Nuclear transport of nicotinamide phosphoribosyltransferase is cell cycle–dependent in mammalian cells, and its inhibition slows cell growth

Nicotinamide phosphoribosyltransferase (NAMPT) is located in both the nucleus and cytoplasm and has multiple biological functions including catalyzing the rate-limiting step in NAD synthesis. Moreover, up-regulated NAMPT expression has been observed in many cancers. However, the determinants and regulation of NAMPT's nuclear transport are not known. Here, we constructed a GFP–NAMPT fusion protein to study NAMPT's subcellular trafficking. We observed that in unsynchronized 3T3-L1 preadipocytes, 25% of cells had higher GFP–NAMPT fluorescence in the cytoplasm, and 62% had higher GFP–NAMPT fluorescence in the nucleus. In HepG2 hepatocytes, 6% of cells had higher GFP–NAMPT fluorescence in the cytoplasm, and 84% had higher GFP–NAMPT fluorescence in the nucleus. In both 3T3-L1 and HepG2 cells, GFP–NAMPT was excluded from the nucleus immediately after mitosis and migrated back into it as the cell cycle progressed. In HepG2 cells, endogenous, untagged NAMPT displayed similar changes with the cell cycle, and in nonmitotic cells, GFP–NAMPT accumulated in the nucleus. Similarly, genotoxic, oxidative, or dicarbonyl stress also caused nuclear NAMPT localization. These interventions also increased poly(ADP-ribosyl) polymerase and sirtuin activity, suggesting an increased cellular demand for NAD. We identified a nuclear localization signal in NAMPT and amino acid substitution in this sequence (424RSKK to ASGA), which did not affect its enzymatic activity, blocked nuclear NAMPT transport, slowed cell growth, and increased histone H3 acetylation. These results suggest that NAMPT is transported into the nucleus where it presumably increases NAD synthesis required for cell proliferation. We conclude that specific inhibition of NAMPT transport into the nucleus might be a potential avenue for managing cancer.

Nicotinamide phosphoribosyltransferase (NAMPT) is located in both the nucleus and cytoplasm and has multiple biological functions including catalyzing the rate-limiting step in NAD synthesis. Moreover, up-regulated NAMPT expression has been observed in many cancers. However, the determinants and regulation of NAMPT's nuclear transport are not known. Here, we constructed a GFP-NAMPT fusion protein to study NAMPT's subcellular trafficking. We observed that in unsynchronized 3T3-L1 preadipocytes, 25% of cells had higher GFP-NAMPT fluorescence in the cytoplasm, and 62% had higher GFP-NAMPT fluorescence in the nucleus. In HepG2 hepatocytes, 6% of cells had higher GFP-NAMPT fluorescence in the cytoplasm, and 84% had higher GFP-NAMPT fluorescence in the nucleus. In both 3T3-L1 and HepG2 cells, GFP-NAMPT was excluded from the nucleus immediately after mitosis and migrated back into it as the cell cycle progressed. In HepG2 cells, endogenous, untagged NAMPT displayed similar changes with the cell cycle, and in nonmitotic cells, GFP-NAMPT accumulated in the nucleus. Similarly, genotoxic, oxidative, or dicarbonyl stress also caused nuclear NAMPT localization. These interventions also increased poly(ADP-ribosyl) polymerase and sirtuin activity, suggesting an increased cellular demand for NAD. We identified a nuclear localization signal in NAMPT and amino acid substitution in this sequence ( 424 RSKK to ASGA), which did not affect its enzymatic activity, blocked nuclear NAMPT transport, slowed cell growth, and increased histone H 3 acetylation. These results suggest that NAMPT is transported into the nucleus where it presumably increases NAD synthesis required for cell proliferation. We conclude that specific inhibition of NAMPT transport into the nucleus might be a potential avenue for managing cancer.
Nicotinamide phosphoribosyltransferase (NAMPT) 2 catalyzes the rate-limiting step in NAD biosynthesis, the reaction between nicotinamide and 5-phosphoribosyl-1-pyrophosphate to form nicotinamide mononucleotide (NMN) (1). NAD is a required cofactor in numerous oxidoreductase reactions (2) and in nonredox processes such as DNA repair and replication, transcription, cell division, and cell death. Two main classes of NAD-consuming enzymes are the poly(ADP-ribosyl) polymerases (PARPs) and sirtuins (SIRTs). Nonredox, NAD-consuming enzymes are found in the nucleus, mitochondria, and cytosol (3). NAMPT has also been identified as a secreted factor with an extracellular function affecting the immune system. Alternate names of the NAMPT are visfatin (4) and pre-B cell colony enhancing factor (5). The fact that NAMPT is both an intracellular enzyme and a secreted hormone is intriguing and would suggest complex trafficking of this protein.
Because NAD is required throughout the cell, NAMPT might be expected to be distributed in multiple cellular compartments. In dividing Swiss 3T3 and PC-12 cells, NAMPT was predominantly cytosolic, but its localization changed to nuclear when the cell division was stopped (6). Translocation of NAMPT from cytosol to nucleus as a result of cell division arrest has also been observed in colorectal carcinoma cells HCT-116 (7). NAMPT was located in the nucleus in prepu-bertal chicken testis cells (8). In nonstimulated human vascular endothelial cells, NAMPT was mainly nuclear; it localized to the cytoplasm after activation by interleukin-1␤ (9). These results suggest that the transport of NAMPT between cytosol and nucleus is a regulated process. However, NAMPT's transport mechanism, relationship to cell cycle, and role in the nucleus are not known.
To characterize NAMPT transport and its regulation, we studied endogenous NAMPT and GFP-NAMPT during the cell cycle, after inhibition of cell proliferation, and under stress conditions. We identified a required nuclear localization signal. Our data suggest that inhibition of NAMPT nuclear transport has potential as a therapeutic target in cancer treatment.

NAMPT is located both in the nucleus and cytoplasm
We studied the cellular localization of NAMPT in three cell types: (i) human hepatoma HepG2 cells, (ii) mouse 3T3-L1 preadipocytes, and (iii) differentiated 3T3-L1 adipocytes. Fusion proteins with either EGFP or HA at the N terminus of NAMPT were constructed. The GFP-NAMPT was transiently transfected, and its localization was analyzed by fluorescence microscopy (Fig. 1A). HA-tagged NAMPT showed similar localization as GFP-NAMPT (Fig. 1B). Fluorescence was detected in both the nucleus and the cytoplasm with a large intercellular variability in nuclear to cytoplasmic proportion. To quantify NAMPT localization, we counted the number of cells with NAMPT located predominantly in the nucleus or cytoplasm according to average fluorescence. In HepG2 hepatocytes, NAMPT was nuclear in 84% and cytoplasmic in 6% of the cells, without clear preferential localization in the rest. In 3T3-L1 preadipocytes, NAMPT was nuclear in 62% of cells and cytoplasmic in 25%. In contrast, in nondividing 3T3-L1 adipocytes, NAMPT localization was nuclear in Ͼ99% of cells. These results demonstrate that NAMPT localizes to both the nucleus and cytoplasm in mul- Figure 1. Localization of NAMPT in HepG2 hepatocytes, 3T3-L1 preadipocytes, and 3T3-L1 adipocytes. A, cells were transfected with plasmid pEGFP-C1-NAMPT, and localization of GFP-NAMPT was monitored for 24 h post-transfection using fluorescence microscopy. B, cells were transfected with plasmid pCMV-HA-NAMPT, and localization of HA-NAMPT was studied by immunostaining using FITC-labeled antibody directed against HA, followed by fluorescence microscopy. Specific nuclear staining was achieved by staining with DAPI. For quantitative evaluation, the proportion of cells containing NAMPT predominantly in nucleus (bar N) or in cytoplasm (bar C) was calculated from a random field of view (0.16 mm 2 ) of five independent cultures (in 3T3-L1 preadipocytes each view field contained 22 Ϯ 7 transfected cells; in 3T3-L1 adipocytes each view field contained 16 Ϯ 4 transfected adipocytes; and in HepG2 hepatocytes each view field contained 20 Ϯ 5 transfected cells). The remaining cells had dual localization (equal cytoplasmic and nuclear) of NAMPT. The data are expressed as means Ϯ S.D. Bar, 10 m.

NAMPT nuclear transport
tiple cell types and suggest that the GFP does not affect NAMPT distribution.

NAMPT localization changes during cell cycle
To further explore NAMPT localization, we produced stably transfected HepG2 cell lines expressing GFP-NAMPT and monitored its localization using live cell imaging microscopy (Movie S1). In cells undergoing mitosis during the initial 24 h, GFP-NAMPT location was scored according to average fluorescence until the next mitosis or up to the end of monitoring, whichever occurred first ( Fig. 2A). Prior to mitosis, NAMPT was always predominantly nuclear, whereas newly formed nuclei were free of NAMPT, and for the subsequent 6.9 Ϯ 1.7 h, the average fluorescence was higher in cytoplasm. This period was followed by 5.3 Ϯ 2.6 h of dual localization, and then NAMPT was observed to be predominantly nuclear for 7.6 Ϯ 3.6 h, until the next mitosis. Live-cell microscopy was also performed in 3T3-L1 preadipocytes stably expressing GFP-NAMPT (Movie S2). Two patterns of NAMPT localization were observed (Fig. 2B). In 60% of cells, localization changes were similar to those in HepG2 (4.9 Ϯ 1.5 h cytoplasmic, 3.3 Ϯ 1.5 h dual, and 8.2 Ϯ 3.2 h nuclear). The other 40% of cells skipped the nuclear localization phase. After mitosis, the NAMPT in these cells was cytoplasmic (6.7 Ϯ 3.2 h) and then dual (8.9 Ϯ 3.5 h) before proceeding into the next mitosis. The intensity of nonmitotic nuclear GFP-NAMPT fluorescence appeared to be stable or increasing in both cell lines.
DNA content was measured by flow cytometry to determine cell cycle phase (Fig. 2C). In the unsynchronized HepG2 cells carrying GFP-NAMPT, 39.5 Ϯ 2.6% were in the G 1 /G 0 phase, 40.0 Ϯ 2.5% were in the S phase, and 20.5 Ϯ 1.3% in the G 2 /M phase. In unsynchronized GFP-NAMPT 3T3-L1 cells, 46.8 Ϯ 1.6% were in the G 1 /G 0 phase, 26.7 Ϯ 1.9% were in the S phase, and 26.5 Ϯ 2.2% were in the G 2 /M phase. In these unsynchronized cultures, the percentage of cells in a phase reflects the relative duration of the phase.
Localization of endogenous NAMPT was determined in unsynchronized cultures of HepG2 cells. These cells were pretreated with a FUCCI cell cycle sensor kit to determine the cell cycle phase (Fig. 2, D and E). Most of cells (79 Ϯ 16%) in the G 1 /G 0 phase had higher NAMPT level in cytosol. 71 Ϯ 18% of cells in the early S phase and all cells in the late S, G 2 , and M phases had NAMPT predominantly in nucleus. This result was similar to a previous experiment utilizing GFP-NAMPT; therefore, overexpression of GFP-NAMPT did not significantly affect cell cycle-associated changes in NAMPT localization. Taken together, these results demonstrate that NAMPT changes its cellular localization in a cell cycle-dependent manner.

Cell cycle arrest and stress conditions influence NAMPT localization
We next analyzed the effect of cell cycle inhibitors on NAMPT localization (Fig. 3, A and B). Aphidicolin inhibits DNA replication and arrests at the S phase (10), RO-3306 inhibits cyclin-dependent kinase 1 and arrests cells at the G 2 /M phase (11), and culture at confluency and differentiation both arrests cells at the G 1 /G 0 phase. Cell cycle inhibition was con-firmed by flow cytometry (Table S1). All four manipulations increased nuclear NAMPT. NAMPT was predominantly nuclear in 22% of control cells versus 86% of aphidicolin-treated cells, 78% of RO-3306 -treated cells, 67% of confluent cells, and 100% of differentiated cells. All of the cell cycle inhibitors increased SIRT activity (Fig. 3C). Confluency reduced cellular NAD levels and NAMPT and NMNAT1 mRNA expression ( Fig. 3D and Tables S2, S3, S4, and S6).
To test the hypothesis that DNA damage and stress conditions affect nuclear transport of NAMPT, we monitored NAMPT localization after DNA damage by UV light or BrdU (12) or induction of oxidative (H 2 O 2 ) or dicarbonyl (methylglyoxal (MG)) stress (Fig. 3E). Each treatment increased number of cells with predominantly nuclear NAMPT: UV light, BrdU, H 2 O 2 , and MG to 70, 67, 38, and 72%, respectively, compared with 22% in controls. Increased PARP activity was also observed in all four conditions (Fig. 3F). UV illumination and H 2 O 2 (the more aggressive DNA-damaging agents) reduced NAD level and mRNA expression of NAMPT and NMNAT1 ( Fig. 3G and Tables S2, S3, and S5). In agreement with these results is NAMPT localization after transient plasmid transfection ( Fig. 1), where NAMPT was nuclear at 62% of cells. Transfection is another example of stress condition associated with activation of nuclear processes.
We further tested effect of SIRT6 and PARP inhibition on NAMPT localization. Trichostatin A (TSA) is an inhibitor of SIRT6 and class I and class II histone deacetylases (13). TSA at 0.75 nmol⅐liter Ϫ1 , which did not affect cell viability, increased the number of cells with cytoplasmic NAMPT localization (Fig.  3H) and increased histone H 3 acetylation (Fig. 3I). High doses of TSA (75 nmol⅐liter Ϫ1 ), which caused ϳ50% reduction in cell viability, had the opposite effect and substantially increased number of cells with predominantly nuclear NAMPT to 66% (data not shown). Inhibitor of PARP 3-aminobenzamide (3AB) reduced PARP activity but did not affect NAMPT localization ( Fig. 3, H, J, and K). NAMPT enzymatic activity inhibitor FK866 increased histone H 3 acetylation but had no effect on NAMPT localization or PARP activity (Fig. 3, H-K). Overall, these results show that NAMPT is transported into nucleus under situations with increased requirement for nuclear NAD.

Nuclear localization signal of NAMPT
Sequence analysis of NAMPT suggested two possible nuclear localization signals, 35 SYFECREKKTENSKVRKVKYEE 56 and 423 KRSKKGR 429 (Fig. 4A). Clustered amino acid substitution ( 50 RKVK to GAVA or 424 RSKK to ASGA) were introduced into GFP-NAMPT, producing GFP-NAMPT GAVA and GFP-NAMPT ASGA . After transient transfection of HepG2 cells, GFP-NAMPT GAVA remained mostly nuclear (nuclear in 74 Ϯ 11% of cells and cytoplasmic in 9 Ϯ 4% of cells), not significantly different from GFP-NAMPT WT . In contrast, localization of GFP-NAMPT ASGA was almost completely cytoplasmic (nuclear in 12 Ϯ 7% of cells and cytoplasmic in 77 Ϯ 12% of cells) (Fig. 4B). The fluorescence intensities of GFP-NAMPT GAVA and GFP-NAMPT ASGA were similar to GFP-NAMPT WT , suggesting that the introduced substitutions did not affect protein expression or stability. Region 424 RSKK is located on the surface of NAMPT and separated from its active NAMPT nuclear transport site (Fig. 4C). These results demonstrate that the 423 KRSK-KGR 429 region in NAMPT contains a required nuclear localization signal.

Inhibition of NAMPT nuclear transport slows HepG2 cell growth
To explore the functional role of NAMPT nuclear localization, we generated stable HepG2 cell lines overexpressing NAMPT WT , NAMPT ASGA , and NAMPT carrying amino acid substitutions in the active site (Asp 219 to Ala, Arg 311 to Ala, and Asp 313 to Ala) producing enzymatically inactive NAMPT (NAMPT KO ) (14,15). The three independent clones expressing NAMPT WT , NAMPT ASGA , and NAMPT KO had similar levels of NAMPT protein expression approximately five times that of the endogenous enzyme (Fig. 5A). Overexpression of NAMPT WT or NAMPT ASGA similarly increased the activity of NAD salvage pathway ϳ6-fold, whereas NAMPT KO overexpression had no effect on this activity (Fig. 5B). These data suggest that the 424 RSKK to ASGA substitution does not affect NAMPT enzymatic activity and verified that NAMPT KO is enzymatically inactive. Exogenous NAMPT mRNA levels had high variability and were slightly higher in NAMPT KO (Fig. 5C). Overexpression of NAMPT WT or NAMPT ASGA produced ϳ2-fold higher cellular NAD concentrations compared with control or NAMPT KO cells (Fig. 5D). Expression of exogenous NAMPT had no effect on the mRNA levels of endogenous NAMPT or NMNAT1 (Fig. 5, E and F). Taken together, these results demonstrate that the NAMPT WT , NAMPT ASGA , and NAMPT KO are comparably overexpressed. Like transiently expressed NAMPT, the stably expressed NAMPT WT was mostly nuclear (nuclear in 82 Ϯ 10% of cells and cytoplasmic in 7 Ϯ 5% of cells), whereas the NAMPT ASGA mutant localization was almost entirely cytoplasmic (nuclear in 5 Ϯ 3% of cells and cytoplasmic in 86 Ϯ 9% of cells) (Fig. 5G).
Overexpression of NAMPT WT accelerated growth, with a 30% decrease in generation time. In contrast, overexpression of NAMPT ASGA , as well as NAMPT KO , did not affect growth (Fig.  5H). Accelerated growth in cells overexpressing NAMPT WT was associated with increased proportion of cells in the G 1 /G 0 phase (Table S7). Overexpression of either NAMPT WT or NAMPT ASGA but not NAMPT KO made cells less sensitive to FK866, a NAMPT enzymatic activity inhibitor. Importantly, NAMPT WT relative to NAMPT ASGA overexpression made cells significantly more resistant, when ϳ10 times higher concentration of FK866 was required to achieve the same effect on cell viability (Fig. 5I). Overexpression NAMPT WT reduced histone H 3 acetylation by 64%, whereas in NAMPT ASGA cells, the reduction was only 38%, and no reduction was observed in NAMPT KO (Fig. 5J). NAMPT WT cells, but not NAMPT ASGA cells, had slightly higher SIRT activity (Table S7). PARP activity was increased by 46% in NAMPT WT overexpressing cells and did not change in NAMPT ASGA or NAMPT KO cells (Fig. 5K), suggesting that nuclear localization and enzymatic activity are required for these effects (Fig. 5, H and I).

Cell cycle-dependent NAMPT nuclear transport
NAMPT is localized in the nucleus in nondividing adipocytes and is present in both the cytoplasm and the nucleus in dividing cells. This suggests that NAMPT is transported between the nucleus and the cytoplasm in a cell cycle-dependent manner. Supporting literature includes the observation that NAMPT translocates from cytosol to nucleus after cell division arrest (6,7). We monitored NAMPT localization continuously in cells that were naturally progressing through the cell cycle. NAMPT was lowest in newly formed nuclei right after mitosis and increased thereafter. This suggests that mitosis regulates nuclear NAMPT levels. Furthermore, NAMPT was present in the nucleus in the early S phase, suggesting a possible role of this enzyme in DNA replication.
The requirement of NAD for nuclear processes is a likely reason for NAMPT nuclear localization. Nicotinamide is product of all nonredox NAD-consuming reactions (3), and in mammalian cells, most of the NAD is synthetized from nicotinamide by a salvage pathway of NAMPT and nicotinamide mononucleotide adenylyl transferase (NMNAT) (16). There is one NAMPT isoform and three NMNAT isoforms: nuclear, NMNAT 1; Golgi surface, NMNAT 2; and mitochondrial, NMNAT 3 (17). Compartmentalization of enzymes of a single metabolic pathway is a general feature of living organisms, which increases the rate and efficiency of synthesis (18). Rapid turnover of NAD occurs in the nucleus (19). Because of nuclear colocalization of NAMPT with NMNAT1, it should be possible to quickly restore NAD from nicotinamide. NMN (product of NAMPT), NAD, and nicotinamide can pass through the nuclear pore by passive diffusion (19,20), but the export of nicotinamide and further import of NMN or NAD would significantly prolong the synthesis.
The two main classes of nuclear enzymes utilizing NAD are poly(ADP-ribosyl) polymerases and NAD-dependent deacetylases, also known as SIRTs. These enzymes regulate DNA repair, chromatin structure, transcription, replication, telom-

NAMPT nuclear transport
ere length, cell division, and circadian rhythm (21)(22)(23)(24)(25)(26)(27) and are dependent on nuclear NAD availability (16). SIRTs are the important histone deacetylases and key factors for chromatin condensation and heterochromatin stabilization (26). Chromosomes are maximally condensed during mitotic segregation, then decondense, and are mostly open during replication (28). To reflect these changes, SIRT should have maximal activity prior to mitosis and lower activity in the postmitotic phase. Our results showing changes of NAMPT localization during naturally progressing cell cycle are in agreement with these cell cycle-dependent changes in chromatin structure. NAMPT nuclear content mimics different requirements of NAD because of SIRT-induced changes in chromatin structure during the cell cycle. Hence, these changes in chromatin structure are possible reasons for different NAMPT localization depending on the cell cycle.

Effect of cell cycle inhibition and stress on NAMPT localization
One of the aims of this study was to find conditions that increase NAMPT nuclear content. Cell cycle inhibition at the G 1 /G 0 (confluency and differentiation), S (aphidicolin), and G 2 /M phases (RO-3306) increased NAMPT nuclear localization. This suggests that NAMPT is transported into nucleus in all phases of cell cycle and/or that cell cycle arrest activates transport of NAMPT into nucleus. Concurrently with increased NAMPT nuclear localization, we found increased SIRT activity after all types of cell cycle inhibition. This indicates that increased SIRT-mediated NAD degradation should be at least partly covered by increased NAD synthesis within the nucleus. TSA, an inhibitor of SIRT6 and class I and class II histone deacetylases slightly increased cytoplasmic NAMPT. SIRT6 is located in the nucleus and is crucial for maintaining genomic integrity (13). Increased cytoplasmic NAMPT after TSA treatment suggests a possible relationship between SIRT6 activity, histone deacetylation, and NAMPT nuclear transport.
We hypothesized that NAMPT is transported into nucleus to provide nuclear NAD, which led us to test NAMPT localization under conditions with extreme demand on nuclear NAD. The most effective NAD consumers are PARP-activated by DNA damage (29). Activation of PARP by various stress conditions caused increased NAMPT nuclear localization. UV illumina-

NAMPT nuclear transport
tion and H 2 O 2 , the more aggressive DNA-damaging agents, also reduced cellular NAD level. Inhibition of PARP had no effect on NAMPT transport or NAD level; however, PARP activity is low without activation by DNA damage (30); therefore, PARP inhibition under basal conditions was not expected to not modify nuclear NAD level. NAMPT inhibition also did not change NAMPT localization, suggesting that NAMPT transport is not affected by its total activity or by total cellular NAD content.
In total, we tested several conditions associated with increased nuclear NAD degradation, and all of them caused increased nuclear localization of NAMPT. Although it is not possible to exclude other reasons for NAMPT nuclear localization (such as regulation of activity by clustering with enzymes and transcription factors) and the mechanisms regulating NAMPT nuclear transport are unknown, we believe that the main reason for NAMPT nuclear localization is NAD synthesis and that NAMPT nuclear transport is one of the mechanisms regulating the nuclear NAD level.

Structure of NAMPT NLS
The identified NAMPT NLS ( 423 KRSKKGR 429 ) fits a classical monopartite class 2 NLS (31), suggesting that NAMPT is transported into the nucleus by a mechanism using importins ␣ and ␤ and the small GTPase Ran (32). The regulated, rate-limiting step in this type of transport is importin ␣ recognition of the NLS (32). Humans have seven isoforms of importin ␣, which differ in cargo specificity. Because cargo specificity is not primarily determined by the NLS sequence, one cannot predict which importin ␣ isoform is responsible for NAMPT nuclear transport (33,34).

Effect of NAMPT nuclear transport inhibition on cell growth
Recently it has been shown that NAMPT is necessary for prevention of cellular senescence (35). It is not known whether NAMPT nuclear localization is required for the cell cycle progression. HepG2 cells overexpressing NAMPT ASGA grew slowly and were more prone to pharmacological inhibition of NAMPT compared with cells overexpressing NAMPT WT . These effects occurred despite the same cellular concentration of NAD. Although nuclear NAD levels were not measured, it is possible that they were selectively reduced. Possible mechanisms of cell death or cell cycle arrest caused by reduction of nuclear NAD have been reported (36). Reduced NAD and SIRT activity decreases deacetylation and activates proteins like tumor suppressor protein p53 or FOXO1. These proteins regulate cell cycle arrest and apoptosis through downstream targets (36, 37, 67). In agreement with this mechanism, we found lower histone H 3 acetylation in NAMPT WT and NAMPT ASGA but not in NAMPT KO cells, suggesting higher SIRT-mediated deacetylation caused by higher levels of SIRT substrate (NAD) in these cells. Importantly, NAMPT ASGA cells had histone H 3 acetylation ϳ2 times higher compared with NAMPT WT , which indicates that nuclear NAD regeneration provides substrate to SIRT more effectively then NAD synthesis in cytoplasm. NAMPT WT cells had also higher PARP activity, which was probably caused by higher nuclear NAD availability, because this effect was not seen in NAMPT ASGA or NAMPT KO cells.
Our data demonstrate that all effects of NAMPT WT overexpression were completely abolished in NAMPT KO cells and were either abolished or substantially reduced in NAMPT ASGA cells. Thus, NAMPT enzymatic activity and nuclear localization is required for the effect of NAMPT overexpression on cell growth, histone deacetylation, and PARP activity, suggesting that NAMPT overexpression affects cell growth by providing NAD for nuclear processes.

Possible utilization of NAMPT nuclear transport inhibition in cancer treatment
The overexpression of NAMPT has been observed in many cancers (38,39) and was associated with increased tumor growth, poor prognosis, metastases, and resistance to therapy (40 -43). The faster growth of HepG2 cells overexpressing NAMPT WT suggests that NAMPT may directly contribute to the aggressiveness of these tumors. Tumor cells are more sensitive to NAD depletion, because of increased NAD demand for redox (glycolysis and oxidative phosphorylation) (44,45) and nonredox (PARP and SIRT) reactions (23,46). Thus, NAMPT inhibitors have been tested as anti-cancer agents (47,48). NAMPT inhibition was shown to cause a massive reduction of NAD, ATP depletion, and cell death (49 -51). However, a threshold level of NAD depletion must be achieved for the cytotoxic effect; otherwise cells survive and regenerate the NAD pool (52). Dose-limiting toxicities did not permit the use of sufficiently high doses of NAMPT inhibitors (38), and no tumor response was reported in phase I clinical trials (48). NAMPT inhibitor toxicities included thrombocytopenia, lymphopenia (53,54), gastrointestinal side effects (48), and retinal toxicity NAMPT nuclear transport (54,55). For this reason, other approaches such as targeted delivery of NAMPT inhibitor into cancer cells or newly developed inhibitors are currently being tested (56,57).
NAMPT activity is essential for mammalian cells to survive (58). Here we show that nuclear transport of NAMPT is a regulated process, with inhibition slowing cell growth without significantly changing total cellular NAD levels. Based on our results and on the information that cancer cells have higher NAD demands for nuclear processes (23,46) and that nuclear NAD content is a key regulator of cell cycle and apoptosis (36, 37), we propose inhibition of NAMPT nuclear transport as a new approach for cancer treatment. This would target nuclear processes requiring NAD, sparing metabolic processes in cytosol and mitochondria and is expected to be less toxic than inhibition of NAMPT enzymatic activity. Our data suggest possible efficacy of this approach, although there are some limitations in our study. We were not able to measure nuclear NAD content, and all our tested cells had preserved production of endogenous NAMPT. Because this is the first study suggesting possible utilization of NAMPT nuclear transport inhibition in cancer treatment, further research for verification or exclusion of this approach will be necessary. Inhibition of NAMPT nuclear transport might be achieved by blocking the interaction between NAMPT and specific isoform of importin ␣. It can be expected that this approach should be the most effective in the treatment of tumors, primarily those overexpressing NAMPT. Furthermore, its efficacy could be increased by combination therapy with drugs increasing the requirement of NAD in the nucleus, such as DNA-damaging agents.

Plasmid vectors
NAMPT cDNA was obtained from 3T3-L1 preadipocytes. The total RNA was extracted using the RNeasy mini kit (Qiagen). Total RNA was used for reverse transcription (AMV First Strand cDNA synthesis kit; New England Biolabs) with reverse NAMPT primer followed by PCR amplification (primer sequences are shown in Table S8). Insertion of cDNA into pCMV-HA (Clontech) vector was carried out using restriction endonucleases that produce cohesive ends. The transfer of NAMPT cDNA (or other DNA fragment) into another plasmid was performed either using direct restriction endonuclease digestion or by PCR amplification with primers containing restriction endonuclease sites, followed by the ligation of the received fragment into another plasmid. The structures of all plasmid vectors and their application are described in Fig. S1. Specific point mutations were obtained by PCR mutagenesis using Phusion Hot start II DNA polymerase (Thermo Scientific) and pEGFP-C1-NAMPT as a template. Primer sequences for PCR mutagenesis are shown Table S8. The PCR product was purified by MinElute Reaction Cleanup (Qiagen), and the template was digested by DpnI (New England Biolabs). Competent cells of Escherichia coli XL1-Blue were used for the amplification of prepared plasmids. The correct structures of all plasmids were verified by restriction digestion followed by electrophoresis and by sequencing (GATC Biotech).
3T3-L1 cells were transfected by electroporation using Amaxa Nucleofector Technology (Lonza). HepG2 cells were transfected by lipofection using TransIT-LT1 transfection reagent (Mirus Bio). Stable cell lines were prepared by transfection of cells with the plasmid mixture pcGlobin2-SB100X containing Sleeping Beauty SB100X transposase and one of pT2HB-CAGGS plasmids. 24 h after transfection, the culture medium was changed for the selection medium containing G418 antibiotics (Sigma) at a concentration of 1200 mg⅐liter Ϫ1 (HepG2) and 750 mg⅐liter Ϫ1 (3T3-L1). Selection was carried out for 2 weeks and was followed by cloning of the cells (a single cell was seeded into one vessel of 96-well plate containing a conditioned medium). The clones producing a required protein were selected 3 weeks after cloning. The verification of properties was carried out by fluorescence microscopy (GFP and GFP-NAMPT producing cells) and by Western blotting (GFP-NAMPT, HA-NAMPT, and NAMPT-overexpressing cells, Fig.  S2). The cells producing GFP as an inert protein were used as a control because of the eventual effect of transfection and selection on studied parameters.
3 ϫ 10 4 cells were seeded into one vessel of a 12-well plate to determine the generation time. The generation time was calculated from the number of viable cells at days 1, 2, and 4 after seeding. The number of cells was determined by hemocytometer after trypsinization and complete resuspension.
The effect of NAMPT inhibition on cell viability was determined by 72-h cultivation of cells with various concentration of FK866 (Sigma). Then the cell viability was determined using cell proliferation reagent WST-1 (Roche).

Pharmacological treatment, cell cycle inhibition, and stress conditions
24 h before the start of an experiment, 3T3-L1 preadipocytes with stable GFP-NAMPT production were seeded into a 12-well plate at a density of 2.5 ϫ 10 4 cells/well for the purpose of fluorescence microscopy, into a 6-well plate at of 6 ϫ 10 4 cells/well for flow cytometry, and into a 10-cm dish at a density of 4 ϫ 10 5 cells/well for NAD(P), mRNA expression, and PARP and SIRT activity determination. At time 0 h, the cells were treated with either one of the cell cycle inhibitors: aphidicolin at 6 mol⅐liter Ϫ1 (Sigma) or RO-3306 at 10 mol⅐liter Ϫ1 (Sigma); one of the stress inducers: BrdU at 1 mmol⅐liter Ϫ1 (Sigma), hydrogen peroxide at 500 mol⅐liter Ϫ1 (Lach-Ner, Czech Republic), or MG at 400 mol⅐l Ϫ1 (Sigma); the PARP inhibitor NAMPT nuclear transport 3AB at 10 mmol⅐liter Ϫ1 (Sigma); the SIRT6 inhibitor TSA at 0.75 nmol⅐liter Ϫ1 or 75 nmol⅐liter Ϫ1 (Sigma); or the NAMPT enzymatic activity inhibitor FK866 at 1 nmol⅐liter Ϫ1 (Sigma). Genotoxic stress was further induced by UV light for 20 s (wavelength, 253.7 nm). Other ways to induce the cell cycle inhibition were 2-day cultivation of fully confluent 3T3-L1 preadipocytes or their differentiation into adipocytes. The localization of GFP-NAMPT by fluorescence microscopy, SIRT activity, NAD(P) concentration, and mRNA expression were analyzed 24 h after the beginning of the experiment. PARPs are activated very quickly by damaged DNA; therefore a shorter time after the initiation of stress conditions was used to obtain this parameter. The PARP activity was analyzed after 120 min of incubation with BrdU, 10 min of incubation with H 2 O 2 , and 20 min of incubation with MG and 60 min after illumination of cells with UV light.

Determination of the cell cycle using flow cytometry
Cells were trypsinized, washed with PBS, and centrifuged at 250 ϫ g for 5 min. The cell pellet was resuspended in 0.5 ml of PBS and fixed with 4.5 ml of 80% ice-cold ethanol for 30 min on ice. Fixed cells were then centrifuged at 1500 ϫ g for 5 min, washed with PBS, and centrifuged again for 5 min at 1500 ϫ g. Fixed cells were treated with 0.5 ml of RNase A (1.5 g⅐liter Ϫ1 ) and stained with 0.5 ml of propidium iodide (0.1 g⅐liter Ϫ1 ) for 90 min at 37°C. Stained cells were analyzed by flow cytometer FACS Aria III (BD) with a 488-nm laser as an excitation source, and the emission was separated by a band pass filter at 575/26 nm. The obtained data were analyzed with BD FACSDiva software.

Determination of NAD metabolism-associated parameters
Activity of NAD salvage enzymes was determined by the previously described method (60). Briefly, the cells were trypsinized, washed with PBS, and lysed by osmotic shock. Reaction mixture contains the cell lysate in reaction buffer (50 mmol⅐liter Ϫ1 Tris-HCl, pH 8.8, 20 mmol⅐liter Ϫ1 MgCl 2 , protease inhibitors), 20 units of alcohol dehydrogenase from yeast (Sigma), 5 mmol⅐liter Ϫ1 5-phospho-␣-D-ribose-1-diphosphate (Sigma), 4 mmol⅐liter Ϫ1 ATP (Sigma), 50 mmol⅐liter Ϫ1 nicotinamide (Sigma), and 1% (v/v) ethanol (Lach-Ner, Czech Republic). The reaction mixture was incubated in 37°C for 16 h, and the fluorescence of NADH was measured every 30 min (excitation wavelength, 355 nm; emission wavelength, 460 nm). SIRT activity was analyzed using a sirtuin activity assay kit (Bio-Vision). Cell lysates containing 40 g of protein were used for analysis. Histone H 3 acetylation was determined by a histone H 3 acetylation assay kit (ab115102; Abcam), and 2 g of histone proteins were used for analysis. PARP activity was analyzed using HT Universal Colorimetric PARP assay kit (Trevigen). Cell lysates containing 40 g of protein were used for analysis. The manufacturer-provided activated DNA was not used for measuring PARP activity in 3T3-L1 cells subjected to genotoxic, oxidative, and dicarbonyl stress, where damaged DNA from the samples serves as PARP activator. The cellular content of NAD(P) was measured by high-performance liquid chromatography with mass spectrometry detection. The method are described in detail in the supporting text; Table S9.

mRNA quantification
Total RNA was isolated using RNeasy mini kit (Qiagen). Possible DNA contamination was eliminated by DNase treatment (Qiagen). The standard amount of 1 g of total RNA was used for reverse transcription using a high-capacity cDNA reverse transcription kit (Applied Biosystems) with random primers. 5 l of cDNA solution was used for quantitative PCR using Pow-erUp TM SYBR TM Green Master Mix (Applied Biosystems) with concentration of primers at 1 mol⅐liter Ϫ1 . Primer sequences are described in Table S8. PCR was run on CFX96 Touch TM (Bio-Rad).

Fluorescence microscopy and immunolocalization
Fluorescent microscopy (including live cell imaging) was performed by inverted microscopy on an Olympus IX-81 with integrated CO 2 incubator (Olympus). The brightness and contrast of figures were adjusted using ImageJ 1.49v software (National Institutes of Health).
The immunolocalization of HA-NAMPT was carried out 24 h after cell transfection with pCMV-HA-NAMPT plasmid. The cells were fixed with 4% formaldehyde (P-LAB) for 15 min and permeabilized using 0.1% Triton X-100 (Pierce) for 20 min. The permeabilized cells were incubated with 3% BSA (Sigma) for 15 min and subsequently with FITC-labeled mouse antibody directed against HA (1:300, H7411; Sigma). After washing, the cells were incubated with 1 g⅐ml Ϫ1 DAPI (Sigma) for 5 min and observed by fluorescence microscopy. The same protocol of cell fixation and permeabilization was used for the immunolocalization of nonlabeled NAMPT. The permeabilized cells were incubated with a primary rabbit antibody directed against NAMPT (1:200, RB-08-0003; RayBiotech) for 2 h and with Alexa Fluor 488 -labeled secondary donkey antibody directed against rabbit IgG (1:600, 111-545-003; Jackson Immuno-Research) for 1 h. After washing, the cells were incubated with 0.5 g⅐ml Ϫ1 DAPI (Sigma) for 5 min and observed by fluorescence microscopy. Specificity of antibodies used to NAMPT immunolocalization was validated by immunofluorescence microscopy and Western blotting (Fig. S3).
Fluorescence microscopy filters for excitation and emission, respectively, were 492 and 530 nm for EGFP, Alexa Fluor 488 and GFP of FUCCI, 350 and 457 nm for DAPI, 620 and 700 nm for Alexa Fluor 647, and 572 and 620 nm for red fluorescent protein of FUCCI. Microscopy images acquired using various filters were merged using ImageJ 1.49v software (National Institutes of Health), whereas DAPI staining or bright field were used to verify position of the nucleus. To quantify NAMPT localization, the proportion of cells containing NAMPT predominantly in nucleus or in cytoplasm was scored. Two independent investigators in each image counted the number of cells that had (i) higher average fluorescence in the nucleus (nuclear localization), (ii) higher average fluorescence in the cytoplasm (cytoplasmic localization), and (iii) similar average fluorescence in the cytoplasm and in the nucleus; it was not possible to distinguish whether fluorescence is higher in nucleus or cytoplasm (dual localization). The proportion of cells with nuclear or cytoplasmic localization was expressed as a percentage of the total number of cells in the field of view.

NAMPT nuclear transport Colocalization of NAMPT with cell cycle markers
Cell cycle phase in HepG2 cells with natural expression of NAMPT was determined by transduction Premo TM FUCCI cell cycle sensor (Thermo Scientific), followed by incubation for 48 h. Immunolocalization of NAMPT was performed by similar procedure described in fluorescence microscopy and immunolocalization. The only difference was in the secondary antibody used for this experiment, which was Alexa Fluor 647-labeled secondary donkey antibody directed against rabbit IgG (1:600, 111-605-003; Jackson ImmunoResearch) for 1 h.

Statistics
The data are expressed as means Ϯ S.D. Student's t test was applied to determine differences between cell samples, with a statistical significance defined as p Ͻ 0.05. Data normality was verified by the Shapiro-Wilk W test.