Enzymes in the NAD⁺ Salvage Pathway Regulate SIRT1 Activity at Target Gene Promoters^*

Plasmids and Protein Expression

The human NAMPT cDNA variant 1 was obtained from American Type Culture Collection (ATCC catalog no. 10327682). The human NMNAT-1 cDNA was kindly provided by Dr. Mathias Ziegler (University of Bergen). The NAMPT and NMNAT-1 cDNAs were PCR-amplified with an amino-terminal FLAG tag and cloned into pQCXIP (BD Biosciences) for retrovirus-mediated expression in mammalian cells. Catalytic mutants and RNA interference-resistant mutants of NMNAT-1 and NAMPT were generated by site-directed mutagenesis following QuikChange protocol (Stratagene). The NMNAT-1 cDNA was also cloned into the following bacterial expression plasmids; 1) a modified pGEX-2TK vector for production of a GST-NMNAT-1 fusion protein and 2) a modified pET-15b vector with an amino-terminal His₆ tag for production of a His₆-NMNAT-1 fusion protein. The recombinant proteins were purified under standard non-denaturing conditions. All plasmid constructs derived from PCR products were confirmed by DNA sequencing. Recombinant mouse SIRT1 with an amino-terminal His₆ tag was prepared as described previously (36).

Antibodies

Rabbit polyclonal antibodies were raised against purified human NMNAT-1 and mouse SIRT1 (Pocono Rabbit Farm and Laboratory). All other antibodies were from commercial sources: anti-NAMPT rabbit polyclonal antibody (Bethyl Laboratories), anti-FLAG M2 monoclonal antibody (Sigma), anti-acetyl-histone H4 (Lys-16) polyclonal antibody (Millipore), and anti-H3 polyclonal antibody (Abcam).

Short Hairpin RNAs (shRNAs)

shRNA sequences for RNA interference-mediated knockdown were designed using the web-based siDESIGN® Center software from Dharmacon and cloned into pSuper.Retro vectors (Oligoengine). Multiple target sequences were tested for each gene, and the most effective sequences were used for the studies described herein: NAMPT, 5′-GAGTGTTACTGGCTTACAA-3′ (#1) and 5′-TAACTTAGATGGTCTGGAA-3′ (#2); NMNAT-1, 5′-AACACAAGATTCTAGTCAA-3′ (#1) and 5′-AACTCACCTACTCTAGAAA-3′ (#2); SIRT1, 5′-TGAAGTGCCTCAGATATTA-3′ (#1) and 5′-AAGCGATGTTTGATATTGA-3′ (#2). An shRNA directed against firefly luciferase (Luc), 5′-GATATGGGCTGAATACAAA-3′, was used as a control (38).

Cell Culture, Ectopic Expression, and shRNA-mediated Knockdown

MCF-7 cells were kindly provided by Dr. Benita Katzenellenbogen (University of Illinois, Urbana-Champaign) and maintained in Eagle's minimum essential medium (Sigma) supplemented with 5% calf serum and antibiotics. Phoenix-Ampho retrovirus producer cells (ATCC) were cultured in Dulbecco's modified Eagle's medium (Sigma) with 10% fetal bovine serum. Retroviruses were prepared following a standard transfection protocol using the Phoenix-Ampho cell line and the retroviral vectors described above (i.e. derivatives of pQCXIP, pSuper.Retro). The recombinant retroviruses were used to transduce MCF-7 cells, which were then selected with 1.0 μg/ml puromycin (Sigma) or 800 μg/ml G418 (Invitrogen) as appropriate for each vector. Expression and knockdown were screened by Western blotting.

NAD⁺ Measurements

The concentrations of NAD⁺ in whole cell extracts from MCF-7 cells were determined using a quantitative HPLC/mass spectrometry (MS) method (HPLC/matrix-assisted laser desorption ionization/MS) with ¹⁸O standards (13, 39).

Expression Microarray and Gene Ontology (GO) Analyses

To knockdown target factors for expression microarray analyses, the following shRNA sequences were used: NAMPT, shRNA #1; NMNAT-1, shRNA #1; SIRT1, shRNAs #1 and #2 (maintained under puromycin and G418 selection). For each factor studied, three independent isolates of MCF-7 cells recently (within 2–3 weeks) transduced with shRNA-expressing retroviruses were used. Studies on NMNAT-1, NAMPT, and SIRT1 were carried out independently, and each had its own matching Luc control. Total RNA was isolated from the various MCF-7 cell lines using Trizol reagent (Invitrogen) followed by RNeasy columns (Qiagen).

Sample labeling and hybridization to Affymetrix Human Genome U133 plus 2.0 and U133A 2.0 arrays were carried out under standard conditions at the Cornell Microarray Core Facility. The raw data were processed using Affymetrix GCOS software to obtain detection calls and signal values. Where applicable, the common probe sets of the two array platforms were normalized by scaling. The data sets were adjusted for batch effects using a parametric empirical Bayes method (40). After this normalization, all values <0.01 were adjusted to 0.01, and the data were log₂-transformed, median-centered for each array, and median-centered for each gene. Only those probe sets having “present” calls in at least two of the three replicates were included for further analysis. A two-tailed Student's t test was applied to the normalized data matrix to identify genes differentially expressed between each knockdown condition and the matching Luc control. A p value cutoff of 0.05 was applied to define the differentially expressed gene set. A -fold change cutoff of log₂ <−0.5 or >0.5 was applied to select significantly regulated gene sets. The gene lists were analyzed for enrichment of GO terms using the Functional Annotation Clustering and Functional Annotation Chart tools from the DAVID Bioinformatics Resources website (david.abcc.ncifcrf.gov). The expression microarray data sets can be accessed from the NIH GEO website (www.ncbi.nlm.nih.gov) using accession number GSE13577.

Reverse Transcription-Quantitative Real-time PCR (RT-qPCR) Assays

Total RNA samples were purified using Trizol Reagent (Invitrogen). cDNA samples were prepared by reverse transcription with an oligo(dT) primer and analyzed by SYBR Green real-time PCR under standard conditions. For data normalization, β-actin was used as the reference gene. The sequences of the primers used are available on request.

Chromatin Immunoprecipitation (ChIP)-Real-time PCR (qPCR) Assays

ChIP assays using MCF-7 cells were performed as described previously (41, 42). ChIP samples were analyzed by SYBR Green real-time PCR under standard conditions. The sequences of the primers used are available on request. Positive ChIP signals for all antibodies used in this study are at least 3-fold above no antibody controls. The ChIP data presented are values after subtraction of the corresponding no antibody control data.

ChIP-Western Assays

ChIP assays were carried out as described above. After washing samples in ChIP wash buffer, the protein A/G beads were resuspended in 2× SDS loading dye and boiled for 10 min. Samples were resolved on 10% SDS-PAGE for standard Western blot analysis.

GST-NMNAT-1 Interaction Assays

GST and GST-NMNAT-1 were expressed in bacteria and purified using standard glutathione-agarose affinity chromatography. The purified proteins were quantified by Coomassie Blue staining on SDS-PAGE gels using BSA as a standard. HeLa cell nuclear extract was incubated with immobilized GST or GST-NMNAT-1, the samples were washed, and the specifically bound proteins were analyzed by Western blotting.

NAMPT and NMNAT-1 Regulate NAD⁺ Biosynthesis and Gene Expression in MCF-7 Breast Cancer Cells

NAMPT and NMNAT-1 catalyze sequential reactions in the nuclear NAD⁺ salvage pathway that re-synthesizes NAD⁺ using NAM (Fig. 1A). To determine the functions of NAMPT and NMNAT-1 in gene regulation, we first examined their roles in NAD⁺ synthesis and gene expression in MCF-7 breast cancer cells. shRNA-mediated stable knockdown reduced NAMPT and NMNAT-1 protein levels to ∼10–20% of the levels in control (Luc) knockdown cells (Fig. 1B). Total cellular NAD⁺ levels in extracts from control and knockdown cells were measured using a quantitative HPLC/mass spectrometry method with ¹⁸O-labeled standards (13, 39). Using this sensitive method, we detected a 30–45% decrease in the total cellular NAD⁺ levels upon knockdown of either NAMPT or NMNAT-1 (Fig. 1C). These results indicate that both NAMPT and NMNAT-1 contribute to NAD⁺ production in MCF-7 cells. Considering that there are two other NMNAT enzymes in cells, one in Golgi (NMNAT-2) and one in mitochondria (NMNAT-3) (11), our results implicate NMNAT-1 as a major contributor to cellular NAD⁺ production.

Expression microarray analyses using the same MCF-7 knockdown cell models indicate that both NAMPT and NMNAT-1 have broad effects on gene expression, each regulating as much as ∼10% of the MCF-7 cell transcriptome at 95% confidence (two-tailed Student's t test) or as much as ∼3% with a log₂ <−0.5 or >0.5-fold change cutoff (Fig. 2, A and B). Overall, the extent and magnitude of expression regulation by NAMPT and NMNAT-1 in our study are in agreement with previously reported data from samples ectopically expressing these enzymes (12, 43).

Interestingly, we only observed an ∼15–20% overlap between the NAMPT- and NMNAT-1-regulated gene sets (Fig. 2, A and B), perhaps reflecting differences between the two enzymes in subcellular localization, interacting factors, and the metabolic intermediates accumulated as a result of enzyme depletion. The regulation of the overlapping genes, however, generally occurred in the same direction (up or down), with similar magnitude and was highly correlated (Spearman correlation coefficient 0.72; p value <10⁻⁵) (Fig. 2C). GO analyses support the same observation. The NAMPT- and NMNAT-1-regulated genes have overlapping, yet distinct cellular and molecular functions (supplemental Table S1). For the commonly regulated NAMPT and NMNAT-1 target genes, GO analysis revealed an enrichment of terms related to neuronal differentiation, cell signaling, and cellular membrane functions (supplemental Table S1), suggesting that these functions are particularly sensitive to perturbation of the nuclear NAD⁺ salvage pathway.

To confirm that expression regulation by the specific shRNAs used in our microarray analyses is not due to off-target effects, we tested an additional shRNA each for NAMPT and NMNAT-1 (supplemental Fig. S1A). In gene-specific RT-qPCR analysis, the two shRNAs targeting each factor showed similar effects on the expression of 38 genes (supplemental Fig. S1B). Furthermore, we observed a good agreement between the efficiency of protein knockdown and the effects on expression by the different shRNAs (supplemental Fig. S1). These results suggest that off-target effects were not a major influence on the identification of target genes in our expression analyses using the NAMPT and NMNAT-1 shRNAs.

For the purposes of this study, we focused on the commonly regulated NAMPT and NMNAT-1 target genes with a log₂ <−0.5 or >0.5-fold change cutoff (37 genes total; Fig. 2B) as they are most likely to represent genes regulated by the intact NAMPT/NMNAT-1 salvage pathway rather than independent actions of NAMPT and NMNAT-1. As expected, gene-specific RT-qPCR analyses confirmed the NAMPT- and NMNAT-1-regulated expression of most of these genes, including TGFB2 (down-regulated by NAMPT or NMNAT-1 knockdown) and ATXN10 (up-regulated by NAMPT or NMNAT-1 knockdown) (Fig. 2D; see also Fig. 4). To test directly if this regulation requires NAD⁺, we examined the effect of adding exogenous NAD⁺ to the cell culture medium, as described previously (44). Treatment with 1 mm NAD⁺ for 24 h restored the expression of TGFB2 and ATXN10 in the NAMPT and NMNAT-1 knockdown cells (Fig. 2D). For some genes (e.g. CAV1), the addition of exogenous NAD⁺ was not able to restore expression in the NAMPT and NMNAT-1 knockdown cells (supplemental Fig. S2), suggesting NAD⁺-independent functions of NAMPT and NMNAT-1 in the regulation of some target genes.

Focusing on the genes responsive to NAD⁺ treatment, we further tested if NAMPT and NMNAT-1 enzymatic activity is required for the regulation of these genes. When RNA interference-resistant forms of the NAD⁺ biosynthetic enzymes were re-expressed in the corresponding knockdown cell lines, the wild type enzymes restored the expression of TGFB2 and ATXN10 in the knockdown cells (supplemental Fig. S3). In contrast, the catalytic mutants had no effect (supplemental Fig. S3). This result demonstrates the essential role of NAD⁺ synthesis in the regulation of these genes. Overall, our NAD⁺ add back, and enzyme re-expression studies indicate that, for a subset of the NAMPT- and NMNAT-1-responsive genes, cellular NAD⁺ level plays a direct role in the regulation of their expression.

SIRT1 Mediates NAMPT- and NMNAT-1-dependent Gene Regulation

Given the NAD⁺ dependence for the expression of TGFB2, ATXN10, and other NAMPT- and NMNAT-1-regulated genes, we considered the possible involvement of SIRT1, an NAD⁺-dependent nuclear deacetylase. SIRT1 regulates a variety of cellular processes, including stress responses, metabolism, and cell differentiation, maturation, and survival (33). In MCF-7 cells, reduced SIRT1 activity promotes re-expression of epigenetically silenced tumor suppressor genes, impairs activation of mitogen-activated protein kinase pathways, and induces senescence-like growth arrest (45, 46). SIRT1 NAD⁺-dependent enzymatic activity can mediate the deacetylation of histones and other nuclear proteins, including a variety of transcription factors (33). Previous studies have shown that SIRT1 regulates cellular events downstream of NAMPT and NMNAT-1 (12, 14, 17, 18, 24). Interestingly, in MCF-7 cells SIRT1 depletion had similar effects on the expression of ATXN10 and TGFB2 as depletion of NAMPT and NMNAT-1 (Figs. 2D and 4). Furthermore, although exogenous NAD⁺ rescued the expression of TGFB2 and ATXN10 in NAMPT- or NMNAT-1-depleted cells, it had little effect on these genes in SIRT1-depleted cells (Fig. 2D). These results indicate that SIRT1 is required for the NAD⁺-dependent regulation of these genes.

To further explore the relationship between gene regulation by SIRT1 and gene regulation by NAMPT or NMNAT-1, we performed an expression microarray analysis upon SIRT1 knockdown versus a control (Luc) knockdown in MCF-7 cells. Depletion of SIRT1 significantly altered the expression of ∼200 genes (log₂ <−0.5 or >0.5-fold change cutoff), with approximately equal numbers of up-regulated and down-regulated genes (Fig. 3, A and B). In gene-specific RT-qPCR analyses, the two shRNA constructs used in our SIRT1 knockdown studies produced similar effects on gene expression (supplemental Fig. S1), suggesting that off-target effects were not a major influence on the identification of SIRT1 target genes in our expression analyses. GO analysis revealed important functions of the SIRT1-regulated genes in cell signaling and metabolism (supplemental Table S1). Many of the NAMPT- or NMNAT-1-regulated genes described above are similarly regulated (i.e. magnitude and direction of regulation) by SIRT1 knockdown, and the expression profiles show significant correlation in the knockdown cells (Fig. 3, C and D). Further analysis of the 37 genes commonly regulated by NMNAT-1 and NAMPT (log₂ <−0.5 or >0.5-fold change cutoff) revealed extensive co-regulation by SIRT1 (supplemental Fig. S4), an observation that we confirmed by RT-qPCR (Fig. 4). Together, these results suggest that SIRT1 is a key regulator of NAMPT- and NMNAT-1-responsive genes.

The genes commonly regulated by NAMPT, NMNAT-1, and SIRT1 are significantly enriched in GO terms including “metabolism” and “cellular membrane” (supplemental Table S1), highlighting the cellular functions that are most critically controlled by the NAD⁺-dependent SIRT1 activity. Careful examination reveals that many of these genes play roles in cell signaling and/or neuronal functions (Table 1). For example, TGFβ2, SOCS2, and PEG10 function in cytokine signaling pathways, but PEG10 and SOCS2 also have neuronal functions (see Table 1 and the references therein). Furthermore, NPY1R and NELL2 have predominantly neuron-related signaling functions. Other factors, including ATXN10, NAT1, TFPI, and TMSNB, are associated with different pathological aspects of the nervous system. These results suggest that the NAD⁺-salvage pathway and NAD⁺-dependent SIRT1 activity may be important regulators of cell signaling and neuronal functions. Interestingly, neuronal protective functions have been reported for NMNAT and SIRT1 (24–30). The functions of these factors in gene regulation in a neuronal cell model is currently under investigation.

NAMPT and NMNAT-1 Regulate SIRT1 Activity at Target Gene Promoters

Next, we examined potential functional interplay between NAMPT, NMNAT-1, and SIRT1 at commonly regulated gene promoters. We restricted our analysis to these genes because they are more likely targets of regulation by the NAD⁺ salvage pathway as well as NAD⁺-dependent SIRT1 activity. Previous studies have reported the binding of SIRT1 to target gene promoters, with subsequent deacetylation of specific target proteins including histones and transcription factors (18, 33, 34, 37). We also observed binding of SIRT1 to the target genes that we identified in our analyses in MCF-7 cells. As shown in the ChIP assays presented in Fig. 5A, SIRT1 bound at the promoters but not upstream regions (∼−10 kilobases) of ATXN10, NAT1, and NELL2. A similar level of SIRT1 binding was not observed, however, at the promoter of a non-target gene, STC1 (Fig. 5, A and C). The SIRT1 ChIP signal was reduced to background levels upon SIRT1 knockdown, confirming the specificity of the signal in this assay (Fig. 5A).

One outcome of SIRT1 activity is the deacetylation of histone H4 acetylated at lysine 16 (Ac-H4K16), a modification targeted exclusively by sirtuin family members (47). Previous studies have shown that Ac-H4K16 levels peak at gene promoters, although they may be somewhat elevated in the coding regions of genes as well (48). Our ChIP-qPCR analysis and previous ChIP-chip analyses indicates that SIRT1 is also found in peaks at target gene promoters (supplemental Fig. S5) (37). Based on these observations, we focused our analysis of Ac-H4K16 levels at promoter regions of target genes as an indicator of SIRT1 activity. shRNA-mediated knockdown of SIRT1 or inhibition of SIRT1 enzymatic activity with the chemical inhibitor sirtinol (49) increased the levels of Ac-H4K16 (normalized to total histone H3 content) at the promoters of the common target genes ATXN10 and NAT1 (Fig. 5B). The same treatment conditions increased expression for ATXN10 and decreased expression for NAT1 (Fig. 5C), suggesting different modes of SIRT1-dependent regulation for these two genes. Interestingly, knockdown of NAMPT or NMNAT-1 had similar effects on ATXN10 and NAT1 with respect to the levels of promoter-directed Ac-H4K16 (Fig. 5D) and expression (Fig. 4). In contrast, NELL2, another gene regulated by SIRT1 knockdown, showed (i) no requirement for SIRT1 enzymatic activity for expression (as determined by sirtinol treatment; Fig. 5C), (ii) no effect of SIRT1 knockdown or sirtinol treatment on the levels of promoter-directed Ac-H4K16 (Fig. 5B), and (iii) no effect of NAMPT or NMNAT-1 knockdown on the levels of promoter-directed Ac-H4K16 (Fig. 5D). As expected, the non-target gene STC1 was largely unaffected by all of the experimental conditions tested (Fig. 5, B–D). As a control, recruitment of SIRT1 protein to these gene promoters was not affected by knockdown of NAMPT or NMNAT-1 (supplemental Fig. S6). Thus, NAMPT and NMNAT-1 control SIRT1-dependent deacetylation of Ac-H4K16 and transcriptional regulation of target genes requiring SIRT1 enzymatic activity (e.g. ATXN10 and NAT1). At target genes regulated independently of SIRT1 enzymatic activity (e.g. NELL2); however, NAMPT and NMNAT-1 may control gene expression through mechanisms other than NAD⁺ production. Alternatively, other NAD⁺-dependent enzymes such as PARP-1 (50) may mediate NAD⁺-dependent transcription regulation of these genes. Together, these results indicate that NAMPT and NMNAT-1 can directly regulate molecular outcomes at some target genes by regulating SIRT1 catalytic activity through the production of NAD⁺.

NMNAT-1 Is Recruited to Gene Promoter Regions through Interaction with SIRT1

Our results suggest a mechanism in which NAD⁺ production by NAMPT and NMNAT-1 regulates gene expression through modulation of SIRT1 activity at target gene promoters. Given that both NAMPT and NMNAT-1 are found in the nucleus (3, 4, 11), we considered the possibility that they might be recruited to the promoters of target genes for localized actions in a manner similar to SIRT1. To test this hypothesis, we examined the binding of NAMPT and NMNAT-1 to target gene promoters in MCF-7 cells expressing FLAG-tagged NAMPT or NMNAT-1. Like SIRT1, NMNAT-1 bound at the promoters of the commonly regulated target genes (i.e. ATXN10, NAT1, and NELL2) but not at upstream and downstream regions of the same genes or at the promoter of a non-target gene (i.e. STC1) (Fig. 6A and supplemental Fig. S5). Under the same assay conditions, we were unable to detect NAMPT binding at the same promoters despite the fact that the FLAG antibody can immunoprecipitate FLAG-NAMPT under our ChIP assay conditions (data not shown). The difference in the chromatin association of NMNAT-1 and NAMPT may mark an important distinction between their cellular functions.

Given the colocalization of NMNAT-1 and SIRT1 to the same promoters, we considered the possibility that SIRT1 might recruit NMNAT-1 to specific sites on chromatin. Indeed, knockdown of SIRT1 reduced the binding of NMNAT-1 to target gene promoters (Fig. 6B). Furthermore, under our ChIP assay conditions SIRT1 co-immunoprecipitated with FLAG-NMNAT-1 and vice versa (Fig. 6C), demonstrating in vivo interaction between the two proteins. In an in vitro binding assay, native SIRT1 in nuclear extracts specifically bound to GST-NMNAT-1 but not GST alone (Fig. 6D). Together, these results indicate that SIRT1 binds to and recruits NMNAT-1 to specific sites on chromatin.