The NR Bible – Research on benefits, dosage, side effects and more

Screenshot 2016-09-05 12.09.38

Recent press releases describing how a single dose of Nicotinamide Riboside has been shown to increase NAD+ in humans have been published by Chromadex, the manufacturer of Niagen, which is the only commercially available Nicotinamide Riboside.

This week, the doctoral thesis was published that was the basis of such press releases.

The thesis, NOVEL NAD+ METABOLOMIC TECHNOLOGIES AND THEIR APPLICATIONS TO NICOTINAMIDE RIBOSIDE INTERVENTIONS; by Samuel A.J. Trammell was done at the Graduate College of The University of Iowa under supervision of Professor Charles Brenner.

Based on the supporting research, the thesis notes many positive conclusions on possible benefits from NR. Some of those noted are:

NR dramatically reduced fatty liver disease in prediabetic and diabetic animals and protected against liver damage

NR reduces weight gain on HFD, hepatic steatosis, circulating cholesterol, HbA1C, and depresses nonfasting and fasting glucose. Overall, NR opposed the deleterious metabolic effects of HFD

Supplementation with NR increases NAD+, and is superior to other NAD+ precursors such as NMN and Niacinamide

NR Increases hepatic (liver) NAD+ and NAAD, protects against AFLD (Alcoholic Fatty Liver Disease)

NR shows ability to fight effects of alcoholism and diabetes

There are also some possible negative impacts noted in the research which need further investigation and are noted below:

– NR increases mortality in type 1 diabetic rats and decreases glucose tolerance in control mice… (NA has been shown to increase insulin resistance)

– injection of NR could cause NAD+ glycohydrolase inhibition in liver (Figure 5.6j) and may negatively impact liver health as consequence

Some other points that I found interesting:

NR increases NAAD in dosage dependent manner and may be an important marker of NAD+ effect on metabolism

NMN appears to be converted to NR extracellularly. Hence,the therapeutic effects of NMN are almost certainly a result of NR, meaning NR is likely a superior therapeutic agent

The summary of the thesis is below. The full dissertation is here.

Samuel A.J. Trammell

University of Iowa

Copyright 2016 Samuel AJ Trammell
is dissertation is available at Iowa Research Online: http://ir.uiowa.edu/etd/3203

 

CHAPTER 1

INTRODUCTION
“Targeted, LCMS-based Metabolomics for Quantitative Measurement of NAD+ Metabolites”*

Samuel A.J. Trammell1,2 and Charles Brenner1,2
1Department of Biochemistry, 2 Interdisciplinary Graduate Program in Genetics, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, USA

*Sections 1.1 and 1.2 and Figure 1.1 of the following chapter are reprinted from a published article in Computational and Structural Biotechnology Journal volume 4 (1). The copyright of the article belongs to the authors. The manuscript was written by Samuel Trammell with guidance from Charles Brenner, PhD.

1.1 Significance of NAD+ and Description of the Need for Improved Technologies for Its Measurement

The essentiality of NAD+-dependent processes in fuel utilization, gene regulation, DNA repair, protein modification, and cell signaling events makes the analysis of NAD+ metabolites central to an understanding of what a tissue is doing. NAD+ is the key hydride transfer coenzyme for a wide variety of oxidoreductases and is also the consumed substrate of sirtuins, poly(adenosine diphosphate ribose (ADPr)) polymerase, mono ADPr transferases, and cyclic ADPr synthases (2, 3). Measurement of NAD+ and related metabolites including several nucleosides and nucleotides (hereafter, the NAD+ metabolome) serves as a powerful indicator of the ability of a cell or tissue to perform processes such as glycolysis, gluconeogenesis, fatty acid oxidation, reactive oxygen species detoxification, among others. Moreover, the state of the NAD+ metabolome can serve as an indication of nutrition, health and disease.

Because NAD+ and related metabolites vary in cellular concentration from ~1 μM to ~1 mM, the analytical procedure must be robust, reproducible, and rapid. Liquid chromatography (LC)-based assays afford the ability to measure multiple metabolites in a timely fashion with thduration of each run ranging from 10 minutes to an hour. However, quantification through HPLC-UV-Vis methods is severely compromised based on the complexity of samples. In complex mixtures, a single peak may contain the metabolite of interest in addition to many other metabolites of identical retention time. In addition, peak shapes are rarely unaffected by complexity. Some investigators use a UV-vis signal at a retention time as the primary means for identification of a metabolite of interest—collected fractions are then subjected to mass spectrometry to confirm (nonquantitatively) the presence of the metabolite. This process leaves a great deal of data in the dark. Because every NAD+ metabolite can be converted to one or more other metabolites, snapshots of the levels of NAD+ , nicotinamide (Nam) or any other NAD+ metabolite without assessment of the NAD+ metabolome on a common scale has the potential to be misleading.

Because of its specificity and sensitivity, LC coupled to mass spectrometry (LC-MS) is a leading analytical method in the measurement of small molecules in complex samples. As with HPLC-UV-vis methods, LC serves to separate compounds of interest and must be optimized in the same way as any HPLC method. Because all LC-MS data contain at least two dimensions of data (retention time plus the mass:charge ratio, termed m/z), LC-MS increases specificity with respect to LC-UV-vis methods that report complex absorbance spectra as a function of retention time or matrix-assisted laser desorption ionization (MALDI)-based methods that report complex m/z data without retention times. Multidimensional MS, i.e., LC-MSn, provides further information because a particular analyte breaks down to component ions at a particular ionization energy. An ideal LC-MS method identifies an optimal extraction and separation method for all molecules of interest, detects the compounds in either negative or positive ion mode MS, and has sufficient LC separation to subject each molecule of interest to MSn analysis. The method is then a series

2

of selective reaction monitoring (SRM)1 protocols in which analytes are identified and quantified by MS as they come off the LC.

Whereas metabolomic discovery projects require high mass resolution instruments, targeted quantitative LC-MS assays can make use of lower resolution tandem mass spectrometers such as triple quadrupoles (QQQ). Here, the multidimensional data (retention time, m/z, and MS2 transitions) are used to distinguish closely related metabolites, such as NAD+ from NADH. Limits of quantification in optimized targeted, quantitative LC-MS assays are in the femtomole range.

Though mass spectrometers offer great analytical power for measuring the NAD+ metabolome, they also present technical challenges not encountered in other analytical techniques. These challenges include development of optimal mass spectrometry conditions, proper separation of metabolites, and best choice of internal standards. Here we discuss NAD+ metabolism and describe an optimized LC-MS2 assay of the NAD+ metabolome.

1.2 NAD+ Transactions

In fungi and vertebrates, NAD+ concentration is maintained by either de novo synthesis from tryptophan (4) or through salvage of nicotinic acid (NA) (5), nicotinamide (Nam) (6), and the recently identified NAD+ precursor vitamin nicotinamide riboside (NR) (7) (Figure 1.1). Some organisms, such as Candida glabrata, lack de novo synthesis (8). Many vertebrate cell types turn this pathway off (3). De novo synthesis proceeds from tryptophan in six steps to produce nicotinic acid mononucleotide (NAMN) and in two additional steps to produce NAD+. When NAD+ is the substrate of an enzyme such as glyceraldehyde phosphate dehydrogenase (GAPDH), fuel oxidation reactions will reduce NAD+ to NADH. In the case of GAPDH, the reaction is reversible, such that NADH is reoxidized to NAD+ in the gluconeogenic direction. NAD+ and NADH can be phosphorylated to NADP+ and NADPH. NADP+ is required for the

1 Multiple SRMs are referred to as multiple reaction monitoring (MRM). 3

pentose phosphate pathway (PPP), which produces NADPH. NADPH is required for detoxification of reactive oxygen species and reductive biosynthesis of lipids and steroids. Just as glucose-6-phosphate oxidation by the PPP produces NADPH, glutathione reactivation and reductive biosynthesis reoxidizes NADPH to NADP+.

Beyond serving as a coenzyme in hydride reactions, NAD+ is a consumed substrate for enzymes such as sirtuins, PARPs, and other ADPr transfer enzymes (2, 3, 9, 10). Though CD38 has an activity on NADP+, at least in vitro (11), the typical activity of an NAD+-consuming enzyme involves NAD+ as the substrate, and products that include Nam and an NAD+-derived ADPr moiety. Thus, to sustain intracellular NAD+ levels, actions of NAD+-consuming enzymes must be accompanied by Nam salvage (2, 3). Nam salvage differs between fungi and vertebrates. In fungi, Nam is hydrolyzed by the PNC1-encoded nicotinamidase to NA (6). NA is then converted by the first enzyme of the Preiss-Handler pathway, the NPT1-encoded NA phosphoribosyltransferase, to form NAMN. The second and third steps of Preiss-Handler salvage correspond to the final two steps of de novo synthesis, whose last step is glutamine- dependent NAD+ synthetase (12). In vertebrates, Nam produced as a product of NAD+- consuming enzymes cannot be salvaged as NA intracellularly. However, if Nam goes through the gut, bacterial nicotinamidases produce NA (13), which circulates and is used via Preiss- Handler salvage.

Intracellular Nam salvage in vertebrates depends on a Nam phosphoribosyltranferase, which entered the scientific literature with the names pre-B cell colony enhancing factor (PBEF) (14) and Visfatin (15). Now termed Nampt, this protein is widely expressed as an intracellular enzyme and also circulates as an active extracellular molecule (16, 17). First predicted to be part of a partially extracellular NAD+ biosynthetic cycle (2) along with CD73, a homolog of bacterial NMN 5’-nucleotidase, extracellular Nampt clearly has enzymatic activity (17). However, extracellular NMN remains controversial in part due to deficiencies in NAD+ metabolite quantification. As a phosphoribosyltransferase, Nampt activity depends on PRPP, an

4

extracellular source of which has not been demonstrated (18). By an HPLC-UV method, which may have been distorted by co-eluting analytes, the abundance of extracellular NMN was reported to be 80 μM (17). However, using LC- MSn, it was reported that PRPP and NMN are virtually absent and, moreover, are unstable in mouse plasma (18). It stands to reason that extracellular Nampt may have activity in local environments and developmental/nutritional conditions in which the substrates, Nam and PRPP, and the ATP activator are at substantial levels. Systemic NMN at 80 μM appears to be implausible, however.

Nam and NA can also be methylated, which would be predicted to block salvage. In plants, NA N-methyltransferase produces a compound known as trigonelline by transfer of the methyl group from S-adenosyl-methionine (19, 20). The corresponding Nam N- methyltransferase (NNMT) has been well characterized in vertebrates (21). Increased NNMT expression has been observed in Parkinson’s Disease (22) with a potential role in disease etiology (23, 24). NNMT is also increased in malignancy (25) and plays an apparent role in cell migration (26). Despite the reported roles in disease, N-methyl Nam (NMNam2) is a natural metabolite in healthy individuals with reported antithrombotic (27) and vasorelaxant (28) activities that is increased in plasma and urine after endurance exercise (29). NMNam is ultimately converted to N1-Methyl-2-pyridone-5-carboxamide and N1-Methyl-4-pyridone-5- carboxamide.

Though the primary breakdown product of NAD+ is Nam and the complete bacterially digested product is NA, nicotinamide riboside (NR) is an additional salvageable precursor that exists intracellularly and in milk (7, 30, 31). The unique NR salvage pathway is via nicotinamide riboside kinases (7). In addition, NR can be split into a Nam moiety and resynthesized to NAD+ via Nam salvage enzymes (32). Nicotinic acid riboside (NAR) is an alternate substrate of

2 In the future, this metabolite is abbreviated MeNam. 5

nicotinamide riboside kinases (33) and purine nucleoside phosphorylase (13) that has been shown to be an intracellular NAD+ precursor (30) but has not been reported to circulate.

Whereas NA is the salvageable precursor of NAD+ that has been exposed to the most digestive enzymes and Nam is the salvageable precursor that is produced by every cell with NAD+-consuming enzymes, the main source of dietary NR is probably partial digestion of NAD+. Depending on one’s nutrition and potentially one’s microbiome, the three vitamin precursors of NAD+ (NA, Nam and NR) and trp should be in circulation (3). The existence of extracellular enzymes with the potential to produce and consume NMN, and which consume NAD+, suggests the circulation of pyridine nucleotides (2). Moreover, NMN supplementation of mice on high fat diet (HFD) increases insulin sensitivity, glucose tolerance, and intracellular NAD+ compared to non-treated mice on the same diet (34). Though extracellular NMN was interpreted to function via direct incorporation of the nucleotide into cells (34), careful examination indicates that extracellular NA, Nam, and NR increase intracellular NAD+ in yeast and vertebrate cells, whereas NMN requires dephosphorylation to NR (35). Consistent with the prediction that the ectoenzyme CD73 has NMN 5’-nucleotidase activity (2), CD73 has the requisite biochemical activity to catalyze NMN dephosphorylation (36). In the yeast system, NR extends replicative longevity in a manner that depends on conversion to NAD+ (32). In mice on high fat diet, NR improves glucose control and insulin sensitivity, while moderating the observed increase in adiposity (37).

In addition to the major difference in Nam salvage between vertebrate and yeast systems, there is a mitochondrial compartmentalization problem in vertebrates. In yeast, transporters Ndt1 and Ndt2 carry NAD+ across the mitochondrial inner membrane (38) and the only mitochondrial NAD+ biosynthetic enzyme is NADH kinase, Pos5 (39). However, in vertebrate cells, the nucleocytoplasm and the mitochondrial matrix constitute distinct pools of NAD+, NADH, NADP+ and NADPH owing to impermeability of the mitochondrial inner membrane to these compounds. Though systems such as the malate-aspartate shuttle and

6

nicotinamide nucleotide transhydrogenase transfer reducing equivalents across mitochondrial membranes, vertebrate mitochondria require a system to import an NAD+ precursor into the matrix for conversion to NAD+. On the basis of localization of NAD+ biosynthetic enzymes, that precursor is NMN (35) 3. Nmnat3, which converts NMN to NAD+, is localized to the mitochondrial matrix. Nmnat3 is one of three vertebrate NAMN/NMN adenylyltranferases—the other two are localized in the nucleus and on the cytosolic face of Golgi. Though one could argue that the ability of Nmnat3 to convert NAMN to NAAD suggests that NAMN or NMN could be the mitochondrial NAD+ precursor, the NAAD product of the NAMN reaction requires glutamine- dependent NAD+ synthetase for conversion to NAD+. Glutamine-dependent NAD+ synthetase is not mitochondrially localized (35).

As shown in Figure 1.1, the implication of NMN as the limiting precursor for vertebrate mitochondrial NAD+ biosynthesis is profound. De novo synthesis and NA-dependent Preiss- Handler synthesis can only supply mitochondria with NAD+ by nucleocytosolic conversion to NAD+ followed by the pyrophosphate-dependent conversion of NAD+ to NMN in a back reaction of Nmnat first demonstrated by Arthur Kornberg in 1948 (44) or by conversion of NAD+ to Nam and subsequent conversion of Nam to NMN. In contrast, Nam and NR can be converted directly to NMN by Nampt and NR kinases, respectively.

In mitochondria that are burning fuel, the redox reactions are largely directional because fuel oxidation converts NAD+ to NADH and complex I of the electron transfer chain reoxidizes NADH to NAD+. Three vertebrate sirtuins, Sirt3-5, reside in mitochondria, where they consume NAD+ in reactions that either modify proteins or relieve protein modifications (45). For the

3 The origin of mammalian mitochondrial NAD+ is controversial. Later investigations revealed Nmnat3 is expressed in erythrocytes, which do not contain mitochondria (40). Further, Nmnat3 deficiency does not alter mitochondrial NAD+ (41) and Nmnat3 knockout animals are viable and capable of maintaining in NAD+ both fractions (42), suggesting the nucleocytoplasmic and mitochondrial pool are continuous. So far, no mammalian mitochondrial NAD+ importer has been identified (43) and the partitioning or lack thereof of NAD+ requires further investigation.

7

mitochondrial sirtuins to work and avoid robbing redox enzymes of NAD+, NMN must be imported from the cytosol.

Completing the major NAD+ transactions, yeast possess two cytosolic NAD+/NADH kinases and the mitochondrial NADH kinase, Pos5 (39). Vertebrate cytosolic NAD+/NADH kinase is related to the yeast enzymes (46), whereas the vertebrate mitochondrial NAD+/NADH kinase was recently identified as a homolog of A. thaliana Nadk3, which can use ATP or polyphosphate as the phosphate donor (47).

1.3 Thesis Goals

The main focus of my thesis research was to develop LC-MS/MS technologies for the quantitation of NAD+ and related metabolites to further our understanding of NR interventions in healthy and diseased states. Previous members of my thesis laboratory focused upon the enzymes related to NAD+ and its biosynthesis from NAR and NR (7, 30, 32, 33, 48). In their investigations, the first NAD metabolome assay was developed and included substrates and products of these enzymes and other enzymes related to NAD+ metabolism (31). From their work, NR was established as a bona fide salvageable NAD+ precursor that could extend life- span of yeast (32) and work from other groups revealed NR extends life-span in C. elegans (49) and acts to oppose metabolic (37, 50-54) and neurodegenerative disorders (55, 56) in rodents. Translating these health promoting effects of NR in the clinic required the improvement and development of novel technologies for accurate and robust quantitation of NAD+ and related metabolites.

In my thesis work, I improved upon the previous assay by including internal standards, adding additional NAD+ related metabolites, and further optimizing extraction procedures for cell and tissues. With these improvements, we were able to produce a detailed metabolic image of the fate of NAD+ metabolism in a variety of biological contexts with and without NR interventions. The technologies described herein allowed for both the confirmation and

8

generation of hypotheses regarding the effect of NR on NAD+ metabolism in the normal and abnormal function of a cell or organism.

In chapter 2, I include a reprint of the remaining publication used to introduce my dissertation in this chapter. Sections are added after section 2.2 Conclusions to describe further method development as necessitated by my thesis work.

Chapters 3, 4, 5, 6, and 7 are demonstrations of the technologies described in chapter 2. Our work in chapter 3 establishes the true B3 vitamin content of bovine milk, uncovers that farming practices may influence the vitamin quality of milk, and reveals that B3 vitamin fortification of bovine milk may be a future route of delivery of NR to populations at risk for developing neurological and metabolic disorders. Specifically, we show that NR represents 40% of the B3 vitamin content of bovine milk from a herd of Bos taurus and in store procured cow’s milk. We uncover that organic milk tends to contain less NR than conventional milk and suggest that, in part, Staphylococcus aureus infection may be responsible. We then uncover that NR is stable in and binds to milk. This chapter illustrates how the same technology utilized to merely elucidate the composition of a food product can be utilized to generate and test hypotheses for how the vitamin content of a food.

In chapter 4, we establish that NR is a superior NAD+ precursor compared to NMN using stable isotope labeling technologies. Work in chapter 3 and 4 were crucial for later work described in chapter 5 where more complex stable isotope labeled experiments were performed.

The work described in chapter 5 is the culmination of my thesis work. This chapter is my perspective and narration of a work that was written by my advisor Dr. Charles Brenner and includes data generated using methods pioneered by me but performed by both myself and Dr. Mark Schmidt, a current staff member of the Brenner laboratory. In this work, we quantified the NAD metabolome in a healthy middle-aged human subject after initial and subsequent supplementation of NR and uncover that NAAD is a potential, non-obvious, accessible

9

biomarker for NR supplementation. NAAD is a non-obvious product of NR since there is currently no known mammalian deamidating NAD+ pathway (Figure 1.1). We then compare the effect of NA, Nam, and NR on the murine hepatic NAD metabolome.

All three precursors increase NAD+ as expected. However, both Nam and NR increase NAAD. Additionally, we report for the first time that NR is a far superior effector in NAD+ metabolism, increasing both hepatic NAD+ and NAAD to a greater extent compared to NA and Nam. Further, we tested whether and confirmed that NR directly contributes to NAAD using stable isotope technologies.

In Dr. Schmidt’s work, we confirm that NAAD positively correlates with NR dosage in a group of healthy human subjects. Together, these works performed in human and murine systems prove NR is superior to other B3 vitamins effecting the NAD metabolome and increasing NAD+ in particular and uncover that NAAD may be a future, clinical biomarker for the effect of NR on NAD+ metabolism.

Chapters 6 and 7 are essentially the phenotypic effects of NR supplementation on metabolic syndromes. In chapter 6, we tested the hypothesis that NR supplementation would prevent alcohol-induced fatty liver disease by increasing hepatic NAD+ and consequently reversing the metabolic damage of alcohol on mitochondrial metabolism. In Chapter 7, we tested the hypothesis that the alteration of the NAD metabolome in diabetic animals is involved in the etiology of diabetic peripheral neuropathy and that supplementation with NR could prevent this devastating complication of diabetes. In this chapter, I present our work with a Type I diabetic animal model and NR and my perspective on my work with a Type II diabetic animal model and its context alongside the work performed by Dr. Mark Yorek and coworkers detailed in Chapter 7.1-4 written by Dr. Charles Brenner, thesis advisor.

10

1.4 Figure

Figure 1.1 NAD+ biosynthesis in yeast and vertebrates.

Intracellular NAD+ is derived from either de novo synthesis from tryptophan or from salvage of NA, Nam, or NR. In yeast, Nam is converted to NA by nicotinamidase Pnc1p (dotted line). In yeast and vertebrates, NA is phosphoribosylated to NAMN, an intermediate in de novo synthesis, and converted to NAD+ by way of NAAD in a step catalyzed by glutamine-dependent NAD+ synthetase (12). In vertebrates, Nam conversion to NMN is catalyzed by Nampt (16). The other source of NMN in yeast and vertebrates is phosphorylation of NR by NR kinases. NR and NAR can be split to the corresponding pyridine bases. NAR phosphorylation yields NAMN. NMN is converted to NAD+ by NMN adenylyltransferase activity, which is reversible. As shown, in vertebrates, NMN must be imported into mitochondria for conversion to NAD+. Enzymatic NAD+ and NADP+ consumption releases the Nam moiety and produces ADPr products. Finally, Nam and NA can be converted to non-salvageable products.

11

CHAPTER 2

NAD METABOLOME ANALYSIS VIA LIQUID CHROMATOGRAPHY MASS SPECTROMETRY

“Targeted, LCMS-based Metabolomics for Quantitative Measurement of NAD+ Metabolites”*

Samuel A.J. Trammell1,2 and Charles Brenner1,2
1Department of Biochemistry, 2 Interdisciplinary Graduate Program in Genetics, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, USA

*The following chapter is a description of the general liquid chromatography and mass spectrometry methodologies employed in all subsequent chapters. Sections 1 – 2 are reprints of the rest of the publication included in Chapter 1.1-1.2 (1) which was written by myself with guidance and editing by CB.

2.1 Quantitative NAD+ Metabolomics

The NAD+ metabolome4, as defined here, includes dinucleotides, nucleotides, nucleosides, nucleobases and related compounds (Table 2.3). The masses of many of the analytes differ by a single Dalton, necessitating optimal separation and careful MS. The current method is an improvement over methods, which measured only select metabolites (17, 57), and more recent methods, which embraced a more complete set of metabolites, but which lacked resolution of several compounds (31, 58). Here we review optimization of all parameters and a solution to the ionization suppression problem that plagued previous methods.

Optimized Extraction
Methods that do not inactivate enzymatic activities upon cell lysis (58) are clearly flawed

and, based upon the amount of time of sample autolysis, cellular NAD+ can be degraded to ~1% of expected values (~10 μM) with elevation of apparent NR concentration to ~100 times

4 After publication of this document, we have referred to the NAD+ metabolome interchangeably with the NAD metabolome.

12

expected values (1 mM) (59). The preferred method of extraction is to use boiled, buffered ethanol (60), which is well validated for NAD+ metabolites (30, 31).

For yeast samples, an ideal cell number is 2 to 4.5 x 107, the midrange of which can be obtained by harvesting 25 ml of cells at an OD600 nm of 0.7. For mammalian cell culture, we typically use 4 to 20 x 106 cells, depending upon the cell type. Yeast cell pellets are extracted directly. Mammalian cell pellets are washed once in ice-cold potassium buffered saline. Cells are resuspended in 300 μL of a 75% ethanol/25% 10 mM HEPES, pH 7.1 v/v (buffered ethanol) solution, preheated to 80 °C. Samples are shaken at 1000 rpm in an 80 °C block for three minutes. Soluble metabolites are separated from particulate by refrigerated microcentrifugation (10 min, 16kg). Though the ethanol-soluble extract contains all the metabolites of interest, the weight of the particulate can be used to determine the optimized resuspension volume for dried metabolites. Thus, both the particulate and soluble metabolites are dried by speed vacuum at 40 °C.

Empirically, we determined that 3.6 mg of yeast or mammalian cell-derived particulate corresponds to a metabolite pellet, which can be resuspended in a 100 μl volume and produce the desired absorbance and LC-MS signals. Thus, the dry weight of each pellet is recorded, divided by 3.6 mg, and multiplied by 100 μl to obtain the initial resuspension volume. Extracts are resuspended in 1% (v/v) acetic acid adjusted to pH 9 with ammonium hydroxide (ammonium acetate buffer). These conditions were chosen to preserve NADH and NADPH prior to analysis5 (61).

Resuspended metabolites (2 μl) are checked in a Nanodrop (ThermoFisher) to determine the OD260 nm, which is typically greater than or equal to 14. The remaining volume is diluted to an OD260 nm of 14 in ammonium acetate buffer to obtain the final resuspension volume.

5 In Chapter 2.4, I describe problems in analysis uncovered after publication which required alterations to the re-suspension solvent and to the metabolites included in the metabolomic assay and the way with which they were dealt.

13

For LC-MS, this material is diluted two-fold into two different metabolite standards and 2.5 μl of the resulting material is injected and analyzed. Because these dilutions convert the total intracellular volume into a known volume of which an effective volume of 1.25 μl is analyzed, it is straightforward to calculate the intracellular volume of cells under analysis.

The calculation of intracellular volume is as follows. For yeast cells, the intracellular volume of a single cell is taken as 70 fl (62). Thus, the calculated intracellular volume is obtained as 70 fl times the cell number. For mammalian cells, we use 2.5 pl as volume of a HeLa cell (63) and calculate the total extracted intracellular volume in the same way as for yeast cells. For example, an extraction of 3 x 107 yeast cells has a calculated intracellular volume of 2.1 μl. If this sample were resuspended into 100 μl and require no further adjustment after checking on the Nanodrop, the 1.25 μl of cell extract in a 2.5 μl injection would represent 1.25% of 2.1 μl = 26 nl. Because the internal standards permit metabolites to be quantified on a mol scale, intracellular metabolite concentrations are determined, in this example, as mol of metabolite divided by 2.6 x 10-8 l.

Optimized Internal Standards
Ionization suppression is the tendency for sample components to dampen the ionization

and detectability of particular analytes. Thus, one cannot reliably depend on the peak height or area of a metabolite in a standard curve of purified metabolites to be on the same scale as its peak size in a complex mixture. In the most advanced previous quantification method for NAD+ metabolites, ionization suppression was a problem for NAD+, inosine and NA (31).

Two internal standard sets are employed. One set is used to quantify 16 metabolites (all analytes except NR, Nam, and NA) and is used in an alkaline separation. The other set is used to quantify NR, Nam, and NA in an acidic separation.

For analytes in the alkaline separation, an extract of Fleischmann’s yeast metabolites is prepared from cells grown in 99% uniformly labeled 13C glucose (Icon Isotopes, Summit, New Jersey). In the course of this culture, the PPP converts 13C glucose to 13C ribose-5-phosphate,

14

such that all of the cells’ nucleic acids, mononucleotides and dinucleotides incorporate 13C. Mononucleotides incorporate 5 additional Da from one ribosyl moiety, whereas dinucleotides incorporate 10 additional Da from two ribosyl moieties6. Complete incorporation is achieved by growing two serial starter cultures in 13C glucose synthetic dextrose complete media, followed by inoculation of a 250 ml volume of 13C glucose media at a starting OD600 of 0.2 and growth to OD600 of 0.8. Cells in 50 ml aliquots are then pelleted and stored at -80 °C prior to extraction with 300 μl buffered ethanol solution. Metabolites are resuspended in 100 μl of the ammonium acetate buffer and have a typical OD260 nM of 100. For LC-MS, Fleischmann’s extract is diluted 1:40 into ammonium acetate buffer. This diluted Fleischmann’s extract is further diluted 1:1 with experimental samples. In the LC-MS analysis, 2.5 μl of the fully diluted material is injected. This material represents 1.25 μl of the experimental extract and an appropriate amount of Fleischmann’s extract for metabolite quantification.

Because the vitamins Nam and NA do not contain a carbohydrate group, they are not labeled by heavy labeled glucose and require the second set of internal standards for accurate quantification. To each sample or standard solution, heavy labeled Nam and NR are added such that the final concentration is 1.5 μM. 18O labeled Nam is prepared by tetramethylguanidine- catalyzed hydrolysis with 3-cyano-pyridine and H218O, as described (64). Though heavy NR is present in the Fleischmann’s yeast extract standard, NR is best separated and quantified in an acidic separation with Nam and NA. Heavy labeled NR is made as described (65).

We prepare stock solutions (typically at 10 mM) of each metabolite to be quantified and then prepare a set of standard solutions containing the whole set of metabolites to be quantified

6 This process creates isotopologues, which are compounds that differ in isomer composition (e.g. the NAD+ produced from yeast supplemented with [13C6]-glucose produce [13C10]-NAD+, the isotopologue to 12-carbon NADH. Isotopologues should not be confused with isotopomers, which are molecules that contain the same isotopic composition but differ in the location of the isotope.

15

at a range of concentrations (0, 0.1, 0.2, 0.6, 2, 6, 20, 60, and 200 μM). For the alkaline separation, these standards are mixed 1:1 with the 1:40 dilution of Fleischmann’s extract so that each quantifiable metabolite in the extract can be set to a pmol amount. For example, if a particular metabolite in the Fleischmann’s sample were to interpolate precisely between the 2 μM and 6 μM standards (1 μM and 3 μM) in the 2.5 μl injected volume, we would calculate there to be 5 pmol of that metabolite in the standard amount of Fleischmann’s extract that will be used for all subsequent samples. This determination is made in technical replicates. Importantly, we cannot expect the 5 pmol peak area of that metabolite to remain constant when Fleischmann’s extract is mixed with experimental yeast extracts because peak shapes are often distorted, and ionization is suppressed due to sample complexity. However, because the 13C metabolites in Fleischmann’s extract will have the same degree of ionization suppression as the 12C metabolites in the experimental extract, the peak area ratios of 12C to 13C metabolites and the known mol amounts of the Fleischmann’s metabolites allow calculation of the amounts of the experimental metabolites. Molar calculation of the Fleischmann’s metabolites are not performed. Mol amounts of metabolites in the experimental samples are converted to molar using the calculation of intracellular volume described above. Because the IMP peak in Fleischmann’s extract is quite small, we use relative peak areas of IMP and NMN in the standard solutions to derive a correction factor that allows IMP in experimental samples to be quantified against the Fleischmann’s NMN peak7. For each metabolite, the peak area used for quantification is that of the MS2 transition.

In theory, inclusion of 1.5 μM 18O labeled Nam and 18O labeled NR should be sufficient to quantify nonlabeled Nam and NR in the experimental extracts. In practice, because label

7 Later analysis revealed that UMP, ADPR, NADP+ internal standards are also of poor signal batch-to-batch. The yeast produced internal standard for NAAD is also found to be weak in signal and sometimes co-eluting with what is believed to be the alpha isomer of NAD+ internal standard. When signal is too poor, NAD+ internal standard is utilized for ADPR, NAAD, and NADP+. CMP internal standard is used for UMP quantitation.

16

incorporation may vary and there is greater accuracy in preparation of 10 mM standards than 1.5 μM radiolabeled standards, we quantify the 18O peaks of Nam and NR against a standard curve of NA, Nam and NR. These analyses result in a mol amount of heavy Nam and NR determined from the light standards. Because the same amount of heavy Nam and NR will be in all experimental samples, relative peak areas allow conversion of experimental Nam and NR peaks to mol and molar. The peak area ratio between heavy Nam and NA in the standard solutions is used to derive a correction factor that allows the heavy Nam peak to calculate the amount of NA in experimental samples.

Optimized Liquid Chromatography
Earlier, we described an assay of the NAD+ metabolome based on hydrophilic

interaction liquid chromatography (31). Encouraged by its track record for separation of nucleosides and nucleotides (66-68), we have since developed an improved separation with the porous graphitic carbon reversed phase material, Hypercarb (Thermo). Resolution of all compounds is done with two different mobile phases on two Hypercarb columns, each used solely for one separation.

In the alkaline separation, solvent A is 7.5 mM ammonium acetate with 0.05% (v/v) ammonium hydroxide and solvent B is 0.1% (v/v) ammonium hydroxide in acetonitrile. The optimized gradient is described in Table 2.1 with a flow rate of 0.08 ml/min and a column temperature of 60 °C. As shown in Table 2.1, the complete run takes 32.2 min on a 1 mm x 100 mm Hypercarb column and must be equilibrated for 20 min prior to first injection.

In the acid separation, solvent A is 10 mM ammonium acetate with 0.1% formic acid and solvent B is 0.1% formic acid in acetonitrile. The optimized gradient is described in Table 2.2 with a flow rate of 0.2 mL/min and a column temperature of 60 °C. As shown in Table 2.2, the complete run takes 23.4 min on a 2.1 mm x 100 mm Hypercarb column. Extracted ion currents for each resolved metabolite are provided in Figure 2.

17

Mass Spectrometry Optimization
The power of triple quadrupole mass spectrometers is the ability to perform multiple

SRM protocols in a single run, i.e. multiple reaction monitoring (MRM). Modern QQQs are equipped with automatic optimization software to detect transitions and optimize ionization. The software can be the best friend and greatest enemy in method development. Since many of the metabolites are of similar structure and mass, specific diagnostic fragments must be identified and optimized. The carboxylic acid versus carboxamide metabolites and the oxidized versus reduced metabolites differ by only one Dalton. The 13C peaks produced from metabolites such as NAD+ and NADP+ would produce crosstalk with NADH and NADPH, respectively (Figure 2.1A). However, all four compounds produce diagnostic fragments, allowing for specific quantification. Current automatic optimization software identifies fragments that are most easily produced and not necessarily those that are diagnostic for the metabolite in the context of structural similarities in the NAD+ metabolome. Online searchable libraries such as Metlin (http://metlin.scripps.edu/) and Massbank (http://www.massbank.jp/?lang=en) provide MS/MS spectra for many metabolites with identified fragment structures (69, 70). Specific fragments for NADH and NADPH not identified by the automatic software were chosen based on these searches. The transitions were optimized manually. Transitions and optimized conditions are detailed in Table 2.3.

The cone voltage must be optimized when measuring NAD+ especially for NR, NAR, NAMN, and NMN to reduce on-source fragmentation. NR and NMN readily produce Nam signal, whereas NAR and NAMN produce NA signal (Figure 2.1B, Nam extracted ion current). Optimization of cone voltage decreases but does not completely remove crosstalk. This unavoidable crosstalk greatly illustrates the need for robust LC separation. After development, overall robustness was determined based on the capacity factors (k’), quantitative range, linear goodness of fit (R2), replicative standard deviation (RSD) of the method, and RSD of the system. Capacity factors were above 2 for all analytes with the exception of CMP (k’ = 1) (Table

18

1). Standard curves were linear from 0.125 picomoles to 250 picomoles with R2 values falling above 0.99 for all but NAR (0.948), inosine (0.974), NR (0.981) and NADH (0.988). RSD of the method was measured with six separate standard solutions at 10 μM concentrations. RSD of the system was measured with four injections of the same standard solution. Method RSDs were below 10% for all but cytidine, NAR, inosine, NAMN, ATP, NAAD, NADH, and NADP. System RSDs were below 10% for all but cytidine and NAR. Limits of quantification (LOQ) were measured repeatedly empirically and defined as the concentration producing a signal-to-noise ratio of 10. LOQ were below 100 fmol for all but NMN (1 pmol), ATP (1 pmol), NA (2.5 pmol), and uridine (3.1 pmol). Moreover, the LOQ using this method was at least 3-fold lower for seven metabolites than the previously used method. The previous method is 3-fold more sensitive for one metabolite (Table 1.3) (31).

Metabolite Measurement Challenges
In our hands, AMP and NADPH cannot be reliably quantified by these methods. ATP

and ADP fragment on source to AMP, similar to NR fragmentation to Nam (Figure 2.1B Nam extracted ion current). Given poor resolution of AMP from both metabolites, detected AMP signal would represent biological AMP as well as that derived from ATP and ADP. Further, the NADP peak may represent the sum of cellular NADP and a portion of cellular NADPH, which has become oxidized. Thus, care should be taken in interpreting the NADP peak. Nam is strikingly membrane permeable. Nam in cells can be easily lost into post-cellular supernatants and we suspect that Nam in organelles can be easily lost into post-organelle supernatants.

Results in Mammalian Cell Line
To test our method on a real sample, we analyzed a glioma cell line, LN428/MPG (a gift

of Dr. Robert Sobol, University of Pittsburgh), which had been grown in MEMalpha (10% FBS HI, gentamycin, geneticin) media in triplicate 150 mm dishes (2 x 107 cells per dish by CASY cell count). The results of the analysis are reported in Table 2.4. As expected, nucleotides, such

19

as ATP, ADP, UMP, and NAD+, are high in abundance compared to NAD+ precursors and biosynthetic intermediates. The NAD+:NADH ratio is ~39 and the NAD+:NADP ratio is 4.6. As expected for cells grown in a type of Dulbecco’s modified Eagle’s media, Nam, but not other vitamins, is detectable. The concentration of NMN is low when compared with yeast, suggesting the NMN pool is converted rapidly to NAD+ (31).

Conclusions
Here, an improved LC-MS method has been developed to quantify the NAD+

metabolome. Its principle features are resolution and quantification of 16 metabolites in an alkaline separation, and resolution and separation of 3 metabolites in an acidic separation, both on a porous graphitic carbon stationary phase. The problem of ionization suppression that plagued earlier methods has been eliminated. Preservation and quantification of NADPH remains a challenge.

Acknowledgements

Work was supported, in part, by grant MCB-0822581 from the National Science Foundation and by NIH Pre-Doctoral Training Program in Genetics T32 grant GM008629.

2.2 Continued Method Development Post-Initial Publication

ATP and ADP: The Other Problem Metabolites
ATP and ADP were found to be labile in the buffered boiled extraction protocol. Since

ATP degrades to ADP, this complicates analysis of both metabolites. Further, ATP and ADP are thought to be at 10:1 ratio but this ratio was not observed in LN428 cells utilizing our method (Table 2.4), suggesting improper metabolic quenching. For this reason, ATP and ADP were not routinely included in future experiments.

Addition of MeNam, Me2PY, and Me4PY to the NAD Metabolomic Assay
In mammals, Nam serves as a precursor to NAD+ and to MeNam (Chapter 1.2). MeNam is further oxidized by aldehyde oxidase 1 (EC: 1.2.3.1) to produce either N-methyl-2-pyridone-5-

20

carboxamide (Me2PY) or N-methyl-4-pyridone-5-carboxamide (Me4PY). As stated in Chapter 1, Section 2, MeNam, Me2PY, and Me4PY have biological effects. Since the writing of that document, we now know that expression of NNMT, the enzyme methylating Nam to MeNam, positively correlates with adiposity in adipocytes (71) and negatively correlates in liver (72) and appears to regulate lipid metabolism in a manner that is tissue dependent and relying upon either methyl metabolism or the MeNam itself (73). MeNam is produced with concomitant demethylation of the primary methyl-donor, S-adenosyl-methionine (SAM), which is also the source of nuclear histone methylation. NNMT activity has been shown to deplete SAM and cause epigenetic changes and contributes to cancer development (74) but also stem cell maturation (75). MeNam itself increases murine hepatic Sirt1 activity by either directly or indirectly stabilizing its protein abundance and has inhibitory effects on cholesterol synthesis (72). Me2PY and Me4PY have also been implicated in calorie mediated increases in C. elegans life-span (76) by increasing catalase activity and promoting resistance of reactive oxygen species damage. Beyond the numerous biological effects of these metabolites or the enzymatic activity synthesizing them, these metabolites are not thought to be NAD+ precursors as there is no known demethylase nor oxidoreductase. In this way, these metabolites can be thought of as an indication of NAD+ precursor wasting as an increased occurrence of MeNam and oxidized derivatives represent a diversion of flux from its synthesis. For these reasons, these metabolites were added to the NAD metabolome.

MeNam, Me2PY, and Me4PY SRM conditions are displayed in Table 2.5 and were produced as described above in Chapter 2.1. These metabolites are readily resolvable (Figure 2.2) on the Thermo Scientific HypercarbTM column with separation as described for the acid separation in Chapter 2.1. Though Me2PY and Me4PY are isomers, little cross-talk is observed (Figure 2.2). Retention times and precision for each metabolite are displayed in Table 2.5. Systematic and analytical variability are below 20% for all metabolites.

21

Internal standard for MeNam was produced by methylating 18O Nam with D3 iodomethane. Briefly, 125 mg of labeled Nam was dissolved in 0.5 ml of ACS grade methanol. 96 μl of labeled iodomethane was added slowly to the solution. The solution was vortexed, then allowed to react at 25 °C with constant shaking at 300 rpm for 24 hours. The reaction produced a yellow needle shaped precipitate and was dried down via speed vacuum for three hours. The product was analyzed using a Waters Premier Q-TOF operated in positive ion mode and the product at m/z 142 was observed alongside smaller amounts of 18O Nam. Any non-methylated labeled Nam was not removed since it would contribute to the internal standard signal and not interfere with analysis of the NAD metabolome. 18O Nam was utilized as an internal standard for Me2PY and Me4PY. MeNam is linear from 0.1 to 100 μM. Me2PY is linear from 0.1 to 30 μM and Me4PY is linear from 0.1 to 100 μM.

Considerations of Quantitative NAD Metabolomics in Mammalian Tissues Metabolomic analysis within any tissue requires optimization of extraction from that

tissue prior to the final experiment. Tissues vary in type and abundance of proteins and metabolites. If the metabolite of interest readily binds to a specific protein, its recovery may be more affected in a tissue that expresses that protein in abundance compared to another. Additionally, some metabolites may be isobaric8 with an analyte and interfere with quantitation if co-eluting and abundant. The results of these differences mean every tissue presents its own unique challenges for the same set of metabolites.

Though the extraction methods may differ, aspects of extraction procedures for proper metabolomic analysis are universal. The extraction process must simultaneously quench metabolism and separate the metabolite(s) from interfering compounds (77). Turnover rates for bioenergetic metabolites such as ATP can be in the 100s of μmol per second range (78) and are greatly influenced by metabolic and genetic interventions. NAD+ turnover is similarly

8 Isobaric metabolites are metabolites of the same mass and/or m/z. 22

affected. Hence, accurate measurement requires rapid and effective quenching of all chemical and enzymatic reactions occurring in the cell, tissue, or fluid.

Early in my thesis work, tissues were prepared as homogenates and no care was taken to quench enzymatic activity. This naïve choice led to clear changes in the NAD metabolome (Table 6.1) exemplified with an extremely high ratio of Nam to NAD+. As described in Chapter 2.1, this high ratio is almost certainly a result of non-quenching of NAD+ consuming activities causing a decrease in NAD+ with simultaneous synthesis of Nam. To this end, we employed freeze clamping9 upon tissue collection at the time of sacrifice to quench enzymatic activity and preserve the NAD metabolome. Unless otherwise stated, all tissues were collected this way prior to storage at -80 °C.

Prior to extraction, tissues were pulverized to a fine powder using a Bessman pulverizer cooled to liquid nitrogen temperatures. The powdering of the tissue was performed to allow for aliquoting of tissue for the complete NAD metabolome analysis and to increase the surface area exposed to the extraction solvent. To continue to slow enzymatic turnover of the metabolites, tissue aliquots were kept at dry ice temperature or lower at all times prior to addition of extraction solvent. Normally, extractions are performed to precipitate protein and other macromolecules and release the metabolites of interest into the soluble fraction. Metabolite extractions are often rely on adjusting the pH and often combined with addition of organic solvents (commonly methanol or acetonitrile) at various temperatures. It is imperative that tissues are lysed as quickly and efficiently to ensure inhibition of enzymatic activity and robust recovery of all metabolites from the sample. Unlike mammalian cell culture samples which easily lyse with repeated pipetting, tissues require mechanical disruption such as bead beating,

9 Freeze clamping of tissue is performed by squeezing tissue between two liquid nitrogen cooled plates for at least 10 seconds. This methodology ensures all enzymatic activity halts due to extreme low temperature.

23

homogenization, or sonication. We found in certain instances that heating of the suspended extract maximized recovery of some metabolites, especially NADP. This effect has been observed by others (79, 80) and has been attributed to greater disruption of protein binding.

Protein precipitation is imperative for accessing metabolites, but further and possibly more importantly, for preserving the life of the analytical column, in our case the Hypercarb column. Loaded proteins can precipitate on column leading to blockage or can adhere to the sorbent10 and interfere with separation of metabolites. But the protein precipitation method can produce a soluble fraction that is incompatible with the liquid chromatograph used for separation. High organic, low aqueous samples disrupt the peak shape and retention time of metabolites eluting in separations such as described in Chapter 2.1. In some instances, the soluble extract is diluted with an appropriate amount of pure or buffered water to ensure proper separation. However, this method obviously dilutes the sample, causing loss of signal for lower abundant metabolites. Here, we employed drying of the sample and re-suspension in an appropriate aqueous solvent. Drying methods themselves can impact recovery of a metabolite (80) and should always be considered when developing a method. These considerations and others were investigated in developing sample extractions protocols.

Quantification of the Oxidized NAD Metabolome in Liver
Liver is the largest mammalian internal organ and potentially the most metabolically

versatile and active. The complete suite of NAD+ biosynthetic routes are expressed in the liver (3, 81), making the liver very responsive to B3 vitamins. We employ the same extraction buffer as for cells and extract 5 – 20 mg of frozen pulverized liver in the following manner. Samples are extracted by addition of 0.1 ml of buffered ethanol (3 volumes ethanol: 1 volume 10 mM HEPES, pH 7.1) at 80 °C. Samples are vortexed vigorously until thawed, sonicated in a bath

10 Sorbents are materials used to absorb and adsorb compounds. It is the material packed into the column that is responsible for separation of the analytes.

24

sonicator (10 sec followed by 15 sec on ice, repeated twice), vortexed, then placed into a Thermomixer® (Eppendorf, Hamburg, Germany) set to 80 °C and shaken at 1050 rpm for five min. Samples are centrifuged (16.1kg, 10 minutes, 4 °C). Clarified supernatants are transferred to fresh 1.5 ml tubes to dry via speed vacuum for two hours. Prior to LC-MS/MS analysis, samples were reconstituted in 40 μl of 10 mM ammonium acetate (>99% pure) in LCMS grade water. Analytes are separated and analyzed as described in Chapter 2.1 unless otherwise stated. Recovery was measured based on internal standard area counts in sample compared to its area counts in reconstitution solvent (Table 2.6). With the exception of NR, all metabolites had ≥ 50% recoveries. NR recovery had only 18%. This may appear to be a problem in quantitation given the importance of NR as a metabolic activator; however, NR is an obligate cation and as consequence is very detectable (LOQ = 10 fmol) (Table 2.3). Further, the ~five- fold deficit in NR recovery should equally affect its internal standard 18O NR. Since quantitation is based upon the ratio of NR to its isotopologue, NR quantitative values are accurate as long as both analyte and internal standard are detected.

Quantification of NAD(P)H and Extraction from Liver
In the published method (Chapter 2.1), NADPH was immediately removed from routine

analysis due to its instability in the ideal conditions for quantitation of the oxidized NAD metabolome. NADH was included since yeast fed labeled glucose produce its isotopologue, which could serve as its internal standard and control for its stability and extraction efficiency. However, over time it was found that the [13C10]-NADH in yeast extract varied in concentration and was difficult to robustly produce. Further, future method development with differing liquid chromatography separations indicated the presence of closely eluting and sometimes co-eluting unknown compounds that produced signal for the NADH internal standard in sample (Figure 2.3). The origin of the multiple internal standard is likely from the yeast extract itself, but this point was never proven since it was unusable in liver. These complications warranted a complete re-working for the analysis of NADH and led to its removal from the routine NAD

25

metabolome assay. Because NADH was removed, considerations for its stability, which negatively impacted the oxidized NAD metabolome, were no longer relevant, the re-suspension solvent for the oxidized NAD metabolome was altered to a more acidic condition to preserve metabolite stability.

Though NADH and NADPH quantitation is difficult, the ratio of these metabolites to their oxidized forms are often reported as indicators for a cell or tissue to perform glycolysis versus gluconeogenesis and/or as a capacity for ROS detoxification (Chapter 1.1-1.2). Further, the action of sirtuins, often thought of as metabolic regulators (82, 83), is inhibited by NADH at physiologically obtainable concentrations (84). For these reasons, we developed and implemented an LC-MS method for the quantitation of these reduced dinucleotides in liver. In so doing, we uncovered a carryover problem with a common liquid chromatography separation for NADH and NADPH (collectively referred hereafter as NAD(P)H) and offer an alternative liquid chromatography separation.

Since NADH and NADPH carry a -2 and -4 negative charge, it was clear at the onset of the liquid chromatography method development that an ion pairing agent was required. Ion pairing agents are additives to the mobile phase11 that interact with the analytes of interest and produce better retention and peak shape. Hydrophobic Ammonium salts carrying alkane chains are often employed for the separation of negatively charged metabolites and have been utilized in other assays for organic acid and NADPH (85). Since ammonium salts can cause ion suppression in positive ion mode (Figure 2.4), negative ion mode was utilized. Single ion monitoring (SIM) mode was selected in lieu of SRM mode to increase sensitivity. In this mode, the mass spectrometer scans only for the intact m/z of the compound of interest at any one time and does not produce diagnostic fragments for confirmation of the identity of the compound. Confirmation of the metabolite of interest in a sample in this method relies upon standard

11 The mobile phase is a liquid or gas that flows through a chromatographic system, carrying compounds at different rates over a stationary phase.

26

addition to the sample and observation of an increase in signal at the suspected retention time. Though SRM mode increases selectivity and confidence within a measurement, NAD(P)H weakly fragments, and, consequently, its signal was greatly diminished in this scanning method. Initially, we utilized10 mM tributylamine (TBA) as the ion pairing agent in the mobile phase and a Waters Acquity BEH C18 column (inner diameter x length: 2.1 x 100 mm) as the stationary phase12 (86). At first, NAD(P)H eluted sharply and were well resolved from each other (Figure 2.4), but over time, the metabolites became increasingly retained on the column and variably eluted, indicating carryover (Figure 2.5). The same was observed using the Phenomenex Synergi Hydro-RP sorbent employed by Fan et al. (2014) (86). The carryover continued despite decreasing the concentration. This carryover could have resulted from either retention of the metabolites in the injector system or on the column. With this separation appearing incompatible with robust quantitation of NAD(P)H, this mobile phase was abandoned without further investigation into the cause of the carryover/increased retention. However, it appears that TBA is too retentive of negatively charged metabolites and should be used cautiously in the analysis of other organic acids.

We hypothesized that triethylammonium (TEA) acetate could replace TBA since it would be charged in the pH of the mobile phase and is less hydrophobic. Indeed, utilization of TEAA at 10 mM decreased retention times compared to TBA (Figure 2.6 versus 2.5) but produced sharp peak shapes and did not suffer the carryover problems observed with TBA. Hence, TEAA was utilized in the mobile phase. To decrease analysis time, flow rate was increased to 0.4 ml/min from 0.2 ml/min and the gradient optimized for separation of NAD(P)H in 13.5 minutes total run time. Gradient and mobile phase conditions are described in Table 2.7. NADPH and NADH eluted at 3.34 and 3.69 minutes, respectively. NADH and NADPH are linear from 9 to 60 μM at

12 A stationary phase is the non-mobile portion of a chromatographic system that interacts with compounds carried in the mobile phase and causes the rate of their flow through the system to vary, thus separating the compounds from a complex mixture.

27

an injection volume of 10 μl and reproducible with a method RSD of 5 and 15%, respectively, indicating the separation is suitably sensitive and reproducible for quantitation of NAD(P)H.

Extraction of these metabolites required the optimization of a different procedure from that described for the oxidized NAD metabolome. First, deoxygenating of the extraction solvent with nitrogen was essential. Second, the pH of the extraction solvent is adjusted to 9 and the extraction solvent is kept at dry ice temperature throughout extraction to increase stability (61). Third, nitrogen drying at ambient temperature is employed rather than drying with speed vacuum. Fourth, multiple rounds of extraction are performed. Inclusion of these changes allows for 100% recovery of the standards in the absence of tissue.

To extract NAD(P)H from liver, ~20 mg of frozen pulverized liver is sonicated (Branson Sonifer 450, output control = 4, intensity = 40%, 10 seconds) in 0.5 ml of -80 °C methanol/25 mM ammonium acetate pH 9 (5:1) in a -4 °C acetone/water bath, then rested on dry ice. The extract is heated at 60 °C using Thermomixer® (Eppendorf, Hamburg, Germany) for three minutes with constant shaking at 1050 rpm then centrifuged (16.1kg, 10 minutes, 4 °C). The supernatant is transferred to a fresh tube and the pellet re-extracted two more times. The supernatants from each round of extraction are combined. Extracts are dried at ambient temperature using nitrogen gas and reconstituted in 50 mM ammonium acetate (>99%) and 0.05% ammonium hydroxide (>99%) in LCMS grade water immediately prior to analysis. The metabolites remain labile in this solvent and must be analyzed within 8 hours post reconstitution. Recovery of NAD(P)H in murine liver was estimated by spiking either water or 1000 pmol of both analytes into an aliquot of liver and calculated with the following equation:

X X 100

where K is the ratio of the weight of the liver aliquot dosed before processing to the weight of the liver aliquot dosed after processing. NADH and NADPH were recovered at 67 and 77%, respectively. To estimate signal suppression from the sample, the added signal of analyte in the

28

dosed after processing sample was compared to signal in fresh standard. Signal recovery was approximately 83 and 93% for NADH and NADPH, respectively. Since no reliable internal standard could be produced for NADH (Figure 2.3) and NADPH at the time of method development, a standard addition method was utilized whereby liver extracts were pooled, then aliquoted equally, and finally dosed with varying concentrations of standard. The pooled liver was aliquoted such that approximately 0.5 mg of liver was loaded at 10 μl injection. This amount of tissue is the same amount loaded from samples at 2.5 μl. This method allowed for control of ion suppression.

Though this method could be improved with the synthesis of internal standards, the work described in this section represents a robust and sensitive framework for future quantitative NAD(P)H assays. Further, through method development, we discovered that TBA, an ion- pairing agent in well-known separations for organic acids (85, 86), causes extreme carryover of NADPH that could not easily be overcome through reduction of the ion-pairing agent concentration in the mobile phase. The applicability of this effect to other metabolites should be explored. In cases where carryover is experienced, TEA may be an adequate substitution.

Quantification of the Oxidized NAD Metabolome in Skeletal Muscle
Muscle depends upon Nam and NR for NAD+ biosynthesis (87). NAD+ increase as a

function of NR opposes myopathy (50, 88). Unlike liver, muscle is a dense tissue containing a tough actinomyosyin network. As a consequence, extracting the oxidized NAD metabolome from this tissue requires greater mechanical disruption and, like with NAD(P)H, requires heating to increase release of the metabolites. To extract quadriceps, ~20 mg of frozen pulverized tissue is sonicated as described for extraction of NAD(P)H in the presence of 0.5 ml ice cold methanol/water (4:5). The extract is heated for 5 minutes at 85 °C using a Thermomixer® (Eppendorf, Hamburg, Germany) with constant shaking at 1050 rpm then centrifuged (16.1kg, 10 minutes, 4 °C). The supernatant is dried using a speed vacuum and reconstituted in 40 μl of 10 mM ammonium acetate (>99%) in LCMS grade water. LC-MS/MS conditions are as

29

described in Chapter 2.1. Recoveries were determined by comparing internal standard area counts in samples in reconstitution solvent (Table 2.8). All recoveries are above 60% except CMP and NAR which are 38 and 31%, respectively.

30

CHAPTER 3

NICOTINAMIDE RIBOSIDE IS A MAJOR NAD+ PRECURSOR VITAMIN IN BOVINE MILK

Samuel A.J. Trammell1,2, Liping Yu1,3, Philip Redpath4, Marie E. Migaud1,4, and Charles Brenner1,2

1 Department of Biochemistry, Carver College of Medicine, University of Iowa, Iowa City, IA
2 Interdisciplinary Graduate Program in Genetics, University of Iowa, Iowa City, IA
3 Nuclear Magnetic Resonance Facility, Carver College of Medicine, University of Iowa, Iowa City, IA
4 Queen’s University Belfast, School of Pharmacy, Belfast, Northern Ireland, UK

3.1 Distribution of Work

CB and I designed the research, analyzed the results and wrote the manuscript. LY performed the NMR measurements. I performed the LC-MS and LC-MS/MS with heavy standards synthesized by myself or by PR and MEM.

3.2 Abstract
Background: Nicotinamide riboside (NR) is a recently discovered nicotinamide adenine dinucleotide (NAD+) precursor vitamin with a unique biosynthetic pathway. Though the presence of NR in bovine milk has been known for more than a decade, the concentration of NR with respect

to the other NAD+ precursor vitamins was unknown.
Objective: We aimed to determine NAD+ precursor vitamin content in raw samples of milk from individual cows and from commercially available bovine milk.
Methods: Liquid chromatography tandem mass spectrometry (LC-MS/MS) and isotope dilution technologies were used to quantify NAD+ precursor vitamin content and to measure NR stability in raw and commercial milk. Nuclear magnetic resonance (NMR) spectroscopy was utilized to test for NR binding to substances in milk.

40

Results: Bovine milk typically contained ~12 μM NAD+ precursor vitamins, of which 60% was present as nicotinamide (Nam) and 40% was present as NR. Nicotinic acid and other NAD+ metabolites were below the limits of detection. Milk from samples testing positive for Staphylococcus aureus contained lower levels of NR and Nam (Spearman R = -0.58 and -0.43, respectively) and NR was degraded by Staphylococcus aureus. Conventional milk contained more NR than milk sold as organic. Nonetheless, NR was stable in organic milk and exhibited an NMR spectrum consistent with association with a protein fraction in skim milk.

Conclusions: The pellagra-preventive activity of bovine milk is likely a function of NR and nicotinamide. Control of Staphylococcus aureus may be important to preserve the B3 vitamin content of milk.

3.3 Introduction

One hundred years ago, pellagra was common in the rural American South. One of the early treatments for pellagra was consumption of a pint and a half to two pints of bovine milk (89). In 1937, nicotinamide (Nam) and nicotinic acid (NA) were identified as pellagra-preventive (PP) factors (90, 91) and tryptophan (trp) was subsequently discovered as a molecule with PP activity (92). Nam and NA, which are collectively termed niacin, contain a pyridine ring that can be salvaged to form NAD+ in two or three enzymatic steps, whereas trp is the de novo precursor of NAD+, requiring 7 enzymatic steps (3). Largely because trp can be incorporated into protein, oxidized as a fuel, and converted to many other metabolites such as serotonin, 50-60 mg of trp is considered the niacin equivalent of 1 mg of Nam or NA. In addition, much of the niacin equivalent in food is, in fact, NAD+ (93).

NAD+ and its phosphorylated and oxidized derivatives, NADP+, NADH and NADPH are essential hydride transfer cofactors in hundreds of oxidoreductase reactions and consumed substrates of several classes of enzymes with activities required for DNA repair, gene expression, regulation of energy metabolism, and calcium mobilization (2). NAD+ is one of the most abundant

41

metabolites in the human body and is turned over at a rapid pace (87), requiring near constant replenishment. Hence, though pellagra is described as niacin deficiency, at a cellular level, pellagra is a disease of NAD+ depletion as a result of diets deficient in NAD+ precursors.

It has long been known that the NAD+ precursors in milk include Nam (94) and Trp (95). More recently, it has been discovered that milk also contains NR, another salvageable NAD+ precursor vitamin (7). Boosting NAD+ levels with NR extends lifespan in yeast (32) and has been shown to prevent and treat metabolic (37, 50, 53, 54) and neurodegenerative (55, 56) conditions in mouse models. Though these studies suggest that dairy products as a source of NR could be beneficial to many aspects of human health, the amount of NR in milk has not been established. In this work, we determine the complete B3 vitamin content in individual and pooled commercial samples of bovine milk using a liquid chromatography tandem mass spectrometry (LC-MS-MS)- based method (1). Our results show that milk from Staphylococcus aureus (Staph a)-positive samples contained lower levels of NR and Nam, and that milk sold as organic milk contained lower levels of NR than conventionally sourced milk. Moreover, we show that NR is stable in milk, is bound by substances in milk, and that approximately 40% of the NAD+ precursor vitamin content of bovine milk is present as NR.

3.4 Methods

Milk Quality and Herd Health Measurements
Milk flows and representative samples were obtained from 19 conventionally raised cows

using a Dairy Herd Improvement Association (DHIA) testing meter. Samples were dispensed into 2 ounce snap cap DHI vials containing liquid bronopol for analysis by Dairy Lab Services (Dubuque, IA) of fat, protein, lactose, other solids, milk urea nitrogen and somatic cells (FOSS, Denmark). Additional aseptic individual milk samples were obtained for bacterial analysis after teats were sterilized with 70% ethanol prior to collecting 3 ml of milk from each teat into 12 x 75 mm culture tubes. All samples were frozen before further analysis. Blood agar culture plates were

42

inoculated with sample, then incubated at 37 °C, and evaluated for bacterial growth at 24 and 48 h. Bacterial growth was characterized by morphology and samples were subjected to confirmatory tests to identify genus and species.

Bovine Milk Sample Acquisition and Preparations
Nineteen milk samples from individual cows plus 8 skim milk samples (4 organic and 4

conventional) purchased in the Iowa City area were analyzed by LC-MS-MS. Two 50 μl aliquots were extracted from each milk sample. Each aliquot was dosed with either solution A (18.75 pmol of [18O1]-NR, 18.75 pmol of [18O1]-Nam, 18.75 pmol of [D3, 18O1]-MeNam and 150 pmoles [D4]- NA) or solution B, a 13C-labeled yeast extract at 1:50 dilution. Aliquots were extracted with 0.5 ml of 1.5% formic acid at room temperature. Each aliquot was vortexed for 10 s and then sonicated for 10 min in a bath sonicator. Aliquots were then centrifuged at a speed of 16,100 x g for 10 min at room temperature. Extracts were transferred to fresh 1.5 ml centrifuge tubes and dried overnight via speed vacuum. Recovery was greater than 90% for all metabolites of interest.

NMR Spectroscopy
NR 1H resonances were assigned with 1H/13C two-dimensional heteronuclear multiple

quantum coherence (HMQC) and heteronuclear multiple bond coherence (HMBC) experiments. NR binding to skim milk was analyzed using water-ligand observed via gradient spectroscopy (WaterLOGSY) (96, 97). To analyze fractions of milk for NR-binding activity, 2 ml of total milk was centrifuged for 1 h at 4 oC at 16,100 x g. The supernatant was termed the soluble fraction, while the pellet resuspended in 2 ml 50 mM sodium phosphate pH 7 was termed the particulate fraction. NMR samples were prepared by adding 150 μl skim milk, skim milk soluble fraction, or resuspended skim milk particulate fraction to 352.2 μl buffer that contained 300 μl 50 mM sodium phosphate (pH 7), 50 μl D2O, and 2.2 μl NR stock, giving a final NR concentration in the NMR samples of 0.3 mM. For the WaterLOGSY experiment, a T2 relaxation filter of 100 ms was used just before data acquisition to suppress signals derived from macromolecules, and a water

43

Nuclear Overhauser Effect (NOE) mixing time of 1 s was used in the experiment. All NMR data were acquired using a Bruker Avance II 800 MHz NMR spectrometer equipped with a sensitive cryoprobe and recorded at 25 oC. The 1H chemical shifts were referenced to 2,2-dimethyl-2- silapentane-5-sulfonate (DSS). NMR spectra were processed using NMRPipe package software (98) and analyzed using NMRView (99).

NR Stability Assays
[18O]-labelled and [13C1,D1]-NR were synthesized as described (65)(Trammell, et. al. in

preparation). [18O]-NR was suspended in conventional and organic milk brands or in water at pH 5, 7, or 11 at a concentration of 10 μM and allowed to sit at room temperature. Twenty μl aliquots were collected at 0, 0.5, 1, 2, 4, and 8 h and extracted as described above.

Staph a Growth Experiments
Strain RN3170 was a kind gift of Patrick Schlievert (University of Iowa) (100). Bacteria

were streaked onto Todd-Hewitt (Becton & Dickinson) 2% agar plates, incubated overnight at 37 °C and then stored at 4 °C. Staph a was then inoculated into Todd-Hewitt media containing 50 mM Bis-Tris pH 6.7 and 10 μM of [13C1,D1]-NR at a starting OD600 nm of 0.1. Non-inoculated medium was used as control for NR stability. All cultures were incubated at 37 °C with constant shaking at 220 rpm. Fifteen ml aliquots were collected at 0, 1, 2, 4, 6, and 8 h. OD600 and pH values were recorded at each time point. Aliquots were centrifuged at 2,060 x g for 30 min at 4 °C at which time, 1 ml of culture medium was collected and snap-frozen in liquid nitrogen. The remainder of culture medium was aspirated and cell pellets were washed with 1 ml of ice-cold PBS and recentrifuged at 16,100 x g for 10 min at 4 °C. PBS was aspirated and pellets were flash frozen. Fifty μl of media were analysed using LC-MS/MS as described below. Cells were extracted using buffered ethanol (3 parts ethanol to 1 part 10 mM HEPES, pH 7.1) heated to 80 °C for 3 min with constant shaking at 1,050 rpm. Extracts were clarified by centrifugation (16,100 x g, 10 min, 4°C). Pellets were extracted again following the same procedure as above. Supernatants

44

from both rounds of extraction were combined and dried via speed vacuum. Extracts were analysed using LC-MS as described below.

LC-MS and LC-MS/MS
Media samples were diluted 1:1 with double distilled H2O. Standard solutions in double

distilled H2O were diluted 1:1 with non-inoculated Todd-Hewitt media containing 50 mM Bis-Tris and [13C1, D1]-NR producing a standard curve with the final concentrations of 0, 0.1, 0.3, 0.5, 1, 3, 5, and 10 μM. Quality control samples at a final concentration of 0.75 and 7.5 μM were also prepared by diluting standard 1:1 with media. Ten μl of media samples, quality controls, and standards were injected and quantified using a Waters TQD mass spectrometer using the acid separation chromatographic conditions described previously (1). Media were quantified using raw peak areas and converted to μM using background-subtracted standard curves.

For bovine milk, standards (final concentrations of 0.08, 0.24, 0.8, 2.4, 8, 24, 80, 120 μM) and two quality control samples (final concentration of 2.5 and 25 μM) were treated in the same manner as the samples and as described above. Five μl of samples, quality controls, and standards containing solution A or 10 μl of those containing solution B were loaded onto the column and quantified using a Waters TQD mass spectrometer according to the procedures previously described (1). Newly quantified metabolites in the acidic separation, MeNam, Me2PY, and Me4PY were assayed with the following transitions: MeNam (137 > 94 m/z), Me2PY (153 > 107 m/z), and Me4PY (153 > 136 m/z). Analyte peak areas were normalized to internal standard peak areas and converted to μM using the standard curve. Staph a cell pellets were suspended in 50 μl of 10 mM ammonium acetate in LCMS grade water. A260 nm values for each sample were measured using a Thermo Scientific 2000c Nanodrop spectrophotometer operated in nucleic acid mode. Samples at 0 and 1 h time points were diluted 1:1 with either solution A or solution B. Samples at all remaining time points were diluted to a final A260 nm value of 14 then diluted 1:1 with solution A or B. All samples were analyzed according to the chromatography protocols previously described (1) and detected and quantified using a Waters Premier QTOF operated in positive full

45

scan mode. The alkaline separation was altered by increasing the flow rate to 0.55 ml/min and shortening the run time to 11.6 min. Separation was performed using a modified gradient with initial equilibration at 3% B, a 0.9 min hold, a gradient to 50% B over 6.3 min, followed by a 1 min wash at 90% B, and a 3 min re-equilibration at 3% B. When performing the alkaline separation, the scanning window was set to m/z = 120 – 800 with a scan rate of 0.1 and an interscan rate of 0.01. When performing the acid separation, the scanning window was set to m/z = 120 – 600 with a scan rate of 0.5 and an interscan delay of 0.05. In both cases, leucine enkephalin was infused and utilized for mass accuracy correction. Analyte peak areas were normalized to internal standard peak areas and converted to μM using the standard curve. Nam concentrations were corrected for the contribution of [12C1]-Nam and [13C1]-Nam to the [18O1]-Nam internal standard area counts. Enrichment for all metabolites was corrected for the natural abundance of the analyte, 13C abundance, and the purity of the doubly labeled NR (95/5% [13C1, D1]-NR/[13C1]-NR). The corrected concentrations of each analyte were converted to intracellular concentrations by calculating the total intracellular volume of Staph a using an intracellular volume of 0.28 fl (101) and an assumption of 1 x 109 cells/ml per OD600 nm (102).

Statistical Analysis
Unless otherwise stated, all values are expressed as mean ± standard deviation. Two-

tailed, unpaired t-tests were performed on all comparisons involving fewer than three groups. Outliers were identified using the ROUT method (103). Two-way, repeated ANOVA followed by Holm-Sidak’s multiple comparisons test was performed on experiments involving Staph a. Media samples were compared to non-inoculated media within time points. Intracellular samples were compared to initial concentrations within condition. Spearmen’s rank correlation coefficient was calculated for the concentration of Nam and NR versus the milk quality and herd health metrics. P-values < 0.05 were considered significant. Statistical analyses were performed using GraphPad Prism version 6.00 for Windows (GraphPad Software, CA).

46

3.5 Results

NR is a Major Component of the B3 Vitamin Content in Bovine Milk
The B3 vitamin content of 19 individual bovine milk samples was determined using LC- MS/MS and isotope dilution techniques. We define the B3 vitamin content as the levels of salvageable NAD+ precursor vitamins (Nam, NA, and NR) plus levels of the higher molecular weight species (NAR, NAMN, NAD+, NAAD and NADP+) from which a vitamin can be released by enzymatic or chemical decomposition. NADH and NADPH are oxidized in extraction, such that

these metabolites, if present, would contribute to the peaks of NAD+ and NADP+.
As shown in Table 3.1, in all 19 samples, Nam and NR and no other NAD+ metabolite were quantifiable. Thus, neither NAD+ nor NA is a PP factor in milk. Excluding one unusual milk sample which contained 24 μM Nam and 27 μM NR (Supplemental Table 3.1), the mean sample contained 7.3 ± 1.5 μM Nam and 4.3 ± 2.6 μM NR. To determine whether other parameters correlate with NAD+ precursor vitamin contents in the 18 remaining samples, breed was recorded and, metrics of the health and milk quality of each cow were measured (Supplemental Table 3.2). As shown in Table 3.2, levels of NR positively correlated with levels of lactose (P = 0.013) and milk urea nitrogen (P = 0.018), whereas Nam negatively correlated with somatic cell count (P = 0.029) and positively correlated with NR (P = 0.011). NR concentration negatively correlated with Staph a infection (P = 0.014). Nam concentration also negatively correlated with Staph a infection but the correlation was not significant (P = 0.09). When we recalculated levels of NR and Nam in the 12 samples without Staph a infection or extremely high levels of NAD+ precursor vitamins, Nam rose to 7.7 ± 1.2 μM and NR rose to 5.1 ± 2.6 μM. Thus, though it was clear from previous work that there is no NA in bovine milk (94), there has been a substantial under-reporting of NAD+

precursor vitamin on account of lack of an assay for NR.

47

Staph a Depletes NR and Nam
Because presence of Staph a was associated with lower levels of NR and Nam, we tested

whether Staph a growth might directly alter the levels of these metabolites in rich media. Before testing stability in the presence of Staph a, we investigated the stability of [18O]-NR in pasteurized bovine milk or in water adjusted to pH values of 5.0, 7.0, and 11.0. As shown in Figure 3.1A, [18O]- NR was stable in pasteurized milk and in water at neutral pH, but exhibited lesser stability at pH 5.0 and pH 11.0, with pH 11.0 producing nearly complete hydrolysis within 1 h. We measured the pH of bovine milk in 4 store-bought milk brands and determined the pH to be 6.72 ± 0.01.

Bacteria might alter levels of NR found in milk by incorporating NR intact into NAD+ and/or by converting NR into Nam or NA, either of which could be subsequently incorporated into NAD+. To distinguish between these possibilities, we synthesized a double-labeled NR containing a 13C in the Nam moiety and a D1 in the ribose. Incorporation of NR into the bacterial NAD+ pool would be accompanied by a 2 Da mass shift (m/z 664 → 666), whereas breakdown of NR to Nam or NA would be accompanied by appearance of 1 Da shifts in the peaks of these metabolites and a 1 Da mass shift in bacterial NAD+.

Three individual colonies of Staph a strain RN3170 were cultured separately in Todd- Hewitt media supplemented with 10 μM [13C1, D1]-NR and buffered at pH 6.7 with 50 mM Bis-Tris. The inoculated media and a non-inoculated medium control were incubated at 37 °C with constant shaking over an 8 h period. Clarified media and cell pellets were collected and analyzed by LC- MS/MS and LC-MS. The pH of the clarified media was also recorded at each time point. The pH consistently remained between 6.5 and 6.7 until between the 6 and 8 h time points, at which time the pH rose to 7.8. As shown in Figure 3.1B, [13C1, D1]-NR was stable in non-inoculated medium over the time course of the experiment. However, Staph a inoculation significantly decreased the concentration of extracellular NR within 1 h and eliminated the presence of NR as an extracellular metabolite within 4 h. As shown in Figure 3.1C, singly labeled Nam appeared in growth media

48

within 1 h. As shown in Figure 3.1D-F, at 4 h, there was a simultaneous rise in singly labeled cellular NAD+, singly labeled cellular Nam, and extracellular NA.

Todd-Hewitt media contains beef heart extract, Nam and NA (104). As shown in Figure 3.1G and 3.1H, non-labeled Nam was exhausted within 4 h while the non-labeled NA slowly declined. Thus, Staph a principally uses NR as an extracellular source of Nam. Consistent with Nam deamidation (105, 106), Staph a can also degrade NR and Nam to NA. Because NR was eliminated by 4 h and the rise in pH occurred after 6 h, pH-mediated mechanisms cannot be responsible for Staph a-mediated NR instability.

NR Content as a Function of Organic Certification
Milk with organic certification requires avoidance of synthetic chemical inputs, irradiation,

genetically modified seed, and adherence to certain standards of feed, housing and breeding (107). Because one or more of these variables could affect the B3 vitamin content of milk, we purchased 4 brands of conventional and 4 brands of organic milk and quantified the NAD+ metabolome. As observed in milk samples from individual cows, only Nam and NR were above the limit of quantification (Table 3.3). In three of four conventional samples and three of four organic samples, the concentration of Nam exceeded that of NR. Moreover, the concentration of Nam was similar in conventional (5.2 ± 3.4 μM) and organic (5.6 ± 2.5 μM) milk. In the samples we obtained, NR tended to have a higher concentration in conventional (3.1 ± 1.6 μM) versus organic (1.9 ± 1.0 μM) milk. We note that only one brand of store-bought milk had B3 vitamin content (8.9 μM Nam plus 5.4 μM NR) that was higher than the mean of Staph a-negative individual cow samples (7.7 μM Nam plus 5.1 μM NR). Though the difference between conventional and organic milk in NR concentration did not rise to statistical significance (P = 0.26), the data suggest that a difference in the feeding and care of cows or milk preparation might depress NR levels in organic store-bought milk.

49

NR is a Bound Metabolite in Bovine Milk
Though a higher level of Staph a infection in organic milk production (32) could potentially

account for lower levels of NR in organic milk, there was no change in Nam and no appearance of NA that would be consistent with bacterial exposure. As shown in Figure 3.1, NR is more stable in milk than in water, suggesting that the metabolite might be complexed to a protective factor.

Organic dairies frequently ultrapasteurize milk at 135 oC for 2 sec, whereas most conventional dairies pasteurize at 72 oC for 15 sec (108). Though ultrapasteurization is employed to kill bacterial spores, it might damage a macromolecule responsible for the stabilization of NR. WaterLOGSY NMR measurements were employed to detect and map protons in NR that are potentially bound by slower rotating macromolecules in milk (96, 97). As shown in Figures 3.2B and 3.2F, when NR was added to conventional or organic skim milk, four aromatic protons (H2, H4, H5 and H6) from the Nam moiety of NR produced positive WaterLOGSY signals consistent with protein-binding from both sources of milk. Interestingly, when conventional and organic milk were separated into soluble and resuspended particulate fractions, the conventional soluble fraction retained more NR-binding activity than did the organic soluble fraction (Figures 3.2C and 3.2G). Consistent with denaturation of an NR-binding protein by heat, the solubilized organic particulate fraction produced stronger NR WaterLOGSY signals than did the solubilized conventional particulate fraction (Figures 3.2D and 3.2H).

3.6 Discussion

The data presented in this paper show that about 60% of the B3 vitamin content of bovine milk is present as Nam, while about 40% is present as NR. Though we occasionally detected bovine milk samples with higher levels of NR than Nam, we did not detect NAD+, NA or any other NAD+ metabolite in any bovine milk sample. In samples from individual cows, presence of Staph a, the most common cause of cattle mastitis (109, 110), was associated with lower concentrations of NR and Nam. We also showed that Staph a degraded NR into Nam and NA and used the Nam

50

as a precursor of intracellular NAD+. These data are consistent with the ability of Staph a to utilize either Nam or NA for NAD+ synthesis (111) with Nam utilization requiring deamidation to NA (106, 112).

Multiple aspects of bovine nutrition are expected to contribute to levels of Nam and NR in milk. In particular, feeding of herds with Nam and NA (113, 114) is likely to produce cows that transmit higher levels of Nam and NR into milk. In addition, it will be interesting to determine whether NR is actively transported by mammary glands (115).

Though PP factors in food include NAD+, NAD+ precursor vitamins and Trp, high doses of NR appear to have some protective activities for metabolic and neurodegenerative conditions. The ability of milk to bind and preserve the integrity of NR makes dairy products potentially good sources of supplemented NR. Further research is needed to maximize NR content of conventional and organic milk and to identify the molecular basis of NR binding to milk.

CHAPTER 4

EFFICACY OF NMN AND NR AS EXTRACELLULAR NAD+ PRECURSORS

Samuel A.J. Trammell1,2, Marcelo Rodrigues3, Allyson Mayer1, Lori J. Manzel1, and Charles Brenner1,2

1 Department of Biochemistry, Carver College of Medicine, University of Iowa, Iowa City, IA 2 Interdisciplinary Graduate Program in Genetics, University of Iowa, Iowa City, IA
3 Queen’s University Belfast, School of Pharmacy, Belfast, Northern Ireland, UK

4.1 Distribution of Work

Experiments were designed by CB and I. The following chapter was written completely by myself. Cell culture work was performed by LM, AM, or MR with guidance from me. All mass spectrometry and data analysis was performed by myself.

4.2 Abstract

The novel NAD+ precursor nicotinamide riboside (NR) opposes age and diet induced morbidities such as obesity and diabetes. Nicotinamide mononucleotide (NMN), phosphorylated NR, is also effective against diabetic effects. Though both increase intracellular NAD+ concentration upon administration to rodents, NMN internalization into cells has not been demonstrated. Mounting genetic and pharmacological evidence suggests NMN is dephosphorylated to NR prior to its intracellular utilization. However, these methods rely upon indirect measurement and does not necessarily exclude whether NMN and NR are kinetically equivalent in supplementing NAD+. In this work, we show through LC-MS/MS and stable isotope enrichment that NMN is dephosphorylated extracellularly to NR and that NR is incorporated into the intracellular NAD+ faster than NMN and as such NR is a more efficacious B3 vitamin in cell culture.

4.3 Introduction

Nicotinamide adenine dinucleotide (NAD+) is a cofactor in hydride transfer reactions and consumed substrate for ADP ribose transferases, poly-ADP ribose transferases, cyclic ADP

57

ribose synthases, and sirtuins (Chapter 1.2). Through these reactions, NAD+ is involved in many cellular processes such as gene expression, fuel utilization, DNA repair, protein modification, and cell signaling. NAD+ is synthesized either through the de novo pathway from tryptophan or salvage of nicotinic acid (NA), nicotinamide (Nam), and nicotinamide riboside (NR) (Figure 1.1) (2, 7). In vertebrates, Nam phosphoribosyltransferase (Nampt/Visfatin/PBEF1) converts Nam to nicotinamide mononucleotide (NMN) using 5-phosphribosyl-1-pyrrophosphate as a phosphoribose donor (16). NR is phosphorylated to NMN by either Nrk1 or Nrk2 (7, 33). NMN is then converted to NAD+ through one of three NMN adenylyltransferases (NMNAT1-3).

Both NR and NMN oppose diet and age induced diabetes (34, 37, 116, 117). The overlapping beneficial effects of these two NAD+ precursors suggests utilization of the same pathway. Indeed, both have been shown to increase intracellular NAD+. Both have been suggested to circulate with NMN at ~80 μM (17) and NR at an unknown concentration (Chapter 5). However, way in which NMN contributes to intracellular NAD+ remains debatable. NR depends upon equilibrative nucleoside transporters (2). Though NMN is a nucleotide and not normally predicted to cross the plasma membrane, NMN has been suggested to enter into cells and across the highly selective blood brain barrier (118). To date, no NMN transporter has been identified but could depend upon a similar uptake mechanism to that found in astrocytes (119). If NMN were directly imported, its utilization would be expected to depend upon NMNAT1-3 but not on the NRK pathway, the pathway converting NR to NMN. However, pharmacological and genetic interventions suggests NMN depends upon NRK (35, 120, 121), CD38 (122), and the predicted NMN extracellular nucleotidase (2), CD73, calling into question whether NMN is extracellular available or simply an NR or Nam prodrug (Figure 4.1).

Though current data is highly suggestive that NMN utilization requires dephosphorylation by CD73 to NR or hydrolysis to Nam by CD38, these studies do not preclude that nor tested whether NMN and NR are equally efficacious as precursors. The rate of NMN dephosphorylation could be non-rate limiting to its utilization, rendering NMN and NR

58

equivalently efficacious. We sought to directly test efficacy of NMN versus NR as salvageable NAD+ precursors using stable isotope labeling and LC-MS/MS. In so doing, we show direct evidence that NR is a more efficacious precursor and confirm metabolite-metabolite relationships in a human hepatocyte cell line.

4.4 Materials and Methods

Compounds
[18O1]-NR was made as described in (65). Labeled NR was phosphorylated to NMN

using Nrk1 and ATP. NMN was purified by HPLC on a strong anion exchange column with a 10 mM to 750 mM gradient of KH2PO4 [pH 2.6]. Isotopic purity was assessed by LC-MS and concentration was assessed using an extinction coefficient (260 nm) of 4200 M-1*cm-1.

Cell Culture Conditions
Human hepatocyte cell line Hepg2 cells were procured from ATCC. For all labeling

experiments, cells were grown in Eagle’s minimum essential medium (MEM) containing glutamine (0.292 g/l) and 10% fetal bovine serum to 75% – 80% confluency. Media was aspirated and replaced with media lacking nicotinamide and fetal bovine serum. Cells were incubated at 37 °C for 24 hours before aspiration and repletion with MEM containing either 10 μM [18O1]-NMN or [18O1]-NR. Media and cells were collected at 0, 0.5, 2, 4, 7, and 24 hours. At collection time, 1 mL of media was collected and the rest aspirated. Cells were trypsinized for 10 minutes at 37 °C, pelleted, washed with ice cold PBS, pelleted, and then flash frozen in liquid nitrogen. All samples were stored at -80 °C until analysis.

Extraction
Cells were extracted as described in Chapter 2.1 using the buffered boiled ethanol

method. Media were diluted 1:1 with LC-MS grade water containing internal standard (10 μM cytidine) and injected without further preparation for analysis.

59

LC-MS/MS
LC-MS/MS was performed as described in Chapter 2.1. Transitions were established for

18O labeled NAD+ metabolites: 18O Nam (125>98 m/z), NA (126>53 m/z), NR (257>125 m/z), NAR (258>126 m/z), NMN (337>125 m/z), NAMN (338>126 m/z), NAD+ (666>428 m/z), NAAD (667>428 m/z), NADH (668>651 m/z), and NADP (746>604 m/z). Enrichment was determined by dividing 18O labeled metabolite peak areas by 16O labeled metabolite peak areas (Tracer/Tracee ratio). The ratios were corrected by subtracting the Tracer/Tracee ratio at each time point by the initial ratio (Tracer/Tracee ratio at time point zero). To ensure that this method was accurate, cells supplemented with the same concentration of non-labeled NR or NMN were grown in parallel with cells supplemented with isotopologues. At each time point, the Tracer/Tracee ratio was corrected by subtracting the ratio found in heavy labeled fed cells by the ratio found in non-heavy labeled cells. These ratios were compared to correcting the ratio with the initial time point. Ratios were not significantly different between conditions (data not shown).

4.5 Results

NMN is Dephosphorylated Extracellularly and Contributes to the Intracellular NAD+ Pool Slower than NR

Hepg2 cells were grown in the presence of either [18O1]-NR or [18O1]-NMN at a concentration of 10 μM after a 24 hour B3 vitamin starvation. Given that labeled NMN nor NR was not detectable at time point zero, any detectable heavy labeled NMN, NR, or other labeled metabolite is derived from the supplemented heavy precursor. Intracellularly, Tracer/Tracee ratios were determined as a corrected ratio of heavy-to-light metabolite peak areas. Overall, the rate of enrichment was greater when cells were fed labeled NR compared to NMN with NAD+ enrichment reaching 2.6 ± 0.3 compared to 0.91 ± 0.04 after 24 hours (P < 0.01), suggesting NR is more efficiently utilized intracellularly to produce NAD+ (Figure 4.2a and 4.2b).

60

Extracellular labeled NMN, NR, NA, and Nam were quantified. Labeled Nam nor labeled NA were not detected upon feeding with NMN or NR (Figure 4.2c and 4.2d), revealing NMN and NR are likely not greatly hydrolyzed in the presence of these cells. NR slowly decreased in concentration from 11 ± 0.84 μM at 30 minutes post incubation to 2.3 ± 1.2 μM after 24 hours (P < 0.01)(Figure 4.2b). Strikingly, NMN dramatically decreased over the 24 hours falling from 10 ± 0.97 to 0.34 ± 0.01 μM (P < 0.001)(Figure 4.2d). Concurrent with this decrease, labeled NR rose in a nearly linear fashion compared to the disappearance of labeled NMN (Figure 4.2d). Intracellular enrichment of NR slowly rose with the appearance of its extracellular pool after NMN feeding, but spiked after NR feeding. Together, these findings reveal NMN is dephosphorylated extracellularly and indicate this process is rate-limiting for its intracellular utilization. NR was kinetically superior in enriching the intracellular NAD metabolome compared to NMN.

4.6 Discussion

NR (32, 37, 50, 52, 54-56) and NMN (34, 123-125) both counter metabolic and age related disorders. NR and NMN are studied as separate pharmacological entities, both augmenting NAD+ through separate pathways but converging due to the action of NAD+ in sirtuin activities. NR is either phosphorylated to NMN through the NRK pathway (7) or phosphorylized to Nam (126). However, NR increases five-fold after NMN injection (34) and other studies indicate NMN depends upon CD38, CD73, and NRK (35, 121, 127) for its utilization and suggest that NMN is metabolized to Nam and NR extracellularly. These studies did not eliminate the possibility that extracellular dephosphorylation is non-rate limiting and that NMN, though biochemically acting as NR, behaves identically as NR. We used LC-MS/MS and stable isotope labeled NR and NMN to measure their kinetic effect on the NAD metabolome. NR contributed to the intracellular NAD metabolome more rapidly than NMN and increased NAD+ by more than 2 fold after 24 hours, indicating NR is kinetically superior to NMN. In line with the

61

intracellular findings, extracellular NMN rapidly and dramatically decreased over 24 hours as extracellular NR rose. No labeled Nam was detected, suggesting hydrolysis, presumably as catalyzed by CD38, was not the primary route of extracellular NMN metabolism in these cells. The relationship of the two labeled compounds was seemingly linear and agreed with genetic evidence that NMN is dephosphorylated before it is salvaged. NR and NMN are not identical, interchangeable entities.

The implication of these data to a biological setting remain to be shown in vivo. The possibility remains that NMN is an endogenous circulating NAD+ precursor as has been suggested (17). An extracellular Nampt (eNampt) is enzymatically active and could produce a constant amount of NMN from Nam. In the same report, NMN was reported to circulate at as high as 80 μM, leading some to suggest that NMN is a type of NR carrier that is dephosphorylated to supply NR around the body (2). However, the ability of eNampt to produce NMN appears unlikely given undetectable levels of its co-substrate 5-phosphribosyl-1- pyrrophosphate (18). Further, our group and others have failed to detect NMN in plasma (Chapter 5)(18). These discordant results may be due to the employment of HPLC versus LC- MS. HPLC is technically more quantitative as the detector is not at the mercy of gas phase reactions such as mass spectrometers are, but HPLC UV-vis methods are incapable of providing selectivity for the metabolite of interest (Chapter 1.1-1.2). This is a very prescient example of the need for improved and accepted analytical procedures for the study of the NAD metabolome.

62

CHAPTER 5

NICOTINAMIDE RIBOSIDE IS UNIQUELY BIOAVAILABLE IN MOUSE AND MAN

Samuel A.J. Trammell1,2, Mark S. Schmidt1, Benjamin J. Weidemann1, Philip Redpath3, Frank Jaksch4, Ryan W Dellinger4, Marie E. Migaud3, and Charles Brenner1,2

1Department of Biochemistry, 2 Interdisciplinary Graduate Program in Genetics, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, USA

3John King Laboratory, School of Pharmacy, Queens University Belfast, Belfast, UK 4ChromaDex, Inc., Irvine, CA 92618

5.1 Distribution of Work

BJW, FJ, RWD, MEM, CB, and I designed the experiments. MSS, BJW, and I performed the experiments and analyzed the data with FJ, RWD, MEM, and CB. PR performed the synthesis under direction of MEM. CB and I wrote the manuscript. All authors edited the manuscript and figures.

5.2 Abstract

Nicotinamide riboside (NR) is in wide use as an NAD+ precursor vitamin. Here we conducted experiments to determine the time and dose-dependent effects of NR on blood and liver NAD+ metabolism in people and mice, respectively. We report that human blood cell NAD+ can rise as much as 2.7-fold with a single dose of NR, that NR elevates mouse hepatic NAD+ with distinct and superior pharmacokinetics to those of nicotinic acid (NA) and nicotinamide (Nam), and that single doses of 100, 300, and 1000 mg of NR provide a dose-dependent increase in the blood cell NAD+ metabolome in the first clinical trial of NR pharmacokinetics. We also report that nicotinic acid adenine dinucleotide (NAAD), which was not thought to be en

65

route for conversion of NR to NAD+, is formed from NR and that the rise in NAAD is a highly sensitive biomarker of effective NAD+ supplementation.

5.3 Introduction

Nicotinamide adenine dinucleotide (NAD+) is the central redox coenzyme in cellular metabolism (128, 129). NAD+ functions as a hydride group acceptor, forming NADH with concomitant oxidation of metabolites derived from carbohydrates, amino acids and fats. The NAD+/NADH ratio controls the degree to which such reactions proceed in oxidative versus reductive directions. Whereas fuel oxidation reactions require NAD+ as a hydride acceptor, the processes of gluconeogenesis, oxidative phosphorylation, ketogenesis, detoxification of reactive oxygen species, and lipogenesis require reduced co-factors, NADH and NADPH, to act as hydride donors (Figure 5.1). In addition to its role as a coenzyme, NAD+ is the consumed substrate of enzymes such as poly-ADPribose polymerases (PARPs), sirtuins and cyclic ADPribose synthetases (128). In redox reactions, the nucleotide structures of NAD+, NADH, NADP+ and NADPH are preserved. In contrast, PARP (130), sirtuins (131) and cyclic ADPribose synthetase (132) activities hydrolyze the glycosidic linkage between the nicotinamide (Nam) and the ADPribosyl moieties of NAD+ to signal DNA damage, alter gene expression, control post- translational modifications, and regulate calcium signaling.

In animals, NAD+ consuming activities and cell division necessitate ongoing NAD+ synthesis, either through a de novo pathway that originates with tryptophan or via salvage pathways from three NAD+ precursor vitamins, Nam, nicotinic acid (NA) and nicotinamide riboside (NR) (129). Dietary NAD+ precursors, which include tryptophan and the three vitamins, prevent pellagra. Though NR is present in milk (7) (Chapter 3), the cellular concentrations of NAD+, NADH, NADP+ and NADPH are much higher than those of any other NAD+ metabolites (1, 31), such that dietary NAD+ precursor vitamins are largely derived from enzymatic breakdown of NAD+. Put another way, though milk is a source of NR, the more abundant

66

sources of NR, Nam and NA are any whole foodstuff in which cellular NAD+ is broken down to these compounds. Human digestion and the microbiome (133) play roles in the provision of these vitamins in ways that are not fully characterized.

Different tissues maintain NAD+ levels through reliance of different biosynthetic routes (87, 134) (Figure 5.1). Because NAD+ consuming activities frequently occur as a function of cellular stresses (130) and produce Nam, the ability of a cell to salvage Nam into productive NAD+ synthesis through Nam phosphoribosyltransferase (NAMPT) activity versus methylation of Nam to N-methylnicotinamide (MeNam) regulates the efficiency of NAD+-dependent processes (135). NAD+ biosynthetic genes are also under circadian control (57, 136) and both NAMPT expression and NAD+ levels are reported to decline in a number of tissues as a function of aging and overnutrition (34, 124, 137, 138).

High dose NA but not high dose Nam has been used by people for decades to treat and prevent dyslipidemias, though its use is limited by painful flushing (139, 140). Though it only takes ~15 mg per day of either NA or Nam to prevent pellagra, pharmacological doses of NA can be as high as 2 – 4 g. Despite the >100-fold difference in effective dose between pellagra prevention and treatment of dyslipidemias, we proposed that the beneficial effects of NA on plasma lipids might simply depend on function of NA as an NAD+ boosting compound (128). According to this view, sirtuin activation would likely be part of the mechanism because Nam is an NAD+ precursor in most cells (87, 134) but is a sirtuin inhibitor at high doses (141).
Based on the ability of NR to elevate NAD+ synthesis, increase sirtuin activity and extend lifespan in yeast (7, 142), NR has been employed in mice to elevate NAD+ metabolism and improve health in models of metabolic stress. Notably, NR allowed mice to resist weight gain on high fat diet (37) and to prevent noise-induced hearing loss (56). Data indicate that NR is a mitochondrially favored NAD+ precursor (35) and, indeed, in vivo activities of NR have been interpreted as depending on mitochondrial sirtuin activities (37, 56), though not to the exclusion of nucleocytosolic targets (143, 144). Similarly, nicotinamide mononucleotide (NMN), the

67

phosphorylated form of NR, has been used to treat declining NAD+ in mouse models of overnutrition and aging (34, 124). Because of the abundance of NAD+-dependent processes, it is not known to what degree NAD+ boosting strategies are mechanistically dependent on particular molecules such as SIRT1 or SIRT3. In addition, the quantitative effect of NR on the NAD+ metabolome has not been reported in any system.

To translate NR technologies to human beings, it is necessary to determine its oral availability and the extent and means by which NR is converted to NAD+ metabolites in different cell types. Here we performed targeted quantitative NAD+ metabolomic measurements on human blood samples in an n=1 human experiment in which a healthy 52 year-old man took 1000 mg of NR chloride daily for 7 days. These data indicated that blood cellular NAD+ rose 2.7- fold after a single dose of NR and that nicotinic acid adenine dinucleotide (NAAD) unexpectedly increased at least 45-fold. To determine the precise time course of oral NR availability to the liver and to determine if NR is converted to NAD+ in a unique manner with respect to Nam and NA, we performed a detailed analysis of 128 mice, who were provided with saline or vitamin gavage in a manner that eliminated the possibility of circadian artifacts. These data indicated that NR boosts hepatic NAD+ and NAD+ consuming activities to a greater degree than Nam or NA and with unique kinetics. As expected, oral NA results in a peak of hepatic NAAD prior to a broad peak of NAD+ accumulation. However, NR and Nam resulted in NAAD peaks coincident with elevated NAD+. Just as NR supplementation produced more hepatic NAD+ than did Nam supplementation, so NR also produced a greater degree of hepatic NAAD.

To address whether the unexpected appearance of NAAD might have been due to inhibition of de novo biosynthesis, we synthesized NR with heavy atoms incorporated into the Nam and ribosyl moieties and discovered that oral NR serves as a biosynthetic precursor of elevated NAAD. Finally, we performed a crossover clinical study with 12 healthy human subjects who took single doses of 100 mg, 300 mg or 1000 mg of NR chloride. This study

68

demonstrated that NR supplementation increases blood cell NAD+ metabolism at all doses and validated elevated NAAD as an unexpected, highly sensitive biomarker of boosting NAD+.

5.4 Methods

Materials and Reagents
NR chloride (NR Cl) was synthesized under GMP conditions. Me2PY and Me4PY were

purchased from TLC PharmaChem Inc. (Vaughan, Ontario, Canada). All other unlabeled analytes were purchased from Sigma-Aldrich (St. Louis, MO, US) at highest purity. Internal standards [18O1]- Nam and [18O1]-NR were prepared as described (64, 65). [18O1-D3]-MeNam was prepared through alkylation of [18O1]-Nam with deuterated iodomethane. 13C-NA and [D4]- NA were purchased from Toronto Chemical Research (Toronto, Ontario, Canada) and C/D/N Isotopes, Inc. (Pointe-Claire, Quebec, Canada), respectively. To prepare [13C, D1]-NR, we first converted 13C-NA to 13C-Nam (145) and D-[2-D1]-ribose (Omicron Biochemicals, South Bend, IN, US) to the labeled D-ribofuranose-tetraacetate (146). The labeled D-ribofuranose- tetraacetate and Nam were then used to synthesize double-labeled NR (147). [13C]-labeled nucleotides and nucleosides were prepared by growing yeast in U-13C-glucose and extracting as described (1).

Mice
Twelve week old male C57Bl/6J mice (Jackson Laboratories, Bar Harbor, ME) were

housed 3-5 mice per cage on a chow diet (Teklad 7013) for one week prior to the experiment. Body weight-matched groups were given either 185 mg NR Cl/kg body weight or equimole amounts of NA or Nam by saline gavage. On each sacrifice day, a saline injection was performed and served as time point zero and an additional saline gavage time course was performed. To avoid circadian effects, time courses were established such that all sacrifices were performed at ~ 2 pm. With protocols approved by the University of Iowa Office of Animal

69

Resources, mice were live-decapitated and the medullary lobe of the liver was freeze-clamped at liquid nitrogen temperature. Tissue was stored at -80 °C prior to analysis.

N of 1 Human Experiment
After overnight fasting, a healthy 52 year old male self-administered 1000 mg of NR

chloride orally at 8 am on 7 consecutive days. Blood and urine were taken for quantitative NAD+ metabolomic analysis. He took 0.25 gram of NA to assess sensitivity to flushing and self- reported painful flushing that lasted 1 hour. No flushing was experienced on NR.

Clinical Trial
A randomized, double-blind, three-arm crossover pharmacokinetic study of oral NR

chloride was performed at 100, 300, and 1000 mg doses. Twelve healthy, non-pregnant subjects (6 male, 6 female) between the ages of 30 and 55 with body mass indices of 18.5 – 29.9 kg/m2 were recruited and randomized to one of three treatment sequences after screening and passing eligibility criteria. Subjects taking multi-vitamins, vitamin B3 in any form, or subsisting on diets that could contain unusually high amounts of NA, Nam or NR were excluded. Complete exclusion criteria are provided in Supplementary Materials.

Overnight fasted subjects received a single morning dose of either 100 mg, 300 mg, or 1000 mg of NR on three test days separated by 7-day periods in which no supplement was given. To evaluate pharmacokinetics, blood was collected and separated into plasma and PBMC fractions for analysis of the NAD+ metabolome at pre-dose and again at 1, 2, 4, 8, and 24 hr. Urine was collected pre-dose and in 0-6 hr, 6-12 hr and 12-24 hr fractions. Safety, vitals, biometrics, complete blood counts and a comprehensive metabolic panel were assessed at time zero and 24 hr after each dose.

The study was reviewed and approved by the Natural Health Products Directorate, Health Canada and Institutional Review Board Services, Aurora, Ontario. Written informed consent was obtained from each subject at the screening visit prior to all study-related activities.

70

Sample Preparation and LC-MS
Two extractions were performed on all samples for quantitative targeted metabolomic

analysis (1). PBMCs were thawed on ice and extracted with three volumes of acetonitrile in the presence of each internal standard mixture. For group B analytes, extracts were passed through Phenomenex Phree© SPE cartridges (Torrance, CA, USA) and combined with a subsequent acetonitrile wash prior to drying. All extracts were dried via speed vacuum. On the day of analysis, samples were re-suspended and placed in a chilled autosampler. Murine liver was processed as detailed in (Chapter 2.2: Quantitation of the Oxidized NAD Metabolome in Liver). Standard curves were prepared in water and processed in the same manner as samples.

Separation and quantitation of analytes was performed with a Waters Acquity LC interfaced with a Waters TQD mass spectrometer operated in positive ion multiple reaction monitoring mode. Enrichment analysis was performed with a Waters Q-TOF Premier mass spectrometer operated in positive ion, full scan mode with the same LC conditions as described for non-enrichment experiments. Details are provided in Supplemental Methods.

Statistical Analyses
Statistical analyses were performed in GraphPad Prism version 6.00 for Windows, (La

Jolla, CA, USA). Murine liver data were analyzed using a two-way ANOVA, whereas human blood cells were analyzed using a repeated, two-way ANOVA. Holm-Sidak and Tukey’s multiple comparisons tests were performed when comparing more than two groups. AUCs in blood cells were calculated after subtracting pre-dose metabolite concentrations of each experimental series. For mouse data, AUCs were calculated as described (148) after subtracting the saline group for that day and propagating error. All other tests are stated in the text. Data are expressed as means ± standard error of the mean.

71

5.5 Results

Oral NR Increases the Blood NAD Metabolome in a Healthy Adult Male
A healthy 52 year old male (65 kg) contributed blood prior to seven days of oral NR

chloride (1000 mg/morning dose). Blood was taken an additional 9 times in the first day and at 24 hours after the first and last oral dose. Blood was separated into a peripheral blood mononuclear cell (PBMC) fraction and a plasma fraction prior to quantitative NAD+ metabolomic analysis by LC-tandem MS (1), which was expanded to quantify methylated and oxidized metabolites of Nam. As shown in Table 5.1, the PBMC NAD+ metabolome was unaffected by NR for the first 2.7 hrs. In six measurements from time zero through 2.7 hrs, NAD+ had a mean concentration of 18.5 μM while Nam had a mean concentration of 4.1 μM and the methylated and oxidized Nam metabolite, N-methyl-2-pyridone-5-carboxamide (Me2PY) had a mean concentration of 2.6 μM. However, at 4.1 hours post-ingestion, PBMC NAD+ and Me2PY increased by factors of 2.3 and 4.2, respectively.

In yeast, deletion of nicotinamide riboside kinase 1 (NRK1) does not eliminate utilization of NR (142). NR can be phosphorylyzed to Nam by purine nucleoside phosphorylase and still contribute to NAD+ synthesis through Nam salvage (126, 142). However, as shown in Table 5.1, Nam concentration in the human subject’s PBMCs merely fluctuated in a range of 2.6 μM to 7.1 μM throughout all 11 observations. The 4.2-fold increase in Me2PY concentration at the 4.1 hour time point suggests that increased cellular NAD+ accumulation is accompanied by increased NAD+ consuming activities that are linked to increased methylation and oxidation of the Nam product.

In the human subject’s PBMCs at 7.7 and 8.1 hours post ingestion, NAD+ and Me2PY reached peak levels, increasing above baseline concentrations by 2.7-fold and 8.4-fold respectively. At these time points, unexpectedly, NAAD, the substrate of glutamine-dependent NAD+ synthetase (12), which is only expected to be produced in biosynthesis of NAD+ from

72

tryptophan and NA (129), was elevated from less than 20 nM to as high as 0.91 μM. Whereas NAAD lagged the rise in PBMC NAD+ by one time point, the rise in PBMC NAD+ was not as pronounced as the spike in NAAD, which was at least 45-fold above the baseline level. Though contrary to expectations, these data suggested that NR might be incorporated into NAAD after formation of NAD+ and chased back to the NAD+ peak as NAD+ declines.

Complete NAD+ metabolomic data from the human subject’s PBMC fraction, blood plasma fraction and urine are provided in Table 5.1 and as supplementary Tables 5.2 and 5.3, respectively. These data show that all of the phosphorylated compounds—NAMN, NAAD, NAD+, NADP+, NMN and ADPR—are found exclusively in PBMCs. Notably, the peak of NADP+, which represents cellular NADP+ plus NADPH that was oxidized in extraction, and the peak of ADPR, which signals an increase in NAD+ consuming activities, co-occur with the peak of NAD+. Using methods that are optimized for recovery of nucleotides, NR was not recovered. The major time-dependent waste metabolite in plasma and urine was Me2PY, which rose about 10-fold from pre-dose to time points after NAD+ peaked in PBMCs.

Oral NR, Nam and NA Elevate Hepatic NAD+ with Distinctive Kinetics
Based on known NAD+ biosynthetic pathways (35), it was difficult to understand how

NAAD increased in human PBMCs after an oral dose of NR. Though NR did not elevate Nam in blood samples at any time during the n=1 experiment, it remained possible that a fraction of NR was converted to Nam prior to salvage synthesis to NAD+. Such conversion to Nam might allow bacterial hydrolysis of Nam to NA by pncA gene products—potentially in the gut (133)—and subsequent conversion to NAD+ through an NAAD intermediate.

We designed an experiment in which a mouse’s daily dose of NR (185 mg/kg)17 and the mole equivalent doses of Nam and NA were provided to mice by oral gavage. To ascertain the

17 This dosage is equivalent to a 65 kg human being ingesting 1 gram NR Cl based upon surface-to-area ratio between a human and mouse (149).

73

time course by which these vitamins boost the hepatic NAD+ metabolome without the complication of circadian oscillation of NAD+ metabolism (57, 136), we sacrificed all mice at approximately 2 pm. Thus, vitamin administration by gavage was performed at 0.25 hour, 1 hour, 2 hours, 4 hours, 6 hours, 8 hours and 12 hours prior to sacrifice. To stop metabolism synchronously, mouse livers were harvested by freeze-clamping. As shown in Figure 5.2, we additionally performed saline gavages at all time points and sacrificed mice for quantitative NAD+ metabolomic analysis to ensure that animal handling does not alter the levels of NAD+ metabolites. The flat time courses of saline gavages established the suitability of this method. Baseline levels of hepatic NAD+ metabolites at 2 pm were 1000 ± 35 pmol/mg for NAD+, 230 ± 29 pmol/mg for Nam, 210 ± 20 pmol/mg for NADP+, 66 ± 13 pmol/mg for ADPR, and < 15 pmol/mg of all other NAD+ metabolites. Hepatic levels of NA, NAR, NAMN, and NAAD have baselines of below 4 pmol/mg. As a point of orientation to quantitative metabolomics in tissue samples, we note that 1000 pmol/mg is ~1 mM, 200 pmol/mg is ~200 μM, and 10 pmol/mg is ~10 μM.

Targeted NAD+ metabolomics (1, 31) allows simultaneous assessment of functionally important and highly regulated metabolites such as NAD+ and NADP+ along with metabolites that could serve as biomarkers of biosynthetic processes, such as NA, NAR, NAMN, NR, NMN and NAAD. In addition, quantification of increases in ADPR, Nam, MeNam, Me2PY and N- methyl-4-pyridone-5-carboxamide (Me4PY) on a common absolute scale with NAD+ allows an assessment of increased NAD+ consuming activities associated with NAD+ precursor vitamin supplementation.

Hepatic concentrations of 13 NAD+ metabolites were quantified in 3 to 4 mice at 7 time points after gavage of saline and after gavage of each NAD+ precursor vitamin. In addition, 4 mice were analyzed after control 2 pm sacrifices without gavage. Each vitamin produced temporally distinct pattern of hepatic NAD+ metabolites. Consistent with the rapid phosphorylation of NR and NAR by NR kinases (33), the only NAD+ metabolites that do not

74

produce hepatic peaks as a function of gavage of NAD+ precursor vitamins are NR and NAR (Supplementary Tables and Figures: Figure 5.5a-b). The accumulation curves of some metabolites as a function of each vitamin were strikingly similar. For example, the accumulation of NMN (Figure 5.2a) is nearly identical to that of NAD+ (Figure 5.2b) and NADP+ (Figure 5.2c) though at a scale of ~1:400:40. In addition, the accumulation of Me4PY (Figure 5.2f) is nearly identical to that of Me2PY (Figure 5.5c).

As shown in Figure 5.2b, NA produced the least increase in hepatic NAD+ but also was 4-6 hours faster than NR and Nam in the kinetics of hepatic NAD+ accumulation. When NA was provided by oral gavage, liver NA peaked (340 ± 30 pmol/mg) in 15 minutes (Figure 5.2g). Hepatic NA appearance was followed by an expected peak of 220 ± 29 pmol/mg of NAAD at 1 hr post-gavage (Figure 5.2i) and a rise in hepatic NAD+ from 990 ± 25 pmol/mg baseline to 2200 ± 150 pmol/mg at 2 hrs (Figure 5.2b). Hepatic NADP+ due to NA (Figure 5.2c) rose in parallel to that of hepatic NAD+. In the hours after gavage of NA, as hepatic NAD+ and NADP+ fell, there was clear evidence of enhanced NAD+ consuming activities with significant rises in ADPR (Figure 5.2j), Nam (Figure 5.2d), MeNam (Figure 5.2e), Me2PY (Figure 5.5c) and Me4PY (Figure 5.2f). Thus, a bolus oral administration of NA doubled hepatic NAD+ from ~1 mM to ~2 mM through expected intermediates and produced an increase in NAD+ consumption and the methylated products, MeNam, Me2PY and Me4PY.

As shown in Figure 5.2g, oral Nam was clearly not used by the liver as NA because it did not produce a peak of NA at any time after gavage. Though there was an increase in hepatic NAD+ 2 hrs after Nam gavage, the Nam gavage drove increased hepatic NAD+ accumulation from 2 to 8 hrs with a peak at 8 hrs (Figure 5.2b). Nam gavage produced two peaks of Nam in the liver (Figure 5.3d). The first peak was at 15 min, consistent with simple transport of the vitamin to the liver. The second broad peak was coincident with elevation of NAD+ and NADP+ (Figures 5.2c-d) and with elevation of the NAD+ consuming metabolomic signature of ADPR

75

(Figure 5.2j), MeNam, Me4PY and Me2PY (Figure 5.2e-f and 5.8 Supplemental Tables and Figures: 5.5c).

Of the metabolites associated with NAD+ consuming activities, ADPR is the only one that must be formed from NAD+ because Nam, MeNam and the oxidized forms of MeNam could appear in liver from the gavaged Nam without conversion to NAD+. Interestingly, of three NAD+ precursor vitamins provided in bolus at equivalent oral doses, Nam provided the least increase in ADPR (Figure 5.2j). Whereas the area under the curve (AUC) of the Nam-driven rise in hepatic NAD+ indicated a ~50% advantage of Nam over NA (Figure 5.2b), there was a >50% deficit in Nam-driven ADPR accumulation versus NA (Figure 5.2j). This is consistent with the idea that high dose NA, though not an ideal hepatic NAD+ precursor, is capable of improving reverse cholesterol transport to a much greater degree than Nam because high dose Nam inhibits sirtuins(128).

As shown in Figure 5.1, Nam is expected to proceed through NMN but not NR, NAR, NAMN or NAAD en route to forming NAD+. Though there was no elevation of hepatic NR or NAR with oral Nam, there was also little elevation of hepatic NMN—this metabolite never reached a mean value of 5 pmol/mg at any time after Nam administration (Figure 5.2a). Surprisingly, as shown in Figure 5.2i, 2-4 hrs after oral Nam, NAAD was elevated to nearly 200 pmol/mg from a baseline of less than 2 pmol/mg. Elevated NAAD occurred during the broad peak of elevated hepatic NAD+ and NADP+ (Figures 5.2b-c). These data suggest that the rise in NAAD is a biomarker of increased NAD+ synthesis and does not depend on the conventionally described precursors of NAAD, namely NA and tryptophan.

As shown in Figure 5.2b, NR elevated hepatic NAD+ by more than 4-fold with a peak at 6 hr post-gavage. NR also produced the greatest elevation of NMN (Figure 5.2a), NADP+ (Figure 5.2c), Nam (Figure 5.2d), NAMN (Figure 5.2h), NAAD (Figure 5.2i) and ADPR (Figure 5.2j) both in terms of peak height and AUC. Importantly, though gavage of Nam produces a peak of Nam in the liver at 15 min, the peak of Nam from NR gavage corresponds to the peak of NAD+, NMN,

76

NADP+ and ADPR. These data establish that oral NR has clearly different hepatic pharmacokinetics than oral Nam. More NAD+ and NADP+ were produced from NR than from Nam. In addition, there was three times as much accumulation of ADPR, indicating greater NAD+ consuming activities. In an accompanying manuscript, we showed that hepatic cells convert NMN extracellularly to NR and that both NMN and NR depend on expression of NRK1 for conversion to cellular NAD+ (150).

As was seen in the n=1 human blood experiment, at time points in which the abundant NAD+ metabolites, NAD+ and NADP+, were elevated by NR by ~2-fold or more, NAAD rose from undetectable levels to approximately 10% of the level of NAD+, thereby becoming a highly sensitive biomarker of increased NAD+ metabolism. Though compounds such as MeNam, Me2PY and Me4PY are also correlated with increased NAD+ synthesis, they can be produced without NAD+ synthesis. While MeNam, Me2PY and Me4PY are waste products that can no longer contribute to elevated NAD+ or NADP+, NAAD is functional NAD+ precursor.

NR Directly Contributes to Murine Liver NAAD
The appearance of hepatic NAAD after murine gavage of Nam or NR, and of hepatic

NAMN after gavage of NR suggested that there is retrograde NAD+ metabolic flux when NAD+ and NADP+ levels are high. Alternatively, high levels of NAD+ metabolites might inhibit glutamine-dependent NAD+ synthetase, thereby resulting in accumulation of NAMN and NAAD derived from tryptophan. To test whether NR is incorporated into the peak of NAAD that appears after gavage of NR, we synthesized NR chloride with incorporation of deuterium at the ribosyl C2 and 13C into the carbonyl of the Nam moiety. This double-labeled NR was provided to 15 mice by oral gavage at an effective dose of 185 mg/kg with the same experimental design used in pharmacokinetic analysis of the three vitamins. The effect of labeled oral NR on the hepatic NAD+ metabolome was first assessed at 2 hours after gavage—a time prior to the rise in the steady-state level of NAD+ (Figure 5.2b).

77

As shown in Figure 5.3a-b, at 2 hrs, 54% of the NAD+ and 32% of the NADP+ contained at least one heavy atom while 5% of the NAD+ and 6% of the NADP+ incorporated both heavy atoms. Because > 50% of hepatic NAD+ incorporates label prior to a rise in NAD+ accumulation, it is clear that the NAD+ pool is dynamic. As shown in Figure 5.3c-d, the majority of hepatic Nam and MeNam following gavage of double-labeled NR incorporated a heavy atom, which is necessarily the 13C in Nam. Because NR drives increased NAD+ synthesis and ADPR production (Figure 5.2), the liberated singly labeled Nam would thereby become incorporated into NMN and NAD+ in competition with double labeled NR, thereby limiting subsequent incorporation of both labels into the NAD+ pool.

Appearance of a peak of NAAD after NR administration could either be due to inhibition of de novo synthesis of NAD+ or from a novel retrograde pathway that produces NAAD from NAD+. If NAAD is not derived from ingested NR, then it should not incorporate heavy atoms. However, if NAAD is derived from ingested NR, then it should incorporate heavy atoms that reflect the rate at which the retrograde reactions occur with respect to NAD+ consuming activities and the degree of heavy atom incorporation into NAD+. As shown in Figure 5.3e, at the 2 hr time point, NAAD contained roughly the same heavy atom composition as NAD+ (Figure 5.3a), i.e., 45% contained at least one heavy atom and 8% incorporated both heavy atoms. Thus, NR is the biosynthetic precursor of NAD+, NADP+ and NAAD. The data suggest that the process that converts NAD+ to NAAD occurs at high NAD+ concentrations at a rate comparable to the rate of NAD+ turnover to Nam.

Incorporation of the Nam and ribosyl moieties of NR into NAAD establishes this metabolite as both a biomarker of increased NAD+ metabolism and as a direct product of NR utilization.

78

NR Increases Blood Cell NAD+ Metabolism in Human Subjects
The n=1 human experiment illustrated the potential of 1000 mg NR to boost human

NAD+ metabolism. We therefore conducted a controlled experiment with 12 consented healthy men and women to determine the effect of three single doses of NR on PBMC, plasma and urine NAD+ metabolites with monitoring of subjects for potential adverse events. Considering that the recommended daily allowance (RDA) of vitamin B3 as Nam or NA is ~15 mg per adult, we chose to test three doses of the higher molecular weight compound NR chloride (100 mg, 300 mg and 1000 mg) that correspond to 2.8, 8.4 and 28 times the RDA. Participants were randomized to receive doses of NR in different sequences with 7-day washout periods between data collection. Participants and investigators were blinded to doses. Blood and urine collections were performed over 24 hours following each dose. Participants were asked to self-report any perceived discomforts.

At 500 mg of niacin, 33 of 33 participants experienced flushing compared to 1 of 35 participants who received a placebo (151). In this study, two individuals self-reported flushing at the 300 mg dose but not at the 100 mg or 1000 mg dose, and two individuals self-reported feeling hot at the 1000 mg dose but not at lower doses. Over the total of 36 days of observation of study participants, there were no serious adverse events and no events that were dose- dependent. To assess whether NR was associated with authentic and dose-dependent episodes of flushing, future experiments will incorporate a validated flushing symptom questionnaire(152).

As shown in Figure 5.4 and Tables A.1 and A.2 (Appendix A), the NAD+ metabolome was quantified in the PBMC and plasma fractions at pre-dose and at 1, 2, 4, 8 and 24 hours after receiving oral NR. Urinary NAD+ metabolites (Appendix A: Table A.3) were quantified in pre-dose, 0-6 hour, 6-12 hour, and 12-24 hour collections.

As shown in Figure 5.2, inbred, chow-fed male mice supplemented with NAD+ precursor vitamins by gavage and sacrificed at the same time of day produced hepatic NAD+ metabolomic

79

data with little variation. However, blood samples from people in a clinical study exhibited a wider degree of variation, apparently due to differing baseline levels of metabolites and variable pharmacokinetics, both of which might be due to genetic and nutritional changes between human subjects (Figure 5.4). In PBMCs, 8 key metabolites were quantified in at least 10 subjects at all time points at each dose. For each of these metabolites, we plotted the average concentration as a function of dose and time, calculated whether NR elevated that metabolite, plotted the averaged peak concentration of the metabolite as a function of dose, and calculated the dose-dependent AUC of the metabolite attributable to NR supplementation.

Collapsing the data into pre-dose versus 24 hr levels of each metabolite at all doses, NR significantly elevated PBMC NAD+ (Figure 5.4b), MeNam (Figure 5.4d) and Me2PY (Figure 5.4e), and significantly elevated PBMC NAAD (Figure 5.4f) at 8 hrs. In contrast, NR did not produce a statistically significant all-dose elevation of NMN (Figure 5.4a) or Nam (Figure 5.4c) at any time point.

The averaged peak concentration of MeNam (Figure 5.4d), Me2PY (Figure 5.4e) and NAAD (Figure 5.4f) increased monotonically with increased doses of NR. Of these metabolites, only NAAD was below the detection limit in individuals before they took NR. Nam (Figure 5.4c) exhibited no tendency toward higher cellular concentrations with higher doses of NR. NMN (Figure 5.4a) and NAD+ (Figure 5.4b) rose to higher concentrations of ~2 μM and 20 μM, respectively, in people given 300 mg and 1000 mg doses of NR than in people given 100 mg doses of NR. Thus, 100 mg supplementation produced an average ~4 ± 2 μM increase in PBMC NAD+, whereas the two higher doses produced average ~6.5 ± 3.5 μM increases in PBMC NAD+.

As was first seen in the n=1 human experiment and in the mouse liver experiments, NAAD is the most sensitive biomarker of effective NAD+ supplementation because it is undetectable in the blood of people prior to dosing. At all doses, the peak shape of NAAD indicated that NAD+ metabolism is most greatly boosted at 8 hrs with significant

80

supplementation at 4 hours and significant supplementation remaining at 24 hours. At the 8 hour peak, the average concentration of NAAD was elevated to 0.56 ± 0.26, 0.74 ± 0.27 and 1.24 ± 0.51 μM in PBMCs from volunteers taking 100, 300 and 1000 mg single doses of NR, respectively.

Finally, we plotted pre-dose-subtracted AUCs of each metabolite as a function of dose of NR. With the exception of Nam, the levels of which were unaffected by NR, NR produced or tended to produce dose-dependent elevation of the entire NAD+ metabolome (Figure 5.4).

In the plasma, levels of MeNam, Me2PY and Me4PY also rose in a dose-dependent manner and were identified at concentrations similar to those in the PBMC fraction. The methylated and oxidized Nam derivatives were accompanied by low levels of NAR, which increased with increased doses of NR. Urinary metabolites were similar to plasma metabolites.

5.6 Discussion

Despite more than 75 years of experience with human use of NA and Nam (153) and more than a decade of preclinical work on NR (7), there has never been a quantitative metabolomic or pharmacokinetic comparison of the three NAD+ precursor vitamins in any system. In terms of elevation of mouse liver NAD+, here we show that NR is more orally bioavailable than Nam, which is more orally bioavailable than NA (Figure 5.2b). The three precursors also differ in the degree to which they promote accumulation of ADPR, a measure of NAD+ consuming activities. As shown in Figure 5.2j, the ability of NR to elevate ADPR exceeded that of Nam by ~3-fold. On a molar basis, this qualifies NR as the favored NAD+ precursor vitamin for increasing NAD+ and NAD+ consuming activities in mouse liver.

NR, Nam and NA each have unique pharmacokinetic profiles in mouse liver, both in terms of the kinetics of NAD+ formation and the population of NAD+ metabolites as a function of time. As shown in Figure 5.2d, Nam is the only vitamin precursor of NAD+ that produces

81

elevated hepatic Nam 15 min after oral administration and, as shown in Figure 5.2g, NA is the only precursor that produces elevated NA 15 min after oral administration.

When PBMCs were analyzed from the first human volunteer taking 1000 mg of NR, NAAD was observed to increase at least 45-fold from a baseline of less than 20 nM to a peak value of nearly 1 μM. This occurred concomitant with a rise in NAD+ from ~18.5 μM to 50 μM. NAAD was also observed to be elevated in the liver when mice were provided with NAD+ precursor vitamins. Surprisingly, NA, the only precursor expected to proceed to NAD+ through an NAAD intermediate, produced the least NAAD. Indeed, though Nam and NR never produced peaks of hepatic NA or NAR, both produced peaks of hepatic NAAD during the periods in which these compounds elevated hepatic NAD+. The temporal basis of the NAAD excursions suggested that elevating NAD+ (Figure 5.2b) not only stimulates accumulation of NAD+ consumption products ADPR (Figure 5.2j), Nam (Figure 5.2d), MeNam (Figure 5.2e) and Me4PY (Figure 5.2f), but also stimulates retrograde production of NAAD (Figure 5.2i) and NAMN (Figure 5.2h). According to this view, when NAD+ is elevated at least 2-fold, a previously unknown activity would deamidate NAD+ to NAAD.

In the mouse liver system, the potential flux of this pathway is quite significant: the NR- driven peak of NAAD amounted to 10% of the NR-attributable peak of NAD+. Production of high levels of NAAD from NAD+ could therefore account for the NR-stimulated peak in NAMN because NAMN adenylytransferase is known to be a reversible enzyme (154).

The hypothesis that NAAD is formed from NAD+ in vivo was tested by administering NR that had been labeled in the Nam and ribosyl moieties. As shown in Figure 5.3, NR stimulates appearance of double-labeled NAAD (8% of total) at the same 2 hr time point in which 5% of NAD+ was double-labeled. The biochemical basis for apparent NAD+ deamidation is not known. However, the glutamine-dependent NAD+ synthetase reaction would appear to be irreversible (12, 155). One intriguing possibility is that NAAD is formed by the long-sought enzyme that forms intracellular NAADP (156). According to this view, an NADP deamidase may be

82

responsible for formation of NAADP—this same activity might deamidate NAD+ at high concentrations, stimulating formation of NAAD. Unlike ADPR and methylated Nam waste products, NAAD is not only a biomarker of elevated NAD+ metabolism but is also a reserve metabolite that contributes to elevated NAD+ over time.

Finally, in the first controlled clinical study of NR, it was established that PBMC NAD+ metabolism is increased by 100 mg, 300 mg and 1000 mg doses of NR without dose-dependent increases in PBMC Nam and without dose-dependent serious adverse events. Though some sporadic thermal responses were self-reported in this study, the experiment was not designed to assess flushing with a validated questionnaire.

In people, as in mice, NAAD is the most sensitive biomarker of boosting NAD+. While NR elevated PBMC NAD+ from ~12 to ~18 μM, NAAD was elevated from below the limit of quantification to ~1 μM. The ability to detect NAAD in human samples is expected to aid conduct of clinical trials of NR and other NAD+ boosting strategies.

The wide availability of over-the-counter nutritional supplements can complicate clinical trials because patients may enroll in order to obtain compounds they expect to bring benefits and such patients may be inclined to take supplements in the case that they are assigned to placebo. Detection of NAAD should therefore be incorporated in phase II and III clinical studies of NR efficacy to eliminate the confounding effects of off-study use of NR.

83

5.10 Perspective on Chapter 5

Introduction
In sections 3-6 of this chapter, we present the first evidence that NR augments the

human NAD metabolome and identify NAAD as a possible non-obvious potential biomarker for NAD+ elevation. During our investigation of NAAD, we compared NR to Nam and NA efficacy in altering the murine hepatic NAD metabolome and found that 1) NR is most efficacious in increasing NAD+ and is uniquely metabolized, 2) NR and Nam increase murine hepatic NAAD, and 3), through the use of stable isotope technologies, prove direct contribution of NR to NAAD. We then performed the first human trial of the effect of NR and found a clear effect of NR on the NAD metabolome in human blood cells and confirmed that NAAD responds to NR supplementation in a dose-dependent manner. The following sections are my perspective regarding what has been written by myself and Dr. Brenner regarding the data that Dr. Mark Schmidt and I generated. In addition, I present and discuss the effects of intraperitoneal (IP) injection on the murine hepatic NAD metabolome and the effect of both IP injection and gavage of double labeled NR on muscle from the same mice.

Results and Discussion
In the above sections, we set out to test the efficacy of NR in altering the NAD

metabolome in a human being and performed the first head-to-head comparison of NR to Nam and NA. We first supplemented a healthy 52 year old, male with 1 g NR Cl over a week. Peripheral blood monocytes (PBMCs) were collected from blood at each time point as well as urine. Time points were taken within 24 hours and after six subsequent dosages. NR effectively increased NAD+ within 24 hours starting 4 hours after ingestion and its level remained increased compared to the pre-dose after six dosages (Table 5.1). NADP+ and NMN were also elevated. Overall, these data represent the first report of the effect of NR on a human subject and reveal

97

NR can increase NAD+ in a human being. Further, this initial experiment revealed blood is an accessible sample for the effects of NR on the NAD metabolome for future clinical trials.

Methylated Nam derivatives MeNam, Me2PY, and Me4PY were expectedly also highly elevated by NR. Regardless of the amount of NR absorbed intact versus as Nam, any contribution of NR to a cellular fraction could convert to Nam through either enzymatic consumption of NAD+ synthesized from NR or direct phosphorolysis via purine nucleoside phosphorylase (PNP) activity (13). These methylated species have been reported as biomarkers for B3 vitamin deficiency and supplementation in urine and blood (157-161) and hence are indicative of cellular uptake of NA and Nam and presumably for NAD+ synthesis. However, these metabolites are a diversion of Nam from NAD+ biosynthesis and theoretically could appear without contributing to NAD+ and, hence, are not truly biomarkers of NAD+ elevation. Others have suggested measurement of NAD+ and NADP+ as indicative of whole body NAD+ status (162) but these measurements may not be indicative (163) possibly due to a buffering capacity of NAD+ concentration and/or the inherent analytical problems in detecting a small change in a very abundant metabolite.

In our study, we identified a rise in NAAD from below the limit of quantification (<0.02 μM) by at least 45-fold in the PBMC fraction (Table 5.1). The appearance of NAAD is unexpected since neither NR nor Nam are thought to contribute to the deamidated pathway (Figure 5.1), but could represent a novel, accessible biomarker for efficacious NAD+ supplementation. Unlike the methylated Nam derivatives, NAAD is expected to contribute to NAD+ as it is the proximal deamidated precursor. Even if NAAD were converted to NAMN and subsequently to NAR intracellularly, the NAR could be recycled back to NAD+ through the NRK pathway or exported and contribute to NAD+ in near or distant cells (164). Indeed, we do observe an increase in NAR in the plasma and urine of subjects ingesting NR (Appendix A: Table A.2 and A.3).

98

In furthering our investigation into the NAAD phenomenon, we turned to the murine model. While performing these experiments, we were also interested in the kinetics of NR compared to NA and Nam after a single dose. The NAD+ biosynthetic machinery differs in a cell and tissue specific manner (87). These biosynthetic processes are also competing with other modifying activities, such as methylation and/or oxidation (Chapter 1.1), which would divert pyridine from NAD+. Hence, the efficacy of NR utilization compared to these other precursors remains an unresolved and crucial question to the future use of NR. We dissected the livers and muscles18 from mice after either gavage19 or intraperitoneal injection (IP) of 185 mg/kg body weight NR Cl or mole equivalent of Nam and NA and analyzed using LC-MS/MS. Saline injections were also performed. The liver NAD metabolome exquisitely responded to all three precursors; however, NR displayed unique and superior effects in increasing NAD+ (four-fold compared to two-fold after Nam and NA) after gavage (Figure 5.2b). This profile is in stark contrast to NAD+ after IP injection, whereby all three precursors produced indistinguishable elevations (Figure 5.6b), suggesting that absorption of these metabolites may be responsible for the differential observed kinetics. The route and kinetics of NR absorption are current active projects in the laboratory. The IP experiments also revealed that the mode of NR delivery could differentially effect the NAD metabolome. NR depressed ADPR concentration (Figure 5.6j) after injection but not gavage (Figure 5.2j), suggesting injection of NR may act as an NAD+ consuming enzyme inhibitor. Elucidation of the effect of NR IP injection on the liver warrants further investigation. Intriguingly, Nam and NR increased NAAD after gavage and IP (Figure 5.2i and 5.6i) and this elevation correlated with the efficacy of each in increasing NAD+ (Figure 5.2b and 5.6b) showing that NAAD correlates with the concentration of NAD+ in a non-accessible

18 See Chapter 2.3: Considerations of Quantitative NAD Metabolomics in Mammalian Tissues for information regarding tissue treatment.
19 Gavage is the process by which a drug or food is delivered directly to the stomach.

99

tissue in clinic. These results coupled with the human n of one experiment indicated that NAAD may serve as a future clinical biomarker for the efficiency of NAD+ supplementation.

The appearance of NAAD in such great quantity from Nam and NR begs the question about how we will re-draw the NAD+ pathway (Figure 5.1). The amidated precursors Nam and

NR could either directly contribute to NAAD through some as yet known mammalian
act to inhibit NAD synthase, the enzyme converting NAAD to NAD+. We tested these possibilities by administering a specially labeled NR that contains a deuterium on the ribose and a 13-carbon on the nicotinamide moiety. If NR contributed directly to NAAD, we expected to observe enrichment of the NAAD pool. Indeed, we observed NAAD enrichment by both gavage (Figure 5.3e) and IP (Figure 5.7e) in liver and in muscle (Figure 5.8e and 5.9f), suggesting the rise in NAAD is an effect of an unknown deamidating pathway that exists in both liver and muscle.

In order to begin to establish NAAD as indicative of efficient NAD+ synthesis, we performed the first clinical trial of NR at several dosages (100, 300, and 1000 mg NR Cl) with a one week wash out period on 12 individuals (6 males and 6 females between 30 and 55 years of age at normal BMI (18.5 – 29 kg/m2). NR increased NAD+ in the blood cell fraction compared to pre-dose when collapsing all dosages (Figure 5.4a). The methylated derivatives rose and did so in a dose-dependent manner (Figure 5.4c-e). Additionally and in agreement with the initial human n of one study and murine study, NAAD increased in a dose-dependent manner, indicating that NAAD may be an applicable biomarker for NR effectiveness.

Direct contribution of NR to NAAD proves the existence of an NR to NAAD biosynthetic route and necessitates the existence of an unknown mammalian deamidase. In the 1960s, Nam deamidating activity was reported in liver (165) but displayed a non-physiological Km at above at least 40 mM (166). At that time, it was thought that all Nam was converted to NA and utilized much as it is in yeast (Figure 1.1); however, later work revealed high dose Nam caused the appearance of non-amidated intermediates and was hypothesized to occur due to bacterial

100

pathway or

deamidases in the gut (167). Our work argues against such a mechanism given that NA was detected at ~320 pmol/mg liver at five minutes after NA gavage/IP but not detected at any time after NR nor Nam. Further, IP administration of the double labeled NR produced a clear enrichment in the M+2 isotopologue of NAAD (Figure 5.7 and 5.9). If NR were hydrolyzed to Nam and the Nam deamidated to NA, enrichment would only occur in the M+1 NAAD isotopologue. To put it another way, the NAAD produced from NR in the labeled experiment was from intact, non-hydrolyzed NR. And though NR could be deamidated to NAR, NAR did not rise after supplementation of any B3 vitamin, suggesting but not completely excluding that NR is not a direct NAR precursor. Together, these data strongly suggest NA is not the initial precursor to the observed NAAD and that the NAAD may be synthesized within the liver which necessitates a mammalian deamidation pathway exists.

But if not NA and likely not NAR, then what? Since both NR and Nam increase NAAD, it must be a shared metabolite downstream of both precursors (NMN or NAD+). We argue that NAD+ is the likely source of NAAD and that its increase by at least 2 fold induces a deamidating pathway. Though this may be the case, the data we have presented does not exclude NMN as the precursor. Indeed, NAMN, the deamidated analog of NMN, increases at the same time as NAAD. Though this NAMN could be a result of reversible NMNAT1-3 activity (44) as suggested above, the “forward” (NMN/NAMN to NAD+/NAAD) direction of the reaction in vivo appears favorable due to limiting inorganic phosphate (87). As it stands, the true route from Nam/NR to NAAD remains to be elucidated in future investigations.

Regardless of the route of NAAD synthesis from the amidated precursors, further investigation is required to establish NAAD as a biomarker in the clinic for the effectiveness of NR in NAD+ synthesis. At present, we are unaware of the correlation between blood NAAD and hepatic NAAD and NAD+ after NR administration. If NAAD is to be established as a robust accessible biomarker for increased NAD+ in an inaccessible tissue in clinic, then tissue NAD+ concentration must correlate with blood NAAD concentration. Future work in model organisms

101

and in humans could be used to examine the correlative value of the relationship between NR supplementation, blood NAAD concentration, and whole body increases in NAD+. Initial results from Chapter 6 may indicate that NAAD indeed correlates with NAD+ abundance in liver.

Perhaps more importantly, the metabolism of NR requires careful and thorough investigation. At present, NR has not been detected in the blood cell fraction nor in plasma due to difficulty in its extraction. As it stands, we have measured that NR is absorbed and circulates in a “shadowy” manner. NR does not simply serve as a Nam precursor given the very distinct overall effects on the hepatic NAD metabolome after gavage. However, hepatic detection of NR varied and displayed no response to NR administration nor that of Nam and NA (Figure 5.5b), but was detected after IP of double labeled NR in liver (Figure 5.7) and muscle (Figure 5.9), revealing NR does circulate. Additionally, the shadow of NR was detected as M + 2 NAD+ in both liver and muscle. In both cases, establishing the accuracy of double label enrichment the nucleotides of the NAD metabolome is difficult to determine without purified standards. These standards are not commercially available and not easily synthesized. To strengthen the claim of intact NR circulation and utilization, loss of function NRK rodent models should be utilized. If NR indeed does not enter a cell intact, then NAD+ synthesis after NR administration would be independent of the NRK pathway and completely dependent upon NAMPT (a route in which NR is hydrolyzed to Nam then utilized). A rodent model lacking NRK activity would be expected to experience a complete loss of enrichment in the M+2 NAD+ isotopologue and complete preservation of the M + 1 isotopologue. A whole body NRK1 knockout mouse has been generated (150) and experiments are underway to definitively test NR intact absorption.

Methods
Unless otherwise stated, all methods are as described in Chapter 5.3: Methods and

Chapter 2.2: Quantification of the Oxidized NAD Metabolome in Skeletal Muscle.

Figure 5.6 IP administration of NR, Nam, and NA produce similar effects on murine liver NAD metabolome.
Either saline (orange) or equivalent moles of NR (black), NA (blue) and Nam (green) were administered to male C57BL6/J mice by IP. Livers were excised and freeze-clamped and then analyzed by LC-MS/MS. In the left panels, the hepatic concentrations of each metabolite are shown as a function of drug or vehicle. In the right panels, the baseline subtracted areas under the curve are shown. Left panels: ‡ p-value < 0.05; ‡‡ p-value < 0.01; ‡‡‡ p-value <0.001 Nam vs NA; † p-value < 0.05; †† p-value < 0.01; ††† p-value < 0.001 Nam vs NR; # p-value < 0.05; ## p-value < 0.01; ### p-value < 0.05 NA vs NR; Right panel: * p-value < 0.05; ** p-value < 0.01; *** p-value <0.001.

103

1000

500

0

‡‡‡ ‡‡‡

200

†† ‡‡‡ ###‡‡‡

NR
NA

Nam

NR
NA

Nam

NR
NA

Nam

NR
NA

Nam

NR
NA

Nam

Nam

NR
NA

NR
NA

Nam

Nam

NR
NA

NR
NA

Nam

Nam

NR
NA

# ††

### †††

ADPR (pmol/mg)

Me2PY (pmol/mg) MeNam (pmol/mg) NMN (pmol/mg)

AUC (pmol*hr/mg) AUC (pmol*hr/mg) AUC (pmol*hr/mg)

a

c

e

5000 4000 3000 2000 1000

400 200 0

b

M+1 M+2

M

Time (hours)

d

Figure 5.7 NR directly contributes to hepatic NAAD after IP injection.

Double-labeled NR was intraperitoneal injected into mice. At indicated times, mice were sacrificed and livers freeze-clamped for isotopic enrichment analysis using LC-MS. a-b. Isotopic enrichment of NAD+ and NADP+ over time at both the M+1 and M+2 mass shifts. c-d. Isotopic enrichment of Nam and MeNam over time at the M+1 mass shift. e. Isotopic enrichment of NAAD over time at both the M+1 and M+2 mass shifts. Similar in effect to gavage, NR directly contributes to hepatic NAAD. NR contributed a much lower amount to NAD+ over the hour than by gavage but much more of it was intact rather than metabolized to Nam.

104

0
15

0
15

30
60

30
60

0
15

30
60

0
15

30
60

0
15

30
60

a b25 20 15 10 5

1.5 1.0 0.5 0.0

cd
150 6

100 4

50 2

00 Time (hours)

Time (hours)

Time (hours)

e
600

400

200

0

Time (hours)

Figure 5.8 NR contributes to muscle NAAD following gavage.

Quadriceps was dissected and freeze-clamped from the same mice that contributed liver after double labeled NR was gavaged. Quadriceps were extracted and analyzed for enrichment using LC-MS. a-b. Isotopic enrichment of NAD+ and NADP+ over time at both the M+1 and M+2 mass shifts. c-d. Isotopic enrichment of Nam and MeNam over time at the M+1 mass shift. e. Isotopic enrichment of NAAD over time at both the M+1 and M+2 mass shifts. The data reveal that NR also contributes to NAAD but only as Nam and not in a consistent manner. Additionally, NAD+ is not as effectively increased following NR gavage as in liver.

105

NAD+ (pmol/mg muscle)

0
2

80
2

4
6

4
6

0
2

4
6

0
2

80
2

4
6

8

4
6

8

8

a

b

800 700 600 500 400

60 40 20

0

M

M+1 M+2

dc

ef

Time (hours)

Figure 5.9 NR contributes to muscle NAAD following IP.

Quadriceps was dissected and freeze-clamped from the same mice that contributed liver after double labeled NR was gavaged. Quadriceps were extracted and analyzed for enrichment using LC-MS. a. Quantitation of double labeled NR. Endogenous NR was non-quantifiable. b-c. Isotopic enrichment of NAD+ and NADP+ over time at both the M+1 and M+2 mass shifts. d-e. Isotopic enrichment of Nam and MeNam over time at the M+1 mass shift. f. Isotopic enrichment of NAAD over time at both the M+1 and M+2 mass shifts. Double labeled NR was detected in muscle following IP injection, meaning NR does circulate to muscle intact. As observed with gavage, NR contributed to NAAD only as Nam and poorly increased NAD+ compared to liver IP. Together, the data suggest NR may be mainly metabolized through first pass metabolism by liver.

106

0
15

30
60

30
60

0
15

30
60

0
15

0
15

0
15

30
60

30
60

0
15

30
60

CHAPTER 6

NICOTINAMIDE RIBOSIDE PREVENTS ALCOHOL INDUCED FATTY LIVER

Samuel A.J. Trammell1,2, Sirisha Ghanta1, Kyle Klingbeil1, Keisuke Yaku1, Nicholas M. Riley3, Joshua J. Coon3, and Charles Brenner1,2

1Department of Biochemistry, 2 Interdisciplinary Graduate Program in Genetics, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, USA

3Department of Chemistry and Genome Center of Wisconsin, Madison, WI 53706

6.1 Distribution of Work

Experiments were designed by SG, CB, and I. All mouse husbandry in the initial experiment was performed equally by myself and SG. In the subsequent experiment, KY and I equally shared all mouse husbandry. Mouse phenotypic data were collected as a joint effort of myself, SG, and KY. Western blotting was performed by KK with guidance from me. All microscopy was performed by SG. All metabolomic data were collected and analyzed by myself. All proteomic data were generated by NMR in JJC’s laboratory. Analysis of proteomic data was performed by myself.

6.2 Abstract

Chronic alcohol consumption can lead to fatty liver disease (alcoholic fatty liver disease (AFLD)) through poorly understood mechanisms. Ethanol metabolism causes a reductive skew in the NAD+/NADH ratio and correlates with mitochondrial protein lysine hyperacetylation in a manner similar to loss of the NAD+ consuming enzyme Sirt3, suggesting NAD+ metabolic dysfunction may be related to the hepatic dyslipidemia. We hypothesized mitochondrial protein lysine hyperacetylation acts to reconfigure carbon metabolism from catabolism to anabolism, leading to unfettered fat accumulation. NR, precursor to NAD+, opposes age-related and

107

metabolic dysfunctions, including non-AFLD. Here, we employed NR as a chemical tool to determine the role of NAD+ metabolism and mitochondrial protein acetylation in the etiology of AFLD. In so doing, we identified NR as a potential anti-AFLD agent and provide evidence that NR could act to reconfigure mitochondria back to respiration. Unfortunately, we observed that NR increased consumption of the control and ethanolic diet, complicating further experimentation.

6.3 Introduction

Alcohol in the form of ethanol is an important part of American culture and often a substance of abuse. Approximately 20% of alcoholics and heavy drinkers develop alcoholic fatty liver disease (AFLD), which with continued alcohol abuse, leads to liver damage (168). The liver is especially vulnerable to chronic ethanol ingestion since it is the site of ethanol metabolism (169). Ethanol is metabolized to acetaldehyde then acetate with two moles of NAD+ reduced to NADH per mole of ethanol. Unlike acute ingestion, chronic ethanol ingestion induces microsomal ethanol oxidizing system which oxidizes NADPH to NADP+ and ethanol to acetaldehyde, which is then converted to acetate as above (170, 171). Acute and chronic ingestion causes a decrease in the NAD+/NADH ratio in the cytoplasm and, due to the malate- aspartate shuttle and mitochondrial acetaldehyde dehydrogenase activity, convey this ratio to mitochondria (172). The reductive effect of ethanol inhibits glycolysis, oxidative phosphorylation (172, 173), and fatty acid oxidation (174) and was thought to explain the accumulation of fat. However, preservation of the NAD+/NADH ratio does not inhibit AFLD (175, 176) indicating other aspects of ethanol metabolism are responsible for fatty liver.

Chronic ethanol ingestion causes hyperacylation (including acetylation) of mitochondrial protein lysines (177-179). Mitochondrial acylation is reversible through the action of Sirt3 and Sirt5. Sirt3 and Sirt5 are NAD+-dependent deacylases with differing substrate specificities. Sirt3 primarily removes acetylation (180, 181) and Sirt5 removes malonyl, succinyl, glutaryl, and

108

other acyl groups (10, 182). Acylation can be enzymatically inhibitory (10, 183-185) or, at least in one case, activating (186) and appears to affect multiple targets within pathways related to amino acid, carbohydrate, and lipid metabolism (187). Indeed, Sirt3-/- mice experience mitochondrial hyperacetylation and decreased lipid catabolism (188), indicating mitochondrial acetylation inhibits and Sirt3 activates lipid utilization.

We propose that mitochondrial protein acetylation is part of the etiology of AFLD. Mitochondrial protein acetylation has been described as a cell intrinsic long lasting satiety program that diverts mitochondria from carbon oxidation to carbon storage (189). In this model, chronic high caloric, low vitamin dense diets induce small molecule metabolic changes (i.e. decreases NAD+/NADH) and, as consequence, leads to increased mitochondrial protein acetylation. Unlike nuclear protein acetylation, mitochondrial protein acetylation appears to be non-enzymatically driven and instead depends upon the mitochondrial matrix pH and Ac-CoA concentration (190-192). Protons from NADH are exported across the inner mitochondrial membrane to create the proton-motive force for ATP synthesis. As consequence, the mitochondrial matrix pH is alkaline (pH 7.6 – 8) (192), allowing for deprotonization of lysines with depressed pKas. As observed with high fat diet (193), a decrease in the NAD+/NADH encumbers the TCA cycle and increases mitochondrial Ac-CoA (190). These chemical conditions promote transfer of the acetyl group from Ac-CoA to a deprotonated protein lysine and, in chronic conditions, can account for hyperacetylation.

Ethanol metabolism recapitulates the conditions of the overfed state and could directly contribute to the protein acetylation. The vast majority of ethanol is oxidized to CO2 in muscle and not stored as fat (194, 195). Fat as a result of ethanol ingestion appears to be from dietary sources and fat mobilization from adipocytes (196-199) indicating ethanol induces a metabolic switch in mitochondria towards fat storage. Though restoring the oxidative balance of NAD+ does not prevent fatty liver, the reductive environment coupled with the increased availability of

109

acetate (Ac-CoA precursor) (200) may be crucial in causing ethanol induced mitochondrial hyperacetylation, resulting in AFLD.

Though loss of Sirt3 activity is certainly involved in ethanol induced hyperacetylation (201), involvement of the other seven sirtuins (Chapter 1.2) cannot be ruled out (202). Additionally, as discussed in Chapter 5, increasing sirtuin enzymatic activity without increasing NAD+ does not necessarily result in increased deacylation. Here we employed novel NAD+ precursor NR to unravel the role of mitochondrial hyperacetylation in the etiology of AFLD. NR is a far superior whole cell (Figure 5.2) and perhaps mitochondrial (Chapter 1.2) NAD+ precursor than Nam and NA and has been shown to oppose non-alcoholic fatty liver disease (53). We predicted that NR would oppose mitochondrial hyperacetylation and, if hyperacetylation is necessary for AFLD, inhibit hepatic dyslipidemia. Here, we utilized a chronic alcoholic mouse model whereby animals derived increasing calories from ethanol over a six week period and remained on a diet of 30% calories from ethanol for two to three weeks. We reveal that NR increases hepatic NAD+ in an alcoholic mouse model and moderately opposes fatty liver development. However, mortality was high in the initial experiment which we had hoped to overcome through alterations of the animal protocol. In so doing, we found that NR increases consumption of liquid diet and confounded the experimental outcomes. We discuss possible improvements to the model in future experiments.

6.4 Materials and Methods

Animal Husbandry and Experimental Design
All animal protocols were approved by the University of Iowa Institutional Animal Care

and Use Committee. Male C57BL/6J mice were purchased from Jackson Laboratories. Control (catalog #: F1259SP) and ethanol (catalog #: F1697SP, Lieber-DeCarli ’82 formulation) diets were purchased from Bio-Serv (Frenchtown, NJ, USA).

110

Initial Experiment

At ten weeks of age, mice were transitioned to a liquid food following manufacturer’s instructions. After transition, mice were split equally into a control liquid diet group, ethanol diet group, and ethanol + NR (0.33 g/l). Those fed ethanol were transitioned from control liquid diet to ethanol following manufacturer’s instructions. Ethanol content was adjusted as follows: 2% (w/v) for two weeks, 3.1% (w/v) for two weeks, and finally 4.2% (w/v) for three weeks. Diet was contained in standard mouse water bottles and changed every 48 hours or as necessary. Volume of diet consumed was measured at each diet change. Mice were weighed once a week to monitor health. Mice were euthanized using CO2 or live decapitation. Livers were dissected and weighed and portioned for metabolomics, proteomics, western blotting, and microscopy. Subsequent Experiment

All parameters remained the same as in the initial experiment except that the animals remained on 4.2% (w/v) diet for four weeks rather than three and standard water bottles were exchanged with feeding tubes (catalog #: 9019, Bio-Serv) that were changed every day.

Mitochondrial Isolation
Mitochondria were isolated as described in (203).Briefly, 150 mg of liver tissue was

homogenized in 1 mL of isolation buffer (70 mM sucrose, 210 mM mannitol, 5 mM HEPES, 1 mM EGTA, 10 mM Nam, and 0.5% (w/v) fatty acid free BSA) on ice. An aliquot of homogenate was saved (~200 μL). The rest of the preparation was then centrifuged at 1,000 x g and 4 °C for 10 minutes. The supernatant containing cytoplasm and mitochondria was removed. The pellet (nuclear fraction) was snap frozen using liquid nitrogen. The supernatant was centrifuged at 10,000 x g and 4 °C for 10 minutes to pellet mitochondria. The supernatant (cytoplasmic fraction) was snap frozen. The mitochondria pellet was washed once with isolation buffer at 10, 000 x g and 4 °C for 10 minutes. The supernatant was discarded and the mitochondrial pellet snap frozen.

111

Western Blotting
Mitochondrial fractions were thawed in lysis buffer (50 mM Tris-HCl, 150 mM KCl, 1 mM

EDTA, 1% NP-40, 1 mM Na butyrate, 5 mM nicotinamide, and Roche complete protease cocktail). Protein was loaded and separated via standard SDS-PAGE and transferred to PVDF membrane. Membranes were blocked with 5% skim milk in TBST for 30 minutes, then washed five times with TBST. Membranes were probed with primary antibodies directed against acetyllysine (Cell Signaling, Boston, MA or PTM Biolabs, Inc., Chicago, IL). The membrane was incubated with secondary antibody for 1 hour. Horseradish peroxidase was applied and membranes were imaged following standard procedures.

Microscopy
A small portion of liver was dissected out and frozen in clear freezing media (purchased

from TBS – A Division of General Data Healthcare). Frozen tissue was sectioned (10 μm) and placed onto positive slide glass and fixed in formaldehyde. Fixed tissue was washed with distilled water and stained with Harris’ hematoxylin for 30 seconds. Slides were then washed with running tap water for five minutes, washed with distilled water, and placed into Oil Red O solution for 10 to fifteen minutes. Slides were washed with distilled water and imaged using an Olympus BX61 upright microscopy with 20X magnification.

NAD Metabolomics
On the day of extraction, 300 mg of liver tissue was homogenized in 1 mL of isolation

buffer lacking Nam (see Mitochondrial Isolation). The preparation was snap frozen and stored at -80 °C until analysis. 50 μL of liver homogenate was extracted with 300 μL of buffered ethanol (75% Ethanol/25% 10 mM HEPES) heated at 80 °C and constant vigorous vortexing for 3 minutes. Extract was separated from particulate through centrifugation (16.1 x g, 10 minutes, 4 °C). Both extract and pellet were dried overnight using speed vacuum. Dry particulate was weighed and used for normalization of mole amounts of metabolite. Extracts were re-suspended

112

in LCMS grade water and then analyzed as described (Chapter 2.1 and 2.3) with the following differences. The acid separation internal standard solution contained 0.75 μM 18O NR, 0.75 μM 18O Nam, and 6 μM D4 NA (purchased from C/D/N Isotopes, Pointe-Claire, Quebec, Canada). Quantitation was performed using internal standards and calibration curve.

Acetylomics

Chemicals and supplies

The Tandem Mass Tags (TMT) reagents were purchased from Thermo-Pierce (Rockford, IL). The BCA assay Protein Assay Kit was purchased from Pierce Biotechnology (Rockford, IL). Trypsin Gold was purchased from Promega (Madison, WI). Sep-Pak tC18 cartridges were purchased from Waters (Milford, MA). A polysulfoethyl A column (9.4 mm x 200 mm, 5 mm, 200Å) was purchased from PolyLC (Columbia, MD). Bridged Ethylene Hybrid (BEH) C18 resin (1.7m diameter particles, 130 Å pore size) was purchased from Waters (Milford, MA). Fused-silica capillary tubing was purchased from Polymicro Technologies (Phoenix, AZ). Formic acid and trifluoroacetic acid ampoules were purchased from Thermo Scientific (Rockford, IL). Pan-acetyl lysine antibody-agarose conjugate was purchased from Immunechem (Burnaby, Canada). Protease (complete mini ETDA-free) and phosphatase (PhosSTOP) inhibitors were purchased from Roche (Mannheim, Germany). All other chemicals were purchased from Sigma- Aldrich (St. Louis, MO).
Sample Preparation

Liver tissue samples were homogenized with 3 strokes of a motorized stirrer at 1000 rpm in a Potter-Elvehjem tissue grinder with 1 mL of buffer (8 M urea, 50 mM Tris pH 8.0, 5 mM CaCl2, 100 mM NaCl, protease inhibitors, and deacetylase inhibitors). Homogenates were then sonicated at 5 W for 30 seconds and centrifuged at 10000xg to clear the lysate of debris. Protein concentrations in lysates were quantified by BCA Protein from each sample (1 mg) was reduced with 5 mM dithiothreitol for 45 minutes at 58 °C and then alkylated with 15 mM iodoacetamide for 45 minutes at ambient temperature in the dark. The alkylation was quenched

113

with 5 mM dithiothreitol. Following dilution to 1.5 M urea with 50 mM Tris pH 8.0, the samples were digested with trypsin (50:1 protein:enzyme) overnight. Additional trypsin (200:1 protein:enzyme) was spiked into the sample the following morning, digestions were quenched by TFA acidification two hours later, and samples were desalted with a tC18 sep-Pak. Desalted material was resuspended in 200 mM TEAB pH 8.5 and labeled with 10-plex TMT (only nine of the ten tags were used, with the lightest channel being omitted). Labeled peptides were combined and desalted. Labeling efficiency was evaluated by analyzing a test mixture by LC/MS/MS for each experiment. Labeling efficiency was > 95%, calculated by the number of N- terminal labeled peptides divided by the total number of peptide identifications.

Fractionation and Enrichment

Labeled peptides were fractionated by strong cation exchange (SCX) on a polysulfoethyl A column (0.4 mm x 200 mm) with mobile phases A: 5 mM KH2PO4 pH 2.7 and 30% acetonitrile; B: 5 mM KH2PO4 pH 2.7, 350 mM KCl, and 30% acetonitrile; C: 5 mM KH2PO4 pH 6.5, 500 mM KCL and 20% acetonitrile; D: water. Peptides were eluted over the following gradient on a Surveyor LC quaternary pump (Thermo) at 3 mL/min: 0-2 min, 100% A; 2-5 min, 0-10% B; 5-35 min, 10-60% B; 35-41 min, 60-100% B; followed by washes with C and D prior to re-equilibration with mobile phase A. Sixteen fractions were collected and desalted. A small portion (5%) of each was retained for protein analysis, while the remaining material was pooled into 6 fractions for acetyl lysine enrichment.

These pooled fractions were dissolved in 50 mM HEPES pH 7.6, 100 mM NaCl, and each fraction was combined with approximately 75 uL pan-acetyl lysine antibody-agarose conjugate. The samples were rotated overnight at 4 °C and then rinsed eight times with cold 50 mM HEPES pH 7.6, 100 mM NaCl. Rinses were followed by elution with 0.1% TFA, and eluted peptides were desalted prior to analysis.

114

LC/MS/MS

All samples were analyzed by reverse phase liquid chromatography on a nanoAcquity (Waters) coupled to an Orbitrap Elite (Thermo). Samples were loaded onto a 75 μm inner diameter analytical column made in-house, packed with 1.7m diameter, 130 Å pore size, BEH C18 particles (Waters) to a final length of 30 cm. The column was heated to 62 °C for all runs. The elution portion of the gradient was 5% to 30% B (A: water/0.2% formic acid; B: acetonitrile/0.2% formic acid) over 140 minutes for acetyl enriched fractions and 80 minutes for protein fractions.

Mass spectrometry instrument methods all started with one MS1 survey scan (resolution = 60,000; 300 Th – 1,500 Th; target value = 1e6) followed by data dependent MS2 fragmentation and analysis (resolution = 30,000) of the fifteen most intense precursors by beam-type CAD (HCD; normalized collision energy = 35%, target value = 5e4). Only those precursors with charge state +2 or higher were sampled for MS2. The dynamic exclusion duration was set to 40 seconds with a 10 ppm tolerance around the selected precursor and its isotopes, and monoisotopic precursor selection was turned on. Isolation width was set to 1.80 Da, precursor injection time was capped at 200 ms, and the first mass value for HCD scans was 120 Th.

Database search, FDR filtering, and acetylation analysis

Spectra were converted to searchable text files using DTA generator. Generated text files were searched for fully tryptic peptides with up to three missed cleavages against a UniProt target-decoy database populated with mouse canonical plus isoforms (downloaded February 2014) using the Open Mass Spectrometry Search Algorithm (v. 2.1.8) (204). Mass tolerance was set to ± 125 ppm for precursors and ± 0.02 Da for fragment ions. Carbamidomethylation of cysteine, isobaric labeling of lysine, and isobaric labeling of the peptide N-terminus were searched as fixed modifications for all samples. Enriched fractions were additionally searched for variable acetylation modifications, in which the acetylation mass shift was set to the

115

difference between an acetyl group and an isobaric label (-187.1523 Da) to allow the isobaric label on lysine to remain a fixed modification even for acetylated peptides. Search results were filtered to 1% FDR at the unique peptide level using the COMPASS software suite (205). TMT quantitation of identified peptides was performed within COMPASS, as previously reported (206). Peptides were grouped into proteins according to previously reported rules and protein identifications were further filtered to 1% FDR (207). Protein quantitation was performed by summing all reporter ion intensities within each channel for each protein.

Acetylation events were localized to specific residues using previously described probabilistic methods (208). Briefly, for each peptide spectral match (PSM) that contains an acetyl modification, every possible peptide isoform was generated and fragmented in silico to produce theoretical fragmentation spectra. Each theoretical spectrum was compared to the experimental spectrum at 10 PPM m/z tolerances; the number of matching peaks was recorded and a p-value was calculated using a cumulative binomial distribution. An AScore (i.e. the difference of p-values) was calculated between every pair of isoforms. A peptide was declared “localized” if all AScores for a particular isoform were larger than the minimum value (AScore = 13, p-value < 0.05) for every comparison. Localized acetylated peptides were grouped together if they share identical modification sites and the reporter ion intensities were summed; peptides with C-terminal acetylation are excluded from quantitation.
Protein Normalization

All reporter ion intensities were log2 transformed and mean normalized for every acetyl isoform and protein. To account for protein abundance differences, the acetyl isoforms were normalized by subtracting the quantitative value of the reporter ion channel for the corresponding protein from the value for each acetyl isoform reporter ion channel. This gives a protein normalized acetylation mean normalized value which is then used to investigate fold changes between conditions. Fold change calculations were made by averaging the protein normalized values for each condition and then calculating the difference of averages.

116

Statistical Analysis
Unless otherwise stated, regular two-way ANOVA with Tukey’s posttest was performed

using GraphPad Prism version 5.00 for Windows, GraphPad Software, San Diego California USA, www.graphpad.com”. All data are displayed as mean ± SEM. P-values less than 0.05 were considered significant. Pathway analysis was performed using DAVID (209, 210) after filtering the data for acetylation sites hyperacetylated by ethanol (fold-change ≥ 2, P < 0.05) but opposed by NR (P < 0.05).

6.5 Results and Discussion

36 male C57Bl/J mice were transitioned to a control liquid diet (F1259SP, Bio-Serv, Frenchtown, NJ, USA). After transitioning, 12 mice were transitioned to ethanol (F1697SP, Bio- Serv, Frenchtown, NJ, USA) and another 12 were transitioned to ethanol containing NR Cl (0.33 g/l). Diets were kept isocaloric by replacing the carbohydrate with ethanol. The ethanol ingesting groups started with 2% (w/v) (14% calories) and remained on it for two weeks. After two weeks, the concentration was increased to 3.1% (w/v) (22% calories derived from ethanol) and remained on it for two weeks before replacement with 4.2% (w/v) (30% calories derived from ethanol) on which the mice remained for two – three weeks. The other 12 mice remained on control diet throughout the experiment. We found that animals on ethanol tended not to gain as much weight as control (data not shown) but ate similar amounts of diet (data not shown). The reason for the lack of weight gain is at odds with the consumption data and may represent an error in measurement of the suspension diet in the water bottles. It was noticed that the ethanol diet tended to “spoil” the diet and cause clogs, disallowing the mice from accessing the diet.

We hypothesized that NR supplementation would oppose AFLD through increased NAD+ and mitochondrial sirtuins activity. Lipid content was noticeably but variably increased in the livers of ethanol fed animals control diet animals (Figure 6.1). NR appears to have moderately opposed ethanol induced lipid accumulation in a non-homogenous manner with

117

cells closer to blood vessels experiencing less staining. This may indicate higher access to NR lessens lipid content, but requires follow up. These findings suggest NR may have positive outcomes for ethanol associated symptoms and partially prevents fat deposition in liver.

The effect of ethanol on the NAD metabolome has been implicated in its associated diseases (172-174). We sought to quantify the NAD metabolome and the effect of NR in in homogenized liver using LC-MS/MS. As discussed in Chapter 2, Nam concentration is much higher than any other NAD+ metabolite, suggesting degradation of the NAD+ (Table 6.1). Non- NAD+ related nucleoside and nucleotides (Cytidine, IMP, Inosine, UMP, and Uridine) were unaffected, establishing that observed changes are specific to NAD+ metabolism. Though NAD+ was not significantly depressed by ethanol feeding (105 ± 16 vs. 94 ± 15 pmol/mg of dry particulate; Control vs. EtOH), NADH was very significantly increased (11± 1.5 vs. 83 ± 22 pmol/mg dry particulate, Control vs. EtOH (P < 0.01)), reducing the NAD+/NADH ratio from 12 ± 3.5 to 1.1 ± 0.18 pmol/mg dry particulate (P < 0.01). Hence, the NAD+/NADH ratio is altered through increases in NADH but constant NAD+. This significant shift in redox state is similar to that observed previously for ethanol feeding (172). If treated as one pool, the amount of NADH + NAD+ increased with ethanol feeding (177 vs. 116 pmol/mg dry particulate), which could suggest NAD+ biosynthesis increased as a function of ethanol. Biosynthetic intermediates ADPR, NAAD, and NMN were decreased as a function of ethanol compared to control (ADPr: 4.7 ± 0.9 vs. 1.6 ± 0.8, Control vs. EtOH; NAAD: 5.6 ± 2.3 vs. 16 ± 5.7, EtOH vs. Control; NMN: 5.9 ± 1.1 vs. 12 ± 3.2, EtOH vs. Control). The detriment to NAAD in the presence of ethanol, suggest NAAD could be a biomarker in some cases for metabolic disease, expanding upon what was found in Chapter 5. NADP+ was significantly depressed by ethanol (NADP: 54 ± 11 vs. 16 ± 3.2, Control vs. EtOH (P < 0.05)). Further, the nicotinamide methylated product, Me4PY, decreased as a function of ethanol (7.6 ± 0.83 vs. 3.5 ± 0.98, Control vs. EtOH (P < 0.05)).

118

NR supplementation significantly increased NAD+ compared to both control and EtOH (179 ± 20 vs. 105 ± 16 vs. 94 ± 15, EtOH + NR Cl vs. Control (P < 0.01), vs. EtOH (P < 0.05)). However, NR supplementation did not alter NADH. NR doubled the NAD+/NADH ratio compared to EtOH (P < 0.05) but was five-fold lower than that of control (P < 0.05). NR increased the concentration of NMN compared to both control and ethanol diets (NMN: 23 ± 3.9 vs. 12 ± 3.2 vs. 5.9 ± 1.1, EtOH + NR Cl vs. Control (P < 0.01), vs. EtOH (p-value < 0.05)). The potential B3 vitamin biomarker, NAAD (Chapter 5), increased dramatically versus ethanol alone and control (NAAD: 29 ± 9.7 vs. 16 ± 5.7 vs. 5.6 ± 2.3, EtOH + NR Cl vs. Control, vs. EtOH), which could suggest that NAAD concentration can be depressed by metabolic dysfunction and further that it may serve as a marker for efficacious NAD+ supplementation.

Ethanol induces a reductive shift in the NAD+ + NADH pool and increases the total NAD(H) pool but did not affect NAD+. Ethanol may increase the rate of NAD+ synthesis in order to continue detoxification. Indeed, metabolic stress can induce expression of the NAD+ biosynthetic machinery (54) and could be a common pathway in opposing certain metabolic dysfunctions. Expression of the NAD+ biosynthetic machinery should be investigated to test for induction. Additionally, stable isotope technologies and mass spectrometry (similar to that employed in Chapter 3 and 5) could be used to elucidate whether ethanol increases NAD+ turnover. NR nearly doubled hepatic NAD+ but did not greatly affect NADH, causing a modest but noticeable increase in the NAD+/NADH ratio. Though this ratio is one of the more striking metrics for the effect of ethanol here and within the literature, improvement in the ratio does not oppose AFLD (175, 176). Our finding that NR opposes fatty liver (Figure 6.1) and raises NAD+ (Table 6.1) supports the hypothesis that ethanol is not really a disease of too much NADH but rather of too little NAD+.

In increasing NAD+, activity of its consumers such as sirtuins may increase and be better able to oppose the ethanol-induced dyslipidemia. We predicted that NR would oppose mitochondrial hyperacetylation. Global mitochondrial protein acetylation was determined in

119

isolated mitochondrial fractions using Western blotting (Figure 6.2a). Little to no acetylation was detected in control samples. Intense hyperacetylation was observed in ethanol fed animals. NR decreased acetylation of higher molecular weight proteins but did not greatly effect global acetylation.

We speculated that the effect of NR on the acetylome could be more site-specific rather than global. To determine site-specificity, we performed an acetylomic analysis of whole liver using LC-MS (performed by Nick Riley in Dr. Coon’s laboratory at the University of Wisconsin— Madison). In all, 4796 total proteins were identified. 605 of these proteins were mitochondrial, indicating very good quantitation of mitochondrial proteome (605 out of 701 confirmed mitochondrial proteins) (211). Of the 3412 acetyl isoforms, a staggering 1449 were mitochondrial, representing 42% of the total liver acetylome, consistent with reported results (211). Overall, ethanol and ethanol + NR Cl fed animals acetylomes were indistinguishable from each other (Figure 6.2b). If the profiles differed greatly, acetylation fold changes of ethanol versus those with NR feeding would not correlate (i.e., fall along the y = x line). These data agree with the western blotting data (Figure 3a) showing that NR feeding did not lead to global deacetylation.

We hypothesized that the site-specific NR mediated alterations to acetylation are responsible for its opposing fatty liver disease. We predicted based on this hypothesis that pathways involved in carbon oxidation and lipid catabolism would be enriched with proteins that are hyperacetylated (sites significantly increased in acetylation compared to control: ≥ 2 fold- change and P < 0.05) as consequence of ethanol but not hyperacetylated with NR treatment (sites non-significantly increased in acetylation compared to control: P < 0.05). We filtered results using these criteria (Appendix B) then performed a pathway analysis using DAVID (209, 210). In agreement with previous results (179, 201), ethanol ingestion caused hyperacetylation of proteins involved in mitochondrial metabolism, specifically amino acid metabolism, lipid metabolism, nitrogen metabolism, the TCA cycle, and the electron transport chain (Table 6.2).

120

Strikingly, 17 of the 24 KEGG pathways enriched for ethanol induced hyperacetylation were opposed by NR treatment (i.e. NR prevented acetylation of proteins in key metabolic pathways) (Table 6.3 and Figure 6.3). NR prevented acetylation of proteins involved in lipid metabolism, amino acid metabolism, and respiration (the TCA cycle and the electron transport chain). Many of these enzymes affected are involved in branch chain amino acid metabolism, which has been implicated in obesity and diabetes (212) and could explain in part how NR opposes high fat diet induced obesity (37). Together, these data suggest that NR could affect mitochondrial lipid metabolism and respiration and that its points of regulation are few but potentially impactful. Further, these sites are very likely in vivo Sirt3 substrates, suggesting these sites are indeed mediators of mitochondrial metabolism. Future investigation into the impact of NR on enzymes in these pathways should be followed up in vitro with protein biochemistry techniques and in vivo with fluxomic analysis. In particular, amino acid metabolism (213) and beta oxidation (214) should be investigated to interrogate whether NR activates these pathways in the presence of ethanol. Additionally, exact quantitation of the sites on a mol-to-mol scale is necessary to establish the sites that are most acetylated. And finally, acylation (carbon > two) should be investigated in terms of NR supplementation (179).

During the course of the experiment, animals fed ethanol experienced increased mortality compared to control (33% compared to 100% surviving) (Figure 6.4). The animals that survived appeared to be of poor health with irregular gaits, constant tremors, and lethargy. NR opposed these effects with 67% of animals surviving and surviving animals were similar in behavior to control. The mortality was not due to impurities in the ethanol as non-denatured ethanol was used. Additionally, this high mortality is not reported in the literature though publication bias may certainly preclude reporting. Compared to the NIAAA guidelines (215), we noticed that the way in which we were delivering the liquid diet and the housing arrangement were incongruent. The NIAAA protocol called for feeding tubes rather than the standard animal housing water bottles and clearly stated that diet should be changed daily instead of every 48

121

hours. Also, the protocol called for no more than two animals per cage rather than the four per cage that we had implemented.

Following these guidelines eliminated mortality differences between groups. Further, animals appeared healthy and did not display the symptoms above. In this experiment we decided to include a control + NR Cl condition so that we may test for effects of NR versus effects of ethanol. Unlike with the water bottles, measurement of the diet consumed daily appeared to be much more accurate with more recovery of the diet out of the feeder tube. Mice on ethanol tended to eat less than mice on the control diet and this effect only increased with increasing ethanol in the diet (Figure 6.4). We found that NR caused increased consumption of diet and confounded the experiment since NR animals did not ingest equivalent calories or gain equivalent dosages of ethanol. The results of this experiment were too confounded for appropriate interpretation of the effects of NR on AFLD.

Overall, we found initially that NR opposed AFLD (Figure 6.1), increased NAD+ (Table 6.1), and protected against mortality (6.4). Further, we found that NR does not greatly prevent global mitochondrial hyperacetylation but does appear to oppose acetylation in a site-specific manner and may effect key pathways in lipid metabolism. In attempting to improve upon the initial results, we uncovered NR increases consumption of diet making this model potentially inappropriate for the study of NR as a preventative measure in AFLD. Preventative studies are often attempted first when exploring the effect of a drug on a disease model, but this sort of regimen is rarely applicable to chronic metabolic disease in clinic. In line with this, treatment experiments whereby mice have become alcoholic and then are transitioned to a non-alcoholic liquid and then solid diet should be attempted. NR could be added at the moment of transition to non-ethanol containing liquid diet. In this way, the effect of NR as a treatment for AFLD could be adequately assessed while uncovering the effect of NR on ethanolic liver metabolism.

122

CHAPTER 7

NICOTINAMIDE RIBOSIDE OPPOSES TYPE 2 DIABETES AND NEUROPATHY IN MICE

Samuel A.J. Trammell1, Benjamin J. Weidemann1, Matthew S. Yorek2, Amey Holmes2, Lawrence J. Coppey2, Alexander Obrosov2, Randy H. Kardon2,3, Mark A. Yorek2,4 and Charles Brenner1,4

Departments of Biochemistry1, Opthalmology3 and Internal Medicine4, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, USA; Iowa City Veterans Administration2, Iowa City, IA 52246

7.1 Distribution of Work

CB, MAY, and I designed experiments. BJW and I performed statistical analyses. Mouse husbandry and dissections were performed by MSY, AH, LJC, AO, RHK, and MAY. Microscopy was performed by MSY and AH. Lipid parameters were measured by LJC. All mass spectrometry was performed by myself. CB wrote the manuscript. BJW and I edited the document.

7.2 Abstract

Male C57BL/6J mice raised on a high fat diet become prediabetic and develop insulin resistance and sensory neuropathy. The same mice given low doses of streptozotocin are a model of type 2 diabetes, developing hyperglycemia, severe insulin resistance and diabetic peripheral neuropathy involving sensory and motor neurons. Because of suggestions that increased NAD+ metabolism might address glycemic control and be neuroprotective, we treated prediabetic and type 2 diabetic mice with nicotinamide riboside in their high fat diets. Nicotinamide riboside improves glucose tolerance, reduces weight gain, liver damage and the

130

development of hepatic steatosis in prediabetic mice while protecting against sensory neuropathy. In type 2 diabetic mice, nicotinamide riboside greatly reduces non-fasting and fasting blood glucose, weight gain and hepatic steatosis, while protecting against diabetic neuropathy. The neuroprotective effect of NR cannot be explained by glycemic control alone. Corneal confocal microscopy was the most sensitive measure of neurodegeneration and this assay allowed detection of the protective effect of nicotinamide riboside on small nerve structure in living mice. The hepatic pool of NADP+ plus NADPH was significantly degraded in prediabetes and type 2 diabetes but was largely protected when mice were supplemented with nicotinamide riboside.

7.3 Introduction

The global epidemic of obesity and diabetes has created severe economic stresses for health systems and intense neuropathic complications for affected individuals. Obesity is frequently associated with prediabetic polyneuropathy (PDPN) (216), while about half of individuals with diabetes will suffer from diabetic peripheral neuropathy (DPN) (217), rendering them insensitive to heat and touch. Severe DPN can progress to foot ulcers and amputations. Few treatments are effective for obesity while nothing has been found to arrest or reverse DPN. Best available care is tight glycemic control, dietary improvement and exercise, and pain medication when DPN is painful (218).

Deficiency in the NAD+ co-enzyme causes pellagra, which was endemic a century ago in the American south in populations subsiding on corn rations and lard (129). Though pellagra has been nearly eliminated, there are indications that supplementation with nicotinamide riboside (NR), a recently discovered NAD+ precursor vitamin (7, 142), can improve metabolic health in overfed mice (37) and function as a neuroprotective agent in conditions involving Wallerian degeneration (219-222). Though the mechanisms accounting for resistance to weight gain and improved glycemic control for mice on high fat diet (HFD) are not fully understood, NR

131

elevates NAD+ levels in skeletal muscle, liver and brown adipose tissue and appears to increase activity of nuclear and mitochondrial NAD+-dependent protein lysine deacetylases, the sirtuins SIRT1 and SIRT3 (37). Two mechanisms have been proposed for neuroprotection: boosting mitochondrial NAD+ to support SIRT3 (221) and preserving axonal NAD+ in the face of damage- induced SARM1 activation, which results in NAD+ degradation (222). In addition, a neuroprotective mechanism has been proposed that depends on both mitochondrial and axonal NAD+ (143). Though NR is not only a precursor of NAD+ but also of NADH, NADP+ and NADPH (129), the NAD+ metabolome has not been investigated in any disease model for which NR prevention or therapy has been tested. In addition, NR has not been tested on DPN.

Because of the potential for NR to improve prediabetic (PD) and diabetic glucose and lipid metabolism while also treating neuropathic complications, we made mice obese and PD with HFD and rendered them type 2 diabetic (T2D) with HFD plus two low doses of streptozotocin (STZ) (223). Here we show that NR improves fasting glucose levels and glucose tolerance of PD mice, while providing resistance to a substantial degree of hepatic steatosis, hypercholesterolemia, liver damage and weight gain. NR greatly lowered fasting and nonfasting glucose of T2D mice, while reducing hepatic steatosis and weight gain. Though hepatic steatosis and hyperglycemia were not fully corrected by NR, supplemented mice have greatly reduced neuropathic symptoms in both models. Remarkably, PD and T2D mice have lower levels of hepatic NADP+ plus NADPH, and T2D mice trended toward lower levels of hepatic NAD+. Upon supplementation, NAD+ was more correctable than was NADP+ plus NADPH, suggesting that maintenance of the latter metabolites is challenged by obesity. Our data also indicate corneal confocal microscopy (CCM) can be used as a minimally invasive and translational assay to monitor NR-dependent improvements in PDPD and DPN in future clinical investigations.

132

7.4 Methods

Mouse Models
Mouse methods were as described with investigators blinded to treatments (223-225).

NR chloride was a gift of ChromaDex, Inc.

NAD Metabolomics
Methods were performed as a revision of established procedures (1) as detailed in

Supplemental Methods of this chapter and as described in Chapter 2.2.

Statistics
Data are presented as mean ± SEM unless indicated otherwise. The effect of treatment,

NR supplementation, and interactions of the two factors were determined by two-way ANOVA with multiple comparisons performed using the Holm-Sidak test. Time dependent measurements (i.e. in GTT and body weight) were analyzed across and within the six groups via two-way repeated measures ANOVA followed by Holm-Sidak tests. P-values of less than 0.05 were considered significant.

Study Approval
All animal procedures were approved by the Iowa City Veterans Administration Animal

Care and Use Committee, which has an Animal Welfare Assurance (A3748-01) on file with the Office of Laboratory Animal Welfare and is fully accredited by AAALAC International.

7.5 Results and Discussion

Sixty male C57Bl/6J mice, housed 3 or 4 per cage, were raised on Teklad 7001 normal chow (NC). At 12 weeks of age, when mice weighed ~23 g, 40 mice were transferred to HFD (Research Diets 12492, 60% calories from fat) to render them PD, while 20 mice remained on NC. After 8 weeks on HFD, 20 of 40 mice were given two low doses (75 mg/kg body weight followed 2 days later with 50 mg/kg body weight) of STZ to induce T2D. PD and T2D mice remained on HFD for the duration of the experiment. Five weeks after STZ administration to

133

create the T2D population, 10 of 20 mice in each of the three groups (NC, HFD and HFD+STZ) were supplemented with 3 g of NR chloride per kg of their diet, thereby creating six groups of 10 mice (NC, NC+NR, HFD, HFD+NR, HFD+STZ, HFD+STZ+NR; 7.9 Supplemental: Figure 7.4). Five weeks before sacrifice, intraperitoneal glucose tolerance tests (GTT) were performed on fasted mice. Seven weeks after the beginning of NR supplementation, one mouse from each group was sacrificed per day for 5 days per week over a 2-week period. PD mice were effectively on HFD for 21 weeks without supplementation or with NR supplementation from week 13 to 21 on HFD. All T2D mice were non-supplemented for five weeks post STZ administration and 10 out of 20 were supplemented with NR from week 13 to 21 on HFD. On the day of sacrifice, mice were subjected to CCM, motor neuron conduction velocity (MNCV) and sensory neuron conduction velocity (SNCV) tests, and assayed for thermal sensitivity. The remaining assays were performed post-mortem (224).

As shown in Figure 7.1a and Supplemental Figure 7.4, during the 21 week experiment, mice on HFD gained ~27 g of body weight while mice in the HFD+STZ treatment group gained ~16 g. Though supplementation was for only 8 weeks, NR blunted weight gain in PD by ~7 g (P < 0.01) and by ~6 g in the T2D group (P < 0.05). As shown in Figures 1b-d, mice on HFD developed severe hepatic steatosis. Whether or not HFD mice were treated with STZ, supplementation with NR strikingly reduced the hepatic oil red O-positive staining area (P < 0.01). NR supplementation reduced oil red O droplet size by two-thirds in PD mice (P < 0.001). As shown in Figures 7.1e and 7.1f, NR significantly depressed circulating cholesterol (P < 0.05) and alanine aminotransferase (ALT) (P < 0.05), a sign of liver damage, in PD mice.

As shown in Figures 7.1g and 7.1h, NR tended to normalize hemoglobin A1c (HbA1c) and significantly improved nonfasting glucose (P < 0.01) in T2D. As shown in Figure 7.1i, NR had a powerful effect on fasting glucose, depressing levels from 210 mg/dl to 161 mg/dl in PD mice (P < 0.01) and from 345 mg/dl to 194 mg/dl in T2D mice (P < 0.001). Finally, as shown in Figure 7.1j and Figure 7.5, NR significantly improved glucose tolerance in PD (P < 0.01) and

134

tended to improve glucose tolerance in T2D. These data indicate that NR has profound effects on whole body metabolism in PD and T2D mouse models. However, mice supplemented with NR are neither hyperactive nor hypophagic (data not shown).

As shown in Figures 7.2a and 7.2b, PD mice retained good MNCV but had significantly depressed SNCV (P < 0.001). This deficit was not evident in mice supplemented with NR. T2D mice had significantly depressed MNCV (P < 0.001) and SNCV (P < 0.001) that were prevented by NR supplementation. Thermal insensitivity, a particularly dangerous aspect of human DPN (225), was strikingly evident in the PD (P < 0.001) and T2D (P < 0.001) models and was significantly reduced by NR in PD (P < 0.01) and T2D (P < 0.001). Consistent with the sensory neuron deficits in both models, as shown in Figures 7.2d and 7.2e, intraepidermal nerve fiber density (INFD) in hindpaws was significantly degraded in PD (P < 0.001) and T2D (P < 0.001). NR significantly protected against these deficits in PD (P < 0.01) and T2D (P < 0.001).

Early small fiber neuropathic changes are difficult to quantify in human populations and this may contribute to a failure to translate potentially effective treatments from DPN animal models to the clinic (226). The cornea is the most densely innervated structure of the human body, containing Aδ and unmyelinated C fibers derived from the ophthalmic division of the trigeminal nerve (227). CCM is gaining establishment as a valid measure of diabetic nerve damage in the clinic (228, 229) that can also be used to monitor diabetic neurodegeneration in rodent models (223, 224, 230). As shown in Figures 7.3a and 7.3b, quantification of sub- epithelial corneal nerves by CCM indicated that corneal nerves are severely degraded by PD (P < 0.001) and T2D (P < 0.001). CCM indicates that NR protects corneal innervation in T2D (P < 0.05) and trends positively in PD. Upon sacrifice, sub-basal corneal innervation was analyzed by staining for class III β-tubulin. This assay, shown in Figures 7.3c and 7.3d, produces the same qualitative results as those obtained from live CCM images.

In cultured dorsal ganglion root neurons, the concentration of NAD+, as determined by LC, was reported to decline in a SARM1-dependent manner in a four hour period after axotomy

135

(222). Because NR affects whole body metabolism, the targets of NR supplementation are not assumed to reside in a single tissue, nor is it assumed that obesity exclusively dysregulates targets of the NAD+ metabolome that depend exclusively on NAD+. Moreover, because sensory nerves die back in DPN, all neuronal metabolites are expected to fall as neuronal tissue declines with disease. We therefore employed LC-MS/MS to measure the NAD+ metabolome on a common pmol scale (1, 31) in freeze-clamped liver samples from freshly euthanized mice. NADPH is oxidized in extraction, such that the obtained NADP+ signal represents the sum of NADP+ plus NADPH.

As shown in Table 7.1, the liver NADP+ plus NADPH pool was significantly depressed in PD and T2D (P < 0.0001) with respect to NC controls. NR supplementation significantly boosted hepatic NADP+ plus NADPH but did not fully correct it. In PD, NAD+ trended down (P = 0.84) and trended down further in T2D (P = 0.11) mice with respect to NC controls. Hepatic NAD+ was fully normalized by NR in both models—the boost in hepatic NAD+ achieved significance in NR- supplemented T2D mice (P < 0.05). Consistent with a challenge to hepatic NADP+ plus NADPH metabolism, nicotinamide (Nam) waste products, 1-methyl nicotinamide (MeNam) and 1-methyl- 4-pyridone-3-carboxamide (Me4PY), were increased in PD (P = 0.0310 and P < 0.001).
It had previously been shown that HFD produces severe hepatic lipid accumulation in mice, which primes them for loss of glycemic control with low doses of STZ (223). Here we show that liver NADP+ plus NADPH is significantly compromised in these PD and T2D models and that NAD+ tends to decline in the mouse model of T2D. NR supplementation is accompanied by substantial resistance to weight gain and improvements in dyslipidemia, liver function and glycemic control in one or both models. Moreover, the PD and T2D mouse models exhibited structural and functional sensory nerve deficits that were not manifested when mice were supplemented with NR for their last 8 weeks on HFD. Though NR lowered hepatic steatosis and weight gain and greatly assisted glycemic control, NR did not normalize any of these metabolic parameters. In addition, neuroprotection cannot be explained by glycemic control alone. For

136

example, T2D mice supplemented with NR have higher nonfasting glucose than PD mice without NR (P < 0.01). Nonetheless, PD mice without NR have SNCV deficits, whereas T2D mice supplemented with NR do not. Thus, NR is presumed to have neuronal and hepatic targets. Finally, the decline in CCM-monitored neuronal density was more severe than any other measure of neuropathy and the protection of corneal innervation by NR was evident in the T2D model.

A large body of work has investigated NAD+-consuming enzymes including poly(ADPribose) polymerases (PARPs) and sirtuins (128). However, the SARM1-dependent factor that degrades axonal NAD+ in Wallerian degeneration is resistant to PARP inhibition and the pool of NADP+ plus NADPH was not investigated (222). Whereas NAD+ is the central hydride-accepting coenzyme for fuel oxidation, NADPH is the key hydride-donating cofactor for detoxification of reactive oxygen species (ROS) (231), a major contributor to insulin resistance (232). Because significant depression of NADP+ plus NADPH occurs in PD and T2D whereas NAD+ only trended down and was easier to correct, we suggest that the overnutritional stresses of HFD specifically challenge maintenance of hepatic NADPH and that this may be central to PD and its progression.

Cellular NADPH is known to be limited by expression of NAD+ kinase (233) and could be depressed by loss of a repair system that restores damaged NADPH (234). In addition, there are reports of an NADP+ phosphatase (235) and NADP+ glycohydrolase activities (236)— induction of such enzymes could be responsible for loss of these metabolites. By diminishing levels of NADPH, any of these mechanisms could lower the capacity of hepatocytes and potentially other cells to detoxify ROS (231) and diminish circadian functions (237), thereby contributing to two major systems depressed in obesity. Ongoing work is designed to test the effect of NR on ROS damage in PD, T2D, PDPD and DPN, to discover the basis for depressed hepatic NADP+ and/or NADPH in PD, and to translate these results to human populations.

137

7.6 Acknowledgments

This work was supported by a pilot and feasibility grant from the Fraternal Order of Eagles Diabetes Research Center, the Roy J. Carver Trust, National Institutes of Health grant DK081147, and grants from the Department of Veterans Affairs, BX001680-01, RX000889-01 and C9251-C.

138

7.7 Figures and Tables for Sections 3-5

Figure 7.1 NR improves metabolic parameters in PD and T2D.

Male C57Bl/6J mice were made prediabetic (PD) with HFD or diabetics (T2D) with HFD and STZ treatment (Figure 7.4). All groups including normal chow (NC) animals were supplemented as described in text. Gross metabolic parameters of PD and T2D were measured as described in methods. (a)

NR reduces weight gain on HFD independent of STZ. (b – d). NR reduces hepatic steatosis in the PD and T2D models. NR reduces circulating cholesterol (e) and alanine aminotransferase (f) in PD. In T2D, NR tends to lower HbA1C (g) and depresses nonfasting glucose (h). NR depresses fasting glucose in both models (i). NR improves GTT in PD (j). Overall, NR opposed the deleterious metabolic effects of HFD and HFD + STZ. Statistics were by two-way ANOVA followed by a Holm-Sidak multiple comparisons test. n = 10. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

139

Figure 7.2 NR opposes PDPN and T2DPN.

Prior to sacrifice, neuropathy was assessed in both PD and T2DPN models. (a) NR protects against a decline in motor nerve conduction velocity (MNCV) in T2D. (b) NR protects against declines in sensory nerve conduction velocity (SNCV) in PD and T2D. (c) NR protects against loss of thermal sensitivity in both models. (d) and (e) NR improves INFD on NC and in both disease models. NR acted to oppose all measured forms of neuropathy in both models. Statistics were by two-way ANOVA followed by a Holm-Sidak multiple comparisons test. n = 10. **, P < 0.01; ***, P < 0.001.

140

Figure 7.3 Activity of NR in DPN can be monitored by corneal confocal microscopy (CCM).
(a) and (b) CCM is a sensitized assay for PD and T2D nerve loss and the protective effects of NR. (c) and (d) By post-mortem class III β-tubulin staining, NR protects against corneal sub- epithelial nerve loss in T2D. CCM is a non-invasive assay that can be performed on live animals. CCM may be an effective tool in assaying neuropathy and the action of anti- neuropathic agents such as NR in future clinical trials. Statistics were by two-way ANOVA followed by a Holm-Sidak multiple comparisons test. n = 10. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

141

Table 7.1 The hepatic pool of NADP+ and NADPH is depressed by PD and T2D and is partially restored by NR.
Livers were excised and analyzed using LC-MS/MS. NAD+ was depressed in PD and T2D animals. NR restored NAD+ to control levels in all cases. NADP+ was significantly depressed in PD and T2D animals and partially restored by NR in both cases. The data suggest that the ROS detoxification system is undermined by HFD and that NR could be protective against its diminishment. Values are expressed as mean ± SEM pmol/mg liver. Underlined concentrations within a treatment are significantly different from NC after collapsing for the effect of supplementation. The effect of treatment and supplementation of NR were analyzed by two-way ANOVA followed by multiple comparisons using the Holm-Sidak test. † P < 0.05, †† P <0.01, ††† P <0.001 for effect of NR within a treatment group (i.e. NC vs. NC+NR, HFD vs. HFD+NR, HFD+STZ vs. HFD+STZ+NR). # P < 0.05, #### P <0.0001 for effect of treatment versus NC within supplementation group (i.e. NC vs HFD, NC vs HFD+STZ, NC+NR vs HFD+NR, NC+NR vs HFD+STZ+NR). *** P <0.001 for effect of STZ vs. HFD+STZ.

NAD+

NADP+

Nam

ADPR

MeNam

NMN

Me4PY

NR

NC

1200 ± 84

230 ± 11

170 ± 9.3

84 ± 17

3.1 ± 0.28

2.8 ± 0.13

2.6 ± 0.25

1.2 ± 0.12

NC+ NR 1500 ± 99

250 ± 16

230 ± 20†

98 ± 19

4.4 ± 0.25

3.7 ± 0.24

7.4 ± 0.64†††

2.0 ± 0.45

HFD

1000 ± 85

150 ± 9.2####

190 ± 12

60 ± 8.6

2.7 ± 0.14

3.7 ± 0.36

6 ± 0.45#

1.6 ± 0.38

HFD+ NR 1200 ± 95

200 ± 17†,#

180 ± 7.8#

70 ± 14

2.9 ± 0.22

3.5 ± 0.14

8.3 ± 1.5

HFD+ STZ 760 ± 72

150 ± 13####

130 ± 9.0

44 ± 10

2.4 ± 0.32

2.2 ± 0.30***

4.2 ± 0.37###

HFD+STZ+ NR 1300 ± 180†

200 ± 11†††

210 ± 17††

84 ± 19

3.8 ± 0.63

3.1 ± 0.21

7.9 ± 0.95†

1.1±0.15 1.0±0.15 2.3±1.3

142

7.8 Supplemental Materials

Sample Extraction for NAD+ Metabolomics
Murine liver obtained by freeze-clamp was pulverized using a Bessman pulverizer (100 –

1000 mg size) (Spectrum® Laboratories, Rancho Dominguez, California) cooled to liquid N2 temperatures. Each pulverized liver sample was aliquoted (5 – 20 mg) into two liquid N2 cooled 1.5 ml centrifuge tubes, which were stored at -80 °C until analysis. Prior to analysis, 13C-labeled yeast extract (1) was added to one tube and an internal standard solution containing 60 pmol of 18O NR, 60 pmol of 18O nicotinamide (Nam), 60 pmol of 18O D3 1-methylnicotinamide (MeNam), and 240 pmol of D4 nicotinic acid (NA) (C/D/N Isotopes, Pointe-Claire, Quebec, Canada) was added to the other tube while on dry ice. Samples were extracted by addition of 0.1 ml of buffered ethanol (3 parts 100% ethanol: 1 part 10 mM HEPES, pH 7.1) at 80 °C. Samples were vortexed vigorously until thawed, sonicated in a bath sonicator (10 seconds followed by 15 seconds on ice, repeated twice), vortexed, then placed into a Thermomixer® (Eppendorf, Hamburg, Germany) set to 80 °C and shaken at 1050 rpm for five minutes. Samples were centrifuged (16.1kg, 4 °C, 10 minutes). Clarified supernatants were transferred to fresh 1.5 ml tubes and dried via speed vacuum for two hours. Prior to LC-MS/MS analysis, samples were resuspended in 40 μl of 10 mM ammonium acetate (>99% pure) in LCMS-grade water. All samples were transferred to Waters polypropylene plastic total recovery vials (Part # 186002639) and stored in a Waters Acquity H class autosampler maintained at 8 °C until 10 μl injections.

LC-MS/MS Analysis for NAD Metabolomics Chromatographic separation was performed using a 2.1 mm X 100 mm Thermo

Scientific HypercarbTM column as described with slight modification to the alkaline separation (1). Specifically, flow rate was increased to 0.55 ml/min and run time shortened to 11.6 minutes.

143

Separation was performed using a modified gradient with initial equilibration at 3% B, a 0.9 minute hold, a gradient to 50% B over 6.3 minutes, followed by a 1 minute wash at 90% B, and a 3 minute re-equilibration at 3% B. Analytes were detected and quantified using a Waters TQD operated in positive ion multiple reaction monitoring (MRM) mode. MRM transitions for all mononucleotides and dinucleotides in the acidic separation were as described (1), though NADH was not quantified due to high variability in the internal standard mixture. Newly quantified metabolites in the acidic separation, MeNam and Me4PY were assayed with the following transitions: MeNam (137 > 94 m/z) and Me4PY (153 > 136 m/z). Samples were electrospray ionized at a capillary voltage of +3.1 kV, a desolvation gas flow rate of 500 l/hr, a cone gas flow rate of 100 l/hr, a desolvation temperature of 350 °C, and a source temperature of 150 °C. Analytes were quantified using a calibration curve containing the same concentration of internal standard as samples with external standard concentrations ranging from 0.1 – 100 μM. When NAD+ concentration exceeded the linear range, the sample was diluted by 10 and samples re-analyzed. All pmol amounts were normalized to g of wet liver weight extracted

144

7.9 Supplemental Figures

Figure 7.4 Experimental design and weight gain.

12 week old, male C57Bl/6J mice were placed on either NC or HFD for 8 weeks. At 8 weeks, half of the HFD fed mice were treated with two low doses of STZ to create the type 2 diabetes (T2D) model. Mice remaining on HFD without STZ treatment are the prediabetic (PD) model. Both groups continued on HFD for the remainder of the study. Five weeks after creation of the T2D group, all conditions were supplemented with NR, creating six conditions (NC, NC + NR, HFD, HFD + NR, HFD + STZ, HFD + STZ + NR). Weight was followed throughout the experiment as shown above. Glucose tolerance tests (GTTs) were performed at 16 weeks of the experiment. Measurements of neuropathy were performed on or after sacrifice. At time of sacrifice, blood and liver were collected for metabolic and LC-MS/MS metabolomic analyses. Mouse age and weeks of experiment are displayed on the x-axis. NR caused resistance to HFD-induced weight gain in both PD and T2D mice.

145

Figure 7.5 GTT primary data used for Figure 7.1 i and j.

Glucose tolerance tests (GTTs) were performed at 16 weeks (Figure 7.4) of the experiment as described in methods. Data from all groups are displayed at left. At right, GTTs of each diet condition +/- NR are displayed with NC at top, HFD in the middle, and HFD + STZ at bottom. A repeated two-way ANOVA was performed to test for an interaction between condition or treatment and time. NR did not significantly GTT over time. Overall, NR tended to improve GTT in PD and T2D mice with significant differences observed as indicated. A post-hoc Holm-Sidak multiple comparison was performed to test for significant differences between no NR and NR within each time point. n = 10. *, P < 0.05; **, P < 0.01; ***, P < 0.001

146

7.10 Perspective on Chapter 7

In the above sections of this chapter, we enumerated the beneficial effects of NR on pre- diabetic and diabetic mouse models. However, before investigating the effect of NR on type 2 diabetes, we evaluated NR as an anti-PDN agent in a type 1 diabetic rat model. Here, I describe our finding in the type 1 diabetic model while discussing how these effects inform our findings in the type 2 diabetic mice. Additionally, I measured NADH and NADPH in the same samples analyzed by LC-MS/MS in Table 7.1 and discuss the findings in light of the PD and T2D effects on the oxidized NAD metabolome.

7.11 Results and Discussion

Type 1 Diabetes in Rat Compared to Type 2 Diabetes in Mouse
At first, we chose to investigate a type 1 diabetic (T1D) model to test the hypothesis that

NR opposes DPN. Male Sprague-Dawley rats were treated with streptozotocin (STZ) as described (238) to induce T1D and began treatment on NR Cl at a dosage of 0.3% (w/w of chow (Harlan Teklad, #7001, Madison, Wi)) after 96 hours post STZ administration. NR Cl treatment lasted on average for six weeks. The animals in this initial experiment were not treated with insulin. Unlike with PD and T2D (Figure 7.2 and 7.9 Supplemental Figure 7.5), NR was unable to oppose the heightened, non-fasting blood glucose experienced by T1D (Table 7.2). Additionally, A1c, indicative of the long term concentration of blood glucose, was elevated in T1D compared to control and was not improved by NR. However, the hyperglycemia experienced in the T1D model appears to have been more dramatic than that of PD and T2D (Holm-Sidak test: T1D vs PD: P < 0.001; T1D vs T2D: P < 0.001) and may have overpowered the modest though significant effects of NR on hyperglycemia. This may also explain the lack of improvement in the blood lipid profile. Essentially, these results established that these animals were severely diabetic and of poor health.

147

Despite the severe diabetes and similar to its effect in T2D, NR improved neuropathy (Table 7.3). Though variability was high in the mechanical test, T1D rats were less responsive to mechanical prodding and NR treatment increased sensitivity. NR significantly prevented thermal insensitivity. Motor nerve conduction velocity and sensory nerve conduction velocity tended to be and was significantly aided by NR, respectively. C fiber density was also maintained by NR administration. Together, these results are perfectly congruent with what was observed with T2D, i.e. NR opposes DPN in a manner independent of hyperglycemia.

Unlike the mouse sciatic nerve, the rat sciatic nerve is large enough for NAD metabolomic analysis. Granted, as stated above (Chapter 7.5), a problem of normalization exists. The diabetic nerves are dying back, meaning DNA, RNA, protein, and other macromolecules are decreasing in mass, indicating that similar masses of sciatic nerve in the diabetic animals may cause an overestimate of the NAD metabolome. Additionally, at this time homogenates containing 10 mM Nam1 were utilized for NAD metabolomic analysis. This addition complicates interpretation of the results as the alterations of the NAD metabolome between conditions would represent in vivo effects on metabolite concentration combined with the ability of any one sample to metabolize Nam. With these in mind, T1D depressed NMN, NADP+, and NAD+ (Table 7.4). NR tended to elevate NAD+ and NADP+ and significantly increased NMN. These results are very similar in profile to the hepatic NAD metabolome in T2D with and without NR, suggesting the murine hepatic NAD metabolome may be representative of the murine sciatic nerve.

The results from the T1D rat model strongly revealed that NR affected DPN independent of glycemic control and that this effect correlated with improvement of NAD+ and related metabolites. Though these results were fairly clear, we decided to extend treatment with NR form six weeks to 8 weeks and also to compare the effects of NR to the effects of Nam and NA.

1 Nam was added to the homogenate in large quantities to maintain the mitochondrial acetylation profile. It was hoped that the same homogenate could be used for multiple analyses.

148

12 week old male Sprague-Dawley rats were allocated into five groups with 10 animals per group and either administered STZ as above or treated with vehicle. Treatment with 0.3% (w/w) of NR Cl or mole equivalent of Nam and NA began 96 hour after STZ or vehicle administration. These animals were treated with insulin every other day. All three B3 vitamins increased mortality starting at ~3 weeks post beginning of treatment. Unfortunately, the times of death were not recorded, disallowing appropriate analysis of survival. What is known, is that at the end of only 6 to 7 weeks, 4, 3, and 2 rats remained in the NR, Nam, and NA treated groups, respectively. Only three rats were sacrificed in the control and STZ groups as the remaining rats were diverted to other experiments. The dramatic increase in mortality is likely due to the poor health of the T1D model further confounded by the insulin resistance effects of Nam and NA (239, 240) which worsened hyperglycemia in these animals (Table 7.5). In the T2D model, glucose intolerance increased in NC animals fed NR (Figure 7.1 j and 7.9 Supplemental Figure: Figure 7.5), which could be a result of insulin resistance. Further work is necessary to measure insulin resistance in NR supplemented animals but is a crucial question to answer before clinical testing.

Though the sample sizes were low and the animals were visibly ill2 at time of sacrifice, B3 vitamins appear to cause differential effects on DPN. T1D and those treated with Nam tended to be less sensitive to mechanical stimuli (Table 7.6). T1D caused significant thermal insensitivity, which as was not experienced by any B3 vitamin treated groups. Interestingly, MNCV dysfunction was prevented only by NR. In this experiment, T1D tended to experience deficit in SNCV (28 ± 1.2 vs 34 ± 2.3 m/s), but Nam caused significant deficit (25 ± 0.2 vs 34 ± 2.3 m/s, P < 0.05) whereas NA tended to do so (27 ± 3.7 vs 34 ± 2.3 m/s). NR was indistinguishable from control (32 ± 1.3 vs 34 vs 2.3 m/s). Together, these data suggest specific effects of NR and not a general effect of B3 vitamins.

2 Animals appeared lethargic and many died after administration of anesthesia. Some animals contained blood in kidney and intestines.

149

In the second T1D experiment, sciatic nerves were snap frozen in liquid nitrogen then prepared in a manner similar to liver (Chapter 2.2) for analysis using LC-MS/MS. Unlike the previous samples where sciatic nerve was prepared as a homogenate, NMN, NADP+, and NAD+ were not depressed by T1D nor augmented by supplementation of any B3 vitamin. Unlike in murine liver (Table 7.1), MeNam and Me4PY were unaffected by T1D but both as well as Me2PY were greatly increased in sciatic nerve after B3 vitamin supplementation, indicating that though NAD+ was not increased, the NAD metabolome was activated by B3 vitamins. These methylated metabolites could be indicative of either increased flux through NAD+ whereby precursor would contribute to NAD+ then degrade to Nam through enzymatic consumption (Chapter 1.2) and then become methylated and subsequently oxidized. As stated in Chapter 5.2, these methylated compounds could also occur without utilization for NAD+, complicating interpretation of these metabolites as a measure of flux. Unfortunately, NAAD was not detected in this experiment (Chapter 5). Using the same technologies described in Chapter 3 – 5, stable labeled NR supplementation could clarify whether NAD+ is effectively elevated by NR in the sciatic nerve with expectation that NAD+ enrichment would increase as a function of T1D after NR supplementation. However, further experiments whereby the concentration of NR was decreased by a third produced similar toxicities to 0.3% NR dosage. Due to the incredible increase in mortality, we abandoned the T1D model for the T2D model.

In total, NR prevented DPN in the T1D model in a similar matter to that of T2D. First, NR effects appear to be independent of glycemic control. Tight glycemic control has been shown to prevent DPN in type 1 diabetic individuals (241). This glycemic control is achieved through an intensive regimen with 3 or more insulin doses per day compared with the conventional treatment of 1-2 per day, which is impractical in the treatment of type 2 diabetics. NR increases glycemic control in T2D, but this effect appears dispensable for its action on DPN given NR acted as an anti-neuropathic in both models of diabetes. These results suggest NR may be efficaciously coupled to other glycemic normalizing drugs such as metformin and insulin to

150

prevent and treat further neuropathy. Future work in rodent models is necessary to establish whether NR is synergistic with other anti-diabetic drugs.

NADH and NADPH Measurement in T2D Murine Liver
T2D mice experience depression in liver NADP+. In our routine analysis of the NAD metabolome as described in Chapter 2, an amount of NADPH oxidizes to NADP+ (61), and

hence, measurement of NADP+ is a proxy measurement for NADPH. Without direct measurement, the effect of HFD on the NADP+ plus NADPH pool could either represent increased reduction to NADPH with concomitant detriment to NADP+, diminishment of NADPH alone, or decrease in both. NAD+ tended to decrease as a result of HFD -/+ STZ, but the fate of its reduced form is unknown. HFD could certainly decrease the NAD+/NADH (189, 193). In order to measure these metabolites and gain further insight into the effect of high fat diet and STZ on the NAD metabolome, a methodology for the extraction and analysis was developed (Chapter 2.2: Quantification of NAD(P)H and Extraction from Liver) and performed on the same liver samples used in the original analysis (Figure 7.4). In development of the NAD(P)H assay, we also determined the amount of NADH and NADPH that became NAD+ and NADP+ through the processing and analysis of the oxidized NAD metabolome. 12% and 60% of NADPH and NADH contributed to the NADP+ and NAD+ quantitation in our routine analysis, respectively. Based upon this finding, more accurate NADP+ and NAD+ was determined by subtracting the oxidized form concentration of each by the product of the reduced forms and the appropriate factor. The trends in NADP+ and NAD+ concentration in Figure 7.4 are exactly as described in table 7.1 with NADP+ depressed by HFD and even more so by HFD+STZ (Figure 7.4d) and NAD+ trending down (Figure 7.4a). NR easily restored hepatic NAD+ and tended to restore NADP+ though not back to control levels. The reductive metabolites presented an interesting pattern. NADH was unaffected in PD and T2D (Figure 7.4b). The NAD+/NADH ratio which is often used a measurement of the metabolic status of a cell decreased not because both metabolites decreased but solely due to NAD+ deficits (Figure 7.4a and c). PD depressed NADPH (P <

151

0.001) compared to NC (Figure 7.4e). T2D depressed NADPH compared to both NC (P < 0.001) and PD (P < 0.05). In both disease models, NR tended to increase NADPH but failed to reach statistical significance and in fact was still significantly depressed compared to control in T2D (P < 0.01). Since both NADP+ and NADPH were depressed in approximately equal amounts (Figure 7.4f), the oxidized-to-reduced ratio remained unchanged regardless of disease or supplementation. These results indicate that NAD+ is specifically affected by PD and T2D and not NADH. Increased acetylation in both the nucleus and mitochondria is observed after HFD (37, 242, 243), which if NAD+ is limiting, could decrease NAD+ through the deacylation action of sirtuins. Further, it implies that modest depletion of NAD+ is non-limiting to oxidoreductase reactions given NADH concentration remained constant. In contrast, NADP+ and NADPH appear to be intimately connected with deficit in one leading to deficit in the other in a nearly one-to-one manner. This presents an interesting problem in measuring the ratio of the two metabolites rather than their individual concentrations. If this ratio were the readout for investigating an effect on ROS detoxification, the consistency in the ratio would lead to a false negative conclusion. In our data, we observe that the capacity of these phosphorylated NAD+ analogs to oppose ROS may be crippled. Regardless of the ratio, if the mole amount of ROS outpaces the mole amount of NADPH, GSH will not regenerate, and the ROS glutathione detoxification pathway will collapse. As stated above, future work is aimed at unraveling the damaging effects of NADP+ and NADPH deficit in both PD and T2D and whether NR opposes said damage.

7.12 Methods

Unless otherwise stated, all methods were performed as described in Chapter 7.2 Methods. NAD(P)H were measured as described in Chapter 2.2: Quantification of NAD(P)H and Extraction from Liver. Sciatic nerve homogenate (0.05 ml) was thawed and combined with appropriate internal standard solutions (Chapter 7.8 Supplemental Materials) and extracted with

152

0.3 ml of buffered boiled ethanol as described Chapter 2.1. Metabolites separated in the alkaline condition were analyzed as described (1). MeNam, Me2PY, and Me4PY were analyzed as detailed in Chapter 7.5 Supplemental Methods of this chapter and as described in Chapter 2.3 Continued Method Development Post-Initial Publication: Addition of MeNam, Me2PY, and Me4PY to the NAD Metabolomic Assay.

153

7.13 Tables and Figures

Table 7.2 Glycemic control, dyslipidemia, and overall health were not improved by NR in T1D.
Male Sprague-Dawley rats made Type 1 diabetic with STZ treatment and immediately began supplementation with NR. Starting and ending weight were recorded for each rat after six weeks of experiencing T1D and NR supplementation. Blood parameters were measured post-mortem. Statistical significance was tested using a one-way followed by a multiple comparisons test with the Holm-Sidak method. A two-way ANOVA was performed to test for statistical significance between starting and ending weight followed by a multiple comparisons tests using the Holm- Sidak method. Results are displayed in parentheses. * P < 0.05, ** P < 0.01, *** P < 0.001 versus Control

Starting weight (g)

Ending weight (g)

Blood glucose (mg/dl)a

Blood triglycerides (mg/dl)
Blood cholesterol (mg/ml) Hemoglobin A1c (%)

Control (n=9)

328±12

467±24

152±31

63.2 ± 10.5

1.13 ± 0.06

6.7 ± 0.4

STZ Treated (n=7)
318 ± 10 257 ± 39*** (0.083)

580 ± 37***

634 ± 175**

4.23 ± 0.88

17.9 ± 1.1

STZ Treated + NR (0.3%) (n=6) 319±7
228 ± 34***

(0.0368)

562 ± 36***

571 ± 127*

4.18 ± 0.66

16.6 ± 0.5

a Blood glucose meters measure at maximum 600 mg/dl, meaning some observations are at or outside of the maximum range.

Table 7.3 NR opposes T1D neuropathy.

Mechanical response threshold, thermal response latency, motor nerve conduction velocity, and sensory nerve conduction velocity were measured prior to sacrifice. C fiber density was measured post-mortem in rat hind paws. Statistical significance was determined using a one- way ANOVA followed by a Holm-Sidak multiple comparisons test. STZ tended to worsen measurements of sensation and nerve conduction velocity as well as paw pad C fiber density. NR tended to ameliorate all measurements. * P < 0.05 versus Control, † P < 0.05 versus STZ

Mechanical response 13.8 ± 4.7 26.8 ± 31.8 15.0 ± 3.6 threshold (g)

Thermal response latency (sec)

Sensory nerve conduction velocity (m/sec)
C fiber density (/mm2)

Control (n=9)

9.8 ± 0.9

20.2 ± 0.9

14.9 ± 1.2

STZ Treated (n=7)

16.3 ± 2.3*

16.6 ± 0.8*

10.6 ± 0.6*

STZ Treated + NR (0.3%) (n=6)

8.8 ± 1.4†

18.4 ± 0.7

14.1 ± 0.4†

Motor nerve conduction 62.6 ± 7.4 40.6 ± 4.2 54.9 ± 5.2 velocity (m/sec)

154

Table 7.4 NR tends to improve STZ induced NAD metabolome defects in sciatic nerve homogenate.
Sciatic nerves were excised on the day of sacrifice from Sprague-Dawley rats and prepared as described in 7.12 methods for LC-MS/MS analysis. Overall, NMN, NADP+, and NAD+ tended to decrease as a function of STZ and increase a function of NR. Metabolites are expressed as mean ± SEM pmoles/mg of dry particulate weight. Significance was determined using a one- way ANOVA followed by a Holm-Sidak multiple-comparisons test. * P < 0.05 compared to STZ

NAD+

NMN

NADP+

Control (n= 6)

270 ± 150

86 ± 26

30 ± 12

STZ (n = 6)

136 ± 36

36 ± 7.9

22 ± 5.7

STZ + NR Cl (0.3%) (n = 6)
390 ± 9

110 ± 18*

44 ± 7.3

Table 7.5 B3 vitamins were ineffective in improving glycemic control and overall health.

Type 1 diabetic rats were treated with either 0.3% (w/w) NR or molar equivalent of Nam and NA for up to 12 weeks after STZ treatment. Starting and ending weight as well as non-fasted blood glucose were measured in these animals. NR did not improve these parameters. Weight data were analyzed using a two-way ANOVA followed by a Holm-Sidak multiple comparisons test. P values between start and end are shown in parentheses. Blood glucose measurements were tested for significance using a one-way ANOVA followed by the same type of multiple comparisons test as above. *** P < 0.001 versus Control

Control (n = 3)

Ending 420 ± 13 Weight (<0.001) (g)

STZ (n = 3)

260 ± 17

STZ + NR (n = 4)

270 ± 20***

STZ + Nam (n = 3)

250 ± 4.3*** (0.0315)

STZ + NA (n = 2)

270 ± 1.5***

Starting 300 ± 4.9 290 ± 7.1 300 ± 6.6 300 ± 0.9 Weight(g)

321 ± 0.5

Blood 150 ± 20 590 ± 10*** 580 ± 11*** 590 ± 9.3*** Glucose
(mg/dl)a

540 ± 65***

a Blood glucose meters measure at maximum 600 mg/dl, meaning some observations are at or outside of the maximum range.

155

Table 7.6 Among the B3 vitamins, NR consistently opposed aspects of T1D neuropathy.

Type 1 diabetic rats were treated with either 0.3% (w/w) NR or molar equivalent of Nam and NA for up to 12 weeks after STZ treatment. NR tended to prevent T1D induced neuropathy in a similar fashion as observed after six weeks of T1D (Table 7.3). Nam and NA were indistinguishable from non-treated STZ. Nam appears to have caused a deficit in sensory nerve conduction velocity. Measurements were tested for significance using a one-way ANOVA followed by the same type of multiple comparisons test as above. The low sample sizes preclude strong conclusions. * P < 0.05, ** P < 0.01 versus Control

Tactile Response 16 ± 2.4 8.0 ± 1.6 14 ± 1.7 16 ± 3.5 9.6 ± 1.3 (g)

Thermal Response Latency (sec)

Sensory Nerve Conduction Velocity (m/sec)

Control (n = 3)

9.9 ± 0.4

34 ± 2.3

STZ (n = 3)

17 ± 1.3*

28 ± 1.2

STZ + NR (n = 4)

11 ± 1.4

32 ± 1.3

STZ + Nam (n = 3)

14 ± 1.6

25 ± 0.2*

STZ + NA (n = 2)

10 ± 0.5

27 ± 3.7

Motor Nerve 60 ± 3.9 38 ± 1.4** 50 ± 4.2 43 ± 1.4* 42 ± 0.05* Conduction
Velocity (m/sec)

Table 7.7 All three B3 vitamins tended to alter the NAD metabolome in
sciatic nerve.
Type 1 diabetic rats were treated with either 0.3% (w/w) NR or molar
equivalent of Nam and NA for up to 12 weeks after STZ treatment. Sciatic nerve was excised at time of sacrifice and flash frozen. Tissue was extracted and analyzed using LC-MS/MS. Measurements were tested for significance using a one-way ANOVA followed by the same type of multiple comparisons test as above.

Metabolites are expressed as mean ± SEM pmoles/mg of dry particulate weight.

NAD+

Nam

MeNam

NADP

NMN

Me-4PY

Me-2PY

Control (n = 3)

49 ± 1.1

5.8 ± 0.68

0.19 ± 0.06

3.8 ± 0.67

1.6 ± 0.55

0.37 ± 0.19

<0.18

STZ (n = 3)

64 ± 11

4.3 ± 0.68

0.15 ± 0.05

4.1 ± 0.45

1.1 ± 0.13

0.36 ± 0.06

<0.18

STZ + NR (n = 4)

61 ± 12

6.1 ± 1.6

1.7 ± 0.39

4.9 ± 1.1

3.2 ± 1.4

0.94 ± 0.32

1.2 ± 0.40

STZ + Nam (n = 3)

67 ± 11

13 ± 8.0

3.2 ± 1.8

4.0 ± 0.30

1.2 ± 0.23

0.45 ± 0.28

0.65 ± 0.57

STZ + NA (n = 2)

80 ± 21

8.0 ± 0.34

2.2 ± 0.62

3.9 ± 1.2

2.6 ± 0.28

1.1 ± 0.93

0.35 ± 0.14

156

Figure 7.6 NADP+ and NADPH were equally depressed by PD and T2D and improved by NR.
Livers were excised at time of sacrifice and prepared for LC-MS/MS or LC-MS as described in methods. (a) NAD+ was depressed by HFD and HFD+STZ regardless of NR. PD and T2D tended to be depressed compared to control and were restored by NR. (b) NADH remained unchanged and consequentially (c) the NAD+/NADH ratio was depressed in HFD and HFD+STZ as a function of NAD+. In contrast, (d) NADP+ was significantly depressed in PD and T2D and was partially restored by NR. (e) NADPH was similarly affected and as consequence (f) the NADP+/NADPH ratio remained unchanged. Overall, the results reveal an effect of HFD that is compounded by STZ and partially restored by NR. Metabolites are expressed as mean ± SEM pmoles/mg wet liver weight. Statistical significance was measured using two-way ANOVA and a Holm-Sidak multiple comparisons test * P < 0.05, ** P < 0.01, *** P < 0.001. Comparisons between groups regardless of NR supplementation are shown below graphs.

157

CHAPTER 8

GENERAL SUMMARY AND FUTURE DIRECTIONS

8.1 General Summary

NAD+ is crucial to the health of a cell and organism. Pellagra, a disease of NAD+ deficiency, was common in the American rural south a century ago and represented one of the first recognized health crises in the United States. Investigators identified Pellagra as a nutritional disease caused by a diet mainly consisting of maize and lard. Pellagra was cured by including milk and animal meat in the diet and today has been eliminated in high income nations. Later investigations, identified the first B3 vitamins NA and Nam as anti-Pellagra agents. We began this century with new health crises related to an aging population consuming low vitamin, high calorie diets and, as consequence, experiencing obesity, diabetes and heart disease at a frequency never before observed. We began this millennium with the identification of sirtuins as anti-aging enzymes in yeast that depend upon NAD+ for activity. Intense investigation into the role of sirtuins and other NAD+ glycohydrolases as targets against age and obesity related morbidities ensued. These investigations continue to reveal these disorders may also be diseases of dysfunctional NAD+ metabolism. Concurrent with these investigation, the Brenner laboratory identified NR as a novel B3 vitamin that acts to oppose aging in yeast. NR was later shown to oppose metabolic and neurodegenerative disorders, implicating NR as a health-promoting agent in rodent models. In this thesis, I developed and implemented LC-MS based NAD metabolomic technologies to assess the extra- and intracellular NAD+ concentration and related metabolites in health and disease. The information gathered here represents a first step in translating the health promoting actions of NR to humans and demonstrates the power of utilizing metabolomic tools in answering biological questions.

In this thesis, I describe current methods for measuring NAD+ and related metabolites and their disadvantages compared to LC-MS. The most common methods, enzyme coupled

158

assays and HPLC, lack specificity and, as consequence, are prone to erroneous quantitation. Further, most investigators report the oxidized versus reductive ratios of NAD+ to NADH and NADP+ to NADPH. These ratios reflect gross abnormalities in the NAD metabolome but fail to elucidate the exact impact of disease/drug/treatment. In my investigations, the ratio of NAD+/NADH was altered upon ethanol ingestion (Chapter 6) in a manner depending upon NADH increase. In Chapter 7, my NAD metabolomic analysis revealed the NADP+/NADPH was unchanged but the absolute quantity of both metabolites decreased in models of prediabetic and diabetic animals. In both cases, NR positively affected the individual metabolites but did not greatly affect the ratio. These examples reveal these ratios are not necessarily indicative of disease nor treatment.

I developed quantitative NAD metabolomics using LC-MS and LC-MS/MS due to the superior sensitivity and specificity. Though these technologies add much needed specificity in measurement, quantitation is influenced by sample induced ion suppression. I improved upon earlier LC-MS based NAD metabolomics by employing isotopologue internal standards to control for ion suppression and sample extraction efficiency. Extraction of these metabolites presented challenges due to the chemical behavior of the metabolites, especially NADH and NADPH, versus the rest of the metabolome. I developed a novel extraction method and LC system to quantitate these reduced metabolites along with the rest of the NAD metabolome. In so doing, I uncovered that a commonly used ion pairing agent (TBA) is incompatible with NADPH quantitation and may impact quantitation of other organic acids. I then reveal that TEA may be an appropriate alternative to TBA.

In quantifying what we refer to as the NAD metabolome, I was able to assess within a dataset the quality of the data. Normally, the Nam/NAD+ ratio is approximately 1:5 (Chapter 5) but inappropriate sample handling can increase this ratio to ~30 (Table 6.1). Additionally, inclusion of metabolites not related to NAD+ (UMP, Uridine, CMP, Cytosine, etc.) allows distinguishing of NAD+ metabolism specific effects versus base, nucleoside, and nucleotide

159

metabolic defects. These improvements in NAD metabolomic analysis may be crucial to investigations of NR as a naturally occurring substance and a therapeutic agent.

NR is a natural precursor found in milk albeit at trace amounts but makes up a majority of its B3 content. In quantifying the abundance of NR, I found that NR concentration negatively correlated with Staphylococcus aureus infection. I employed stable isotope technologies to determine the way in which NR is metabolized by these microbes. In so doing, I found that these microbes hydrolyze NR via an unknown enzyme resulting in NR serving as a Nam precursor to this bacterium. To my knowledge, this is the first indication of NR hydrolysis by this bacterium and requires further investigation to identify responsible enzyme(s). But to a greater extent, the technology exemplifies a procedure to uncover unknown pathways, which proved instrumental in subsequent work.

The health promoting effects of NR are known to overlap with its phosphorylated form, NMN. These findings are obvious if both depend upon NAD+ elevation for their therapeutic action, but to date, no one has directly measured the efficacy of NR versus NMN as a precursor to NAD+. The metabolism of NMN has become a source of great debate over direct import versus extracellular metabolism to NR. I compared the kinetics of NR versus NMN using stable isotope technologies and revealed NR is a superior precursor to NMN and that NMN appears to be converted to NR extracellularly. Hence, the therapeutic effects of NMN are almost certainly a result of NR, meaning NR is likely a superior therapeutic agent.

The therapeutic effects of NR are achieved at high dosage in mice, receiving 400 mg/kg body weight per day from ad libitum feeding. This design indicates that these mice are receiving a near constant lower dose over the entire day, which is nearly impossible to translate to a treatment regimen for a chronic disease. As a proof-of-principle, a healthy 52 year old male weighing about 63 kg ingested 1 g of NR to assess whether NR could alter the NAD metabolome. I then analyzed the blood and urine of this individual collected over a day and a week after NR supplementation and found robust increases in NAD+ and associated

160

metabolites. These data are the first indication that the health promoting effects in rodents could translate to a human being in a controlled dose strategy.

Since the health promoting effects of NR depend upon its ability to increase NAD+ and possibly other metabolites (73), measuring NR mediated effects on the NAD metabolome in the target tissue could serve as an indicator of efficacy for NR in opposing disease. Many targets are inaccessible in clinic, warranting identification of accessible biomarkers for NAD+ status. In performing the first human trial of NR supplementation, I noticed a large increase in NAAD that occurred around the same time as NAD+ elevation. NAAD is part of the deamidated pathway and is thought to be synthesized from tryptophan and NA, not from NR. The increase in NAAD could result from a negative feedback mechanism that inhibits NAD synthase, the glutamine requiring enzyme that converts NAAD to NAD+ or from direct conversion of NR or subsequent metabolite (Nam, NMN, NAD+) through an unknown deamidase pathway.

To test this hypothesis, I compared the kinetics of non-labeled NR, Nam, and NA and employed the same isotope labeling technology as with the Staphylococcus aureus experiment. Based on these experiments, NR is a superior liver precursor to NAD+ compared to Nam and NA and the increase in NAD+ correlated with NAAD. I proved that NR directly contributes to murine hepatic NAAD and that the isotopic distribution of NAAD correlated with NAD+, indicating the presence of a deamidating pathway and that NAAD may truly serve as both a precursor of and biomarker for NAD+. We tested whether NAAD could act as a biomarker for efficacious B3 vitamin supplementation by performing the first multi-subject human trial of NR at several one- time dosages. Blood NAD+ was variable in the subjects and mostly did not respond to NR; however, NAAD responded in a dose dependent manner, supporting NAAD as a biomarker. In studying alcoholic fatty liver disease (AFLD), I uncovered that NAD+ is mildly abrogated whereas NAAD decreased by half but qualitatively correlated with the decrease in NAD+, suggesting NAAD may serve as a biomarker for not only NAD+ elevation but of NAD+ deficit

161

(Table 6.1). In this same experiment, NR increased NAAD, agreeing with the human results and further establishing NAAD as an NAD+ biomarker.

The work presented in Chapters 3 – 5 establishes that NR is a naturally occurring metabolite that is by far the superior precursor to NAD+ in vitro and in vivo. With this in mind, we turned our attention to NR in treatment of metabolic disease, specifically, AFLD (Chapter 6) and diabetic peripheral neuropathy (Chapter 7). Chronic ethanol ingestion leads to fatty liver disease, but the mechanism of fatty deposition remains unclear. Ethanol metabolism causes a dramatic reductive skew in the NAD+ pool, supplies an acetylating precursor to the mitochondria, and fosters a pro-protein acetylating milieu in said mitochondria leading to mitochondrial protein hyperacetylation. Mitochondrial protein hyperacetylation causes reduction in respiration and lipid metabolism. We hypothesized mitochondrial protein acetylation causes fatty liver disease and that NR would oppose AFLD by restoring the NAD+/NADH ratio and decreasing mitochondrial acetylation through the action of sirtuins. We found that NR could serve as an anti-AFLD agent and positively affected NAD+ but did not restore the NAD+/NADH ratio. Preliminary work seemed to indicate that NR diminished acetylation in a site-specific manner but not globally, which could mean that most acetylation is non-regulatory or non- responsive to sirtuins. The NR responsive sites may be regulatory of mitochondrial metabolism and did appear to enrich in pathways that effect lipid metabolism.

Like AFLD, diabetic peripheral neuropathy is a disease of NAD metabolomic alteration. Increased NAD+ glycohydrolase activity by activation of SARM1 leads to axonal death. Protection of NAD+ through supplementation of NR opposes axonal death in vitro. Further, the NMNAT1 overexpressing Wlds mouse resists NAD+ depletion and in several models of neuropathy, suggesting NAD+ decrement is part of the etiology of neuropathy. We hypothesized NR would oppose diabetic neuropathy by increasing available NAD+. We found that NR protected against neuropathy in both a type 1 and type 2 diabetic model. NR did not oppose hyperglycemia in the type 1 diabetic model but did so in the type 2 diabetic model, suggesting

162

that neuropathy and the positive effect of NR does not depend upon increased glycemic control.

Unlike in the AFLD model, NR dramatically reduced fatty liver disease in prediabetic and diabetic animals and protected against liver damage. Further, prediabetic and diabetic mice experienced a deficit in the hepatic NAD metabolome, which was opposed by NR. NR tended to increase the sciatic NAD metabolome in type 1 diabetic rats, agreeing with the hypothesis that in tissue NAD+ elevation protects against neuropathy.

These projects are an expansion of tools and knowledge generated by the Brenner laboratory upon my arrival. This laboratory was the first to report NR as a mammalian B3 vitamin (7) and a potential therapeutic agent (32) in yeast and the first to develop an NAD metabolomic assay (31). The works in this thesis provide improvements on the original LC-MS methodology, greatly establish NR as a superior B3 vitamin in vivo, and add to our knowledge of NR as a potential therapeutic in metabolic disease.

8.2 Regulation of the NAD+ Metabolome: a Future Avenue of Inquiry

Age related and obesity induced metabolic disorders have been described as diseases of NAD+ insufficiency (82, 144). However, NAD+ concentration rarely decreases by more than 30% (50) and indeed appears to be maintained by as yet unknown mechanisms (Table 6.1). Others have reported up-regulation of NAD+ biosynthetic enzymes, such as NRK2, and down- regulation of ADPR transferases following cellular damage (54, 220). These findings highly suggest the existence of mechanisms maintaining NAD+ concentration. Indeed, SIRT1 appears to control expression of the NAD+ biosynthetic enzyme NAMPT (57, 136), indicating NAD+ regulates its own synthesis. The induction of NRK2 and suppression of NAD+ consuming enzymes may follow a similar pathway, whereby activity of a nuclear localized sirtuin decreases in activity, resulting in increased acetylation of a key protein(s). Future work should be designed to elucidate the molecular entities monitoring NAD+ abundance and the subsequent mechanisms that are induced to maintain its normal concentration. In this future work, gene

163

expression of all known NAD+ biosynthetic and degradative enzymes could be assayed by qPCR. Additionally, I predict that there exists a nuclear acetylomic program which controls expression of both types of enzymes and could be elucidated using quantitative LC-MS based proteomics.

8.3 Future Investigations of NR as a Health Promoting Agent

How NR mediates its near miraculous effects remains to be elucidated. We hypothesized NR would oppose ethanol induced mitochondrial hyperacetylation; however, NR appeared to oppose particular sites of acetylation. These sites could be regulatory for mitochondrial function. As of now, the proteomic data presented here is relative quantitation in the form of fold changes. Acetylation at any one site could be greatly affected by ethanol and responsive to NR but represent only 1% of the lysine at that site. In order to identify the importance of each site, determination of mol-to-mol occupancy of acetylation in a site-specific manner before and after NR is necessary. Current efforts in the laboratory are aimed at developing LC-MS/MS technologies for such quantitation.

Orally delivered NR increases the human blood NAD metabolome, indicating that the effects of NR in ad libitum and controlled dose rodent studies may translate to a human population. Age and obesity related diseases normally develop gradually after chronic metabolic insult. Treatments of these diseases require constant administration and in many cases become less effective as the disease progresses. Though NR does possesses remarkable properties against diabetic peripheral neuropathy, NR increases mortality in type 1 diabetic rats and decreases glucose tolerance in control mice (Figure 7.1j). Decreased tolerance could be a result of increased insulin resistance. Indeed, NA has been shown to increase insulin resistance (240). Further, despite higher innervation than control, NR causes lower thermal sensitivity in non- diabetic mice (Figure 7.2c). This could indicate that healthy individuals could be negatively impacted by NR supplementation. Further work is necessary to uncover whether NR increases

164

insulin resistance in healthy mice and to establish the long term effects of NR over time on both healthy and diseased animal models. If NR proves efficacious over time in opposing metabolic insult, at risk groups for developing metabolic disorders could be supplemented in a preventive regimen.

With the exception of acute noise induced hearing loss (56), all studies thus far have implemented a prevention arm alone and delivered NR in food or water. Though injection of NR appears to treat noise-induced hearing loss, we present the first evidence that injection of NR could cause NAD+ glycohydrolase inhibition in liver (Figure 5.6j) and may negatively impact liver health as consequence. Investigation into the appropriate delivery method of NR is currently underway. In most cases, the metabolic disorders aided by NR in rodent studies occur over time with most symptoms appearing after the disease has developed. NR as a treatment rather than prevention warrants study.

Finally, though we and others are NAD+-centric, we cannot rule out that NR acts through non-NAD+ related metabolism. NR increases MeNam and its oxidized derivatives Me2PY and Me4PY (Figure 5.2 e and f and Figure 5.5 c). As mentioned in Chapter 1.2: NAD+ Transactions and Chapter 2.2: Addition of MeNam, Me2PY, and Me4PY to the NAD Metabolomic Assay, these metabolites carry their own biological activities and should not be merely interpreted as waste products. These biological activities could indicate that NR mediated effects are at least in part independent of NAD+. If the effects of NR are phenocopied by MeNam or Me2PY or Me4PY or some combination thereof, then the biologically relevant activity of NR may be less about contributing to NAD+ and more about producing these methylated and oxidized metabolites. However, the effect of NR could be a result of simultaneous elevation of both NAD+ and these methylated metabolites, indicating that NR is a superior and distinct actor from these other pharmacological agents. Supplementation of MeNam and its oxidized derivatives should be undertaken in a head to head comparison to NR in a metabolically challenged or neurologically stressed rodent model.

165