NMN: The NAD precursor at the intersection between axon degeneration and anti-ageing therapies

The past 20 years of research on axon degeneration has revealed fine details on how NAD biology controls axonal survival. Extensive data demonstrate that the NAD precursor NMN binds to and activates the pro-degenerative enzyme SARM1, so a failure to convert sufficient NMN into NAD leads to toxic NMN accumulation and axon degeneration. This involvement of NMN brings the axon degeneration field to an unexpected overlap with research into ageing and extending healthy lifespan. A decline in NAD levels throughout life, at least in some tissues, is believed to contribute to age-related functional decay and boosting NAD production with supplementation of NMN or other NAD precursors has gained attention as a potential anti-ageing therapy. Recent years have witnessed an influx of NMN-based products and related molecules on the market, sold as food supplements, with many people taking these supplements daily. While several clinical trials are ongoing to check the safety profiles and efficacy of NAD precursors, sufficient data to back their therapeutic use are still lacking. Here, we discuss NMN supplementation, SARM1 and anti-ageing strategies, with an important question in mind: considering that NMN accumulation can lead to axon degeneration, how is this compatible with its beneficial effect in ageing and are there circumstances in which NMN supplementation could become harmful?


Introduction
In the 19th century, neurophysiologist Augustus Volney Waller described, for the first time, that nerves degenerate following a physical injury (Waller and Owen, 1850).Waller's first nerve transection experiments marked the beginning of a research field that has only recently identified a conserved, genetically encoded, and druggable pathway of programmed axon death activated by injury (Wallerian degeneration) and in disease (Coleman and Höke, 2020) (Fig. 1A).In humans, disease-associated mutations in key regulators of programmed axon death have recently been identified in ALS (amyotrophic lateral sclerosis) patients (Gilley et al., 2021;Bloom et al., 2022) and rare cases of polyneuropathies (Huppke et al., 2019;Lukacs et al., 2019;Dingwall et al., 2022), fuelling significant investments from the pharma industry in this research area (Krauss et al., 2020).
The discovery of a neurotoxic function of NMN was somewhat surprising, given that strategies to enhance NAD production with supplementation of NMN and other NAD precursors have recently been in the spotlight as anti-ageing therapeutic approaches (Yoshino et al., 2018;McReynolds et al., 2020).Studies in rodents show that boosting NAD production, especially with NMN or NR (nicotinamide riboside, a form of vitamin B 3 that is converted into NMN intracellularly, Fig. 1B ( Bieganowski and Brenner, 2004)), is beneficial in several disease models and counteracts age-related functional decline (Mills et al., 2016;Yoshino et al., 2018).
The prospect of promoting longevity and healthy ageing by simply taking vitamins and derivatives thereof explains why this research area rapidly gained the attention of media and consumers.The nutraceutical industry markets NMN and other NAD precursors as food supplements, avoiding the strict regulations for the approval and the launch of a therapeutic drug on the market.Today NMN and other NAD precursors can be easily purchased by the public, with many people taking these supplements daily.While clinical trials are ongoing, the significant inflow of NMN-based and NMN-related products on the market makes it particularly important to rapidly determine the safety profile of NAD precursors (Nadeeshani et al., 2022), especially in light of the newly recognised role of NMN in axon degeneration.
This important research topic was recently reviewed by (Nadeeshani et al., 2022).In this perspective, we extend the discussion focusing on the relationship between NMN and SARM1 and, based on what we learned from axon degeneration studies, we examine potential scenarios in which these supplements could lead to SARM1 activation and possible neurotoxicity in humans.We also highlight important aspects to consider when designing future experimental studies and clinical trials with NAD precursors.

The role of NMN in axon degeneration
In 2015, we showed that a rise in the NAD precursor NMN promotes axon degeneration after injury (Di Stefano et al., 2015).The "NMN hypothesis" predicted that the loss of short-lived NMNAT2 in transected axons leads to a failure of NMN conversion into NAD and toxic accumulation of NMN.This model was supported by strong genetic and pharmacological evidence showing that limiting NMN accumulation is axon protective.For example, ectopic expression of NMN deamidase, a bacterial enzyme that converts NMN to its deamidated form NAMN (nicotinic acid mononucleotide), and pharmacological inhibition of the NMN-synthesising enzyme NAMPT (nicotinamide phosphoribosyltransferase) (Fig. 1B) with FK866, both of which limit NMN accumulation, confer robust protection against Wallerian degeneration (Di Stefano et al., 2015, 2017).In injured axons, where NMNAT2 levels are low, exogenously administered NMN and NR cannot be converted into NAD, leading to NMN accumulation, reversal of the protection conferred by FK866, and neurotoxicity (Di Stefano et al., 2015).
We followed up with in vivo studies providing further support for the "NMN hypothesis" in mice, zebrafish and Drosophila, as well as in several non-injury models of axon degeneration (Loreto et al., 2015(Loreto et al., , 2020(Loreto et al., , 2021;;Di Stefano et al., 2017;Llobet Rosell et al., 2022).We also showed that NMN only induces axon degeneration in the presence of SARM1, placing the toxic accumulation of NMN upstream of SARM1 on a common pathway of axon death (Loreto et al., 2015).Back then, however, SARM1 was not known to possess any enzymatic activity, and identifying a clear mechanism or a direct action of NMN on SARM1 was not straightforward.

SARM1 enzymatic activity is regulated by NMN and NAD levels
The "NMN hypothesis" remained controversial for several years.In 2016, a study showed that high NMN levels are compatible with axon protection in mouse primary neurons but, notably, the combination of genetic and pharmacological manipulations used to elevate NMN also lead to a concomitant increase of baseline NAD levels (Sasaki et al., 2016).A neurotoxic role for NMN was also considered unlikely given the plethora of studies reporting that supplementation with NMN and other NAD precursors has therapeutic effects in animal models of disease and is beneficial for several physiological functions that decline with age (Mills et al., 2016;Yoshino et al., 2018;McReynolds et al., 2020).
In 2017, a key study from the Milbrandt and DiAntonio laboratories showed that SARM1 is an NADase which cleaves NAD into ADPR (ADP Ribose), cADPR (cyclic ADPR) and NAM (nicotinamide) (Essuman et al., 2017).Following studies extended the original findings showing that SARM1 is a multi-functional enzyme which also cleaves NADP (NAD phosphate) and possesses base exchange activity (Fig. 1A) (Essuman et al., 2018;Zhao et al., 2019;Horsefield et al., 2019;Angeletti et al., 2022).The discovery of SARM1 enzymatic activity was a major turning point which made it possible to test whether the toxic role of NMN (Di Stefano et al., 2015, 2017) is to directly activate SARM1, which was indeed demonstrated in 2019 when Zhao and colleagues reported that NMN increases SARM1 NAD(P)ase activity both in an enzymatic assay on purified SARM1, and in cells (using a cell-permeable analogue of NMN) (Zhao et al., 2019).These findings have now been replicated by several groups and are supported by solid structural data showing that NMN binds an allosteric pocket in SARM1 ARM domain at concentrations within the physiological range, inducing a conformational change that increases SARM1 activity (Zhao et al., 2019;Bratkowski et al., 2020;Figley et al., 2021;Gilley et al., 2021;Loreto et al., 2021;Angeletti et al., 2022;Shi et al., 2022).Additional studies also suggest a role for NAD in the regulation of SARM1 catalytic activity, indicating that the inactive conformation of SARM1 is stabilised by the binding of NAD, which competes for the same allosteric site bound by NMN in the ARM domain (Jiang et al., 2020;Sporny et al., 2020;Figley et al., 2021;Angeletti et al., 2022).At high concentrations, NAD exhibits an inhibitory effect (Jiang et al., 2020;Figley et al., 2021;Angeletti et al., 2022), which might partly explain seemingly contradictory results in the field (Sasaki et al., 2016).
Together, these studies provide compelling evidence that NMN is an endogenous activator of SARM1 and that the intracellular NMN/NAD ratio is important for the regulation of SARM1 activities, including NADPase and base exchange (Angeletti et al., 2022).Importantly, the pro-degenerative action of NMN is now widely accepted in the field.

Supplementation of NMN and other NAD precursors as an anti-ageing therapy: is it safe?
NAD is a key molecule for cellular health, well-known for its role in major metabolic pathways such as glycolysis, the Krebs cycle, and oxidative phosphorylation.NAD is also a substrate and co-substrate for many enzymes, including PARPs (poly ADP-ribose polymerases), sirtuins, CD38, and now also SARM1.Preserving NAD homeostasis is, therefore, essential for life (Verdin, 2015).
The concept that NAD biology is central to controlling ageing and longevity in mammals is not new.Studies in rodents suggest that tissue levels of NAD decline with age, likely due to a combination of changes in the expression/activity of enzymes involved in NAD synthesis and degradation (Camacho-Pereira et al., 2016;Schultz and Sinclair, 2016;McReynolds et al., 2020).Data in humans are less convincing; some reports (but not all) observed a declining trend of NAD with age, at least in some tissues (Massudi et al., 2012;Guest et al., 2014;Zhu et al., 2015;Elhassan et al., 2019;McReynolds et al., 2020), but whether this is a consequence of ageing alone is a matter of debate and more studies are required to convincingly show that NAD levels are lower in the aged population (Peluso et al., 2021).However, there is a general consensus that perturbed NAD metabolism, even without changes in total NAD levels, contributes to age-associated functional decay, and that maintaining NAD homeostasis could therefore be a viable anti-ageing strategy.Supplementation with NMN or NR is the most widely studied approach to achieve this; both precursors efficiently increase tissue NAD levels in animal studies, with therapeutic effects in several rodent models of disease and promising data on delaying age-associated functional decline (Trammell et al., 2016;Mills et al., 2016) (for an extensive review see (Yoshino et al., 2018)).There are also promising preclinical and clinical data suggesting that administration of NAM (another form of vitamin B 3 that is converted into NMN intracellularly, Fig. 1B) could be an effective treatment for glaucoma and other eye disorders (Williams et al., 2017a;b;Hui et al., 2020;Tribble et al., 2021;De Moraes et al., 2022).
The extensive data from preclinical studies rapidly pushed the field toward human studies, but the pharmacology, safety profile and efficacy of NMN and other NAD precursors are still being assessed in clinical trials (Reiten et al., 2021).The fact that these molecules are easily available to the public and are already used by many people leads to an important question: in the absence of comprehensive and long-term data from clinical trials, are there circumstances in which NMN itself, or other NAD precursors that also raise NMN levels, could be harmful to people given that NMN is neurotoxic in certain conditions?

Efficient conversion of NMN into NAD is key to preventing SARM1 activation and neurotoxicity
A key point in the context of SARM1 activation and neurotoxicity is whether NMN is efficiently converted intracellularly.In healthy cells, NMNATs, which are not rate-limiting enzymes for NAD synthesis (Revollo et al., 2004), should rapidly convert any NMN produced into NAD (Fig. 1B) and, unless saturating doses of NMN (or other precursors) are administered, this prevents the toxic accumulation of NMN.To our knowledge, there is no evidence that supplementation of NMN or other NAD precursors cause neurodegeneration through SARM1 when NMNATs are active and normally expressed.There are, instead, several reports of neuroprotective effects of NAD precursors in disease models, also against SARM1-dependent neurotoxicity (Sasaki et al., 2016;Yoshino et al., 2018;Schöndorf et al., 2018;Gilley et al., 2021).But what about circumstances in which NMNATs are compromised?
Recent studies have identified LoF (loss-of-function) and partial LoF mutations in NMNAT2 in patients with polyneuropathies (Huppke et al., 2019;Lukacs et al., 2019;Dingwall et al., 2022).LoF mutations in NMNAT1 are also known to cause Leber Congenital Amaurosis (Falk et al., 2012), with animal studies suggesting SARM1 involvement in the pathogenesis of the disease (Sasaki et al., 2020b, p. 1).Furthermore, there is some evidence that NMNAT2 mRNA levels vary widely between people (Ali et al., 2016), and preclinical data indicate that low NMNAT2 levels increase axon vulnerability (Gilley et al., 2019) and that levels of the NMNATs decline with ageing (Camacho-Pereira et al., 2016).Animal models of disease also suggest that many pathological processes including mitochondrial dysfunction, impaired axonal transport and decreased protein synthesis (Gilley and Coleman, 2010;Conforti et al., 2014;Summers et al., 2014Summers et al., , 2020;;Geisler et al., 2019b;a;Loreto et al., 2020;Merlini et al., 2022) reduce NMNAT2 levels.Finally, a study in mice indicates that axonal transport of NMNAT2 declines with age (Milde et al., 2015), which could reduce NMNAT2 levels in the axon where it is most needed to prevent SARM1 activation.
Taken together, this suggests that individuals in the human population could have an inherently higher or lower capacity for converting NMN into NAD, which could also change throughout life and in disease.People with NMNAT2 LoF mutations or who express low levels of NMNAT2 may have increased susceptibility to axonal damage and react differently to supplementation with NMN and other NAD precursors (Fig. 2).It will be important to determine whether there is a threshold at which NMNAT2 activity/levels become so low that the suggested risk begins.Establishing the frequency of coding and non-coding mutations affecting NMNAT2 function in the human population and further investigating the variation in expression levels of NMNAT2 (Ali et al., 2016) in health and disease should be priorities in the field.Similar questions should be asked for NMNAT1, as it is becoming clear this NMNAT isoform is involved in SARM1 regulation, with already existing links to human disease (Falk et al., 2012;Zhao et al., 2019;Sasaki et al., 2020b).
While encouraging results from clinical trials suggest that supplementation of NAD precursors is generally safe, the current data originate from relatively short-term studies with low participant numbers (Reiten et al., 2021).These studies are not designed to identify interpersonal variability, which may need large numbers, or the effects of long-term administration (years) of these supplements, especially in the elderly and diseased, where levels of NMNATs are more likely to change.Furthermore, while acute neurotoxicity is likely uncommon given it has not been reported yet in any of the current trials, adverse effects such as the vulnerability of a specific subset of neurons or slow, progressive axon loss, which could occur for many years before symptoms appear, are unlikely to be detected in relatively short-term trials.

NMN vs NMN precursors
It is important to clarify which, among the NAD precursors currently used to boost NAD levels in humans, could lead to SARM1 activation.Theoretically, NMN and any precursor that could increase intracellular NMN levels are potentially neurotoxic.In mammals, NMN is synthesised from NAM and NR (Fig. 1B) (Bogan and Brenner, 2008), two forms of vitamin B 3 that are currently used in preclinical and clinical studies (Hui et al., 2020;Reiten et al., 2021;De Moraes et al., 2022).In our axon degeneration assays, both NMN and NR revert FK866 protection of injured axons where NMNAT2 is depleted, causing axon degeneration (Di Stefano et al., 2015), indicating that both NMN and its precursors, if not efficiently converted, lead to a build-up of NMN, with consequent SARM1 activation and neurotoxicity.A recent study also suggests that simply administering NR in healthy primary mouse neurons leads to SARM1 activation (Figley et al., 2021), although no neurotoxicity is observed as these neurons express normal levels of NMNATs, which likely compensate for the NAD depletion caused by SARM1 activation, restoring the NMN/NAD ratio.NR administration can even counteract neurotoxicity caused by constitutively active SARM1 (Gilley et al., 2021), confirming a "double-edged sword" effect of these molecules, which can be neurotoxic when not efficiently converted into NAD, and neuroprotective when rapidly metabolised (Fig. 2).
Things, however, are much more complicated in the context of a systemic administration of NAD precursors in humans.Regulation of NAD metabolism is complex and NMN, NR and NAM supplementation, despite theoretically feeding into the same branch of NAD biosynthesis, elicit distinct effects on downstream metabolites in humans (Trammell et al., 2016;Reiten et al., 2021;Brakedal et al., 2022;Okabe et al., 2022).Their mechanisms of absorption in the gut and how these NAD precursors enter the cells (which is still a matter of debate) will play a critical role (Ratajczak et al., 2016;Grozio et al., 2019;Schmidt and Fig. 2. Hypothesised effects on axonal health following supplementation of NAD precursors.In healthy individuals, NMNAT2 rapidly converts NMN into NAD, preventing NMN toxic accumulation and maintaining healthy neurons and axons.However, when NMNAT2 levels or activity are low, NMN levels increase, leading to SARM1 activation, increased axonal vulnerability and/or axon degeneration.Created with Biorender.com. A. Loreto et al. Brenner, 2019;Yaku et al., 2021).Simplistically, in the context of SARM1 activation and neurotoxicity, it will depend on whether, and how rapidly, these precursors lead to NMN accumulation in neurons, and how efficiently NMN is converted into NAD by NMNATs (Fig. 2).

Impact of NMN/NAD ratio on SARM1 activity
We previously suggested that SARM1 activity is regulated by a ratio between NMN and NAD (Di Stefano et al., 2017), and recent data suggest that a concomitant increase in NAD levels keeps SARM1 in an inactive state, despite high NMN (Sasaki et al., 2016;Jiang et al., 2020;Figley et al., 2021).There is an argument that NMN supplementation might not affect SARM1 activity status, as it would lead to both NMN and NAD increase.While this could be true in healthy cells, there are two considerations to be made: First, following up on what was discussed above, inefficient conversion of NMN into NAD because of compromised NMNAT activity is the most likely scenario in which NMN can become toxic.Under these conditions, NMN would accumulate whereas NAD would not rise significantly, shifting the ratio strongly towards NMN.Second, changes in NMN levels close to the physiological concentration have a much more significant impact on SARM1 activity than NAD levels (Figley et al., 2021;Angeletti et al., 2022).When NMN rises, partial inhibition of SARM1 is only seen at high concentrations of NAD.Preclinical studies show that a twofold increase in NAD levels is not sufficient to delay axon degeneration after injury (Sasaki et al., 2009), and even higher NAD levels only temporarily delay axon degeneration (Sasaki et al., 2016).
Thus, while data from clinical trials indicate that supplementation with NAD precursors does increase NAD levels, the extent of this increase is unlikely to be sufficient to keep SARM1 in an inactive state in the presence of high NMN (Reiten et al., 2021).

SARM1 activity and intracellular signalling pathways
Recent developments in the field are unveiling a potential role for SARM1 that goes beyond axon degeneration, with indications that SARM1 can be activated at sub-lethal levels.For instance, administration of the NMN analogue CZ-48 or NR with or without overexpression of NRK1 (nicotinamide riboside kinase 1) leads to SARM1 activation in mouse DRG neurons without causing axon degeneration (Zhao et al., 2019;Figley et al., 2021).SARM1 activation with CZ-48 has also been reported in other cell types (Zhao et al., 2019;Li et al., 2021).While the effects of this sub-lethal activation of SARM1 on cellular physiology are unclear, products of SARM1 enzymatic activity, such as cADPR and NAADP (nicotinic acid adenine dinucleotide phosphate) (Fig. 1A), are known calcium-mobilising agents and have important roles in signalling pathways (Angeletti et al., 2022), and a role for SARM1 in innate immunity has also been suggested (Carty et al., 2006;Belinda et al., 2008;Carlsson et al., 2016;Hopkins et al., 2021).It is reasonable to expect that sub-lethal/chronic SARM1 activation could increase axonal vulnerability or have a significant impact on NAD homeostasis and important intracellular signalling pathways in neurons as well as other cell types (Fig. 2).Data from clinical trials suggest that supplementation with NAD precursors in humans leads to changes in NAD metabolism which are compatible with an increase in both the rate of NAD synthesis and consumption (Trammell et al., 2016;Brakedal et al., 2022;Okabe et al., 2022).It is possible that sub-lethal SARM1 activation occurs following supplementation with NAD precursors and contributes to some of the observed changes in NAD metabolites.
We have only scratched the surface of the impact that SARM1 activity has on cellular physiology.It is hard to predict whether sub-lethal activation of SARM1 is beneficial or not for cellular behaviour and for maintaining NAD homeostasis, but it is important to take this into account in future studies involving the supplementation of NAD precursors.

Can we detect SARM1 activation?
Developing methods to detect SARM1 activation in humans would make it possible to reveal neurotoxic or sulethal activation of SARM1 that might cause axon degeneration or make axons more vulnerable in response to NMN or its precursors, and thus be a way of monitoring whether these are safe in specific individuals.The rapid development in the understanding of SARM1 biology means we now have potentially useful, measurable biomarkers of SARM1 activation, at least in animal studies where tissues are readily accessible.In neurons, an increase in the levels of SARM1 product cADPR is a reliable biomarker of SARM1 activation (Sasaki et al., 2020a).This, together with measuring NMN and NAD levels as an indication of NMNAT2 loss, can thus be used as markers of SARM1 activation status (Fig. 1A) (Merlini et al., 2022).However, these might not be practical in clinical studies of supplementation of NAD precursors, at least not in nervous system tissue.The development of minimally invasive methods to identify SARM1 activation would be a game changer, especially in the context of neurodegenerative diseases.Accessing neuronal tissue in human studies is a major complication, but a specific set of SARM1-dependent metabolic changes in the CSF (cerebrospinal fluid), serum or other readily accessible tissues would represent an important step forward.
General markers of neurodegeneration, such as an increase in NfL (neurofilament light chain) levels in CSF and serum (Khalil et al., 2018), could also be useful to monitor any unexpected neurodegenerative events and should become routine in clinical studies involving the supplementation of NAD precursors.Although NfL is not a specific marker of SARM1 activation, its levels change as a result of SARM1-dependent neurodegeneration (Sasaki et al., 2020a;Hughes et al., 2021;Bosanac et al., 2021).Encouragingly, a recent clinical trial on NR supplementation in PD (Parkinson's disease) patients included measurements of NfL levels in both serum and CSF, with no changes detected (Brakedal et al., 2022).

Conclusions
Extensive data from preclinical studies unquestionably show the potential of supplementing NAD precursors as an anti-ageing therapeutic strategy.Slowing down or reversing certain aspects of ageing are among the biggest health challenges of our times and it would be a tremendous breakthrough if simple strategies such as supplementation of vitamins and derivatives thereof have a major impact on ageing in humans.However, it is also clear that there is the potential for harmful effects in some people and this needs more investigation before general application of such interventions.While the number of clinical trials is rapidly growing, there is an urgent need to develop viable approaches to monitor adverse effects after long-term administration of NAD precursors, possibly starting with individuals that have already been taking these supplements for years.
Realistically, NAD precursors are likely to be safe for most people, but there is a risk that these supplements could cause SARM1 activation and neurodegeneration in people with compromised NMNAT activity, either caused by mutations, impaired axonal transport, or a decline in expression levels throughout life (Fig. 2).We still do not know how many people fall in this category and where the threshold is after which supplementation of NAD precursors could switch from being harmless to becoming harmful.The daily dosage is also likely to be important, and while current trials suggest that even high doses of NAD precursors are safe in most people in the short term (Reiten et al., 2021), the fact that these molecules are taken as food supplements gives little control over how much each person takes.The addition of markers of SARM1 activation and general markers of neurodegeneration in future preclinical and clinical studies is key, and further understanding of the role of SARM1 in axon degeneration and normal cellular physiology is needed to guide future trials.
While human studies are confirming that supplementation of NAD precursors increases NAD levels and/or stimulates NAD metabolism, striking beneficial effects on other parameters of human physiology and age-related deficits, backed by sufficient data from well-controlled clinical trials, are yet to be reported.While boosting NAD production could help in some people, it may be unnecessary in many and potentially harmful in a few, and more work is required to work out who is in which category.

Declaration of interests
M.P.C. holds funding jointly provided by AstraZeneca for academic research and consults for Nura Bio, neither of which is directly relevant to this review.