Nicotinamide mononucleotide (NMN) deamidation and indirect regulation of the NAD metabolome


 Treatment with nicotinamide mononucleotide (NMN) is a prominent strategy to address the age-related decline in nicotinamide adenine dinucleotide (NAD+) levels for maintaining aspects of late-life health. It is assumed that exogenous NMN is directly incorporated into the NAD+ metabolome in mammals via the canonical recycling pathway. Here, we show that NMN can undergo direct deamidation and incorporation via the de novo pathway, which is in part mediated by the gut microbiome. Surprisingly, isotope labelling studies revealed that exogenous NMN treatment potently increased the endogenous production of unlabelled NAD metabolites, suggesting that exogenous NMN impacts the NAD metabolome through indirect means, rather than through its direct incorporation. This included a striking increase in endogenous production of the metabolites nicotinic acid riboside (NaR) and nicotinamide riboside (NR) which was amplified in antibiotics treated animals, suggesting the production of endogenous NaR/NR through altered metabolic flux, enzyme kinetics and/or an as-yet unidentified pathway that interacts with the gut microbiome.

Nicotinamide adenine dinucleotide (NAD + ) is an essential redox cofactor central to metabolic processes such as glycolysis, the tricarboxylic (TCA) cycle and fatty acid oxidation 1,2 . NAD + is also consumed by enzymes such as the sirtuins 3 and poly(ADP-ribose) polymerase (PARP) enzymes 4 which are mediators of genome stability 5 and DNA repair 6 . Given the essential role of this metabolite, the decline in NAD + that occurs during biological ageing [7][8][9][10][11][12] and disease states [13][14][15] has gained attention as a target for therapeutic intervention 16 . Strategies to boost NAD + levels through supplementation with NAD precursors such as nicotinamide mononucleotide (NMN) and nicotinamide riboside (NR) are emerging as promising therapeutics 12,[16][17][18][19][20][21][22][23] . Historically, dietary supplementation with the NAD precursors nicotinic acid (Na) or nicotinamide (Nam) was used to prevent chronic NAD deficiency, which causes pellagra. When these micronutrients are replete, the step converting Nam into NMN by the enzyme nicotinamide phosphoribosyltransferase (NAMPT) is rate limiting in NAD synthesis 24 , and the use of NAD precursors that occur after this step, namely NMN and NR, have gained prominence as a strategy to raise NAD + .
One surprising aspect of this strategy is the striking appearance of the deamidated metabolite nicotinic acid adenine dinucleotide (NaAD) following oral delivery with the amidated metabolite NR 25 . NR is phosphorylated into NMN by NR kinases (NRK1/2) 26,27 , and then adenylated into NAD + by NMNAT enzymes (NMNAT1-3) [28][29][30][31][32][33] , effectively bypassing NaAD, which is an intermediate of the Preiss-Handler or de novo pathway 34,35 . In contrast, bacteria have a well-characterised NMN deamidase enzyme, PncC 36 that prevents the accumulation of NMN, which inhibits the bacterial DNA ligase [37][38][39] . One theory to explain the increase in NaAD with NR treatment 25 could be that NMN and NR assimilation follows a non-canonical route that combines steps of both microbial and mammalian processes, whereby NMN is deamidated into NaMN or NaR prior to its uptake into mammalian tissue, and then assimilated into NAD + via the intermediate step of NaAD. This could explain the appearance of NaAD following NR supplementation 25 , however an unexplained aspect is that the delivery of labelled NR results in the formation of unlabelled NaAD 25 .
Here, we use targeted metabolomics to trace the in vitro and in vivo metabolism of strategically designed NMN isotopologues to answer these questions. We show that NMN can be incorporated following its deamidation and metabolism via the de novo route, which is in part mediated by the microbiome. We further show that ablation of the microbiome by antibiotic treatment increases the uptake and conversion of orally delivered NMN into the NAD metabolome, and that isotope labelled NMN overwhelmingly presents in intestinal tissue in the form of NaR and NR. Contrary to the assumption that exogenous NMN treatment raises NAD + levels solely through its direct incorporation into the NAD metabolome, we show that treatment with isotope labelled NMN increases the levels of endogenous, unlabelled NAD metabolites. Overall, our results provide unique insights into the assimilation of orally delivered, exogenous NMN into gastrointestinal tissue, and raise questions around how exogenous precursors alter the NAD metabolome.

NMN treatment alters the de novo arm of NAD + synthesis
According to canonical models of mammalian NAD homeostasis, the metabolism of NMN, an amidated intermediate in the recycling pathway, does not intersect with the de novo pathway, which utilises deamidated intermediates. Unlike mammals, bacteria present in the gut microbiome do encode deamidase enzymes such as PncC, which deamidates NMN into nicotinic acid mononucleotide (NaMN) for metabolism via the de novo pathway 36 . To test whether the gut microbiome alters the in vivo metabolism of orally administered NMN, we used mice that were exposed to a course of antibiotics to ablate the gut microbiome (Supp. Fig. 1). These animals received a bolus of NMN (500 mg/kg) by oral gavage, and four hours later, animals were sacrificed and tissues rapidly preserved for targeted metabolomic analysis (Fig. 1). We focused our analyses on the gastrointestinal tract (GIT) and the liver, as these two tissues have high levels of NAD synthetase (NADS) activity 40 , and are the primary sites of uptake and metabolism for orally delivered compounds. In agreement with previous work 25 , NMN treatment increased the abundance of the deamidated metabolites NaR and NaMN in both the gastrointestinal tract (GIT) (Fig. 1a, b) and liver ( Fig. 1d, e), while NaAD was increased in the liver (Fig. 1f), matching previous findings for NR 25 .
Interestingly, this was completely abolished in antibiotic treated animals, where NMN treatment instead led to a spike in the amidated metabolites NR (Fig. 1g, j) and NMN (Fig. 1h, k), and abolished the increase in liver NaAD (Fig. 1f). To highlight the inverse relationship between amidated and deamidated metabolites during antibiotics treatment, the abundance of each deamidated metabolite was expressed as a ratio of its amidated counterpart (Fig. 1m-r), highlighting a profound role for the microbiome in dictating the roles of the de-amidated and amidated arms of NAD metabolism.

Strategic isotope tracing of NMN metabolism
We next sought to carefully test whether exogenous NMN was indeed undergoing direct deamidation prior to its incorporation into the NAD metabolome using isotope tracing studies. We designed two separate isotopologues of NMN that were strategically labelled at positions that would answer our hypothesis of NMN deamidation. The first of these, designated as NMN1, was 13 C labelled at all five carbon positions of the ribose moiety for an M+5 mass shift, and 15 N labelled at the pyridine ring for an overall M+6 mass shift (Fig. 2a, Supp. Fig. 2). In the last step of the de novo pathway, the enzyme NAD synthetase (NADS) amidates the carboxylic acid of NaAD using an ammonia intermediate derived from the amide group of glutamine, yielding glutamate (Fig. 2). By delivering the NMN1 (M+6) isotope in the presence of 15 N-amide labelled glutamine (M+1), the presence of M+7 labelled NAD + with an additional mass shift from the nicotinyl amide would indicate that the original amide N atom had been lost during deamidation and replaced by the 15 N amide from 15 N-Gln, indicating incorporation of NMN into NAD + via prior deamidation and the de novo pathway (Fig. 2a). To complement this experiment, we designed a second isotope, designated as NMN2, where all five carbons of the ribose moiety were 13 C labelled, and both the pyridine ring and primary amide positions were 15 N labelled, for an overall M+7 mass shift (Fig. 2b, Supp. Fig. 2). When delivered in a separate experiment, if NMN2 (M+7) underwent deamidation prior to its incorporation, the 15 N amide would be lost and replaced by an unlabelled amide from the endogenous glutamine pool, resulting in M+6 labelled NAD + . By comparing the ratios of M+7 and M+6 labelled NAD + in each experiment, we could quantify the proportion of NMN that had been incorporated into NAD + following deamidation and assimilation by the de novo pathway. This would be supported by comparing the ratios of M+1 and M+2 labelled Nam, which is released by NAD + consuming enzymes, however this interpretation would be complicated by the recently described role of the bacterial nicotinamide (Nam) deamidase PncA in systemic mammalian NAD + homeostasis 41 . By using triple-quad mass spectrometry and multiple reaction monitoring (MRM) for targeted metabolomics, we could further refine these data to determine where mass shifts occurred, including whether Nam was labelled at the pyridine ring or amide positions, and whether M+6 or M+7 labelling of NAD was from the NMN rather than the adenosine phosphate moiety.

N-Gln labelling of NAD + biosynthesis
To test whether this scheme would lead to labelling of the NAD pool as anticipated, we first used primary rat hepatocytes grown in vitro, to avoid contributions from the microbiome. Hepatocytes were treated for 24 hr with 15 N-glutamine (M+1) in the presence or absence of NMN1 (M+6), or with NMN2 (M+7) (Fig. 3). Cell lysates were subject to targeted metabolomic analysis to assess the degree of isotope incorporation into each metabolite ( Fig. 3a-e). Delivery of each of these isotopes yielded the expected M+6 and M+7 mass shifts of NMN (Fig. 3b) as well as its de-phosphorylated counterpart NR (Fig. 3a), which is consistent with the indirect transport of NMN 26 , though these data do not exclude the direct transport of NMN via the putative transporter SLC12A8 42 -for this reason, the data in this investigation could be interpreted as evidence for deamidation of NMN and/or NR, rather than NMN alone. To test the strategy of using 15 N glutamine to label NAD + synthesis, we compared the ratio of M+1 (nicotinamide labelled) to M+0 (endogenous) NAD + (Fig. 3f). As expected, 15 N-Gln treatment increased M+1 labelling of endogenous NAD + , with M+1 NAD + labelling in untreated samples due to baseline levels of naturally occurring isotopes. High levels of M+1 NAD + labelling ( Fig. 3f) were observed in samples treated with NMN1, likely due to recycling of the M+1 labelled Nam moiety (Fig. 3g) following the breakdown of NAD + (Fig. 2), or by cleavage of the NMN glycosidic bond between the ribose and nicotinamide groups. As expected, treatment with NMN1 (M+6) and NMN2 (M+7) led to M+6 and M+7 labelling of NAD + (Fig. 3c). While we had hypothesised that in vivo treatment with NMN1 (M+6) and 15 N-Gln (M+1) would lead to M+7 labelled NAD + due to the deamidation of NMN by the gut microbiome, in these primary hepatocytes we observed that 15 N-Gln co-treatment with NMN1 (M+6) increased the formation of M+7 labelled NAD + , when compared to NMN1 (M+6) alone (Fig. 3h). In line with the expected recycling of labelled Nam from NAD + (Fig. 2a), this increased formation of M+7 labelled NAD + during NMN1 (M+6) and 15 N-Gln co-treatment was matched by an identical increase in M+2 labelling of free Nam ( Fig. 3i), which was re-incorporated into the nicotinyl moiety of NAD + (Fig. 3j). 15 N-Gln treatment increased M+1 labelling at the amide position of Nam (Fig. 3i), but not the base N atom of the pyridine ring (Nambase, Fig. 3k), which does not undergo substitution by NADS, with NMN1 ( Fig. 2a) treatment serving as a positive control for labelling at this position. Overall, these data verified our system of labelling, and demonstrated the specificity of our targeted analytical approach, based on triple quadrupole mass spectrometry and MRM targeted metabolomics. As mammals do not encode a known NMN, NR or Nam deamidase enzyme, we next sought to measure the incorporation of labelled NMN into the NAD + metabolome of bacteria, which can deamidate NAD precursors including NMN 36 and Nam 43 .

NMN deamidation by bacteria
Unlike mammals, bacteria rely on an NAD + dependent DNA ligase that is inhibited by NMN, the product of its own reaction [37][38][39] , resulting in the accumulation of intracellular NMN during exponential growth 44 . This NMN is salvaged through the bacterial NMN deamidase PncC, yielding NaMN as a substrate for NAD + synthesis by the Preiss-Handler pathway 36 . To model whether extracellular NMN would undergo deamidation by bacteria, growth phase E. coli cultures were supplemented with NMN1 (M+6) (Fig. 2a) and subjected to targeted metabolomics of both cell lysates and extracellular culture media (Fig. 4). Consistent with the role of PncC in NMN metabolism in bacteria, treatment with labelled NMN resulted in the rapid incorporation of isotope labels into NaMN, with vastly increased labelling of NaMN compared to NMN (Fig. 4). Similarly, a role for the Nam deamidase PncA is strikingly reflected in the abundance of nicotinic acid (Na) compared to nicotinamide (Nam) in the cell pellet compared to the culture supernatant, where the ratio of Nam to Na in growth media was completely reversed. Overall, the avid uptake of NMN, followed by its rapid shunting into deamidated metabolites such as NaMN (Fig. 4) supported our hypothesis that the gut microbiome could contribute to the metabolism of orally administered NAD precursors such as NMN.

Antibiotic treatment alters NMN deamidation in vivo
To directly trace whether the increase in deamidated metabolites following NMN administration ( Fig.   1) was indeed due to the direct deamidation and incorporation of these metabolites, we next delivered our strategically designed isotopes into animals that had similarly been treated with antibiotics to deplete the gut microbiome, as confirmed by reductions in faecal DNA concentration (Supp. Fig. 1a), full-length 16S rRNA sequencing (Supp. Fig. 1c) and reduced alpha diversity (Supp. Fig. 1e-r).
In tissues from animals treated with NMN1 (M+6), the deamidation of NMN could be quantified by comparing the ratio of M+6 NAD + , which would assume incorporation following the canonical route, to M+7 NAD + , which had incorporated an extra mass shift from co-treatment with 15 N-Gln (M+1) (Fig. 2a, 7a). In this experiment, an increased ratio of M+7 to M+6 labelled NAD + would indicate the deamidation of NMN. The reason for using ratios, rather than the overall amounts of each isotope ( Fig. 5, Supp. Fig. 3, 4), is that they internally control for differences in bioavailability within each animal. From the intact labelling of NAD + in the GIT from NMN1 treatment, around 13% was M+7 labelled (Fig. 7b, c). Importantly, these data likely underestimates incorporation via the deamidated route, as this scheme relied on the availability of exogenous 15 N-Gln relative to the endogenous pool of unlabelled Gln, which composed only 9-13% of the total plasma Gln pool at the 4 hr timepoint ( Fig. 7k). Consistent with our hypothesis, the ratio of M+7 to M+6 labelling in the GIT was reduced in antibiotic treated animals (Fig. 7b), suggesting reduced deamidation of orally administered NMN when contributions from the microbiome were reduced. This was reflected by a reduction in the ratio of M+2 to M+1base labelled Nam in the GIT, liver and plasma ( Fig. 7d-f), however this change also likely reflected reduced contributions from the bacterial nicotinamide de-amidase PncA 43 following antibiotic treatment 41 . This change in Nam labelling carried into the M+2 to M+1 ratio of NMN (Fig.   7g, h) and NAD + (Fig. 7i, j), with reduced labelling ratios of these recycled isotopes during antibiotic treatment likely reflecting a combination of possible NMN/NR deamidation (Fig. 7a, b), and contributions from the bacterial Nam deamidase PncA 41 .
To complement this approach, in the NMN2 (M+7) experiment (Fig. 2b), we would anticipate that deamidation by the microbiome would result in loss of the 15 N amide label, resulting in the formation of M+6 NAD + at the expense of M+7 NAD + (Fig. 2b, 7l). In contrast to the previous NMN1 experiment, the ratio of M+7 to M+6 NAD + would instead decrease as the rate of deamidation increased. Further, interpretation of deamidation in this NMN2 experiment was not limited by the availability of exogenous 15 N-Gln relative to a large, endogenous pool of Gln, as was the case with the NMN1 experiment (Fig. 7a, k). In this experiment, the ratio of M+7 to M+6 NAD + was around 3:1 (Fig. 7m, n), suggesting that around 25% of orally administered NMN undergoes deamidation prior to its intact incorporation into NAD + . In agreement with the previous experiment, the ratio of M+7 to M+6 labelled NAD + was increased in the GIT and liver of antibiotic treated animals ( Fig.   7m, n). This was similarly matched by an increased ratio of M+2 to M+1base labelled Nam in the GIT (Fig. 7o), liver (Fig. 7p) and plasma (Supp. Fig. 7q) of antibiotic treated animals, reflecting decreased incorporation following deamidation, though this could instead be due the deamidation of Nam rather than NMN. As in the previous experiment, these labels were recycled into M+2 labelled NMN (Fig.   7r, s) and NAD + (Fig. 7t, u). In addition to differences in the isotope labelling of NAD + (Fig. 7), these experiments replicated the inverse relationship between NaMN and NMN levels following antibiotic treatment (Fig. 5b, e, Fig. 6b, e) observed in our earlier experiment with unlabelled NMN (Fig. 1b, h, n). Overall, these data from two complementary isotope labelling approaches support the concept that orally delivered NMN or NR can undergo deamidation prior to incorporation, and a role for the microbiome in mediating this. While these data could in part explain the spike in the de-amidated metabolites NaMN and NaAD following treatment with the amidated precursors NR 25 or NMN ( Fig.   1), it is important to note that when measured as a proportion of the overall NAD + pool, the contribution of both M+7 and M+6 intact labelled NAD + was small. Partially labelled NAD + (M+2) was around 10-fold more abundant than intact labelled NAD + (M+7) (Fig. 5d, 6d, Supp. Fig. 3-6), indicating either cleavage of the labile glycosidic bond of NMN, or rapid recycling of NAD +45 .
Following cleavage of the glycosidic bond to release free Nam, its deamidation in the GIT 46,47 by the bacterial enzyme PncA 41 also likely contributes to these changes.

Exogenous NMN boosts the endogenous NAD + metabolome
The abundance of partially labelled NAD + with labelling at the Nam position only is consistent with previous findings 45 that orally delivered NMN and NR undergo cleavage at the glycosidic bond to release free Nam, with only a small proportion of orally delivered material being incorporated into tissues intact. Given that NMN and NR are overwhelmingly incorporated in the form of free Nam 45 , why do the downstream biological effects of NR or NMN differ from the delivery of Nam 25,48-50 , a widely available nutrient present in dietary sources? One basic assumption is that exogenous NAD precursors increase levels of NAD + and other metabolites due to their direct incorporation, as expected by classic mass-balance models. Following the delivery of near 100% isotope labelled material, we found that the increase in metabolites such as NR includes the increased production of endogenous, unlabelled metabolites. For example, treatment with NMN1 or NMN2 in antibiotics treated animals increased levels of unlabelled NaR (Fig. 5f, 6f, Supp. Fig. 3f, 5f) in the GIT, and increased unlabelled NR (Fig. 5b, 5p, 6b, 6p Supp. Fig. 3b, 5b) and NaAD (Fig. 5n, 6n, Supp.

Host-microbe interactions in the bioavailability of orally delivered NAD + precursors
Another unexpected aspect of these data was the overall increase in levels of these metabolites as a result of antibiotics treatment alone, which more than doubled the labelling of the metabolites NMN, NR, NAD + and Nam ( Fig. 1g-i, 5a-d, 6a-d, Supp. Fig. 3h-k, 4h-k, 5h-k, 6h-k). This increase even occurred in unlabelled metabolites in animals that did not receive exogenous NMN (Supp. Fig. 3a-d, 5a-d). When NMN1 (M+6) was delivered, the incorporation of exogenous labels into NAD + metabolites was vastly increased in antibiotic treated animals, in the case of NR in the gut, by an order of magnitude ( Fig. 5c, 5p, Supp. Fig. 3i), a trend that was recapitulated in a separate cohort of animals receiving the NMN2 (M+7) isotope ( Fig. 6b, 6p, Supp. Fig. 5i) and in animals that received unlabelled NMN (Fig. 1g). The overwhelming abundance of NR as the dominant NAD metabolite in the GIT, especially following NMN delivery in antibiotics treated animals is worthy of later investigation, as the abundance of this single metabolite was greater than all other NAD + metabolites combined, including NAD + itself (summarised in Fig. 5p, 6p). Together, the striking increase in the uptake and overall abundance of both labelled and unlabelled NAD metabolites in antibiotics treated animals suggests that the microbiome could be in competition with mammalian tissue for the uptake of orally administered, exogenous NAD precursors, and the uptake of NAD precursors from dietary sources. Future studies should measure isotope labelling of NAD + metabolites in faecal contents of mice to confirm whether these compounds are being utilised by the microbiome, rather than being excreted via other mechanisms, and should use animals in which the microbiome has been reconstituted to control for the effects of antibiotics treatment.

Evidence for NMN uptake following dephosphorylation into NR
NMN uptake can occur following the dephosphorylation of NMN into NR by the cell surface enzyme CD73, prior to uptake by ENT nucleoside transporters and re-phosphorylation into NMN inside the cell by NRK1/2 ( Fig. 8) 26,27 . Alternatively, the solute carrier protein SLC12A8 has been described as a dedicated NMN transporter 42 . As with CD73 and ENT, SLC12A8 is located on the apical side of the intestinal tissue. As both mechanisms could co-exist, the question is the degree to which each mechanism contributes to the uptake of NMN 51 . If the direct route via SLC12A8 prevailed, we would expect to see high levels of labelled (M+7 or M+6) NMN, with lesser uptake of labelled NR. In contrast, if the indirect transport of NMN following its dephosphorylation into NR was dominant, there would be a higher levels of NR labelling. In primary hepatocytes ( to displace the endogenous NMN pool, combined with the surge of labelled NR, suggests that NMN uptake bypasses direct transport, and would instead support the dephosphorylation of NMN into NR to facilitate its intestinal absorption (Fig. 8). If direct transport of NMN does occur, it (along with NR) is in competition with the microbiome, as even when M+6 or M+7 labelling of NMN was observed at low levels, this only occurred in antibiotic treated animals (Fig. 5b, 6b). An important caveat of this interpretation is that limited availability of isotope labelled material meant that this study used a single time point, rather than a time course which also encompassed very early timepoints, possibly missing the minute-order kinetics of direct NMN transport that were previously reported 20,52 .

Chronic NMN treatment does not alter microbial diversity
Given the potential role for host-microbiome interactions in the metabolism of orally delivered NMN, it might be expected that NMN treatment would alter the composition of the microbiome. To test this, we treated aged, 97-week old animals with chronic NMN dosing via addition to drinking water (~400 mg/kg/day) for 8-10 weeks prior to cull 7,9,11,53 . 16S rRNA long-read sequencing revealed no overall change in the alpha diversity of the gut microbiome ( Fig. 9), however there were a number of changes among individual species, available as a Supplementary File.

DISCUSSION
Together, this work provides evidence for the partial incorporation of exogenous NMN into the NAD metabolome via the deamidated route, and for contributions of the gut microbiome to the metabolism of exogenous NMN. This is in line with recent findings around the role of Nam deamidation by bacteria 41 , however a role for the microbiome in the uptake of exogenous NAD precursors has not been described. It will be interesting to determine whether this relationship persists for other precursors, or is unique to NMN. Rather than being evidence for a "competition" relationship, differences in the uptake of exogenous NMN could reflect its role as an inhibitor of bacterial DNA ligase [37][38][39] , and bacterial mechanisms to prevent its accumulation. In addition to NMN deamidase enzymes, this could include a role for bacterial NAD glycohydrolase enzymes, and/or bacterial SARM-like enzymes 54 .
While intravenous delivery of NR or NMN results in a small degree of intact assimilation into peripheral tissues such as the liver, kidney and muscle, oral delivery of NMN results in hepatic cleavage at the glycosidic bond yielding free nicotinamide due to the action of the liver 45 . Our results were in close alignment with those findings 45 , where the ratio of intact M+7 or M+6 to M+0 unlabelled NAD + was around 2%, whereas the ratio of M+2 labelled to M+0 unlabelled NAD + , presumably as a result of incorporation of free Nam, was over 10% (Fig. 5d, k; 6d, k). Given this evidence for the decomposition of NMN into free Nam prior to its uptake, a key question is why downstream precursors in NAD + synthesis such as NMN and NR lead to different outcomes compared to Nam alone [48][49][50] . A surprising aspect of these results was that treatment labelled NMN led to an increase in unlabelled NAD metabolites. In the case of both NR and NaR, treatment with NMN1 (M+6) or NMN2 (M+7) led to a stark increase in endogenous (M+0) levels, particularly in antibiotics treated animals (Fig. 5c, f, 6c, f, Supp. Fig. 3b, f, 5b, f). We also observed that exogenous NMN increased liver NaAD levels ( Fig. 1f, 5g, n, 6g, n), as was previously reported for exogenous NR treatment 25 , however the majority of this increase was from unlabelled NaAD (Supp. Fig. 4g, 6g).
These results were similar to those of Trammel et al 25 , where isotope tracing of double-labelled NR showed that while total NaAD levels increased by over 40-fold following NR treatment, only around 45% of this NaAD was isotope labelled -with the endogenous origin of the remaining 55% remaining unexplained. We argue that these findings run against the assumed model that exogenous NAD + precursors raise NAD + levels through their direct incorporation into the NAD metabolome, and instead could suggest that treatment with exogenous precursors could indirectly trigger endogenous NAD + biosynthesis. The mechanism for this is not yet clear, though given the profound effect of antibiotic treatment, in particular for the overwhelming abundance of NR in the gut (Fig. 5c, p, 6c, p, Supp. Fig. 3b, 5b), it is likely to involve interplay with the gut microbiome. One possibility for the changes in endogenous NAD + metabolites following exogenous NMN/NR treatment could be that exogenous NMN or NR treatment triggers unknown signalling pathways that indirectly alter endogenous NAD metabolism, rather than the direct incorporation of exogenous material.
Another explanation is that increased substrate levels alter the in vivo kinetics of NAD biosynthetic enzymes, increasing the utilisation of endogenous substrates. An important question regarding NR in particular is how its endogenous production is increased by exogenous NMN. NR is available from dietary sources 27 , and is an intermediate in the uptake of extracellular NMN 55,56 . NMN accumulation in neurons can trigger cell death through the NADase SARM1 57 , and disposal of NMN through its adenylation into NAD + can protect against neuronal death 58 . It is possible that exogenous NMN triggers pathways that degrade endogenous NMN into NR, which could act as a reservoir for NAD precursors, however CD73, the NMN ectonucleotidase that carries this out, sits on the extracellular face of the plasma membrane 55,56 rather than the cytosol. In addition, the sheer molar quantity of unlabelled NR that we observed in the gut (Fig. 1g, 5c, 6c) relative to other metabolites (Fig. 5p, 6p) challenges this idea. NAD homeostasis is tightly maintained within a defined range 59 , and the activity of NMNAT enzymes that carry out the last step of NAD biosynthesis is reversible 60 . It is possible that exogenous NAD precursors push the equilibrium of this step in the opposite direction, increasing endogenous NMN production from NAD + , though how intracellular NMN could be dephosphorylated to an NaR/NR reservoir in mammals is unknown. This concept of increased NAD breakdown during treatment with exogenous NMN in young animals is also supported by the increased formation of unlabelled Nam in plasma (Fig. 5o, 6o). Further, while labelled NMN treatment in vitro (Fig. 3) results in the formation of intact labelled NAD + , this occurs at the cost of unlabelled NAD + , for a net zero change in total NAD levels ( Fig. 3c), suggesting that when NAD + is replete, the utilisation of exogenous NMN into newly synthesised NAD + results in a commensurate breakdown of existing material -potentially explaining the formation of unlabelled metabolites such as NR during treatment with exogenous, labelled NMN.
Rather than degrading or recycling existing metabolites into NR, another possibility is that exogenous NMN or NR could trigger a currently unknown step in mammals that leads to endogenous NR production. While NR can be produced by the reversible phosphorolysis of Nam and ribose-5phosphate by a purine nucleophosphorylase (PNP) in E. coli 61 , this step is irreversible in mammals 56,62,63 , and other potential steps involved in endogenous NR production in mammals are unknown. Further work is needed to understand how this occurs, for example, whether it is due to the acute up-regulation of NAD + biosynthetic enzymes, the time-scale by which this increase occurs, and a direct comparison of different isotope labelled NAD + precursors to identify which metabolites trigger the production of endogenous metabolites under normal circumstances and during depletion of the microbiome. Regardless, the ability to trigger the production of endogenous NAD metabolites could explain why exogenous NR and NMN treatment lead to differences in pharmacokinetics, metabolite production and therapeutic effects when compared to Nam alone [48][49][50] , despite their rapid metabolism into free Nam by the liver 45 (Fig. 4, 5).
Another speculative idea is the existence of a signalling pathway in the GIT that is sensitive to both exogenous NAD precursors and to microbial metabolites, which can mediate endogenous NAD metabolism. Metabolite sensing members of the G-protein coupled receptor (GPCR) family are putative candidates for this role as there are already GPCRs known to respond to extracellular nicotinic acid and to NAD + itself 64 . One possible candidate is GPR109a, which acts as a receptor both for nicotinic acid 65 and for butyrate, released from the microbial fermentation of dietary fibre 66 . This could link the observations including the deamidation of orally derived NAD precursors into their acid equivalents, a role for microbiome depletion in triggering the production of endogenous NAD metabolites, and evidence for the poor incorporation of intact NR or NMN into the NAD metabolome.
Identifying cell surface receptors and downstream signalling pathways in the gut that are sensitive to both microbial metabolites and exogenous NAD + precursors will be a key goal of testing our hypothesis around the role of exogenous metabolites in endogenous NAD + metabolism. Together, these findings regarding the deamidation of NMN and its effects on the endogenous NAD metabolome have profound importance for the therapeutic development of NAD precursors.

Methods
Methods are available in Supplementary Material, with raw data available on our Mendeley data site.

Acknowledgements
Funding was from the National Health and Medical Research (NHMRC) of Australia as a Career Development Fellowship APP1122484 to LEW, and sponsored research from Jumpstart Fertility. We wish to thank anonymous donors for philanthropic support.                   Figure 4 of main text, animals were treated with antibiotics (Abx) and M+6 isotope labelled NMN1 followed by metabolomics analysis of intestinal tissue as described in Figure 1 of main text. Animals As in Figure 4 of main text, animals were treated with antibiotics (Abx) and M+6 isotope labelled NMN1 followed by metabolomics analysis of intestinal tissue as described in Figure 1 of main text. Animals were treated with antibiotics (Abx) and M+7 isotope labelled NMN2 followed by metabolomics analysis of GIT tissue as described in Figure 6 of main text. Each data point represents Animals were treated with antibiotics (Abx) and M+7 isotope labelled NMN2 followed by metabolomics analysis of liver tissue as described in Figure 6 of main text. Each data point represents *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, ns=not significant.

Synthesis of isotope labelled NMN
The isotopes used here were generated through a two-step process starting with the custom synthesis of nicotinamide labelled with 15  with addition of sucrose (3 g/L; Bundaberg Sugar) to increase palatability for 4 days, and switched to ampicillin (1 g/L) with sucrose (3 g/L) for an additional week, which can reduce gut bacterial density by 1000-fold 67 . During treatment with the VNAM combination there was a reduction in water consumption (below), which was the reason for the subsequent switch to ampicillin alone. Sucrose treatment (3 g/L) was maintained as a vehicle control in animals that did not receive antibiotic treatment.

Above: Water intake and body weight of mice during antibiotics treatment.
To maintain consistency between the three cohorts presented here (unlabelled NMN - Fig. 1 isotopes at 50 mg/kg, or for unlabelled NMN (Fig. 1) at 500 mg/kg, with water vehicle used as a control. Four hours later, animals were placed under anaesthesia, and blood was obtained by cardiac puncture, followed by euthanasia by cervical dislocation, rapid dissection and snap freezing of tissues.
Gavages were staggered between mice in alternating treatment groups to avoid any experimental bias.
On the day of cull mice were all 5-6 weeks old. Differences in NMN dosing between unlabelled and isotope labelled NMN were due to limited availability of isotope labelled NMN.

Blood plasma collection and preparation for mass spectrometry
Approximately 1 (Fig. 3).

Preparation of NAD + metabolite standards
NAD + metabolites were serially diluted starting from a concentration of 100 µM to 0.39 µM. The same volume (500 µL) of extraction buffer (acetonitrile:methanol:water) was added and vortexed before centrifuging and transferring to new tube ready to be dried down as above. The subsequent steps were the same as preparing the tissue samples as above. All standards and samples were processed on the same day to reduce any experimental bias or variability. Standard curves used to calculate absolute concentrations are shown on the following page (Methods Fig. 2) and are available in supplementary raw data files.  All data are presented as mean ± standard deviation (s.d.). Statistical significance was performed using a two-way ANOVA with a Sidak's multiple comparisons test to determine differences between groups after removing outliers using the ROUT method (Q=1%). Data in Fig. 7 were analysed by Mann-Whitney U-test between NMN isotope treated groups due to the absence of detection of labelled metabolites in animals that did not receive NMN isotopes. All statistics were performed on GraphPad Prism software (version 8.2.1). P values less than 0.05 were considered statistically significant. All data analyses are available as an .xml file available on our Mendeley data site. For in vivo experiments (Fig. 1, 5-8, Supp. Fig. 4-7), each data point represents tissues from a separate animal, while each data point for in vitro experiments (Fig. 2, 3) represents an independent biological replicate.

DNA extraction from faecal pellets
Solid faecal pellets taken from the colonic and rectal region of the gastrointestinal tract were stored in -80℃ until further processing. DNA was extracted from frozen faecal pellets using the QIAamp® PowerFecal® DNA kit (Qiagen, Cat. No. 12830-50) according to the manufacturer's protocol. DNA concentration was determined using the NanoDrop™ (DeNovix®, DS-11 FX) and the purity of double-stranded DNA (dsDNA) was also determined by measuring the 260/280 ratio. All DNA extracts were stored at -80℃ until further processing by 16S rRNA sequencing.