Nicotinamide Mononucleotide Adenylyltransferase Is a Stress Response Protein Regulated by the Heat Shock Factor/Hypoxia-inducible Factor 1α Pathway*

Stress responses are cellular processes essential for maintenance of cellular integrity and defense against environmental and intracellular insults. Neurodegenerative conditions are linked with inadequate stress responses. Several stress-responsive genes encoding neuroprotective proteins have been identified, and among them, the heat shock proteins comprise an important group of molecular chaperones that have neuroprotective functions. However, evidence for other critical stress-responsive genes is lacking. Recent studies on the NAD synthesis enzyme nicotinamide mononucleotide adenylyltransferase (NMNAT) have uncovered a novel neuronal maintenance and protective function against activity-, injury-, or misfolded protein-induced degeneration in Drosophila and in mammalian neurons. Here, we show that NMNAT is also a novel stress response protein required for thermotolerance and mitigation of oxidative stress-induced shortened lifespan. NMNAT is transcriptionally regulated during various stress conditions including heat shock and hypoxia through heat shock factor (HSF) and hypoxia-inducible factor 1α in vivo. HSF binds to nmnat promoter and induces NMNAT expression under heat shock. In contrast, under hypoxia, HIF1α up-regulates NMNAT indirectly through the induction of HSF. Our studies provide an in vivo mechanism for transcriptional regulation of NMNAT under stress and establish an essential role for this neuroprotective factor in cellular stress response.

Stress responses are cellular processes essential for maintenance of cellular integrity and defense against environmental and intracellular insults. Neurodegenerative conditions are linked with inadequate stress responses. Several stress-responsive genes encoding neuroprotective proteins have been identified, and among them, the heat shock proteins comprise an important group of molecular chaperones that have neuroprotective functions. However, evidence for other critical stressresponsive genes is lacking. Recent studies on the NAD synthesis enzyme nicotinamide mononucleotide adenylyltransferase (NMNAT) have uncovered a novel neuronal maintenance and protective function against activity-, injury-, or misfolded protein-induced degeneration in Drosophila and in mammalian neurons. Here, we show that NMNAT is also a novel stress response protein required for thermotolerance and mitigation of oxidative stress-induced shortened lifespan. NMNAT is transcriptionally regulated during various stress conditions including heat shock and hypoxia through heat shock factor (HSF) and hypoxia-inducible factor 1␣ in vivo. HSF binds to nmnat promoter and induces NMNAT expression under heat shock. In contrast, under hypoxia, HIF1␣ up-regulates NMNAT indirectly through the induction of HSF. Our studies provide an in vivo mechanism for transcriptional regulation of NMNAT under stress and establish an essential role for this neuroprotective factor in cellular stress response.
When faced with abnormal conditions, such as heat, oxidative stress, hypoxia, or accumulation of aberrant proteins, cells implement a stress response program to protect themselves and ensure survival. Stress conditions can cause protein unfolding, misfolding, or aggregation, and the consequent inadequate response to stress can lead to developmental defects, shortened lifespan, and neurodegenerative conditions (for review, see Ref. 2). The increased synthesis of molecular chaperone heat shock proteins (HSPs) 2 is central to the stress response because they function to prevent protein misfolding and aggregation to maintain protein homeostasis (1,2). It is thought that elevated expression of HSPs is sufficient to protect cells from a wide range of cytotoxic conditions (1,3,4). However, it is unclear whether the stress response network includes essential genes other than HSPs. Although a few metabolic enzymes have been found to be up-regulated upon stress (5,6), it is unclear whether metabolic enzymes are an integral part of the stress response network.
Heat shock factors (HSFs) are the master stress transcription factors of heat shock response, with one HSF in invertebrates and multiple HSFs in plants and vertebrates (7)(8)(9). Through their roles in mediating transcriptional activation of HSP genes, HSFs function in maintaining protein homeostasis and integrating cellular response to stress and development (10). Upregulation of HSPs by HSF1 is triggered by a variety of acute and chronic stress conditions and disease states (11). When faced with other stress conditions including hypoxia and oxidative stress, additional transcription factors such as hypoxia-inducible factor 1␣ (HIF1␣) play a significant role in ensuring cell survival (12). Interestingly, recent work has demonstrated cross-regulation of HSF by HIF1 during hypoxia (13), revealing that more complex signaling and transcriptional circuits are involved in the regulation of stress-responsive genes.
Recent studies in Drosophila have uncovered the protective effects of the NAD synthase NMNAT against neuronal excitotoxicity-or polyglutamine protein-induced neurodegeneration, and injury-induced axonal degeneration (14 -16). Studies in mammalian neurons have also shown that overexpression of NMNAT protects against injury-or stress-induced axonal degeneration (17)(18)(19)(20). The neuroprotective function of NMNAT is in part mediated through its chaperone activity independent of its NAD synthesis function (15). Although the neuronal maintenance and protective effects of NMNAT have been demonstrated, it is unclear how the nmnat gene is regulated in vivo. Interestingly, it has been observed that when human Ataxin-1 with an 82 polyglutamine expansion (hATX1[82Q]) was expressed in Drosophila brain, endogenous NMNAT protein was up-regulated and recruited to the hATX1[82Q] aggregates (15), indicating that the nmnat gene may respond to proteotoxic stress. Understanding the stress response of NMNAT and the regulatory mechanisms of nmnat gene expression will be important for the design of neuroprotective strategies.
In this study, we investigate the role of NMNAT in stress, show that NMNAT is a stress protein that is transcriptionally up-regulated upon various stress conditions including heat shock and hypoxia, and further characterize the transcriptional circuits involved in the regulation of neuroprotective NMNAT upon induction of stress. Our studies provide an in vivo mechanism for transcriptional regulation of NMNAT under stress and extend the stress response network to include a metabolic enzyme that is critical for neuronal maintenance and protection.

Stress Paradigms
Heat Shock Stress-Adult flies collected and sexed 24 -48 h post-eclosion were subjected to two 1-h heat shocks at 37°C with a 45-min room temperature period of acclimatization between shocks (21). Following shock, the flies were collected at various time points (0 -24 h after heat shock). Fly extracts were prepared to examine the induction of NMNAT at both the mRNA level (quantitative real time PCR) and protein level (by Western blotting).
Oxidative Stress-Adult flies collected and sexed 24 -48 h post-eclosion were fed Paraquat-containing medium. Multiple concentrations of Paraquat (0, 5, 10, and 20 mM) in 1% agar, 5% sucrose medium were used in this study. Approximately 100 male or female flies were used at each drug concentration.

Western Blot Analysis
Proteins were extracted from fly heads with a homogenizing buffer (20 mM HEPES, pH 7.5, 100 mM KCl, 5% glycerol, 10 mM EDTA, 0.1% Triton, 1 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, protease inhibitor mixture (Sigma)) and resolved by SDS-PAGE. Western blot analysis was performed with infrared dye-conjugated secondary antibodies, IR700 and IR800 (LI-COR Biosciences); the blots were imaged and processed on an Odyssey infrared imaging system. The IR700 and IR800 signals appear in green and red, respectively.

RNA Extraction and Quantitative PCR
Total RNA was extracted from fly heads by TRIzol reagent (Invitrogen), according to the manufacturer's protocol. For each extraction, RNA concentration was determined spectrophotometrically at 260 nm, and 2 g of RNA was used for reverse transcription reaction with a SuperScript first-strand synthesis system (Invitrogen). Quantification of nmnat and hsp70 mRNA was performed using an ABI PRISM 7000 sequence detection system (PE Applied Biosystems, Weiterstadt, Germany) and TaqMan universal PCR master mix (Applied Biosystems). A housekeeping gene, rp49, was used as an internal control to standardize mRNA expression. The amplification mix (25 l) contained 100 ng of cDNA template, 12.5 l of TaqMan universal PCR master mix, and 1 l of genespecific TaqMan probe-primer set. The samples were amplified by a PCR program of 40 cycles of 10 s at 95°C, 15 s at 55°C, and 1 min at 72°C. The C T value was defined as the number of cycles required for the fluorescence to exceed the detection threshold, and the data were analyzed using the 2 Ϫ⌬⌬Ct method to quantitatively the assess relative changes in gene expression (22).

Thermotolerance Assay
A thermotolerance assay was carried out as described (23). Briefly, 1-2-day-old flies were sexed and put into empty vials in groups of 20. The flies were given a mild heat shock in a 35°C water bath for 30 min and immediately transferred to a 39°C water bath. Starting from this time, the flies were examined at 10-min intervals under a light microscope, the number of paralyzed flies was counted, and then quickly returned to the 39°C water bath. Flies were considered paralyzed if they did not move any parts of their bodies, even after the vials were tapped. The incubation at 39°C continued until all of the flies were paralyzed.

Lifespan Assay
Adult flies of various genotypes were collected and sexed 24 -48 h after eclosion. Cohorts of 200 -300 flies were placed in groups of 25 individuals. Each group was placed in vials with no drug or with 2 mM Paraquat-supplemented food. Every day, the number of dead flies in each vial was recorded, and every third day, the flies were transferred to new vials containing fresh food with or without drug. This process was followed until all of the flies died, and the percentage of flies alive at each time point was graphed (24). The results were statistically tested with the Kaplan-Meier analysis with a semiparametric log-rank test (25), and the analysis was performed in Microsoft Excel.

Image Acquisition and Processing
Images from fluorescently labeled specimens were taken on an Olympus FV1000 confocal microscope and processed using FluoView10-ASW (Olympus) and Adobe Photoshop CS3.

ChIP Assay
ChIP assay was performed on chromatin obtained from S2 cells using an EZCHIP kit (Upstate), according to the manufacturer's protocol. For HSE binding ChIP, S2 cells were heatshocked at 37°C for 15 min. For HRE binding ChIP, S2 cells were transfected with HA-Sima and HA-Tango (Gift of Dr. Pablo Wappner) and 48 h later were treated with 100 M DFO overnight (26). Following cross-linking, the chromatin was sheared using a Branson Sonifier 150, with seven 20-s bursts at high setting (5) with a 1-min hold in ice after shearing. Immunoprecipitation was followed with antibodies for either anti-HA (Millipore), anti-HSF (a gift of Dr. J. T. Lis), anti-RNA PolII (Upstate), and mouse IgG (Upstate). Reverse cross-linking and elution were performed according to the manufacturer's protocol.

RESULTS
NMNAT Is Up-regulated under Stress-To test whether the housekeeping enzyme NMNAT is a stress protein, we first examined the effects of various stress conditions on NMNAT expression in wild-type flies. We used three stress paradigms: heat shock, oxidative stress, and hypoxic stress. For heat shock stress, wild-type adult flies of 2-3 days of age were subjected to two 1-h heat shocks at 37°C, with a 45-min room temperature period of acclimatization between shocks (21). Oxidative stress was induced by feeding flies with the free radical-producing agent 1,1Ј dimethyl-4 -4Ј-bipyridynium dichloride (Paraquat). Paraquat is a quaternary nitrogen herbicide, a very toxic substance leading to acute poisoning and death (27). The toxicity of Paraquat has been attributed to the generation of the superoxide anion leading to the synthesis of more toxic reactive oxygen species such as hydroxyl radicals and hydrogen peroxide (28). In our experiments, 2-day-old flies were fed media containing 5, 10, or 20 mM Paraquat. Hypoxic stress was induced by exposing adult flies to 5% O 2 for various durations.
To monitor the induction of stress, we used heat shock protein Hsp70 as a positive control. As shown in Fig. 1, heat shock, oxidative stress and acute hypoxia up-regulated Hsp70 at both the protein (Fig. 1A) and the mRNA transcript level (Fig. 1C), whereas the level of actin stayed constant during stress (Fig.  1A). Hsp70 is a heat shock protein that is not expressed under normal conditions (29); however, we detected low protein signals at prestress conditions (Fig. 1A). This is likely because of antibody cross-reactivity to Hsc70 protein, a constitutively expressed protein (29). To test the specificity of the Hsp70 antibody, we probed brain lysates from nonstressed hsp70 null flies (Df(3R)Hsp70A, Df(3R)Hsp70B) that lack all 12 copies of the Hsp70 gene, and detected a low signal at 70 kDa (supplemental Fig. S1), suggesting the antibody has slight cross-reactivity to Hsc70 and the low level signals we detected at prestress conditions correspond to Hsc70.
The level of NMNAT expression was determined both at the protein level by quantitative Western blotting using a NMNAT-specific antibody (14) and at the transcript level by quantitative real time PCR using TaqMan probes. Interestingly, NMNAT protein (  Fig. S2). Acute exposure to hypoxia (5% O 2 ) for 3 h resulted in a 3.1-fold up-regulation in NMNAT transcript level (Fig. 1B). Note that additional protein bands above the NMNAT band (29 kDa, arrowhead in Fig. 1A) were detected in protein samples from heat shock-and hypoxiatreated groups (Fig. 1A). Because the antibody has been shown to specifically recognize NMNAT (29 kDa) (14) and the nature of these additional bands was unclear, we included only the specific 29-kDa NMNAT band in the quantification (supplemental Fig. S2). Collectively, these results show that NMNAT was up-regulated under heat stress, hypoxia, or oxidative stress in a time-or dose-dependent manner similar to the stress protein Hsp70 (Fig. 1) as reported previously (6,30,31), suggesting that NMNAT may be transcriptionally up-regulated under a variety of stress conditions.
Transcription Factors HSF and HIF1␣/Sima Regulate nmnat Gene Expression under Stress-To understand the transcriptional regulation of nmnat gene under stress, we analyzed the promoter region of the Drosophila nmnat gene to locate transcription factor-binding sites using the bioinformatics program MatInspector (32) and identified several consensus binding elements for transcription factors HSF and HIF1␣ (see below and Fig. 5A), suggesting possible regulation of NMNAT via transcription factors implicated in stress. To examine whether Drosophila HSF directly regulates the transcription of nmnat upon heat shock, we measured the expression and induction of nmnat upon heat shock in HSF gain-or loss-of-function flies. For HSF gain-of-function, we used the pan-neuronal driver elav-GAL4 to overexpress UAS-HSF in the nervous system. For HSF loss-of-function, we used a temperature-sensitive mutant allele hsf 4 , where HSF is inactivated at 37°C and unable to mediate transcriptional activation upon heat shock (33). As shown in Fig. 2, in HSF-overexpressing flies (HSF OE), NMNAT protein was induced to a higher level in the brain upon heat shock, and both NMNAT and Hsp70 proteins were present at a moderate level prior to heat shock ( Fig  Wild-type flies were used in these stress paradigms. Heat stress consisted of shocking flies in a 37°C water bath for 1 h, twice, with a 45-min period of room temperature acclimatization between treatments. Oxidative stress was induced by feeding flies 5, 10, or 20 mM Paraquat. Hypoxic stress was induced by exposing adult flies to 5% O 2 for various durations. Actin was used as a loading control, and Hsp70 was used as a positive control for these stress conditions. Western blot analysis was performed with infrared dye-conjugated secondary antibodies IR700 (green signal) and IR800 (red signal); the blots were imaged and processed on an Odyssey infrared imaging system. The arrowhead indicates the higher molecular weight bands that NMNAT antibody recognized but not included in the quantification in Supplemental Fig. S1. B and C, nmnat (B) and hsp70 (C) transcription levels were significantly increased post-stress as measured by quantitative real time RT-PCR. rp49 was used as a housekeeping control for normalization. The values are expressed as the means Ϯ S.E. (n Ն 4). The significance level was established by one-way analysis of variance post hoc Scheffe test. *, p Ͻ 0.05; **, p Ͻ 0.001. condition ( Fig. 2A), which is likely due to antibody cross-reactivity to Hsc70 protein (supplemental Fig. S1).
The effect of HSF on nmnat expression was also evident at the transcript level, as measured by quantitative real time PCR. As shown in Fig. 2C, nmnat was transcriptionally up-regulated upon heat shock in wild-type (yw) flies, and this up-regulation upon heat shock was enhanced in HSF-overexpressing (HSF OE) flies and was diminished in loss-of-HSF (hsf 4 ) flies. These results suggest that heat shock transcription factor HSF is both required and sufficient for the up-regulation of NMNAT under heat shock.
In addition to heat shock, we have shown that hypoxic stress also induced nmnat gene expression in wild-type flies (Fig. 1). We next examined the requirement for HIF1␣ for the up-regulation of NMNAT under hypoxic stress. We measured the expression and induction of nmnat upon hypoxia (5% O 2 ) in HIF1␣ gain-or loss-of-function flies. For loss of function, we used a mutant allele of Sima (34), the Drosophila homologue of HIF1␣ (35). For gain of function, we used elav-GAL4 driver to overexpress UAS-Sima in the nervous system. When Sima was overexpressed, we observed moderate levels of NMNAT and Hsp70 protein and mRNA levels in fly brain extracts prior to hypoxia and enhanced up-regulation of NMNAT upon hypoxic stress (Fig. 2, B and D, and supplemental Fig. S3B). In contrast, flies heterozygous for Sima ϩ/Ϫ failed to up-regulate NMNAT or Hsp70 upon hypoxia. The diminished up-regulation is specific for the hypoxic stress pathway, because both NMNAT and Hsp70 were up-regulated in Sima ϩ/Ϫ mutant flies following heat shock stress, suggesting that the heat shock pathway was intact in these flies ( Fig. 2B and supplemental Fig. S3C). These results suggest that hypoxia transcription factor HIF1␣ is required to induce NMNAT under hypoxia.
To assess stress responses at the cellular level, we used mosaic analyses to examine the effects of HSF or Sima on NMNAT expression. Using a photoreceptor-specific driver GMR-Gal4 (36), we overexpressed HSF or Sima together with GFP specifically in the posterior half of the eye primordium in third instar (L3) larval eye imaginal discs. As shown in Fig. 3, in the eye primordium cells demarcated by GFP where either HSF or Sima was overexpressed, a greater induction of NMNAT was observed when L3 larvae were subjected to either heat shock or hypoxia, respectively. These results suggest that HSF or Sima induces NMNAT under stress in a cell-autonomous manner.
HSF Is the Central Transcription Factor for nmnat Gene Expression under Heat Shock and Hypoxia-Our results so far suggest that the transcription factors HSF and HIF1␣/Sima can regulate nmnat gene expression under stress. To understand the molecular mechanism underlying the transcriptional regulation, we next examined whether HSF or Sima can directly bind to the promoter region of nmnat and induce transcription upon stress and whether the predicted consensus HSEs and HRE in nmnat promoter are functional. In silico analysis revealed two possible HSF-binding sites and one possible HIF1␣-binding site (Fig. 4, A and B). We named the two HSEs as HSEd for HSE distal and HSEp for HSE proximal, referring to their relative distance to the transcription start site. HSEd consists of a single pentameric nGAAn element, whereas HSEp consists of three contiguous inverted repeats of the pentameric element, comparable with those found in the proximal promoter region of HSP genes (37, 38) (Fig. 4A). Because HSF trimerization is required for its binding to HSE (39), the HSEd with a single pentameric nGAAn is less likely than HSEp to be functional.
Using cultured Drosophila S2 cells, we carried out a ChIP assay to detect promoter occupancy for both HSF and Sima transcription factors. As shown in Fig. 4C, upon heat shock, HSF bound to both HSE sites, with stronger binding at the proximal HSE site. A similar enrichment of HSF occupancy was also observed for the HSE-containing region of Drosophila hsp70 (40). To carry out the ChIP analysis of the HRE element, we transfected S2 cells with HA-tagged HIF1␣ (HA-Sima) and its transcriptional co-factor HA-Tango (26,34) and subjected S2 cells to hypoxia induced by treatment with 100 M DFO for 16 h (41). When chromatin was precipitated with anti-HA anti- , or Sima (F and G) was expressed in the L3 larval eye imaginal disc with a GMR-Gal4 driver. L3 larvae were subjected to heat shock, hypoxia, or normal conditions, and eye discs were dissected and stained for NMNAT (red), Hsp70 (magenta), and DAPI (blue). A-C, heat shock or hypoxia caused an overall induction of NMNAT and Hsp70 throughout the eye disc. D and E, HSF-eGFP overexpression caused NMNAT and Hsp70 up-regulation in the GMR-expressing region upon heat shock. F and G, Sima overexpression caused an increase of both NMNAT and Hsp70 levels after hypoxia in the GMR-expressing region.
body, promoter occupancy of Sima at the predicted HRE-containing region of the nmnat gene was not detected (Fig. 4D), although Sima occupancy was detected in the genomic region containing an HRE of a known HIF1␣ target Hph (13). Interestingly, we also detected the binding of Sima to an HRE in the second intronic region of hsf gene upon hypoxia (Fig. 4, B and  D), confirming the previous report that HSF is transcriptionally up-regulated by HIF1␣ during hypoxia (13). Collectively, these results indicate that heat shock factor HSF can bind to the heat shock elements in nmnat promoter, preferentially to the proximal HSE upon heat shock. However, hypoxia-inducible factor Sima/HIF1␣ does not bind to nmnat promoter directly upon hypoxia. It is likely that Sima up-regulates nmnat transcription upon hypoxia indirectly through up-regulation of HSF, which in turn directly binds to nmnat promoter and initiates transcription.
To test whether HSF is up-regulated upon hypoxia, we determined the protein level of HSF under heat shock and hypoxic stress in wild-type flies. As shown in Fig. 5A, heat shock did not alter the protein levels of HSF, although both NMNAT and Hsp70 levels were induced (Fig. 5A, lanes 1-5). However, hypoxia causes a steady increase in HSF protein levels, concomitant with an increase in NMNAT and Hsp70 (Fig. 5A, lanes  6 -9). Interestingly, this up-regulation in HSF levels is dependent on HIF1␣ levels, because overexpression of HIF1␣/Sima caused a more robust induction in HSF levels, consistent with its targets, NMNAT and Hsp70 (Fig. 5A, lanes 10 -13). To exclude the possibility that NMNAT may influence the expression of either HSF or HSP proteins, we examined the levels of HSF and Hsp70 in flies with reduced or increased NMNAT levels (NMNAT heterozygous or overexpression), and compared with levels in wild-type flies. As shown in Fig. 5B, changes in NMNAT levels have no effect on HSF or Hsp70 levels.
To further examine the functionality of these transcription factor-binding sites in nmnat promoter, we employed a dual luciferase reporter assay (Promega) to measure the transcriptional activity of luciferase reporter constructs that contain either the entire nmnat promoter region (nmnat Pro ) or regions containing wild-type or mutant HSEs (HSEpϩd or mHSE) (Fig.  6A). S2 cells that were transfected with constructs containing two HSEs (HSEpϩd) or the proximal HSE (HSEp) showed a significant induction of luciferase activity post-heat shock (Fig.  6B). When two of the residues in the second pentameric repeat in HSEp were mutated (mHSEp) to abolish the binding of HSF trimer, the heat shock-induced luciferase activity was diminished. In contrast, mutating HSEd had little effect on luciferase induction post-heat shock, suggesting that the distal HSE site was nonfunctional. These results suggest that HSEp is necessary and sufficient to mediate HSF binding and nmnat induction upon heat shock.
We carried out similar analysis for hypoxic stress. When hypoxia was induced by 100 M DFO, S2 cells transfected with a luciferase construct containing the full-length promoter (nmnat Pro ) showed significant induction of luciferase activity, to a level comparable with that of positive control cells transfected with a construct containing a functional HRE in the promoter of mouse VEGF gene (42) (Fig. 6C), indicating the transcriptional activation of nmnat upon hypoxia. However, this transcriptional activation was abolished when both HSEs sites  were mutated in this construct, leaving only intact HRE (HREϩmHSEs), and proximal HSE alone was necessary and sufficient to restore the transcriptional activity upon hypoxia (Fig. 6C). Collectively, these results show that the predicted HRE element in the nmnat promoter region was nonfunctional, and instead, the integrity of the proximal HSE was required for nmnat up-regulation upon hypoxia. Our analyses therefore indicate that the proximal HSE is the key promoter element mediating transcriptional up-regulation under both heat shock and hypoxic conditions.
Our results so far suggest that stress transcription factor HSF is required for the up-regulation of nmnat under heat shock and that HIF1␣ is required for the up-regulation of nmnat under hypoxia. However, the promoter element in nmnat gene that mediates the transcription under both heat shock and hypoxia is the same HSE. We also showed that the level of HSF is induced upon hypoxia (Fig. 5). Therefore, under hypoxia HIF1␣ induces the up-regulation of nmnat indirectly through activation of HSF. This would predict the requirement for HSF for the up-regulation under hypoxia. To test this possibility, we subjected hsf 4 mutant flies to hypoxia (24 h at 5% O 2 ) and examined the protein levels of NMNAT or Hsp70. As shown in supplemental Fig. S4, NMNAT or Hsp70 was not up-regulated in HSF loss-of-function flies under hypoxia, suggesting that HSF is required for the induction of NMNAT or Hsp70 upon hypoxia.
To further evaluate the cellular requirement for HSF in NMNAT stress response in vivo, we used mosaic analyses to examine the cellular effects of either gain or loss of function of HSF under stress conditions. Using bbg-Gal4 driver (43), we overexpressed HSF-eGFP in a patch of cells in the third instar (L3) larval wing imaginal disc that will become the future wing margin (43). As shown in Fig. 7, A-L, the GFP signal demarcates cells overexpressing HSF, where a greater induction of NMNAT was observed upon either heat shock or hypoxia (Fig.  7). In contrast, when HSF was down-regulated in these cells using HSF-RNAi driven by bbg-GAL4, a reduction in NMNAT was observed in the cells with no HSF, compared with the surrounding wild-type cells upon both heat and hypoxic stress (Fig. 7, M-X). These results further suggest that HSF is required for the induction of NMNAT under both heat shock and hypoxic stress in a cell-autonomous manner.
Our in vivo analysis on nmnat promoter and gene expression suggest the following model for nmnat transcriptional regulation upon stress: upon heat shock, HSF directly binds to nmnat proximal HSE and induces transcription; upon hypoxia, HIF1␣/Sima binds to the HRE element in the HSF intronic region and up-regulates HSF (13), which subsequently induces transcription via nmnat HSE.
NMNAT Is Required for Thermotolerance and Prolongs Lifespan During Chronic Oxidative Stress-To directly determine the functional role of NMNAT during acute stress, we examined the requirement of NMNAT for thermotolerance in adult flies. Wild-type flies are able to tolerate high temperatures (e.g. 39°C) when preconditioned with a mild heat shock (e.g. 35°C for 30 min) to stimulate the synthesis of HSPs (23). In contrast, flies with reduced levels of HSPs such as Hsp70 have significantly reduced thermotolerance (23). To examine the requirement for NMNAT and HSF, we first compared the thermotolerance between wild type and Hsp70, NMNAT, or HSF loss of function flies. As shown in Fig. 8A, Hsp70 heterozygous flies (Hsp70/ϩ) had a greatly shortened time to paralysis upon exposure to high temperature (39°C) compared with wild type, consistent with a previous report (23) (Fig. 8A, line 3). Interestingly, loss of one copy of nmnat (nmnat/ϩ) caused a more severe reduction of thermotolerance compared with Hsp70 heterozygous flies, suggesting the essential role of NMNAT in thermotolerance (Fig. 8A, line 2). Flies heterozygous for both Hsp70 and nmnat (Hsp70/nmnat) showed even greater reduction in thermotolerance (Fig. 8A, line 1), suggesting that the effect of NMNAT and Hsp70 can be additive. Heterozygous HSF flies (hsf/ϩ) showed a mild reduction of thermotolerance (Fig. 8A, FIGURE 6. The proximal HSE in nmnat promoter region is necessary and sufficient for transcription induction under heat shock and hypoxia. A, analysis of the consensus binding elements for both HSF (red) and Sima (blue) in Drosophila nmnat promoter. B and C, the functionality of nmnat promoter elements was tested using dual luciferase reporter assay. The promoter region of nmnat with various lengths and mutations in A was cloned into a minimal firefly luciferase vector (pGL4). Renilla luciferase was used to normalize for transfection efficiency. S2 cells were transfected with luciferase constructs and treated with heat shock (B) or hypoxia (C). Constructs containing two HSEs (HSEpϩd) or the proximal HSE (HSEp) showed a significant induction of luciferase activity post-heat shock (B). mHSEp diminished the heat shock-induced luciferase activity, but mHSEd had little effect on luciferase induction post-heat shock (B). Upon DFO-induced hypoxia, nmnat Pro and mHSEd showed induction of luciferase activity similar to positive control, whereas HREϩmHSEs with only the HRE intact failed to induce luciferase activity upon hypoxia (C). line 4), suggesting that there was enough HSF protein present to induce transcriptional up-regulation of protective factors upon heat stress. These results suggest that similar to Hsp70, NMNAT plays an essential role in thermotolerance, and the effect of NMNAT and Hsp70 can be additive. Next, we determined the effect of NMNAT overexpression on thermotolerance. Compared to wild type, overexpressing NMNAT either in the brain (elav-GAL4/UAS-NMNAT) or in the whole body (tubulin-GAL4/UAS-NMNAT) greatly increased thermotolerance (Fig. 8A, lines 7 and 9). In addition, NMNAT overexpression in hsp70 or hsf heterozygous background (elav-GAL4/ UAS-NMNAT;hsp70/ϩ and hsf/ϩ;elav-GAL4/UAS-NMNAT) rescued the reduced thermotolerance associated with loss of Hsp70 or HSF and extended the thermotolerance beyond that of wild type (Fig. 8A, lines 6 and 8). These results suggest that NMNAT functions physiologically as a stress protein and that its expression levels are important determinants of stress tolerance in flies.
To investigate the role of NMNAT in chronic oxidative stress response, we examined the lifespan under oxidative stress induced by Paraquat feeding. Increased Paraquat resistance has been correlated with increased lifespan in long-lived mutants of Caenorhabditis elegans, Drosophila, and mouse (44 -46). Hence, we measured the resistance to dietary Paraquat to determine the role of NMNAT in resistance to oxidative stress and prolonging lifespan. We raised newly eclosed adult flies on 2 mM Paraquat and measured lifespan. As shown in Fig. 8B, Paraquat exposure severely reduced lifespan in wild-type flies (yw), from a 50% survival rate of 32.3 days under 0 mM to 13.7 days under 2 mM Paraquat. However, overexpressing either Drosophila NMNAT or a human homologue HsNMNAT3 in the central nervous system with elav-Gal4 driver right-shifted the lifespan curve of flies exposed to Paraquat. Furthermore, nmnat/ϩ heterozygous flies displayed a shortened lifespan compared with wild-type flies, suggesting that NMNAT contributes to the organismal resistance to oxidative stress. These results support the role of NMNAT and its homologues as a stress protein and that its expression levels are important determinants of oxidative stress tolerance and lifespan in flies. The wing discs were dissected at 12 h after treatment and stained for NMNAT (red), HSF (magenta), and DAPI (blue). GFP demarcated HSF overexpression. The NMNAT level was up-regulated in HSF overexpressing cells under heat shock and hypoxia conditions. M-X, HSF-RNAi was overexpressed in the L3 imaginal disc with bbg-GAL4. The wing discs were dissected from L3 larvae at 12 h after no stress (M-P), heat shock (Q-T), or hypoxia (1% O 2 ) (U-X) and stained for NMNAT (red), HSF (magenta), and DAPI (blue). NMNAT level was not up-regulated in HSF-RNAioverexpressing cells, and the level was lower than that in surrounding cells under heat shock and hypoxia conditions. Scale bar in A, 20 m.

FIGURE 8.
A, NMNAT is required for thermotolerance in Drosophila. Flies of different genotypes were preconditioned for 30 min in a 35°C water bath and then exposed to 39°C. The time until paralysis onset was measured. The values are expressed as the means Ϯ S.E. of 8 -10 experiments, where each experiment consisted of 20 flies. B, NMNAT expression significantly improved survival in flies exposed to Paraquat-induced oxidative stress. The flies were maintained on either a 0 or 2 mM Paraquat diet (in 5% sucrose, 1% agar) from 1 day post-eclosion. The flies were transferred to fresh food every 2 days. The survival rates were recorded for each genotype. Wild-type flies had severely compromised lifespan upon 2 mM Paraquat exposure. NMNAT heterozygous flies had further compromised lifespans. Overexpressing either Drosophila or human NMNAT in the CNS with elav-Gal4 driver significantly prolonged the lifespan of flies upon exposure to Paraquat.

DISCUSSION
Our studies in this report show that NMNAT is a stress response protein essential for thermotolerance and mitigation of shortened lifespan in flies subjected to Paraquat-induced oxidative stress. NMNAT is up-regulated in vivo under various stress conditions including heat stress, hypoxia, and oxidative stress. The stress transcription factor HSF is the central regulator of nmnat transcription upon stress. The transcriptional response of NMNAT under stress is consistent with its chaperone function (15), further suggesting that the housekeeping enzyme NMNAT can respond transcriptionally to stress conditions by means of increased protein levels and mitigation of cellular proteotoxicity. Our findings expand the cellular stress network beyond the realm of heat shock proteins to include a metabolic enzyme. Although a few other enzymes have been found to be up-regulated upon stress (5,6), our studies provide a first example of a metabolic enzyme as an integral part of the stress response inasmuch as partial loss of nmnat significantly reduces stress tolerance.
NMNAT May Represent a Class of Stress Response Proteins That Are Distinct from Inducible Heat Shock Proteins-The upregulation of NMNAT under stress shares significant similarity to heat shock protein Hsp70. However, important differences were observed between NMNAT and Hsp70. First, under normal conditions, in contrast to Hsp70, which is not expressed (29), NMNAT is expressed at moderate levels because of its function as a housekeeping enzyme. Second, although both NMNAT and Hsp70 are directly regulated by transcription factor HSF, a significant difference in the relative protein levels between Hsp70 and NMNAT was observed in HSF-overexpressing flies. For example, NMNAT was expressed at a high level in HSF OE flies prior to heat shock and was induced further upon heat shock, whereas Hsp70 was expressed at high levels prior to heat shock but was maintained at the same expression level upon heat shock ( Fig. 2A). This is likely due to the negative feedback mechanism that regulates Hsp70 expression (4,39), where excess Hsp70 protein binds to HSF, reducing the DNA binding capacity of HSF, thereby attenuating transcription of Hsp70 and other HSPs (4,39,47). Therefore in HSF OE flies, the level of Hsp70 prior heat shock is already at a high enough level to inhibit further induction upon heat shock. In contrast to Hsp70, NMNAT is a housekeeping enzyme and is constitutively expressed at a moderate level under normal conditions (14). The up-regulation of NMNAT under stress in HSF-overexpressing flies was persistent without decline (Fig. 2), indicating that NMNAT may not be under the same negative feedback regulation as Hsp70. These differences in transcriptional regulation between NMNAT and heat shock proteins suggest that NMNAT, as a housekeeping metabolic enzyme, may represent a different class of stress response proteins. Inasmuch as housekeeping enzymes are readily available under normal conditions, they can be the first responders to a stress condition and thereby reduce the resultant proteotoxicity.
The Evolutionarily Conserved Transcriptional Regulation of NMNAT under Stress-Several microarray and expression profiling studies in different organisms have indicated the stress-related change in the expression of NMNAT homologues. For example, treating parental salt-sensitive rats with mild hypoxia (12% oxygen) revealed a 1.96-fold up-regulation of NMNAT1 transcripts (p Ͻ 0.0005) (48), and treating the quadriceps of mice in hypoxic conditions for 2 weeks induced a 2.23-fold up-regulation of NMNAT1 transcript levels (p Ͻ 0.0079) (49). In Saccharomyces cerevisiae, a 45-min exposure to anoxic conditions induced a robust up-regulation of NMNAT transcription (fold change, 1.8; p Ͻ 0.04) (50). Such conservation in NMNAT regulation upon stress is also evident in humans, where up-regulation of NMNAT may be an adaptive response to cope with stress from natural environment as seen in a study in Andeans living in high altitudes with chronic hypoxia who have higher levels of NMNAT1 compared with control subjects who live in lower altitudes (51). Interestingly, the protein levels appeared to go down within an hour when the human subjects were brought down to sea level, suggesting that the elevated NMNAT1 levels at high altitude were a result of transcriptional control upon hypoxia (51). The evolutionarily conserved transcriptional regulation of NMNAT expression further suggests the essential role of NMNAT in stress response.
HSF Is the Central Transcription Factor for nmnat Gene Expression-Our genetic studies indicated that stress transcription factor HSF is required for NMNAT up-regulation under heat shock and HIF1␣ is required under hypoxia (Fig. 2). Our genomic analysis of the promoter region of nmnat gene further revealed consensus binding sites for HSF and HIF1␣. However, our subsequent functional analyses indicated that the HIF1␣-binding element is not functional and that the up-regulation of NMNAT under hypoxia is indirectly mediated by HSF. It is known that HIF1␣ mediates the stress transcription during hypoxia by dimerizing with HIF1␤ and binding to the consensus binding site A/GCGTG present in the HREs of many oxygen-regulated genes (52). It has been suggested that the HRE core sequence, (A/G)CGTG, is necessary but not sufficient for gene activation because functional HREs require flanking sequences with important DNA binding elements for additional transcription factors essential for transcription initiation. For example, HIF1 cooperates with the ATF-1/CREB-1 factor in lactate dehydrogenase A gene transcription (53), with AP-1 binding factors in the VEGF gene transcription (54) and with the orphan receptor hepatic nuclear factor 4 in the erythropoietin HRE (55). In all these cases, the interaction of HIF1␣ and other transcription factors is mediated by binding of HIF1 complex with p300 (56). In addition, multimerization of the core consensus sequence is required to form a functional HRE in some genes. For example, in glucose transporter gene and several glycolytic enzymes more than two adjacent core sequences form functional HREs (57). In our promoter sequence analysis, we identified only a single consensus putative HRE site without flanking p300-binding sites or other possible transcription factor-binding sites in the close vicinity in the nmnat promoter. Therefore, it is likely that a lack of additional transcription factor-binding sites or a lack of multimerization of the core sequence renders this putative HRE nonfunctional.