Comparative Metabolomic Profiling Reveals That Dysregulated Glycolysis Stemming from Lack of Salvage NAD+ Biosynthesis Impairs Reproductive Development in Caenorhabditis elegans*

Background: Loss of NAD+ salvage biosynthesis causes a reproductive delay in C. elegans. Results: Loss of salvage biosynthesis causes cytoplasmic, but not mitochondrial, NAD+ deficiency-related metabolic deficits. Conclusion: Salvage synthesis is required for glycolysis, and in the absence of glycolysis, reproductive progression cannot be sustained. Significance: We elucidate the developmental roles of NAD+ biosynthetic pathways and the effects of interventions targeting NAD+ salvage biosynthesis. Temporal developmental progression is highly coordinated in Caenorhabditis elegans. However, loss of nicotinamidase PNC-1 activity slows reproductive development, uncoupling it from its typical progression relative to the soma. Using LC/MS we demonstrate that pnc-1 mutants do not salvage the nicotinamide released by NAD+ consumers to resynthesize NAD+, resulting in a reduction in global NAD+ bioavailability. We manipulate NAD+ levels to demonstrate that a minor deficit in NAD+ availability is incompatible with a normal pace of gonad development. The NAD+ deficit compromises NAD+ consumer activity, but we surprisingly found no functional link between consumer activity and reproductive development. As a result we turned to a comparative metabolomics approach to identify the cause of the developmental phenotype. We reveal widespread metabolic perturbations, and using complementary pharmacological and genetic approaches, we demonstrate that a glycolytic block accounts for the slow pace of reproductive development. Interestingly, mitochondria are protected from both the deficiency in NAD+ biosynthesis and the effects of reduced glycolytic output. We suggest that compensatory metabolic processes that maintain mitochondrial activity in the absence of efficient glycolysis are incompatible with the requirements for reproductive development, which requires high levels of cell division. In addition to demonstrating metabolic requirements for reproductive development, this work also has implications for understanding the mechanisms behind therapeutic interventions that target NAD+ salvage biosynthesis for the purposes of inhibiting tumor growth.


Temporal developmental progression is highly coordinated in
Caenorhabditis elegans. However, loss of nicotinamidase PNC-1 activity slows reproductive development, uncoupling it from its typical progression relative to the soma. Using LC/MS we demonstrate that pnc-1 mutants do not salvage the nicotinamide released by NAD ؉ consumers to resynthesize NAD ؉ , resulting in a reduction in global NAD ؉ bioavailability. We manipulate NAD ؉ levels to demonstrate that a minor deficit in NAD ؉ availability is incompatible with a normal pace of gonad development. The NAD ؉ deficit compromises NAD ؉ consumer activity, but we surprisingly found no functional link between consumer activity and reproductive development. As a result we turned to a comparative metabolomics approach to identify the cause of the developmental phenotype. We reveal widespread metabolic perturbations, and using complementary pharmacological and genetic approaches, we demonstrate that a glycolytic block accounts for the slow pace of reproductive development. Interestingly, mitochondria are protected from both the deficiency in NAD ؉ biosynthesis and the effects of reduced glycolytic output. We suggest that compensatory metabolic processes that maintain mitochondrial activity in the absence of efficient glycolysis are incompatible with the requirements for reproductive development, which requires high levels of cell division. In addition to demonstrating metabolic requirements for reproductive development, this work also has implications for understanding the mechanisms behind therapeutic interventions that target NAD ؉ salvage biosynthesis for the purposes of inhibiting tumor growth.
NAD ϩ , which is synthesized primarily from vitamin B 3 , is a key cellular metabolite. It is well recognized for its role in redox metabolism and has garnered renewed attention as an obligate co-substrate for NAD ϩ consumer enzymes that mediate critical cell stress responses and signaling events (for recent reviews, see Refs. 1 and 2). There is widespread interest in pharmacological manipulation of NAD ϩ consumer activities because of their roles in health span and disease. As a result, NAD ϩ biosynthetic processes, which promote activity of NAD ϩ consumers, have been examined as drug targets (3)(4)(5). There is interest in both boosting NAD ϩ biosynthesis for therapeutic benefit in age-related diseases (6) and inhibiting NAD ϩ biosynthesis for therapeutic benefit in cancer (7)(8)(9). However, we have not fully elucidated how perturbing the availability of specific NAD ϩ biosynthetic precursors or the availability of NAD ϩ in various tissues or cellular compartments impacts organism physiology, especially beyond the activity of the NAD ϩ consumers. For example, although we have largely eradicated pellagra, the deadly disease resulting from dietary deficiency of vitamin B 3 , via supplementation of our food supply, we still lack a basic understanding of the etiology of the hallmark skin, digestive, and nervous system pathologies that are presumably due, at least in part, to a deficiency in NAD ϩ biosynthesis (10,11). Furthermore, the mechanisms leading to toxicity upon treatment with inhibitors of NAD ϩ salvage biosynthesis are not well elucidated (9), and effects on development are largely unexamined.
We use the genetically tractable Caenorhabditis elegans system as a model to probe the roles of NAD ϩ biosynthetic pathways and metabolites in whole organism development and physiology. We have revealed that the reproductive system as well as muscle development and function are particularly sensitive to loss of ability to synthesize NAD ϩ via a salvage pathway from nicotinamide and that these phenotypes are caused by the tissue-specific effects of both nicotinamide and nicotinic acid levels (12)(13)(14). In this study we demonstrate that salvage synthesis has a mild global negative effect on NAD ϩ levels but that mitochondrial function is preserved when salvage synthesis from nicotinamide (NAM) 2 is blocked. Although the overall effect on NAD ϩ levels upon loss of salvage biosynthesis is mild, we show that efficient glycolysis depends on salvage synthesis from NAM and that the reproductive development phenotype in the NAD ϩ salvage synthesis mutant results from the block in glycolysis.
In eukaryotes, NAD ϩ is synthesized from dietary-derived and salvaged forms of vitamin B 3 such as nicotinamide and nicotinic acid and from tryptophan (15,16) (Fig. 1A). Nicotinic acid is processed by the highly conserved Preiss-Handler (17,18) pathway to synthesize NAD ϩ . Dietary nicotinamide riboside (NR) or nicotinic acid riboside are used as substrates by nicotinamide riboside kinase to produce intermediates that are processed by Preiss-Handler pathway enzymes to synthesize NAD ϩ (19,20). The hydrolysis of NAD ϩ by NAD ϩ consumer enzymes produces NAM, which is used to resynthesize NAD ϩ via the salvage pathway. In vertebrates, salvage biosynthesis requires the enzyme nicotinamide phosphoribosyltransferase (NAMPT), which produces nicotinamide mononucleotide (NMN) from NAM. NMN is then converted directly to NAD ϩ by nicotinamide mononucleotide adenylyltransferase (NMAT), the penultimate enzyme of the Preiss-Handler pathway (21). However, invertebrates, including C. elegans, use an alternative salvage pathway whereby NAM is converted to nicotinic acid (NA) by a nicotinamidase, encoded by the pnc-1 gene in C. elegans, and fed into the beginning of the Preiss-Handler pathway (Fig. 1A) (21). Although NAMPT and nicotinamidase are distinct enzymes and co-exist only in select organisms (15), they have biologically comparable functions in consuming NAM and producing NAD ϩ . Human NAMPT can partially compensate for the loss of the C. elegans nicotinamidase pnc-1, demonstrating an equivalent biological role for the two enzymes (12). Finally NAD ϩ is produced from the essential amino acid tryptophan through the kynurenine pathway, which produces quinolinic acid that is converted to nicotinic acid mononucleotide via quinolinic acid phosphoribosyltransferase (21). The C. elegans genome does not encode an apparent homolog of quinolinic acid phosphoribosyltransferase (12,15). Thus, whether tryptophan can be used to synthesize NAD ϩ in C. elegans is unclear.
NAD ϩ is required as an enzyme co-factor in the cytoplasm and the mitochondria. It also has signaling roles in these compartments as well as in the nucleus and the extracellular space. Recognition of these various roles has created new questions about the bioavailability of NAD ϩ in various compartments and how the molecule is partitioned. Subcellular localization patterns of NAD ϩ biosynthetic enzymes suggest that different cellular compartments rely on different precursors for NAD ϩ biosynthesis. It appears that NAM, NA, NR, and nicotinamide riboside can be used for NAD ϩ biosynthesis in the cytoplasm, whereas NMN may be the sole precursor for mitochondrial NAD ϩ biosynthesis (22). The presence of distinct biosynthetic pathways combined with the lack of movement of NAD ϩ between compartments suggests that cells may separately regulate NAD ϩ bioavailability in a particular compartment and/or respond in a distinct manner to a perturbation in NAD ϩ bioavailability in one compartment versus another (23). Compartment-specific differences in NAD ϩ levels have been observed. For example, mitochondria in cardiac and other myocytes are capable of storing a much higher concentration of NAD ϩ than the cytoplasm or nucleus, and neurons have been shown to have more equal proportions between the mitochondria and the rest of the cell (23-26) Compartmentalization of NAD ϩ homeostasis also provides a mechanism for cells to link distinct metabolic activities to distinct NAD ϩ signaling activities (27). In this study, we present evidence that compromised salvage synthesis preferentially affects NAD ϩ availability outside the mitochondrial compartment.
Mutation of the C. elegans pnc-1 nicotinamidase gene causes a variety of developmental and physiological defects (12,13,28). These phenotypes comprise three distinct classes: those that can be mimicked by supplementation of wild-type animals with the PNC-1 substrate NAM, those that are rescued by supplementation of mutants with the PNC-1 product NA, and those where both lack of NA production and accumulation of substrate make a contribution to the phenotypic outcome (12,13). In particular pnc-1 mutants have a delay in development of the gonad relative to the soma; the gonad develops more slowly than expected and no longer displays the characteristic synchrony with somatic development that is expected in the highly invariant C. elegans developmental program. Providing NA as a supplement to the pnc-1 mutant cultures rescues this delay in development, suggesting that the lack of NAD ϩ biosynthesis underlies the gonad developmental phenotype. Here we further investigated this hypothesis by measuring metabolite levels in wild-type and mutant animals as well as in animals treated with conditions predicted to alter NAD ϩ levels. Our data support the hypothesis that a specific lack of NAD ϩ bioavailability is the underlying cause of the reproductive developmental delay and reveal that this deficit appears to affect the nucleo-cytoplasmic compartment preferentially. We use a metabolomics approach to identify perturbations specifically linked to the change in NAD ϩ availability and to investigate the mechanism causing the reproductive delay.

C. elegans Culture and Strains
Strains were maintained under standard conditions with Escherichia coli OP50 as food at 20°C (29) unless otherwise specified. N2 is the wild-type strain. UV-irradiated OP50 plates were prepared as described (12), and killing was confirmed by streaking to LB agar.

Metabolomics
We performed global LC-MS and GC-MS analysis by contracting with Metabolon, Inc. 30 -50 l of worms were collected in M9 buffer (5 biological replicates per sample), flashfrozen in liquid nitrogen, and stored at Ϫ80°C before shipping to Metabolon on dry ice. Samples were processed using the automated MicroLab STAR system from Hamilton Co., where recovery standards were added to each sample, and a proprietary series of organic and aqueous extractions was conducted. Extracts from each sample were divided into two fractions for LC-MS and for GC-MS. Extracts were then frozen and dried under a vacuum. A Waters ACQUITY UPLC and a Thermo-Finnigan LTQ mass spectrometer were used for LC-MS. For acidic conditions, extracts were eluted with 50/50 water/methanol containing 0.1% formic acid, whereas the basic extracts used 50/50 water/methanol with 6.5 mM ammonium bicarbonate. GC-MS was performed on a Thermo-Finnigan Trace DSQ fast-scanning single-quadrupole mass spectrometer using electron impact ionization. The samples were re-dried and derivatized under dried nitrogen using bistrimethyl-silyl-triflouroacetamide (BSTFA). The GC column was 5% phenyl, and the temperature ramp was from 40 to 300°C in a 16-min period. Compounds from LC-MS and GC-MS were identified by comparison to Metabolon's library of standards. After correcting metabolite levels to protein concentrations determined by Bradford assays, minimum observed values were impute for each metabolite. In the data analysis, metabolite levels in all samples were normalized to the average of N2 samples under standard conditions. Welch's two sample t test was used to identify metabolites that differed significantly between experimental groups.
We performed targeted LC-MS metabolomics analysis with the Metabolomics Core Facility at Penn State to measure NAD ϩ and glycolytic intermediates. ϳ50 l of worms were collected in double distilled H 2 O, flash-frozen in liquid nitrogen, and stored at Ϫ80°C. 15-l samples were extracted in 1 ml of 3:3:2 acetonitrile:isopropyl alcohol:H 2 O with 1 M chlorpropamide as the internal standard. Samples were homogenized using a Precellys™ 24 homogenizer. Extracts from samples were dried under a vacuum, resuspended in HPLC Optima Water (Thermo Scientific), and divided into two fractions, one for LC-MS and one for BCA protein analysis. Samples were analyzed by LC-MS using a modified version of an ion-pairing reversed phase negative ion electrospray ionization method (31). Samples were separated on a Supelco (Bellefonte, PA) Titan C18 column (100 ϫ 2.1-mm 1.9-m particle size) using a water-methanol gradient with tributylamine added to the aqueous mobile phase. The LC-MS platform consisted of Dionex Ultimate 3000 quaternary HPLC pump, 3000 column compartment, 3000 autosampler, and an Exactive plus orbitrap mass spectrometer controlled by Xcalibur 2.2 software (all from ThermoFisher Scientific, San Jose, CA). The HPLC column was maintained at 30°C and a flow rate of 200 l/min. Solvent A was 3% aqueous methanol with 10 mM tributylamine and 15 mM acetic acid; solvent B was methanol. The gradient was 0 min, 0% B; 5 min, 20% B; 7.5 min, 20% B; 13 min, 55% B; 15.5 min, 95% B, 18.5 min, 95% B; 19 min, 0% B; 25 min 0% B. The orbitrap was operated in negative ion mode at a maximum resolution (140,000) and scanned from m/z 85 to m/z 1000. Metabolite levels were corrected to protein concentrations determined by BCA assay (Thermo Fisher).

NAD/NADH Ratio
300 l of mixed stage N2 or pnc-1 animals (cultured on live or UV-killed OP50) were collected in M9, snap-frozen in liquid nitrogen, and stored at Ϫ80°C. 50 l of thawed sample was added to duplicate wells of a black 96-well plate. NAD and NADH measurements were performed using the Elite Fluorimetric NAD/NADH Ratio Assay kit (eENZYME, LLC, Gaithersburg, MD) according to the manufacturers' instructions.

Phenotypic Analysis
Gonad Development-Mid-L4 stage animals with an open lumen in the vulva and the uterus are normal. "Delayed" animals are those that do not yet have an open uterine lumen when the vulval lumen achieves its characteristic mid-L4 morphology (12).
Mitochondrial Sensitivity-For the paralyzing assay we treated L4 animals that were aged 24 h with 20 mM sodium azide (Sigma) in a depression slide at time 0. The number of animals that were moving was recorded every 5 s until all animals were paralyzed. For the recovery assay we treated the animals with 5 mM sodium azide for 15 min (all animals were paralyzed). We replaced the sodium azide with M9 buffer at time 0 and recorded the number of animals that had started moving again every 5 min until all worms recovered. All assays were repeated at least three times. Thrashing assays were performed as previously described (13).
O 2 Consumption and Heat Production Methods-Synchronized day 2 adults were collected and washed in S-basal medium. 400 -600 l of worm suspension (dependent on worm number) was used for live oxygen consumption (928 sixchannel oxygen system, Strathkelvin Instruments) and heat production measurements (2277 Thermal Activity Monitor, Thermometric). To prevent death of the pnc-1 mutants from internal rupture due to inability to lay eggs, 300 M fluorodeoxyuridine (FUdR) was supplemented to pnc-1 mutants and N2 controls beginning at the L4 stage. Heat production and oxygen consumption data were normalized to total protein determined by BCA assay of a parallel aliquot of worm suspension. All essays were repeated three times independently.
sod-3::GFP Expression-L4 animals were placed on fresh plates and imaged on a Nikon SMZ1500 stereoscope. Images were collected and analyzed using NIS-Elements software from Nikon. The pharynx of each animal was selected, and the mean GFP signal for each animal was calculated.

Statistical Analysis
Fisher's exact test was carried out to determine p values in gonad development, thrashing, and sodium azide treatment assays. For quantification of sod-3::GFP intensities, oxygen consumption, and heat production assays, p values were calculated using Student's t test. In LC-MS and GC-MS analysis we used Welch's two sample t test to calculate p values. In all figures: §, 0.05 Ͻ p Ͻ 0.1; *, 0.01 Ͻ p Ͻ 0.05; **, 0.001 Ͻ p Ͻ 0.01; ***, p Ͻ 0.001.

Loss of Salvage NAD ؉ Biosynthesis Resulted in Minor Global Depletion of NAD ؉
We used LC-MS to investigate if salvage biosynthesis of NAD ϩ from NAM is affected by loss of pnc-1 function using the pnc-1(pk9605) allele (12). pnc-1(pk9605) mutants showed a 19-fold increase in substrate NAM levels and an 11-fold decrease in NA product levels relative to wild-type animals ( Fig.  1, B and C). The C. elegans genome encodes a second nicotinamidase called PNC-2. Kinetically, PNC-2 is Ͻ10% as efficient as PNC-1 (34). Moreover, RNAi of pnc-2 did not produce any obvious phenotypes in wild type or any synthetic phenotypes in pnc-1(pk9605) nor did it exacerbate any pnc-1(pk9605) phenotypes (Ref. 12 and data not shown). These observations combined with the metabolite measurements demonstrating dramatic changes in NAM and NA in the pnc-1 mutant support the conclusion that PNC-2 makes little contribution to NAD ϩ salvage biosynthesis and that the pnc-1(pk9605) allele effectively prevents conversion of NAM to NA, blocking C. elegans NAD ϩ salvage biosynthesis from NAM.
To evaluate the effect of loss of NAD ϩ salvage biosynthesis on global levels of NAD ϩ , we measured and compared NAD ϩ levels in wild-type and mutant whole animal lysates. pnc-1(pk9605) have an ϳ30% decrease in NAD ϩ levels compared with wild type (Fig. 1D). We also measured NAD/NADH ratios and found that pnc-1 mutants have no difference in ratio relative to wild type (Fig. 1E). We conclude that NAD ϩ salvage biosynthesis from NAM makes a distinct but minor contribution to global NAD ϩ production in whole C. elegans animals but did not affect the NAD/NADH ratio.

Manipulation of NAD ؉ Levels Affected Reproductive Developmental Progression
We previously demonstrated that supplementation of pnc-1 mutant cultures with NA, NMN, or NAD ϩ as well as transgenic expression of human NAMPT in the mutants, which provides an alternative route for salvage of NAM, could rescue the pnc-1 gonad developmental delay (12,13). These results suggest that a lack of NAD ϩ availability is causative of the reproductive delay. However, our measurements revealed only a mild deficit of NAD ϩ in the pnc-1 mutants. If a deficit in NAD ϩ biosynthesis is causative of a gonad developmental delay, we predict that treatments that increase NAD ϩ levels would rescue the phenotype. In contrast, treatments that reduce NAD ϩ levels are predicted to exacerbate the gonad delay of pnc-1 or even cause a gonad delay in otherwise normal animals.
We investigated the consequences of increasing NAD ϩ levels in the pnc-1 mutant using two strategies. First, we supplemented cultures with a precursor for NAD ϩ biosynthesis that is processed via a pathway other than NAM salvage. Supplementation experiments were conducted on UV-killed E. coli OP50 strain as a food source to prevent metabolism of the supplement by the E. coli. Supplementation of cultures with 1.25 mM NR effectively rescued the reproductive delay in pnc-1 mutants ( Fig. 2A), and we confirmed that NR-supplementation elevates NAD ϩ levels ( Fig. 2B) as previously reported (35,36). We also measured NAD ϩ levels in pnc-1 mutants that were supplemented with NA, which rescues the gonad delay (12). We found a trend toward an increase in NAD ϩ levels with an average 70% increase; however, there was no significant difference upon statistical analysis (p ϭ 0.147) (Fig. 2B). Second, we used a genetic strategy to reduce consumption of NAD ϩ in vivo. PARPs are major consumers of NAD ϩ , and an increase in NAD ϩ has been observed in pme-1/PARP loss-of-function mutants (36). The reproductive delay associated with pnc-1(pk9605) is alleviated by the presence of a pme-1 mutation (Fig. 2C), and we confirmed higher NAD ϩ levels in the pme-1; pnc-1 double mutants (Fig. 2D).
We next tested the consequences of decreasing NAD ϩ availability. We predicted that knockdown of other enzymes in the NAD ϩ salvage/Preiss-Handler pathway should both reduce NAD ϩ availability and mimic the pnc-1 gonad development phenotype. RNAi of nprt-1 or qns-1 each recapitulate a pnc-1like gonad developmental delay in a portion of the population (Figs. 3A and 4). Although the NAD ϩ levels trend toward lower average levels (Fig. 3B), the differences were not statistically significant.
We also tested the effects of reducing nicotinamide mononucleotide adenyltansferase activity. A predicted null allele of nmat-2, tm2905, has a severe gonad development phenotype (Fig. 3C). We could not measure NAD ϩ levels in the nmat-2(tm2905) mutant because the homozygous animals are sterile (data not shown), precluding our ability to collect the thousands of animals required for LC-MS analysis. However, we used this allele to demonstrate that the NR-mediated rescue of gonad developmental delay is because of activity as an NAD ϩ biosynthetic precursor. We repeated the NR supplementation experiment in the nmat-2(tm2905) mutant, which should be compromised in ability to use NR to synthesize NAD ϩ . As expected, the nmat-2 mutants are not rescued upon supplementation with NR (Fig. 3C).
The correlation between a mild deficit in NAD ϩ bioavailability and a gonad developmental delay is further supported by the identification of culture conditions that simultaneously increase the penetrance of the gonad delay phenotype and exacerbate the NAD ϩ deficiency in the pnc-1 mutant. We previ-FIGURE 1. Loss of PNC-1 function affects NAM, NA, and NAD ؉ levels. A, NAD ϩ biosynthetic pathways in C. elegans. Salvage NAD ϩ biosynthesis starts with the nicotinamidase PNC-1 (red), which converts NAM to NA. NA is further converted to NAD ϩ through the Preiss-Handler pathway (orange). Nicotinamide riboside kinase converts NR to NMN (green), which is processed by NMAT-1 and/or NMAT-2 to produce NAD ϩ . De novo synthesis produces NAD ϩ from Trp (blue). It is unclear if C. elegans produces NAD ϩ de novo from Trp because of the lack of a quinolinic acid phosphoribosyltransferase homolog in the genome. NaMN, nicotinic acid mononucleotide; NaAD, nicotinic acid adenine dinucleotide. ously reported that culturing animals on UV-killed E. coli exacerbates the pnc-1 gonad developmental defects relative to growth on live food (e.g. compare phenotype penetrance in Fig.  2, A and C) (12), and we show here that the UV-killed conditions also further depress NAD ϩ levels (Fig. 1D), still without affecting the NAD/NADH ratio (Fig. 1E). Together our exper-iments are consistent with the conclusion that lack of NAD ϩ salvage biosynthesis from NAM results in a mild deficit in NAD ϩ bioavailability, and this deficit in the availability of NAD ϩ itself and not, for example, other intermediate metabolites in the biosynthetic pathway, is causative of a developmental delay in the reproductive system. Increasing NAD ؉ levels in pnc-1 mutants rescues gonad developmental delay. Supplementation of 1.25 mM NR to pnc-1(pk9605) mutants effectively rescues the gonad developmental defects (A) and raises NAD ϩ levels (B). pme-1(ok988) loss-of-function also rescues the gonad developmental defects of a significant percentage of pnc-1 mutants (C) and moderately increases NAD ϩ levels (D). NAD ϩ levels were measured using LC-MS. In histograms, error bars are S.E. ***, p Ͻ 0.001, calculated using Fisher's exact test. Box plots are as described in Fig. 1. *, 0.01Ͻp Ͻ 0.05; ***, p Ͻ 0.001, calculated with Welch's two sample t test. Animals in A and B were cultured on UV-killed OP50 plates. Animals in C and D were cultured on standard live OP50 plates. FIGURE 3. Inhibiting NAD ؉ biosynthesis causes a gonad developmental delay. A, RNAi of nprt-1 or qns-1 causes gonad developmental delay. B, NAD ϩ levels in nprt-1 or qns-1 RNAi animals showed a mild decreasing trend compared with animals on control RNAi; however, the difference is not statistically significant. C, nmat-2 mutants have a delay in gonad development, which cannot be rescued by 1.25 mM NR supplementation. In histograms, error bars are S.E. ***, p Ͻ 0.001, calculated using Fisher's exact test. Box plots are as described in Fig. 1. Animals in A and B were cultured on E. coli strain HT115 for RNAi. Animals in C were cultured on UV-killed OP50 plates.

PNC-1 Loss-of-function Inhibited Sirtuin Activity but This Effect Did Not Cause the Developmental Phenotype
We next considered the question of how a decrease in biosynthesis of NAD ϩ might result in a reproductive delay. We investigated several hypotheses to relate NAD ϩ bioavailability to the developmental phenotype. First, we predicted that lack of NAD ϩ bioavailability would compromise NAD ϩ consumer activity (37) and that reduced consumer activity might underlie the phenotype. We specifically examined sirtuin NAD ϩ consumer activity by using a sod-3::GFP reporter as a readout of SIR-2.1 activity. Expression of sod-3::GFP was elevated by overexpression of SIR-2.1 (Fig. 5, A and B, and Berdichevsky et al. (30)). As predicted, SIR-2.1-mediated up-regulation of sod-3::GFP required pnc-1 (Fig. 5, A and B), supporting the conclusion that SIR-2.1 activity is reduced in the absence of PNC-1.
To look for a functional relationship between reduced sirtuin activity and reproductive development, we examined animals carrying a sir-2.1 deletion allele but found no gonad developmental delay (data not shown). We also functionally examined sir-2. . We have already shown that a reduction in PME-1/ PARP1 activity can ameliorate the gonad delay instead of mimicking the phenotype (Fig. 2C), suggesting that a lack of PARP activity is not causing the phenotype. Thus, although our experimental results are consistent with a reduction of SIR-2.1 activity, which could reflect impaired NAD ϩ bioavailability in the nucleo-cytoplasmic compartment or increased levels of sirtuin inhibitor NAM (38 -40), we found no evidence that reduced NAD ϩ consumer activity underlies the reproductive phenotype of pnc-1 mutants.

Mitochondrial Activity in pnc-1 Mutants Is Not Impaired
Mitochondrial NAD ϩ represents a substantial portion of the cellular NAD ϩ pool (23,24) where it is used for important energetic processes such as the TCA cycle and oxidative phosphorylation. We next sought to examine mitochondrial NAD ϩ -mediated processes for evidence of limited NAD ϩ bioavailability in the pnc-1 mutants. Simultaneously, we aimed to test our second hypothesis to explain the reproductive delay of pnc-1 mutants, which is that gonad development is an energetically demanding process, and a lower rate of energy production, perhaps due to a lack of mitochondrial NAD ϩ availability, could negatively influence gonad development. We directly tested oxygen consumption as a measure of oxidative phosphorylation and found no difference between pnc-1 mutants and wildtype animals (Fig. 6A). Heat production is also unchanged (Fig.  6B), suggesting that the overall metabolic rate is unchanged. Thus, a simple lack of energy provided by the mitochondria to sustain gonad development is not an adequate explanation of the reproductive developmental delay.

Metabolomics Analysis of pnc-1 Mutants
NAD ϩ is a major hub in the metabolic network (15), and our data are consistent with a mild, but functionally relevant, deficit of NAD ϩ bioavailability in the nucleo-cytoplasmic compartment. We hypothesized that multiple processes in addition to NAD ϩ consumer activity would be affected by a deficit in NAD ϩ availability. To seek further evidence of perturbation of NAD ϩ biosynthesis and to reveal specific changes that might influence gonad development, we used a metabolomics approach to measure steady-state levels of hundreds of metabolites in wild type and pnc-1 mutants. The loss of pnc-1 activity is expected to have metabolic changes beyond those directly affected by compromised NAD ϩ production. For example, we have demonstrated phenotypic consequences to the increase in NAM (12,13). To allow us to identify those perturbations specifically relevant to compromised NAD ϩ biosynthesis and to the delay in gonad development, we also profiled the mutant under two other conditions. We profiled pnc-1 mutants supplemented with NA; we would expect to see rescue of relevant perturbations in this condition. We also profiled pnc-1 mutants on UV-killed E. coli where we would expect to see exacerbation of relevant perturbations.
We profiled five biological replicates for each condition. 361 named metabolites were identified in our analysis (supplemental Table S1). Of these, 85 (p Ͻ 0.05) were altered by mutation of pnc-1 and another 38 trended toward a change (0.05 Ͻ p Ͻ 0.10) (Table 1, Fig. 7). These 123 metabolites ranged across all subtypes of metabolites examined and are involved in a wide variety of metabolic pathways ( Table 1). Supplementation of mutant animals with NA reversed the pnc-1-induced change of 62 of these metabolites (Table 1, Fig. 7).

Metabolomics Results Are Consistent with Functional Experiments Involving Mitochondria and Sirtuins
In support of our conclusion that mitochondrial respiration is not compromised in the pnc-1 mutant, TCA cycle metabolites are not perturbed (supplemental Table S1). The only change in TCA cycle metabolites upon loss of pnc-1 was a trend toward more ␣-ketoglutarate (0.05Ͻp Ͻ 0.1), but this phenotype did not correlate with a lack of NAD ϩ biosynthesis or the reproductive development phenotype because it was neither rescued by supplementation with NA nor exacerbated by UVkilled E. coli (Table 1).
Consistent with the detected reduction in sirtuin activity, acetyllysine levels are elevated in pnc-1 mutants (Table 1 and Fig. 5C). Interestingly, just as reduced sirtuin activity does not functionally correlate with gonad development, the increase in acetyllysine does not correlate with lack of NAD ϩ biosynthesis or the gonad development phenotype because it was neither rescued by supplementation with NA nor exacerbated by UVkilled food (supplemental Table S1).

A Minority of the Metabolic Changes Associated with pnc-1 Correlate Well with the Reproductive Development Phenotype
Amino acid and Dipeptide Metabolites-Loss of pnc-1 function altered approximately one-third of the detected amino acid-related metabolites (supplemental Table S1). The changes reversed by supplementation with NA included positive and negative changes in various amino acid metabolic pathways ( Table 1). Four of these changes (saccharopine, ␣-hydroxyisocaproate, 2-hydroxybutyrate, and N-acetylornithine) were exacerbated by UV-killed E. coli, but no particular pathway was prominent (Table 1). Thus, amino acid metabolites did not become a primary focus for investigating the reproductive delay. Twenty dipeptides out of 101 examined were altered by loss of pnc-1 (supplemental Table S1), and NA rescued 15 of these, which were all altered in the same direction, displaying an increase upon loss of pnc-1 (Table 1). However, none of these changes were exacerbated by UV-killed E. coli (supplemental Table S1).
Fatty Acid Metabolism-Notably, inactivation of pnc-1 caused increased levels of 8 of 10 detected acylglycerophosphocholines as well as choline phosphate (supplemental Table S1). The changes in choline phosphate as well as three of eight acylglycerophosphocholines were rescued by supplementation with NA (Table 1), suggesting a possible link to NAD ϩ biosynthesis. However, these effects were not significantly exacerbated by UV-killed E. coli (supplemental Table S1) and thus do not meet our criteria for candidates with the most direct functional association to the reproductive delay phenotype.
Of the 33 changes in fatty-acid-related metabolites associated with loss of pnc-1, NA reversed 10 (Tables 1 and supplemental Table S1). And only three (the long chain fatty acid oleate, 2-hydroxyglutarate, and myo-inositol) had changes that were exacerbated by UV-killed food (supplemental Table S1). Thus, no specific aspect of fatty acid metabolism stood out as a strong candidate for association with the gonad developmental delay.
Nucleotide Metabolism-Numerous nucleotide metabolites were both increased and decreased by loss of pnc-1, and strikingly, most of these changes were reversed by supplementation of cultures with NA, suggesting an association with NAD ϩ biosynthesis (Tables 1 and supplemental Table S1). However, only one metabolite, 3-aminoisobutyrate, had a pnc-1-mediated change that was exacerbated on UV-killed E. coli. Thus, we did not focus on nucleotide-related metabolites in our search for a cause for the developmental delay.
Carbohydrate Metabolism-Carbohydrate metabolism is the category in which the largest percentage of measured metabolites were altered by pnc-1, and more than half of these changes were restored by NA supplementation (Tables 1 and supplemental Table S1). pnc-1 mutants have accumulation of a number of mono-, di-, and tri-saccharides (Tables 1 and supplemental Table S1), which are glycogen breakdown intermediates used for energy storage. Glycolysis was the only specific pathway where more than one of the changes induced by loss of pnc-1 was both rescued by NA and exacerbated by UV-killed

TABLE 1 List of metabolites with altered levels in pnc-1 mutants relative to wild type
Metabolites highlighted in green are those whose levels are rescued in pnc-1 mutants with NA supplementation. Metabolites highlighted in blue are those whose levels are both rescued in NA-supplemented pnc-1 mutants and exacerbated in pnc-1 mutants cultured on UV-killed OP50 plates. Blue metabolites met all criteria to be considered candidates for investigating a relationship to the reproductive phenotype. *, p Ͻ 0.05; §, 0.05 Ͻ p Ͻ 0.1; calculated with Welch's two sample t test. (Table 1). Thus, we investigated the functional relationship between carbohydrate metabolism and gonad developmental delay.

Perturbations in Glycolysis Cause the Gonad Developmental Defects in pnc-1 Mutants
Loss of pnc-1 function caused accumulation of glucose as well as glucose 6-phosphate and dihydroxyacetone phosphate (DHAP), which are glycolytic intermediates before the step that requires NAD ϩ (Table 1 and Fig. 8). In contrast, we observed a decrease in 3-phosphoglycerate (3PGA) (0.05 Ͻ p Ͻ 0.10), subsequent to the NAD ϩ -dependent step (Table 1 and Fig. 8). Each of these perturbations is rescued by supplementation with NA (Table 1 and Fig. 8), suggesting that the detected disruption to glycolysis can be rescued by restoring salvage NAD ϩ biosynthesis. Although exacerbation by UV-killed E. coli was statistically detected for only DHAP (Table 1 and Fig. 8), each of these glycolytic intermediates shows trends in the appropriate direction for association with the reproductive delay on UV-killed E. coli (Fig. 8). Pyruvate is an exception to the pattern; its levels are increased in the pnc-1 mutant (Table 1, see "Discussion"). As further evidence for the relevance of the changes in glycolytic metabolite levels, we also compared levels in pnc-1 and NR-supplemented pnc-1. We found that NR supplementation had similar effects to NA supplementation in lowering the levels of a metabolite before the NAD ϩ -dependent step and raising the levels of subsequent glycolytic intermediates (Fig. 9).
To investigate the functional phenotypic consequences of the depletion of the late glycolytic metabolites, we added 3PGA and the subsequent metabolite PEP as supplements in the culture media. Both 3PGA and PEP significantly increased the percentage of the pnc-1 population with normal gonad development relative to supplementation of glucose as a control (Fig.  10A). These results suggest that efficient glycolytic output is sufficient to permit normal progression of gonad development in the pnc-1 mutant.
To further investigate the relationship of glycolysis to gonad developmental progression, we also examined the effects of experiments designed to diminish the glycolytic rate. For this experiment we used RNAi of genes encoding the glycolytic enzymes phosphofructokinase (pfk-1.1 and pfk-1.2) and triosephosphate isomerase (tpi-1). Further inhibition of glycolysis would be expected to exacerbate the gonad developmental defects of pnc-1 mutants and/or cause a gonad developmental delay in wild-type animals. We found that RNAi of each of the three glycolytic enzyme genes causes a gonad developmental delay only in the pnc-1 mutant background (Fig. 10B). These results support a role for compromised glycolysis in contributing to the gonad delay of pnc-1 mutants and suggest that although restoration of glycolysis is sufficient to rescue pnc-1, blocking glycolysis is not sufficient to cause a gonad developmental delay in an otherwise wild-type background.

pnc-1 Mutants Are More Sensitive to Disruption of Mitochondrial Function
Our data show that although the pnc-1 mutants have reduced glycolytic output, mitochondrial function is maintained, suggesting that other metabolic changes compensate for reduced glycolytic output to provide fuel for mitochondria. This working model suggests that the pnc-1 mutants would be hypersensitive to disruption in mitochondrial function relative to wildtype animals, which maintain the flexibility of producing energy via glycolysis. We tested this hypothesis using sodium azide to disrupt mitochondrial function in pnc-1 mutants and wild-type animals and via RNAi of genes required for oxidative phosphorylation. We found that pnc-1 mutants paralyze more quickly in response to sodium azide and recover more slowly from sodium azide treatment compared with wild-type animals (Fig. 11, A and B). Muscle function of pnc-1 mutants is also more sensitive than wild type to RNAi of genes encoding electron transport chain and ATP synthase subunits (Fig. 11C).

A Deficit in NAD ϩ Biosynthesis Results in a Reproductive
Developmental Delay-NAD ϩ biosynthesis via the salvage pathway is critical for temporal progression of reproductive development in C. elegans. Because knock-out of Nampt is embryonic lethal, loss of salvage biosynthesis in mice likely has developmental phenotypes as well (41). We investigated the physiological mechanisms underlying the developmental phenotype of loss of salvage NAD ϩ biosynthesis in C. elegans. We were surprised to find that NAD ϩ levels were only mildly reduced in the pnc-1 mutants. However, our results are consistent with a minor deficit of NAD ϩ bioavailability as a cause of the gonad phenotype based on two lines of investigation. First, every condition predicted to boost NAD ϩ levels ameliorates the gonad delay phenotype, including supplementation with NA, NR, NMN, or NAD ϩ as well as replacing pnc-1 with NAMPT or reducing PARP function (data presented here and Vrablik et al. (12)). We also specifically show that supplemen-tation with NR and loss of pme-1/PARP1, which each rescue the gonad delay, simultaneously boost NAD ϩ levels. Second, experiments designed to lower NAD ϩ levels recapitulate the phenotype. Reducing the activity of each of the other enzymes in the salvage/Preiss-Handler pathway mimics the phenotype, suggesting that loss of NAD ϩ itself causes the phenotype as opposed to perturbation of any other specific intermediate metabolite. We also measured NAD ϩ levels in these experiments designed to deplete NAD ϩ , and although we did not find a significant difference in average levels, the averages did trend lower, and the minimum and maximum measurements were lower as well. The lack of a significant difference could reflect the experimental protocol involving RNAi that is expected to reduce as opposed to remove gene function, which is achieved with the pnc-1 null allele (12). Nonetheless, it suggests that the reproductive system is sensitive to even a minor decrease in NAD ϩ availability.
Role of Salvage Synthesis in Global NAD ϩ Production-Our data show that loss of pnc-1 function blocks the recycling of  NAM produced by NAD ϩ consumers and drastically reduces the production of the precursor NA for NAD ϩ biosynthesis. Nonetheless, the levels of NAD ϩ are only mildly affected. We conclude that salvage biosynthesis makes a significant but minor contribution to the global levels of organismal NAD ϩ in C. elegans. The NAD/NADH ratio is not perturbed by this mild deficit in global levels of NAD ϩ . It remains possible that salvage from NAM may make a substantial contribution to NAD ϩ production, but in its absence, compensatory production via other pathways or regulatory feedback on NAD ϩ consumption might maintain NAD ϩ levels. No increase in the mononucleotide forms of NAM or NA were detected in pnc-1 mutants (supplemental Table S1), but these steady-state measurements may not give a full picture of compensatory mechanisms that could be in play. Moreover, other pathways for NAD ϩ biosynthesis in C. elegans have not yet been well defined. For example, whether C. elegans have a functional de novo pathway is not yet clear (12,15). Thus, these questions remain to be answered, and metabolic flux experiments will be required to investigate hypotheses about compensatory mechanisms. In mammalian systems the need for various NAD ϩ biosynthetic pathways is usually discussed in tissue-specific terms, and the ability of one pathway to compensate for others is also not explored well at the cellular or organism levels. Technical considerations prevent us from measuring NAD ϩ levels in a tissue-specific manner. However, we do not favor a model where the gonad in particular has less NAD ϩ than other tissues. We have previously shown that restoration of NAD ϩ salvage biosynthesis in limited numbers of cells outside the gonad in the pnc-1 mutant can at least partially rescue the gonad delay (14). These results suggest that NAD ϩ biosynthetic intermediates are likely shared liberally between tissues and that the gonad is adept at collecting them if available.
pnc-1 Mutants Have a Decrease in NAD ϩ Bioavailability Specifically Outside the Mitochondria-We suggest that the decrease in NAD ϩ levels in pnc-1 mutants specifically excludes the mitochondria because we detected perturbations in nucleocytoplasmic NAD ϩ -dependent processes but no perturbation in mitochondrial functions. This conclusion is consistent with previous studies demonstrating that NAD ϩ levels can be regulated independently in specific cellular compartments (22,42). The cytoplasmic NAD ϩ deficit produced by a lack of salvage FIGURE 10. Effects of supplementation with glycolytic intermediates on reproductive development in pnc-1 mutants. A, supplementation of late glycolytic metabolites restores gonad development in pnc-1 mutants. 3PGA and PEP supplementation shows significantly higher efficiency in the rescue of gonad development than glucose (***, p Ͻ 0.001) calculated using Fisher's exact test to compare with ϩglucose. Animals were cultured on UV-killed OP50 plates. B, RNAi of glycolysis genes pfk-1.1, pfk-1.2, and tpi-1 does not cause gonad developmental defects in N2 animals (gray bars) but exacerbates gonad developmental delay in sensitized pnc-1 mutants (white bars) cultured on E. coli strain HT115 where the phenotype is not normally expressed. Error bars are S.E. *, 0.01 Ͻ p Ͻ 0.05; **, 0.001 Ͻ p Ͻ 0.01; calculated using Fisher's exact test.
biosynthesis of NAD ϩ affects the efficiency of glycolysis, likely by reducing the activity of the NAD ϩ -dependent enzyme glyceraldehyde-3-phosphate dehydrogenase. However, this effect is not carried through to the TCA cycle. Similar to our findings, salvage biosynthesis has cytoplasmic-specific effects in cancer cells. Inhibition of NAMPT activity in human cancer cells inhibits glycolysis at the NAD ϩ -dependent step but appears to have little impact on the TCA cycle (7). We conclude that the mitochondria are protected from the effects of loss of salvage NAD ϩ biosynthesis and that NAD ϩ salvage biosynthesis is not critical for maintaining functional levels of NAD ϩ in the mitochondria. The maintenance of mitochondrial activity likely occurs as a result of compensatory metabolic changes that direct carbon for oxidation to the mitochondria through other routes. Consistent with our model of a block in carbohydrate utilization and a potential switch to use of alternative fuels, the major carbohydrate storage molecule, trehalose, is drastically increased in pnc-1 mutants (Table 1).
How Is Mitochondrial Activity Preserved?-Unlike the other late glycolytic metabolites, pyruvate levels are not reduced in pnc-1 mutants relative to wild-type animals ( Table 1). The metabolomics data suggest two explanations for the lack of a decrease in pyruvate even in the presence of reduced glycolytic output. First, lactate levels are reduced ( Table 1). Given that cytoplasmic NADH is required to convert pyruvate to lactate, pyruvate levels are likely maintained because of less conversion to lactate, preserving pyruvate to feed mitochondrial metabolism. Second, the metabolomics data are consistent with a potential increase in amino acid catabolism, supplying carbon to the mitochondrial TCA cycle via pyruvate and potentially other metabolites. Both ␣-ketoglutarate and glutamate, which are used in conversion of glucogenic amino acids to pyruvate, are increased (Table 1). Also consistent with measurements indicating no depletion of pyruvate, the addition of pyruvate as a culture supplement does not rescue the gonad delay (data not shown), suggesting pyruvate is not limiting. Regardless of the mechanism by which mitochondrial function is preserved, the metabolic shift makes the animals hyper-reliant on their mitochondria and thus hypersensitive to disruption in mitochondrial function. Our results suggest that a shift away from glycolysis is not compatible with a normal pace of reproductive development.
Why Is the Reproductive System Sensitive to Reduced Glycolytic Output?-Supplementation with late glycolytic intermediates rescues the pnc-1 mutant gonad delay, suggesting that efficient glycolysis is required for rapid progression of reproductive development. Alternatively or in combination, the secondary effects of inefficient glycolysis must be avoided to support normal reproductive development. For example, inefficient glycolysis combined with an increase in amino acid catabolism to fuel the mitochondria for normal metabolic activity could be incompatible with the need for a high level of biomolecules to maintain the rate of cell division required to form a gonad and populate it with germ cells. Blocking glycolysis is not sufficient to induce a gonad delay phenotype in wildtype animals, suggesting that another effect caused by a lack of NAD ϩ bioavailability contributes to the phenotype as well. For example, inhibition of glycolysis alone may not be sufficient to induce the proposed increase in amino acid metabolism in a normal metabolic background. However, in pnc-1 mutants, which have changes in lipid and nucleic acid metabolism as well and may have perturbations in the pentose phosphate pathway (supplemental Table S1), inhibition of glycolysis is not compatible with a fast pace of reproductive development. We propose that the pnc-1 animals delay the pace of reproductive development in response to the need to compensate for reduced glycolytic output. By shifting resources for cell division to basal metabolic function, the pace of reproductive development cannot be maintained. Also, our data demonstrating that diet can have a large effect on reproductive developmental progression in the FIGURE 11. pnc-1 mutants show increased sensitivity to mitochondrial disruption. A, pnc-1 mutant animals (blue lines) paralyze faster than N2 animals (red lines) when treated with 20 mM sodium azide. B, N2 animals (red lines) that were fully paralyzed by pre-treatment with 5 mM sodium azide have a faster recovery time course after transfer to M9 buffer than pnc-1 mutants (blue lines). Error bars in A and B are S.E. *, 0.01 Ͻ p Ͻ 0.05; **, 0.001 Ͻ p Ͻ 0.01; ***, p Ͻ 0.001, calculated using Fisher's exact test. C, RNAi of electron transport chain and ATP synthase subunits have more pronounced effects on pnc-1 mutant muscle function than that of N2 animals. Thrashing rate of pnc-1 relative to wild type is plotted. Error bars are 95% confidence intervals. Actual ratio and sample sizes are indicated on the graph. Control samples were cultured on RNAi feeding E. coli strain HT115. Other samples were treated with RNAi targeting cyc-1 (complex III), nuo-1 (complex I), atp-2 (complex V) and atp-5 (complex V).
pnc-1 mutant is consistent with this model. For example, growth on the normal live HT115 E. coli strain results in a poorly penetrant reproductive delay, whereas growth on conditions with UV-killed food where some nutrients, including NA produced by the E. coli, are likely destroyed or eliminated results in a penetrant phenotype. It is worth noting that embryogenesis, another highly active period in terms of cell division, is not affected by loss of salvage NAD ϩ biosynthesis, suggesting that either other pathways for biosynthesis of NAD ϩ are sufficient to maintain NAD ϩ levels during embryonic development, which we have not yet tested, or the metabolic requirements in terms of the need for efficient glycolysis differs embryonically and post-embryonically. Experiments to examine metabolic profiles at developmental stages will be required to answer these questions.
In summary, we found that the effects on basal metabolic pathways mediate the developmental phenotype in the reproductive system upon the inhibition of salvage NAD ϩ biosynthesis. This result was surprising because the effects on NAD ϩ consumer activity have been the primary focus for probing the functional consequences of manipulation of NAD ϩ bioavailability (43)(44)(45)(46). Such studies have demonstrated the importance of regulation of NAD ϩ consumers, and our experiments also suggest that manipulation of salvage synthesis functionally impacts sirtuin activity. However, the phenotypic effects on reproductive development were not mediated by regulation of NAD ϩ consumers. Instead, upon loss of salvage NAD ϩ biosynthesis in C. elegans perturbation of the elegant interplay of basic metabolic pathways is the underlying cause of the reproductive development phenotype. From a therapeutic standpoint, our results are consistent with efforts aimed at using salvage NAD ϩ biosynthesis as a target to compromise metabolism in cancer cells, where cytoplasmic metabolic pathways are often relatively more important than mitochondrial metabolism.  Fig. 11, A and B. W. H.-R. wrote the paper. All authors reviewed the results and approved the manuscript.