki et al., 2016; Gagne et al., 2012). It has also been reported that mtPARP1 interacts with mitochondrial specific DNA base excision repair enzymes. Interestingly, nuclear PARP1 has a positive effect on DNA repair, while mtPARP1 was observed to negatively affect base excision repair enzymes EXOG and DNA polymerase gamma (Polg) (Rossi et al., 2009; Szczesny et al., 2014). In contrast, Rossi and colleagues reported that mtPARP1 is important for normal mitochondrial DNA ligase III function (Rossi et al., 2009). The same study also showed that mitofilin, a mitochondrial inner membrane (IM) protein, plays an important role in regulating the mitochondrial localization of PARP1 (Rossi et al., 2009).

Normally, mitofilin is involved in keeping the cristae membranes connected to the inner boundary membrane and promoting protein import through the mitochondrial intermembrane space (IMS) assembly pathway via being coupled to the outer membrane (OM) (von der Malsburg et al., 2011). Interestingly, mtPARP1 and DNA ligase III co-occupy the D-loop region in mtDNA, and mitofilin depletion results in decreased mtPARP1 and impaired binding of DNA ligase III to mtDNA (Rossi et al., 2009). These studies have led to Brunyanszki and colleagues hypothesizing that mtPARP1 is normally bound to mitofilin but under oxidative stress is released, consuming mitochondrial NAD+, and PARylating mitochondrial proteins, including the electron transport chain complexes (Brunyanszki et al., 2016).

Conversely, parthanatos is cell death induced by nuclear PARP1 overactivation, as a result of extensive nuclear DNA damage (Fatokun et al., 2014). Excessive PARP1-dependent PARylation causes some PAR chains to translocate to the cytoplasm and bind to apoptosis inducing factor (AIF) on the mitochondrial OM (Wang et al., 2011). AIF is predominately located in the mitochondrial IMS and attached to the IM; although it has been reported that about 30% of AIF is loosely associated with the cytosolic side of the OM (Yu et al., 2009). The binding of PAR to AIF helps induce the release of AIF from the mitochondria and it’s translocation to the nucleus (Fatokun et al., 2014; Wang et al., 2011). This translocation of AIF to the nucleus causes chromatin condensation, and large scale DNA fragmentation resulting in cell death (Yu et al., 2002).

PARP1 overactivation is also associated with NAD+ and ATP depletion (Gerace et al., 2014). PARP1 has been reported to inhibit ATP generation by glycolysis via PAR translocating from the nucleus into the cytoplasm and binding to hexokinase 1 (HK1) on the mitochondrial OM (Andrabi et al., 2014; Fouquerel et al., 2014). The binding of PAR to HK1 can allosterically decrease HK1 activity and cause HK1 to migrate to the cytoplasm, resulting in the inhibition of glycolysis and the reduction of mitochondrial ATP production (Fouquerel et al., 2014).

5. Sirtuins/OAADPR
The sirtuins are class III histone deacetylases (HDACs) and consist of seven isoforms SIRT1-SIRT7, with SIRT3SIRT5 being localized in the mitochondria. Sirtuins require NAD+ for their activity and modulate protein signaling and function by removing acetyl groups attached to lysine residues (Jesko and Strosznajder, 2016; Jesko et al., 2016). In the mitochondria, SIRT3 is the major lysine deacetylase; reports show SIRT3 Knockout (KO) mice and cells have hyperacetylation of mitochondrial proteins, while the same effect was not seen in SIRT4 or SIRT5 KO animals (Lombard et al., 2007; Sol et al., 2012). Sirtuins remove acetyl groups by first cleaving Nam from NAD+ (forming ADPr) and then transferring the acetyl group from the substrate to ADPr, resulting in the formation of O-acetyl-ADP-ribose (OAADPr) (Borra et al., 2004). Uniquely, SIRT5 hydrolyzes malonyl and succinyl lysines, forming O-malonyl-ADPr and O-succinyl-ADPr (Du et al., 2011).

Macro domain (macroD) proteins have been reported to hydrolyze OAADPr to ADPr (Chen et al., 2011). Normally these proteins bind to NAD+ metabolites (e.g. PAR), and are involved in a broad range of biological functions, including DNA repair (Han et al., 2011). ARH3, which normally plays a role in degrading PAR to ADPr, was also found to hydrolyze OAADPr to ADPr in a Mg2þ dependent mechanism in cells (Ono et al., 2006). This finding suggests ARH3 could play a role in both OAADPr and ADPr regulation. Interestingly, compared to PARP1, SIRT1 has a lower affinity (higher Km) for NAD+ and is less effective at breaking down NAD+ (lower Km/Kcat). This suggests PARP1 activation could deplete NAD+ pools at a faster rate then SIRT1 and consequently reduce the function of the NAD+ dependent sirtuins (Canto et al., 2013).

6. CD38/CD157
The membrane proteins CD38 and the less abundant CD157 are multifunctional ecto-enzymes that use NAD+ as a substrate to form cyclic ADP-ribose (cADPr), or under acidic conditions use NADPþ as a substrate to form nicotinamide acid ADP (NAADP). Additionally, CD38 and CD157 can act as glycohydrolases and hydrolyze cADPr to ADPr and under acidic conditions convert NAADP to ADPrphosphate (ADPrP) (Malavasi et al., 2008; Quarona et al., 2013). Normally, cADPr, ADPr, ADPrP, and NAADP play roles in cell signaling pathways and regulating cytoplasmic Ca2þ fluxes (Malavasi et al., 2008; Quarona et al., 2013; Sumoza-Toledo and Penner, 2011). CD38 and CD157 are commonly located on the membranes of immune cells but can be found throughout the body; playing a role in the immune response, hormone secretion, cell activation, egg fertilization, and muscle contraction (Malavasi et al., 2008).
In CD38KO mice, brain tissue NAD+ levels have been observed to be up to 10 fold higher than in wild type mice, although the rate of NAD+ consumption by CD38 may vary greatly depending on the brain region (Aksoy et al., 2006; Long et al., 2016). Interestingly, Camacho-Pereia and colleagues showed CD38 levels not only increase with age but also correlate with a decline of NAD+, suggesting CD38 might be the major cause of the decline of NAD+ during normal aging (Camacho-Pereira et al., 2016). Surprisingly, our lab found CD38KO mice to have dramatically higher PAR levels (particularly in neurons) and decreased PARG activity when compared to WT mice. These findings suggest CD38KO mouse studies might be difficult to interpret due to complexities derived from alterations in several enzymes involved in NAD+ catabolism as a result of knocking out the constitutively expressed CD38 gene (Long et al., 2016).

7. PARG/AHR3
Uncontrolled PARP1 activation has been proposed to deplete intracellular NAD+ and ATP (Fouquerel et al., 2014; Szabo and Dawson, 1998). To regulate PARP1 activation, PARG (nuclear/cytoplasmic) and AHR3 (predominantly mitochondrial) degrade PAR, which can have toxic effects (Dumitriu et al., 2004; Mashimo et al., 2013). It has been suggested that the free PAR polymers is the major toxic molecule generated by the combined activity of PARP1 and PARG (Andrabi et al., 2006). Several reports suggested that many types of PAR structures could play a crucial role in stress-dependent signaling processes in vivo (Dawson and Dawson, 2004; Hong et al., 2004). However, most free or protein-associated PAR polymers are rapidly degraded in vivo to ADPr (Jacobson et al., 1983). In the nucleus and cytoplasm, PARG is the predominant PAR glycohydrolase, reportedly acting as an endoglycohydrolase, hydrolyzing off chunks of PAR from the protein, as well as an exoglycohydrolase, releasing ADPr (Wang et al., 2014).
Recently, Niere and colleagues discovered that ARH3 and not PARG is responsible for degrading PAR to ADPr in the mitochondrial matrix (Niere et al., 2012). ARH3 has also been postulated to play a protective role against parthanatos in the cytoplasm and nucleus (Mashimo et al., 2013). In this theory, PARG acts as an endoglycohydrolase, releasing PAR fragments from proteins in the nucleus, allowing translocation of PAR from the nucleus to the cytoplasm. To counteract this, ARH3 hydrolyzes the detached PAR, releasing ADPr, preventing PAR induced AIF translocation to the nucleus and subsequent cell death via parthanatos (Mashimo et al., 2013).

8. NUDIX hydrolase, enzyme controlling cellular and mitochondrial ADP-ribose levels
The intracellular levels of ADPr are tightly controlled by specific ADPr hydrolases, which hydrolyze ADPr to adenosine monophosphate (AMP) and D-ribose 5-phosphate; thereby, limiting free ADPr accumulation (Fernandez et al., 1996; Ribeiro et al., 1995). This family of enzymes catalyze the hydrolysis of a nucleoside
diphosphate linked to another moiety x, hence the acronym “NUDIX” (Mildvan et al., 2005). For NUDIX hydrolases to be active, Mg2þ or a similar cofactor (e.g. Zn2þ) must bind to NUDIX (Zha et al., 2006). Cloning and expression of human cDNA coding for proteins with NUDIX motifs have revealed two human ADPr hydrolases: NUDT5, an ADP-sugar pyrophosphatase (Gasmi et al., 1999), and NUDT9, an ADPr pyrophosphatase (Perraud et al., 2003). While NUDT9 is highly specific for ADPr, NUDT5 only has a preference for ADPr and can also hydrolyze other ADP-sugar conjugates (Zha et al., 2008).
The human NUDT9 gene gives rise to two alternatively spliced mRNAs, NUDT9a and NUDT9b (Li et al., 2002). NUDT9a possesses a putative mitochondrial targeting sequence and accumulates in mitochondria, while NUDT9b is a cytosolic enzyme (Lin et al., 2002; Perraud et al., 2003). Thus, NUDT9 can play an important role in cellular bioenergetic metabolism, particularly under stressed conditions that can lead to high levels of mitochondrial ADPr.

This is important since free ADPr has been reported to be a competitive inhibitor of NADH oxidation at complex I of the electron transport chain (Zharova and Vinogradov, 1997). Similarly, in rat ventricular myocytes ADPr was reported to inhibit ATP-sensitive Kþ channels (Kwak et al., 1996). Additionally, excessive cytosolic ADPr, cADPr, NAADP, OAADPr, and ADPrP are potential agonists along with ROS, and Ca2þ of the transient receptor potential melastatin 2 (TRPM2) ion channel by binding to the channels NUDT9-H domain. Opening of the TRPM2 ion channel is important for Ca2þ signaling; although, overactivation can cause an influx of Ca2þ, leading to inflammation through chemokine recruitment, and ROS (Grubisha et al., 2006; Shimizu et al., 2013; Toth et al., 2015; Yamamoto and Shimizu, 2016). Interestingly, TRPM2-mediated neuronal death in ischemic brain injury is dimorphic, with TRPM2 knockdown only protecting male brains (Shimizu et al., 2013).

NUDIX hydrolases have also been shown as potential metabolizing enzymes of the byproduct of sirtuin activity, OAADPr, producing AMP and O-acetylated ribose 5’-phosphate (Tong and Denu, 2010). This OAADPr hydrolysis was shown using the NUDIX hydrolases YSA1 (yeast), NUDT5 (mouse), and NUDT9 (human). The same study reported YSA1 and NUDT5 to have a similar affinity to hydrolyzing OAADPr or ADPr, while NUDT9 was 500 fold more efficient hydrolyzing ADPr compared to OAADPr (Rafty et al., 2002).

9. Mitochondrial NAD+ catabolism
The presence of NUDT9a in mitochondria suggests that enzymes generating ADPr are also localized in the intra-mitochondrial compartment. As discussed earlier, several studies have reported that PARP1 is localized not only in the nucleus, but also in the mitochondria (Du et al., 2003; Lai et al., 2008; Pankotai et al., 2009; Rossi et al., 2009). Furthermore, ARH3 has been reported to control mitochondrial PAR hydrolysis (Niere et al., 2012). Concerted activity of mitochondrial PARP1 and ARH3 results in increased levels of matrix ADPr. Intra-mitochondrial ADPr is then metabolized by the NUDT9a, generating AMP and D-ribose 5-phosphate (Perraud et al., 2003). This NUDT9 dependent intra-mitochondrial AMP generation from the hydrolysis of ADPr has been hypothesized to trigger an AMP-dependent mitochondrial failure due to inhibition of the adenine nucleotide translocase (ANT) (Formentini et al., 2009). The primary role for ANT is to exchange ADP/ATP across the mitochondrial IM, thus releasing ATP into the cytosol and importing ADP into the matrix to be phosphorylated via oxidative phosphorylation (Liu and Chen, 2013). In the cell, AMP is generally converted to ADP by adenylate kinases (AKs) (Panayiotou et al., 2014).

10. Adenylate kinases
Adenylate kinases (AKs) are found throughout the body and are involved in homeostasis of cellular adenine nucleotides by transferring a phosphate from a donor, usually ATP, to AMP, resulting in ADP. There are 9 known AKs (AK1-AK9), which differ by subcellular localization, tissue distribution, phosphate donor, and substrate. While AK2, AK3, and AK4 are all mitochondrial, only AK3 and AK4 are found in the brain (Panayiotou et al., 2014). Noma and colleagues reported AK3 was specific to the mitochondrial matrix, found in most tissues, and specifically can only use GTP or ITP as a phosphate donor; making AK3 a possible major consumer of mitochondrial GTP. AK4, is also mitochondrial but is found predominately in the kidney and has low enzymatic activity, with a slight preference to GTP over ATP as a phosphate donor (Noma et al., 2001; Panayiotou et al., 2010).

11. Mitochondrial dynamics and downstream NAD+ metabolites
Recently, it was recognized that an imbalance in mitochondrial fusion and fission could result in deterioration of mitochondrial bioenergetics (Knott et al., 2008; Stetler et al., 2013). Both fusion and fission have physiologic functions (Knott and Bossy-Wetzel, 2008). For example, mitochondrial fission can have a protective role by segregation of damaged and inactive mitochondria and through facilitating autophagic clearance (Barsoum et al., 2006). When mitochondria in cells are stressed, they undergo fission, a process that is reversible since small fragmented organelles can fuse and regain the normal pre-insult morphology (Barsoum et al., 2006; Knott and Bossy-Wetzel, 2008; Owens et al., 2015). The fission of mitochondria is frequently observed in neurons exposed to mitochondrial toxins or excitotoxic levels of glutamate (Barsoum et al., 2006).

Reduction of mitochondrial length compared to control tissue was observed in the brain following focal ischemia, suggesting post-insult mitochondrial fragmentation (Barsoum et al., 2006). We detected extensive mitochondrial fission following global ischemic insult in both hippocampal neurons and astrocytes (Owens et al., 2015). Fission and fusion are regulated by dynamin family proteins that act as GTPases. Fusion is controlled by mitofusin-1 and -2 (MFN1 and MFN2), which are localized in the mitochondrial OM, and by the mitochondrial IM optic atrophy protein (OPA1) (Hoppins et al., 2007; Song et al., 2009). Since these enzymes are GTP binding proteins the efficiency of fission and fusion depends on the cytosolic and mitochondrial GTP levels. Therefore, Intra-mitochondrial GTP depletion could inhibit OPA1, while cytosolic GTP depletion could inhibit the MFNs and Drp1. Interestingly IM fusion requires high GTP levels while OM fusion has less demanding GTP requirements (Escobar-Henriques and Anton, 2013). Our previous studies suggests that there is dramatic fragmentation of mitochondria following ischemic insult and only mitochondria in ischemia resistant cells re-fuse at a later recovery time (Owens et al., 2015). Furthermore, genotoxic stress-induced mitochondrial NAD+ catabolism leads to significant reduction of GTP pools in mitochondrial matrix, suggesting a causal link between downstream NAD+ degradation products and GTP metabolism (unpublished data). Similarly, Dagher reported using cell culture that cellular GTP levels are severely depleted during chemical anoxia (Dagher, 2000).

12. Intra-mitochondrial GTP metabolism
GTP molecules in mitochondria are generated by nucleoside diphosphate kinase (NDPK) or succinyl-CoA synthetase (ligase)
(SCS) in the citric acid cycle (Sanadi et al., 1954) (Fig. 1). NDPK converts GDP to GTP by using ATP as a phosphate donor. SCS generates succinate and CoA from succinyl-CoA, and GDP, as a cofactor, is phosphorylated to GTP in the presence of inorganic phosphate. GTP is involved mainly in energy transfer within the cell. In mitochondria, GTP with ATP is utilized during protein translocation into the mitochondrial matrix (Sepuri et al., 1998). As mentioned above, AK3 is located in the mitochondrial matrix and uniquely uses GTP instead of ATP as a phosphate donor to phosphorylate AMP (Noma et al., 2001). Downstream NAD+ catabolism leads to generation of intra-mitochondrial AMP due to ADPr hydrolysis by NUDT9a. AMP is then converted to ADP by transferring high-energy phosphate from intra-mitochondrial GTP by AK3. Concurrently, low levels of intra-mitochondrial NAD+ also compromise TCA cycle enzyme activities, including a-ketoglutarate dehydrogenase that produces succinyl-CoA. Reduced succinyl-CoA compromises GTP production by SCS, which further depletes mitochondrial GTP and can inhibit mitochondrial fusion (Fig. 1).

13. AMPK/MFF/Drp1 mitochondrial fission
AMP-activated protein kinase (AMPK) is a serine/threonine kinase that is activated by the depletion of energy levels and is important in cellular metabolism by phosphorylating numerous cellular targets. AMPK is predominately located in the cytosol and is heterotrimeric, consisting of three subunits, abg; a is catalytic, b is scaffolding, and g is regulatory. AMPK is activated by phosphorylation of Thr172 in the a subunit by liver kinase B1 (LKB1) or calmodulin-dependent protein kinase kinase-b (CaMKKb) (Kim et al., 2016; Li et al., 2015). Cellular AMP: ATP ratios are reported to be the main regulator of AMPK, either with inhibition of mitochondrial ATP production or high levels of AMP acting as activators (Hardie et al., 2012).

Ekholm and company reported that after 15 min of ischemia AMP levels increase 25 fold, while concurrently ATP levels decrease 25 fold (Ekholm et al., 1993). AMP is able to allosterically activate AMPK by binding to the g subunit, which stimulates LKB1 phosphorylation of Thr172 on the a subunit (Fig. 2). Conversely, ATP reduces AMPK activation by competing with AMP for g subunit binding (Gowans et al., 2013). Furthermore, AMPK can also be activated in a Ca2þ dependent pathway by CaMKKb phosphorylation of Thr172 on the a subunit (Woods et al., 2005).

Interestingly, AMPK activation has recently been directly linked to mitochondrial fission (Fig. 1). Toyama and colleagues reported activated AMPK is able to phosphorylate mitochondrial fission factor (MFF) at Ser155 and Ser172 on the mitochondrial OM. MFF is a dominant receptor for the mitochondrial fission protein DRP1 (Toyama et al., 2016). The phosphorylation of MFF by AMPK has been shown to catalyze mitochondrial fission by causing the translocation of DRP1 from the cytoplasm to the mitochondrial OM. Drp1 wraps around constriction sites on the mitochondrial OM, causing fission, and can eventually lead to mitophagy (Toyama et al., 2016; Wang and Youle, 2016).

14. Conclusion
Pathophysiology of neurodegenerative diseases and acute brain injury encompass DNA damage with subsequent increases in nuclear and mitochondrial PARP1 activation and the depletion of cellular and mitochondrial NAD+ pools. Mitochondrial NAD+ depletion can cause inhibition of oxidative phosphorylation and TCA cycle enzymes activity. PAR is degraded in the nucleus and cytoplasm by PARG/AHR3 and by ARH3 in the mitochondria. Consequently, high PARylation will significantly increase in ADPr levels, which could inhibit complex I in the mitochondria or bind the NUDT9-H domain on TRPM2 receptors, causing influxes of Ca2þ
A. Long et al. / Neurochemistry International xxx (2017) 1e9 5

6 A. Long et al. / Neurochemistry International xxx (2017) 1e9
Fig. 1. Schematic diagram linking mitochondrial NAD+ catabolism with the depletion of mitochondrial GTP and mitochondrial fragmentation in a neurodegenerative disease/acute brain injury model. Localization of enzymatic reactions are color-coded: Blue (mitochondrial), red (cytoplasmic), green (nuclear and mitochondrial), purple (insult or outcome). Red arrows show changes in enzymatic activity or substrate levels as a result of the disease/injury insult. Mitochondrial and nuclear poly (ADP-ribose) polymerase 1 (PARP1) leads to generation of Poly (ADP-ribose) (PAR) as a result of disease or insult. ADP-ribose (ADPr) is generated by hydrolyzing PAR by NAD+ glycohydrolases; ADP-ribosyl hydrolase-3 (ARH3) and poly (ADP-ribose) glycohydrolase (PARG), and then is further degraded to AMP by NUDIX hydrolases NUDT9a and NUDT9b.

Accumulated AMP is then phosphorylated by adenylate kinse 3 (AK3) to ADP by transferring high-energy phosphate from GTP. Depletion of NAD+ will inhibit the TCA cycle enzyme succinyl-CoA synthetase (SCS)-linked GTP generation. GTP can also be generated by nucleoside diphosphate kinase (NDPK), which phosphorylates GDP by using ATP as phosphate donor. However, since the TCA cycle and mitochondrial respiration is inhibited due to low NAD+ levels, the NDPK activity will be limited.

This depletion of mitochondrial GTP can inhibit fusion by reducing the functionality of GTP dependent mitochondrial fusion enzymes, such as optic atrophy protein (OPA1). Elevated cytosolic AMP can also shift the cellular AMP: ATP ratio, which leads to liver kinase B1 (LKB1) phosphorylation and activation of AMP-activated protein kinase (AMPK). Once activated, AMPK can phosphorylate mitochondria fission factor (MFF), which causes an increase in cytosolic Drp1 migration to the mitochondrial OM and subsequent fission. Thus, depletion of NAD+, reduction of mitochondrial GTP, and elevated AMP can lead to excessively fragmented mitochondria.

Fig. 2. Simplified schematic diagram of the overall mechanism linking NAD+ catabolism to mitochondrial fragmentation. Insult caused from disease or acute injury causes NAD+ depletion and PARP1 activation. This leads to high levels of ADPr and AMP in the cell and mitochondria. Mitochondrial AMP phosphorylation and low NAD+ levels cause depletion in mitochondrial GTP and inhibition of GTP dependent mitochondrial fusion proteins. High levels of AMP also increase AMP: ATP ratios, which leads to the activation of AMPK and subsequent mitochondrial fission.

Additionally, mitochondrial NUDT9a and cytosolic NUDT9b/NUDT5 regulate ADPr levels by hydrolysis, producing AMP and D-ribose-5-phosphatate.
We propose a possible mechanism for elevated mitochondrial AMP levels to cause a shift to mitochondrial dynamics to fission (Fig. 2). First, high AMP levels have been reported to inhibit ATP production by inhibiting ANT. Decreasing ATP production along with the already high levels of AMP will result in a spike in the AMP: ATP ratio, thus activating AMPK. AMPK has been shown to regulate mitochondrial fission by phosphorylating MFF, resulting in the translocation of Drp1 to the mitochondria and Drp1 induced mitochondrial fission. Additionally, high AMP levels will increase AK3 activity, which consumes mitochondrial GTP. Depletion of mitochondrial GTP levels can disrupt GTPases, such as the mitochondrial fusion proteins.

Thus, in stress conditions, NAD+ depletion can significantly deplete mitochondrial GTP levels by inhibiting the TCA cycle and through high AK3 activity, resulting in excessively fragmented mitochondria (Fig. 1).
The link between NAD+ catabolism and mitochondrial fragmentation is an understudied area with translational potential. Understanding mitochondrial PARP1 activity and finding ways to regulate cellular ADPr and AMP levels in disease/acute injury states through studying the NUDIX hydrolases could be a target to reduce mitochondrial fragmentation. Additionally, better understanding of intra-mitochondrial GTP metabolism and the complexities of AMPK in relation to mitochondrial dynamics could be targets for future studies.