NAD Kinase Levels Control the NADPH Concentration in Human Cells*

NAD kinases (NADKs) are vital, as they generate the cellular NADP pool. As opposed to three compartment-specific isoforms in plants and yeast, only a single NADK has been identified in mammals whose cytoplasmic localization we established by immunocytochemistry. To understand the physiological roles of the human enzyme, we generated and analyzed cell lines stably deficient in or overexpressing NADK. Short hairpin RNA-mediated down-regulation led to similar (about 70%) decrease of both NADK expression, activity, and the NADPH concentration and was accompanied by increased sensitivity toward H2O2. Overexpression of NADK resulted in a 4–5-fold increase in the NADPH, but not NADP+, concentration, although the recombinant enzyme phosphorylated preferentially NAD+. Surprisingly, NADK overexpression and the ensuing increase of the NADPH level only moderately enhanced protection against oxidant treatment. Apparently, to maintain the NADPH level for the regeneration of oxidative defense systems human cells depend primarily on NADP-dependent dehydrogenases (which re-reduce NADP+), rather than on a net increase of NADP. The stable shifts of the NADPH level in the generated cell lines were also accompanied by alterations in the expression of peroxiredoxin 5 and Nrf2. Because the basal oxygen radical level in the cell lines was only slightly changed, the redox state of NADP may be a major transmitter of oxidative stress.

alase) as well as the NADPH oxidase, which catalyzes the "oxidative burst" as part of the immune response, depend on NADPH. Interestingly, the redox properties of the NAD ϩ / NADH and NADP ϩ /NADPH couples are similar, but their functions are largely divergent. Apparently, a major reason for this separation is the possibility to maintain one pool, namely NADP, in its reduced form to assure an immediate regeneration of the defense systems following oxidative assaults. In this role, NADPH is of vital importance, because survival following oxidative stress, which accompanies a multitude of pathological states such as inflammatory processes or ischemia/reperfusion injury, depends primarily on the capacity of the defense systems. Nevertheless, surprisingly little is known regarding the generation and maintenance of the cellular NADP pool, especially in mammalian cells.
The only reaction known to yield NADP is the phosphorylation of NAD by NAD kinase (NADK). 2 Accordingly, NAD kinase has been found to be an essential activity for cell survival (4 -9). In yeast and plants, three NADK isoforms have been identified and characterized (10). Both organisms possess at least one cytosolic isoform (4,(11)(12)(13). In yeast, a mitochondrial NADK appears to be vital, and its deletion caused a drastically higher sensitivity toward oxidative stress (14,15), whereas deletion of a chloroplast-specific isoform in Arabidopsis dramatically affected chlorophyll synthesis and photosynthesis (16,17). It would appear, therefore, that the presence of NADK within the cytosol and certain subcellular organelles is physiologically important. However, unlike in yeast and plants, only a single gene encoding NADK has been found in mammals (18). As the only animal NADK so far, the molecular identity of the human enzyme was established, and its activity kinetically characterized (18). Its counterpart in mouse has so far only been referred to as unpublished work, which indicated that its deletion is lethal (9).
The subcellular location of NADK in animals has not been determined, although earlier work suggested these enzymes to be cytosolic (19). In any case, the existence of only a single NADK isoform in humans would raise the problem as to how at least two separate pools, the nuclear/cytosolic and the mitochondrial, might be established. That is, both pools contain a substantial amount of the cellular NADP, but there is no evidence for a physiological exchange of NADP across the inner mitochondrial membrane, except in plants (20). * This study was supported in part by Deutsche Forschungsgemeinschaft, Several NAD kinases from lower organisms exhibit specificity toward either NAD ϩ or NADH as substrate. The direct conversion of the reduced form to NADPH might be advantageous, as no further reduction of NADP ϩ by specific dehydrogenases would be required. On the other hand, the necessity of a specific reduction step might be an important prerequisite to maintain opposite redox ratios for NAD ϩ /NADH and NADP ϩ /NADPH. So far it is unknown whether the human enzyme is selective for the reduced or oxidized form of NAD. It was observed, however, that this enzyme is highly specific for ATP as phosphoryl donor and does also not tolerate modifications of the pyridine moiety (18). Although information on the human form has remained scarce, the alteration of this possibly single NADK activity would be expected to greatly influence the cellular level of NADP and thereby NADP-dependent pathways including anti-oxidative defense systems.
To understand the functional significance of the known human NADK we have generated cell lines stably overexpressing human NADK or small hairpin RNA targeting the NADK mRNA. Overexpression of NADK resulted in an almost 200fold increase of both the mRNA level, the protein amount and the catalytic activity. Knock-down resulted in an about 3-fold lower mRNA level, which was accompanied by a similar decrease of the cellular NADK activity. Human NADK was found to reside in the cytoplasm and preferentially accepted NAD ϩ as substrate. However, the modulation of NADK expression was primarily reflected in the alteration of the NADPH, and not NADP ϩ , levels. Our results are consistent with an essential role of the cytoplasmic protein in maintaining the cellular level of NADPH.

Construction of Expression
Vectors-For eukaryotic expression of N-terminally FLAG-tagged human NADK the cDNA was amplified by PCR using the primer 5Ј-GCGGGAAT-TCAATGGAAATGGAACAAGAA and 5Ј-CCGCGGTAC-CGCCCTCCTCCTCCTC and subsequently ligated into pFLAG-CMV-4 (Sigma) using EcoRI and KpnI sites. An additional sequence encoding a C-terminal 6ϫ histidine tag was subsequently introduced via KpnI and BamHI sites using rehybridized synthetic oligonucleotides. For eukaryotic expression of C-terminally FLAG-tagged human NADK the cDNA was amplified using the primer pair 5Ј-GCGGGAATTCCACCAT-GGAAATGGAACAAGAAAAAATG and 5Ј-GGCGGGATC-CGCCCTCCTCCTCCTCCTCC following ligation into pFLAG-CMV-5a (Sigma) via EcoRI and BamHI sites.
The generation of shRNA-encoding vectors was performed according to Ref. 21. In brief, DNA sequences encoding sh-RNAs against human NADK (targeted against positions 313-335 and 334 -356 in mRNA) or eGFP (targeted against positions 122-143 in mRNA) (22) were ligated into pRev-H1-RNA vector via BamHI and HindIII sites using rehybridized synthetic oligonucleotides. Subsequently, the shRNA expression cassettes were subcloned into the pCMV/Myc/Cyto (Invitrogen) via its KpnI and XhoI sites (see Fig. 3A).
Generation of Stably Transfected Cell Lines-HEK293 cells cultivated in Dulbecco's modification of Eagle's minimal essential medium (DMEM) supplemented with 10% fetal calf serum, penicillin (50 units/ml), streptomycin (50 g/ml), and 2 mM L-glutamine were transfected with the plasmid vectors described above by the calcium phosphate method. Colonies were isolated after 2 weeks in media containing 0.55 mg/ml G-418 (PAA Laboratories) followed by another round of subclonal selection. Stably transfected, monoclonal HEK293 lines were maintained in medium supplemented with 0.1 mg/ml G-418.
Treatment of Cells with Oxidants-All experiments were performed on actively growing cells. The medium was removed, and cells were incubated in phosphate-buffered saline supplemented with 1 mM CaCl 2 , 0.5 mM MgCl 2 , and 1 g/liter D-glucose containing 1 mM H 2 O 2 (Merck) or 100 M menadione sodium bisulfite (Sigma) for 30 min at 37°C. Thereafter, fresh medium was added, and the cells were incubated for the indicated period of time.
Measurement of Cell Viability-24 h before treatment, HEK293 cells were seeded at equal density on poly-L-lysinecoated 96-well cell culture plates. Cells were treated with H 2 O 2 (1 mM) or menadione (100 M) in PBS supplemented with 1 mM CaCl 2 , 0.5 mM MgCl 2 , and 1 g/liter D-glucose for 30 min. After 21 h, the culture medium was replaced by medium containing 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyltetrazolium bromide (MTT, 0.3 mg/ml), and incubation continued for 3 h. The MTT solution was aspirated and the generated formazan dissolved in 100 l Me 2 SO. The absorbance was determined using a microplate reader (FLUOstar OPTIMA, BMG-Labtech) at 600 nm. Cell viability is expressed as percentage of the absorbance obtained using untreated control cells.
Determination of Intracellular Reactive Oxygen Species Levels-Intracellular accumulation of ROS was determined using 5-(and-6)-chloromethyl-2Ј,7Ј-dichlorodihydrofluorescein diacetate, acetyl ester (CM-H 2 DCFDA, Invitrogen). This nonfluorescent dye is permeable to cells where it is hydrolyzed to DCFH and oxidized by ROS to the fluorescent DCF. The cells were seeded at equal density on poly-L-lysine-coated 96-well cell culture plates and grown overnight. Cells were then loaded with 10 M CM-H 2 DCFDA in PBS containing 1 g/liter D-glucose, 1 mM CaCl 2 , and 0.5 mM MgCl 2 for 30 min at 37°C. After washing the cells with PBS, they were treated with 1 mM H 2 O 2 or 100 M menadione in PBS supplemented with 1 g/liter D-glucose, 1 mM CaCl 2 , and 0.5 mM MgCl 2 for 30 min. The solution was replaced by PBS, and the DCF fluorescence measured in a fluorescence microplate reader (FLUOstar OPTIMA) at an excitation wavelength of 495 nm and an emission wavelength of 520 nm. Fluorescence was normalized to the protein content, and data are represented relative to untreated control cells.
Preparation of Cell Lysates and Western Blotting-Cells were scraped in ice-cold lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1 mM dithiothreitol, 1 mM PefablocSC, Roth), incubated 5 min on ice and pulled through a 23-gauge needle. After centrifugation for 5 min at 13,200 rpm and 4°C, the supernatant was subjected to separation in 10% SDS-polyacrylamide gels. Following electrophoretic transfer onto nitrocellulose membranes (Schleicher & Schuell) the blots were immunostained using monoclonal antibodies directed either against human NADK (Abnova Corporation), the FLAG epitope (Sigma), or ␤-tubulin (Sigma), and the appropriate horseradish peroxidase-conjugated secondary antibody (Pierce). The rabbit polyclonal antibody directed against PRDX5 was a kind gift from Dr. N. V. Tomilin (St. Petersburg). The blots were developed using enhanced chemiluminescence (ECL, Pierce).
Indirect Immunofluorescence Microscopy-HEK293 and HeLa cells grown on coverslips were washed with PBS, fixed with 4% formaldehyde in PBS, permeabilized with 0.2% Triton X-100 in PBS, and blocked using EMEM containing 10% fetal calf serum. The primary antibodies were anti-FLAG (Sigma) or anti-NADK (Abnova Corporation) and secondary antibodies were AlexaFluor 488-or 594-conjugated IgGs (Invitrogen). Images were acquired using an inverted Leica DMI 6000B microscope.
Measurement of NAD Kinase Activity-NAD kinase activity in cell lysates was assayed by a two-step procedure described previously (14) with minor modifications. First, 30 g of cell lysate was added to a reaction mixture containing 50 mM Tris-HCl, pH 7.8, 10 mM MgCl 2 , 5 mM NAD ϩ , 10 mM ATP in a final volume of 100 l following incubation for 5 min at 30°C. The amount of NADP ϩ produced was then determined, using a cycling assay, by transferring the mixture into 900 l of 50 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 5 mM glucose-6-phosphate, 2 units of NADP-specific yeast glucose-6-phosphate dehydrogenase, 0.5 mM MTT, and 1.5 mM phenazine methosulfate (PMS). Reduction of MTT was monitored at 600 nm for 5 min using a BioMate3 Spectrophotometer (Thermo Electron Corporation), and the amount of NADP quantified by comparison to known concentrations of NADP ϩ standards. NADK activity was calculated by subtracting the amount of NADP produced without added substrates, and one unit was defined as the amount of enzyme producing 1 mol of NADP in 1 min at 30°C. For measurements of NAD ϩ or NADH phosphorylation by the bacterially expressed recombinant NADK, the enzyme was isolated as described previously (18) and nucleotide conversion analyzed by HPLC (see below).
Quantitative RT-PCR-Total RNA was isolated from the indicated cell lines using the RNeasy Mini kit (Qiagen) according to the manufacturer's protocol and used for cDNA synthesis by reverse transcriptase (Mu-MLV, Fermentas) according to the manufacturer's instructions. Realtime PCR probes and primers specific for the indicated cDNAs were selected using the Universal Probe Library (Roche Applied Science). Aliquots of cDNA were amplified in a LightCycler 480 Real-Time PCR System (Roche Applied Science) using LightCycler 480 Probes Master (Roche Applied Science) in triplicate reactions. PCR cycling conditions were: 95°C for 5 min and 40 cycles of 95°C for 10 s, 60°C for 30 s and 72°C for 1 s. Transcript levels were calculated relative to ␤-actin mRNA levels as endogenous control. Relative expression was calculated (23). Analyses of Nucleotides and Redox Status-Nucleotides were extracted by adding 0.5 M KOH to dishes of cells, which were then scraped off and pulled through a 23-gauge needle. After incubation for 3 min on ice, the extracts were neutralized by adding phosphoric acid. After centrifugation, the supernatants were filtered through a 5,000 Da cut-off membrane using microcentrifugation devices (Millipore). The nucleotides in the filtrates were then immediately separated by reverse-phase ionpairing HPLC and quantified by peak integration based on analyses of standard nucleotides. Besides retention times, nucleotide peaks from cell extracts were verified by their UV spectra (200 -400 nm). The HPLC system consisted of a LC-20AB solvent delivery module, an SPD-M20A photodiode array detector and a SIL-20AC autosampler (Shimadzu). Nucleotides were separated in a 125 mm ϫ 2 mm (ID) Nucleodur C18 gravity, 3-m pore size, column (Macherey & Nagel) at a flow rate of 0.3 ml/min. The column was equilibrated with buffer A consisting of 10 mM potassium phosphate, pH 7.0, and the ion-pairing reagent tetrabutyl ammonium bromide (TBA, 2 mM). Following injection of the sample, the column was washed with buffer A for 3 min. Thereafter, nucleotides were eluted in a 5-min gradient to 25% buffer B (10 mM potassium phosphate, pH 5.5, 2 mM TBA, and 50% acetonitrile), followed by a 15-min gradient to 40% buffer B. The column was then rinsed by increasing the concentration of buffer B to 95% for 5 min followed by re-equilibration with buffer A for 5 min before the next injection.

Human NADK Is a Cytoplasmic, NADP ϩ -generating Enzyme-
To establish the subcellular location of human NADK, we first analyzed the primary structure for the existence of potential amino acid motifs that would indicate a targeting sequence. The PSORT II algorithm (24) predicts NADK to be cytoplasmic. The experiments presented in Fig. 1 are consistent with this prediction: independent of whether a FLAG tag was attached to the N terminus (FLAG-NADK) or C terminus (NADK-FLAG), the protein was detected within the cytoplasm, but not mitochondria, as visualized by the concomitant staining of HeLa and HEK293 (not shown) cells with mitotracker. During the original isolation of NADK from mammalian tissues it was noted that the enzyme activity is rather low because of low protein expression rather than a low specific activity (18). Indeed, immunostaining of untransfected cells using a commercial monoclonal NADK antibody did not reveal any specific staining (not shown). The recent identification and characterization of NAD kinases from various species has established that several of these enzymes can use both NAD ϩ and NADH as substrate, whereas some appear to be selective for NAD ϩ (reviewed in Refs. 10,25). As shown in Fig. 2 (A and B), the human enzyme readily accepted NAD ϩ as substrate, whereas a conversion of NADH to NADPH was rather slow at a substrate concentration of 5 mM. Note that the incubation time of NADK with NAD ϩ ( Fig. 2A) was only 10 min compared with 60 min when NADH was used as substrate (Fig. 2B). However, at lower concentrations (50 -150 M) and particularly when NAD ϩ was concomitantly present as substrate, NADH became far more efficiently phosphorylated (Fig. 2C). Nevertheless, NAD ϩ was clearly the preferred substrate (Fig. 2C): in the absence of the other substrate, at 100 M, NAD ϩ was 10-fold faster phosphorylated than NADH. These observations indicate that NADH may become inhibitory at very high concentrations (mM) and that its phosphorylation is stimulated in the presence of NAD ϩ .
According to a previous report (26), the "relaxed" selectivity of NAD kinases is associated with the presence of a glycine or a polar amino acid in the position corresponding to residue 312 (glutamine) in the human enzyme (Fig. 2D). It was shown, for example, for the NAD ϩ -specific NADK from Escherichia coli that a substitution of the corresponding arginine 175 residue by glycine resulted in the capability of the enzyme to phosphorylate NADH (26). Indeed, there is a polar residue (glutamine) in the human enzyme (Fig. 2D) consistent with its capability to use NADH as substrate.
Modulation of NAD ϩ Kinase Activity in HEK293 Cells-We constructed HEK293 cells stably overexpressing human NADK, which we termed NADK(ϩ) cells. Moreover, to decrease the NADK activity, HEK293 cells stably expressing shRNAs targeting the NADK mRNA were also generated. Two different shRNAs targeting two different parts of the NADK mRNA were chosen (shRNA1-NADK and shRNA2-NADK, see Fig. 3A). Control cells were transfected with a vector mediating the expression of an unrelated shRNA, which was directed against the mRNA of enhanced green fluorescent protein (eGFP). The efficiency of the selected shRNAs was analyzed by co-transfection of a pFLAG-NADK plasmid and the respective shRNA-NADK construct followed by immunoblot analysis using an anti-FLAG antibody. As shown in Fig. 3B, co-expression of the shRNAs reduced FLAG-NADK expression indicating their suitability for down-regulation of endogenous NADK. The shRNA1-NADK exhibited a notably higher efficiency compared with the shRNA2-NADK construct.
Stable transformants for the recombinant NADK and shRNA constructs were generated as described under "Materials and Methods" and characterized for their NADK content with regard to catalytic activity, mRNA and protein levels. As shown in Table 1, both shRNA-NADK-expressing cell lines exhibited substantially less NADK activity compared with control cells, whereas cells overexpressing NADK showed a dramatic increase in the specific activity, and this was similar for all selected clones initially analyzed (not shown).
Quantitative RT-PCR analyses revealed that the respective decrease of or increase in NADK activity was closely paralleled by the NADK mRNA transcript levels ( Fig. 3C and Table 1). Furthermore, immunoblot analyses using a monoclonal anti-   NADK antibody showed similar alterations on the protein level with a marked decrease of NADK in the knock-down cells (Fig.  3D). Based upon the more efficient down-regulation compared with shRNA2-NADK, subsequent experiments were conducted using the cells expressing shRNA1-NADK, which were designated NADK(Ϫ) cells. None of the expressed constructs (NADK or shRNA1-NADK) caused any visible phenotypic changes of the cells (Fig. 3E) nor was the proliferation rate affected (not shown).
Human NAD Kinase Activity Controls the Cellular Concentration of NADPH-We next investigated the influence of the NADK expression level on the cellular pyridine nucleotide state. The nucleotides were extracted by alkali treatment.
Under these conditions the reduced nucleotides are stable, while the oxidized ones tend to hydrolyze. The experimental protocol was optimized such that about 75-80% of the oxidized and almost 100% of the reduced nucleotides were recovered (not shown). The filtered cellular extracts were analyzed by reverse-phase ion-pairing HPLC. The nucleotide spectrum (including pyridine nucleotides, ATP and ADP, see Fig. 4A and Table 2) was surprisingly similar in all cell lines studied. However, overexpression of NADK caused a 4 -5-fold increase in the NADPH content, whereas the concentration of this nucleotide was about 3-fold decreased in the NADK(Ϫ) cells (Fig. 4, B and C). These observations were somewhat unexpected, because human NADK would rather phosphorylate NAD ϩ (see above). Because the NADP ϩ concentration did not prominently change, the excessive NADPH generated in the NADK(ϩ) cells must have arisen from the reduction by NADP-dependent dehydrogenases. At least within the cytosol, among the major NADP ϩ -reducing enzymes are glucose-6-phosphate dehydrogenase (G6PD), 6-phosphoglucono-␦-lactonase (6PGL), and 6-phosphogluconate dehydrogenase (6PGD), enzymes of the pentose phosphate pathway. We measured the relative mRNA levels encoding these enzymes. As can be inferred from Fig. 5, there was a slight decrease of expression of all three mRNAs in the NADK(ϩ) cells, whereas in the NADK(Ϫ) cells the expression was slightly elevated. Measurements of the G6PD activity confirmed, that the dramatic differences of NADK activity in the established cell lines had no significant influence on the activity of this dehydrogenase (not shown).
Interestingly, there was no detectable change in the NADK(Ϫ) cells regarding pyridine nucleotide redox ratios or the ATP/ADP ratio ( Table 2) indicating little metabolic effects of the ϳ3-fold reduced NADPH level. Conversely, overexpression of NADK and the concomitant increase of the NADPH concentration were accompanied by a higher reduction state of NADP and an increased NAD ϩ /NADH ratio ( Table 2). The ATP/ADP ratio in the NADK(ϩ) cells was also slightly enhanced.
The Accumulation of Intracellular ROS Reflects NAD ϩ Kinase Expression-Because of the critical role of NADPH in oxidative defense mechanisms it was of interest to establish the consequences of altered NADK expression and the ensuing changes in the NADPH concentration on the amount of intracellular ROS. The relative level of ROS in the HEK293 cell lines    (Fig. 6). When challenged with the oxidizing agents H 2 O 2 or menadione the amount of ROS detected in the NADK(Ϫ) cells exceeded that found in the controls, whereas overexpression of NAD kinase reduced the DCF fluorescence signal significantly (Fig. 6). Increased NAD ϩ Kinase Expression Protects from ROS-induced Cell Death-We then investigated whether the cellular NADK or NADPH level affected the survival of cells following exposure to oxidative stress. The different cell lines were exposed to H 2 O 2 (1 mM) or menadione (100 M) for 30 min. Cell viability was determined 24 h after the treatment. As shown in Fig. 7, NADK overexpression protected the cells and significantly increased the survival of the cells, although to lesser extent than the changes in the NADPH contents might have suggested. On the other hand, HEK293 cells were rather resistant to the menadione treatment. For both challenges (H 2 O 2 and menadione) the survival of NADK(Ϫ) cells was similar to that of control cells. The down-regulation of NADK by shRNA expression resulted in sensitization toward menadione as indicated by a small (10%) decline in survival (Fig. 7).
The NAD ϩ Kinase Expression Level Modulates PRDX5 and Nrf2 Expression-Shifts in the cellular oxidant state influence the expression level of a number of genes. One of the mechanisms governing oxidant-dependent gene expression involves nuclear respiratory factors (Nrfs). These transcription factors activate expression, for example, of genes encoding proteins with anti-oxidative functions (27). It should be noted that most commonly, cells are exposed to acute oxidative stress and tran-  The trace denoted 340 nm shows the specific absorbance of reduced pyridine nucleotides at this wavelength. B, representative nucleotide spectrum of a cell extract obtained from control cells. Absorbance was monitored at 259 nm. The asterisk indicates the NADPH peak, which corresponds to line 1 in panel C. C, NADPH peaks from representative chromatograms of cell extracts generated from the same number of control, NADK(ϩ) or NADK(Ϫ) cells.

TABLE 2 Redox and adenine nucleotide ratios in HEK293 cells modulated in NADK expression
The nucleotide contents of the indicated cell lines were determined by HPLC as scriptional responses are transient and brought about, for example, by phosphorylation events or recruitment of transcription factors from the cytosol to the nucleus. The changes in expression described below are, however, not transient and probably originate from stable shifts in the NADP redox ratio.
We first analyzed the expression of PRDX5, because it is a rather abundant peroxiredoxin, which has been originally localized to the mitochondria and peroxisomes (28,29), but also found in the cytosol (29,30) and, to lesser extent, in the nucleus (31). Therefore, its expression might be influenced by redox changes throughout the cells. As shown in Fig. 8 (A and  B), the PRDX5 expression level varied depending on the expres-sion of NADK. When NADK, and consequently NADPH, was up-regulated, the PRDX5 mRNA content was significantly decreased to about 50%. Down-regulation of NADK with a concomitant reduction of the cellular NADPH concentration led to an increase of the PRDX5 transcript by about 50% (Fig. 8A). The changes of PRDX5 expression were similarly reflected in Western blots using a PRDX5-specific antibody (Fig. 8B).
We also tested whether the NADK level influenced the expression of NF-E2-related factor 2 (Nrf2), a transcription factor that regulates the expression of genes encoding anti-oxidant proteins by interaction with ARE (anti-oxidant response element) present in their promoter regions (32). Whereas overexpression of NADK had no significant influence on the Nrf2 mRNA level, in the NADK(Ϫ) cells Nrf2 expression was significantly enhanced (Fig. 8C).

DISCUSSION
This study demonstrates a key role of the known NADK in animal cells for the maintenance of the cellular NADPH pool. Fig. 9 illustrates that this enzyme is essential for NADP generation, indicates the key functions of this nucleotide and highlights the major observations of this study.
In addition to the lack of any other recognizable NADK homologue within the human genome databases, 3 our experimental observations also indicate that potential other genes coding for NADK activities in human cells would probably make only a minor contribution to the NADP pool. Both up- 3 We found a cDNA sequence (GenBank TM : AF250320) potentially encoding a protein identical to the human NADK described here which contains two additional inserted stretches of 105 and 41 amino acids. However, despite multiple attempts using different human cDNAs including brain (the data base sequence was generated from a hypothalamic cDNA), we were unable to amplify this potential transcript. Moreover, at least for the cell lines used here, Western blot analyses did not indicate the presence of an additional NADK signal corresponding to the predicted size. and down-regulation of NADK expression was directly paralleled by the measured NADK activity as well as the NADP levels. In particular, the NADK(Ϫ) cells exhibited an about 70% reduction of all directly related parameters: the NADK mRNA level, the protein level, the NADK activity and even the NADP concentration. Previously, a partially purified NADK from human neutrophiles was characterized (33). The molecular and kinetic properties of this enzyme were slightly different from the purified recombinant NADK (18). The neutrophil preparation exhibited NADK activity, which was stimulated by calcium and calmodulin, a property also known for partially purified NADK activities or crude extracts of plants and sea urchin eggs (reviewed in Ref. 10). Because none of the three recombinantly expressed NADK isoforms of A. thaliana is activated by calcium/ calmodulin, the effect is likely to be mediated by additional factors present in the partially purified preparations, but not the recombinant enzymes. Therefore, it cannot be excluded that the enzyme described for human neutrophiles corresponds to the NADK investigated in this study.
The analyses of the subcellular location of human NAD kinase support the conclusion that the enzyme resides in the cytoplasm. Overexpression of the cDNA encoding NADK followed by immunocytochemistry led to the detection of the protein in this compartment. The possibility exists that the human NADK gene encodes splice variants, which would be equally affected by the shRNAs, but could be targeted to other subcellular compartments. This notion could be invoked, for example, to account for the generation of the mitochondrial pool of NADP. In previous studies of animal NAD kinases the activity was localized to the cytosolic fraction (19). There is so far no evidence for an exchange of NADP between the cytosol and the mitochondria in mammalian cells, although it has been described for plant cells (20).
On the other hand, cytosolic NADK expression levels and the according cellular NADPH content did have an influence on PRDX5 expression, a protein, which is primarily located within mitochondria, but also in other compartments (28,29,31). The increase in the cellular NADPH concentration brought about by NADK overexpression caused a stable down-regulation of PRDX5 expression, while the decrease of NADPH resulted in an increased expression level of PRDX5. These observations indicate that PRDX5 might serve as a redox buffer, which is adjusted according to the reservoir of the ultimate regeneration source of the oxidative defense systems, NADPH. It will be interesting to establish the mechanism linking NADK, that is, NADPH levels, to PRDX5 expression. There is evidence for a direct, NADH-dependent regulation of transcription. For example, the transcriptional corepressor CtBP mediates the expression of the NAD ϩ -dependent protein deacetylase SIRT1 depending on whether it has NADH bound or not (34). However, no such mechanism has been reported for NADPH. Besides PRDX5, the NADPH-dependent changes of Nrf2 expression also point to a direct relationship between the total cellular NADPH content and activation of anti-oxidative mechanisms. Nrf2 has been established as a transcription factor which, under conditions of oxidative stress, is recruited to the nucleus to specifically enhance expression of ARE-driven genes (32). Transcriptional up-regulation of Nrf2 itself is considered only a secondary event during oxidative stress. However, upon constitutive NADPH depletion by NADK knock-down, with only a minimal increase of ROS levels, Nrf2 expression was found to be significantly enhanced. It will be interesting to explore the mechanism linking NADPH levels and Nrf2 expression and also whether a shift in the NADP redox ratio itself is sufficient to trigger Nrf2 activation including nuclear translocation. In fact, Nrf2 may regulate PRDX5 expression (35), which is in line with our observations, at least in the NADK(Ϫ) cells. In this regard, the availability of the NADK(ϩ) and NADK(Ϫ) cells provides a unique tool to study gene expression which depends solely on the shift of the NADP redox state, in the absence of any side effects incurred when using chemicals or irradiation to induce oxidative stress.
An important conclusion relates to the direct influence of NADK on the NADPH concentration and thereby the maintenance of anti-oxidative defense mechanisms. Overexpression or down-regulation of NADK influenced almost exclusively the content of NADPH in the respective cells, without notably affecting the rather low NADP ϩ concentration or any other nucleotide analyzed (NAD ϩ , NADH, ATP, ADP). While this relationship might have been expected for NAD kinases that prefer NADH as substrate, we found that the human enzyme is rather unlikely to do so under physiological conditions. The enzyme converted NADH to NADPH, but far less efficiently than it phosphorylated NAD ϩ . Moreover, the cellular content of NAD ϩ by far exceeds that of NADH (see Table 2). It appears likely, therefore, that NADP ϩ generated by NADK is immediately reduced to NADPH by specific dehydrogenases. However, the expression of genes encoding the three NADPH-generating enzymes of the pentose phosphate pathway hardly changed both at an almost 200-fold elevated or 3-fold lower expression level of NADK. This observation suggests that the capacity of the dehydrogenases is sufficient to convert the generated NADP ϩ . On the other hand, the increase of NADPH under these conditions was only about 4 -5-fold suggesting that this amount represents an upper limit. It appears possible that, similar to the enzyme from Salmonella enterica (6), human NADK might be specifically inhibited by NADPH preventing further phosphorylation of NAD ϩ or NADH. However, other factors may also reduce NADK activity when the NADPH concentration exceeds physiological levels.
Stable overexpression of NADP ϩ -dependent dehydrogenases resulted in an up to 4.5-fold and about 3-fold increase in activity for mitochondrial isocitrate dehydrogenase (36) and G6PD (37), respectively. Compared with these levels, the almost 200-fold increase of NADK activity in the generated overexpressing cell lines is unusually high and did not seem to have any severe consequences. One would expect that NADK activity should be enhanced under conditions of oxidative stress to strengthen the cellular defense mechanisms. In support of this notion, irradiation or treatment with hydrogen peroxide of plant cells induced up-regulation of the AtNADK-1 mRNA and corresponding protein levels (38). Moreover, yeast cells deleted for ScNADK-3 were highly sensitive toward oxidative stress (14,15). However, no increase of human NADK activity or expression could be detected when untransfected HeLa or HEK293 cells were exposed to oxidizing agents (not shown).
Quite surprisingly, the rather strong overexpression of NADK in HEK293 cells provided only moderate protection, even though the NADPH concentration had been increased 4 -5-fold. A plausible explanation would be that, when exposed to oxidative challenges, the cells rather rely on the existing NADP pool and enhance the capacity of keeping it in a reduced state by increasing the activity of NADP-dependent dehydrogenases. Indeed, NADP-dependent dehydrogenases play a prominent role in the response to oxidative stress. They regenerate NADPH from NADP ϩ , the product of the reductase activities, which provide the reduced pool of GSH, thioredoxins and, in turn, peroxiredoxins, for examples. Several studies have revealed that these dehydrogenases are up-regulated under conditions of oxidative stress in human and other cells (36, 39 -42). Still, the reason to not additionally increase the total NADP concentration is not obvious. Our results demonstrate that an increase of the NADP ϩ synthesizing capacity leads to a considerable increase in NADPH without providing a similar degree of resistance toward oxidative stress. It may well be that the rather moderate extent of protection is caused by the unchanged presence of the NADP-dependent dehydrogenases in the NADK(ϩ) cells. That is, the rate of NADPH regeneration is more critical than the actual concentration of the pre-existing NADPH. Only when significantly diminished, the size of the NADP pool becomes the limiting factor, as indicated by the higher sensitivity of the NADK(Ϫ) cells toward hydrogen peroxide.