NR and NMN: the NAD⁺ pathway, cellular transport, and clinical limits

The question “which raises NAD⁺ more effectively—nicotinamide riboside (NR) or nicotinamide mononucleotide (NMN)?” has several different experimental answers. An increase in whole-blood NAD⁺, incorporation of a labeled precursor into the hepatic pool, and a change in muscle function measure different processes. None by itself establishes that the compound reached a particular cell intact, restored its free nuclear or mitochondrial NAD⁺, and improved an outcome meaningful to people. Comparing the two molecules requires following chemical form, the intestinal lumen, plasma, extracellular enzymes, the membrane, and intracellular use in sequence; at these transitions, the source text’s firm conclusions most often run ahead of the evidence.

Two roles of one molecule: the redox cycle and NAD⁺ consumption

The oxidized form of nicotinamide adenine dinucleotide accepts a hydride in glycolysis, pyruvate oxidation, the tricarboxylic acid cycle, and β-oxidation, while NADH donates reducing equivalents to the respiratory chain. In the redox cycle, the coenzyme changes its oxidation state but its pyridine backbone is not cleaved: $\mathrm{NAD}^{+}+2e^{-}+2H^{+}\rightleftharpoons\mathrm{NADH}+H^{+}$. This cycle supports catabolism and ATP synthesis. Separately, NAD⁺ is consumed as a substrate in protein deacylation by sirtuins, ADP-ribosylation by PARPs, and hydrolysis by CD38/CD157. This releases nicotinamide (NAM), which may return to a synthesis pathway or undergo catabolism. NAD⁺ concentration is therefore determined by synthesis, consumption, reoxidation of NADH, exchange between compartments, and the cellular composition of the sample—not precursor intake alone.

Measured “total NAD⁺” differs from the free ratio $[\mathrm{NAD}^{+}]/[\mathrm{NADH}]$, which sets the direction of reversible dehydrogenase reactions. Cytosolic and mitochondrial-matrix redox potentials differ: the cytosolic ratio remains relatively oxidized, whereas matrix dehydrogenases and the respiratory chain establish a different state. NAD⁺ and NADH should not be pictured as a single, well-mixed pool. Nuclear, cytosolic, and matrix pools are connected by metabolite flux and transport, but enzymes encounter locally available substrate. Estimates of free concentrations from indicator equilibria or fluorescent sensors depend on method and tissue; one fixed ratio cannot be applied to every cell. The stability of the mitochondrial pool cannot be inferred solely from the presence of NMNAT3: NAD⁺ transport across the inner membrane has been described in mammals, and the contribution of local synthesis and transport depends on the model.

Sirtuins share a requirement for NAD⁺ but differ in reaction and localization. SIRT1, SIRT6, and SIRT7 are predominantly nuclear; SIRT2 is mainly cytosolic; SIRT3, SIRT4, and SIRT5 are associated with mitochondria. NAD⁺-dependent deacetylation yields a deacetylated protein, NAM, and O-acetyl-ADP-ribose. SIRT4/5 also prefer other protein modifications and cannot be reduced to one common “deacetylation rate.” PARP1/2 respond to DNA damage and build poly(ADP-ribose) chains that recruit repair complexes, including XRCC1. With excessive PARP1 activation, NAD⁺ consumption can contribute to energetic injury, but physiological DNA repair is not a pathological “drain.” CD38, whose catalytic domain is oriented predominantly extracellularly but also has intracellular topologies, cleaves NAD⁺ to NAM and ADP-ribose; cyclic ADP-ribose links this pathway to calcium signaling. The related CD157 is not equivalent to CD38 in activity or distribution.

In mouse models, an age-related increase in CD38-positive inflammatory cells is associated with lower tissue NAD⁺; Cd38 knockout attenuated this shift. The SIRT1–HIF-1α–impaired nuclear–mitochondrial coordination model also demonstrates a causal chain in experimental systems. In humans, age-related NAD⁺ trajectories vary across tissues, extraction methods, and cellular composition. Thus, a “universal decline by a specified percentage” followed inevitably by “mitochondrial collapse” is not an established clinical law. A fall in NAD⁺ may result from inflammation or another disease and may also participate in its mechanism; direction of causality requires intervention data.

Three entry routes to NAD⁺ and the local cost of using them

The de novo pathway begins with L-tryptophan: TDO or IDO converts it to N-formylkynurenine, followed by kynurenine and 3-hydroxyanthranilate; HAAO then forms unstable ACMS. ACMSD diverts this intermediate, competing with spontaneous cyclization to quinolinate. How much substrate reaches QPRT depends on flux and local conditions, not one reference $K_m$. QPRT uses phosphoribosyl pyrophosphate (PRPP) to form NAMN. NMNAT then adenylates NAMN to NAAD, and NADSYN1 completes NAD⁺ synthesis using glutamine as the amide-nitrogen donor. This route requires sources of tryptophan, PRPP, and ATP; the response to a precursor dose is not equivalent to availability of these resources in every tissue. IDO’s role in inflammation further links NAD biosynthesis to the kynurenine pathway, but increased IDO does not guarantee higher NAD⁺: intermediates are also consumed by other branches.

Nicotinic acid (NA) enters through NAPRT: $\mathrm{NA}+\mathrm{PRPP}\xrightarrow{\mathrm{NAPRT}}\mathrm{NAMN}+\mathrm{PP_i}$, followed by NMNAT and NADSYN1. This is the Preiss–Handler pathway; calling all tryptophan-derived synthesis by that name is inaccurate. Cellular NAPRT expression varies, and microbial deamidation of NAM can provide additional substrate to this branch. Clinical differences between niacin forms and their limitations are discussed in the vitamin B₃ article. Pharmacological doses of NA may activate HCAR2 and cause prostaglandin-mediated flushing; the absence of flushing after NR or NMN characterizes tolerability but proves nothing about aging or tissue delivery.

In the salvage pathway, NAMPT joins NAM to PRPP to form NMN: $\mathrm{NAM}+\mathrm{PRPP}\xrightarrow{\mathrm{NAMPT}}\mathrm{NMN}+\mathrm{PP_i}$. NMNAT then uses ATP: $\mathrm{NMN}+\mathrm{ATP}\xrightarrow{\mathrm{NMNAT}}\mathrm{NAD}^{+}+\mathrm{PP_i}$. NAMPT often acts as a regulated step in a given cellular state, but it is not invariably the sole “bottleneck”: NAM and PRPP concentrations, ATP availability, feedback, and NMNAT activity all alter flux. Histidine-residue phosphorylation of NAMPT described in enzymology matters for catalysis; a single measured $K_m$ cannot show that the enzyme is always fully saturated in living tissue. Extracellular NAMPT is discussed both as a secreted protein and a potential regulator, but its functions cannot automatically be equated with the intracellular reaction because its substrates and compartment differ.

NMNAT1 is predominantly nuclear; NMNAT2 is associated with the cytosol, Golgi, and axonal transport; NMNAT3 is linked to mitochondrial and some tissue-specific pools. After axonal injury, the short-lived NMNAT2 can decline, allowing NMN to accumulate and NAD⁺ to fall; the altered metabolite ratio participates in SARM1 activation, whose NADase activity accelerates axonal degeneration. This does not mean oral NR prevents such degeneration in humans. Nor does NMNAT1’s localization near nuclear consumers establish obligatory direct “channeling” of every newly formed NAD⁺ molecule to PARP1 or SIRT1. A compartment map is needed to understand why the same change in a whole-blood marker may accompany different local responses.

NR and NMN at the cell boundary: what charge and experiments can tell us

NR is a pyridinium ribonucleoside, whereas NMN also carries a 5′ phosphate. It is therefore chemically inaccurate to describe NR as strictly “electroneutral” and NMN as a universal dianion: ionization state reflects both the pyridinium center and the phosphate, and salts include a counterion. The phosphate increases hydration and the energetic cost of passive passage through a lipid bilayer. The Born equation, $\Delta G\propto z^2(1/\varepsilon_{\mathrm{lipid}}-1/\varepsilon_{\mathrm{water}})/r$, illustrates an electrostatic contribution but does not determine permeability across a real membrane containing proteins, conformational changes, and possible endocytosis. Low passive permeability of NMN does not mean that “intact NMN can physically never enter cells.”

One well-studied route begins in the extracellular space: ecto-5′-nucleotidases remove phosphate from NMN, the resulting NR crosses the plasma membrane, and intracellular NRK phosphorylates it back to NMN. The reactions can be written as $\mathrm{NMN}_{ext}+H_2O\xrightarrow{\mathrm{ecto\text{-}5′\text{-}nucleotidase}}\mathrm{NR}_{ext}+P_i$ and $\mathrm{NR}_{int}+ATP\xrightarrow{\mathrm{NRK1/2}}\mathrm{NMN}_{int}+ADP$. CD73/NT5E is a candidate for the first step: it is a GPI-anchored ectoenzyme for which AMP is a well-established substrate. In Ratajczak’s experiments with primary hepatocytes and cell lines, isotope labeling and NRK1 knockout supported extracellular NR formation. However, a specific result at particular concentrations, enzyme profiles, and incubation times does not establish the one universal NMN entry route or a fixed CD73 contribution across all human tissues. Different ectophosphatases and serum enzymes may have overlapping roles; distinguishing hydrolysis before the membrane from intracellular processing requires a label on the phosphate and pyridine backbone, not merely total cellular NAD⁺.

Nucleoside transporters, especially ENT1/SLC29A1 and ENT2/SLC29A2, have been implicated in NR transport; transport and its pharmacological inhibition depend on cell line and inhibitor. Equilibrative transport follows the gradient that intracellular NR phosphorylation helps maintain. ENT2 expression in muscle and ENT1 in endothelium makes these proteins biologically relevant, but does not yield a single tissue transport rate from a general expression profile. ENT4/PMAT should not be included in a proven list of NR transporters without direct demonstration of NR transport; its behavior at acidic pH with other substrates does not provide that demonstration.

NRK1 (NMRK1) is widely expressed and has an important role in cellular responses to exogenous NR; NRK2 (NMRK2) is particularly relevant in striated muscle and may be induced after injury in animal models. Both use ATP, and NMNAT1–3 then adenylate the resulting NMN. Tissue distribution differences between NRK1 and NRK2 cannot be turned into a clinical “rescue therapy” for cardiomyopathy without human intervention outcomes. Comparing tabulated $K_m$ values for NRK1 and NRK2, or CD73 and a post-dose concentration, says little about which route predominates unless local free substrate, expression, $V_{max}$, competitors, and compartmental distribution are known.

The SLC12A8 controversy requires a balanced account. Grozio and colleagues reported sodium-dependent NMN uptake and altered responses after Slc12a8 manipulation in mouse intestinal models. Schmidt and Brenner challenged both the interpretation and the plausibility of assigning an NMN-specific transporter function to an SLC12-family protein; isotope experiments in other models identified a substantial contribution from prior processing. Critics’ commentary is not independent proof that any direct transport is impossible, just as one positive experiment does not prove SLC12A8 dominates in the human intestine. Appearance of label in blood, a post-dose change in NAD⁺, and altered Slc12a8 expression are also not equivalent to proof that intact NMN crossed a particular membrane. The question remains quantitatively unresolved in humans for choosing a clinical form.

From gut to tissue: why a label changes what “bioavailability” means

An oral molecule passes through the intestinal lumen, brush-border enzymes, microbial community, enterocyte, and hepatic first pass. NR may be cleaved to NAM, while NMN may first be dephosphorylated to NR; fate depends on route of administration and microbial environment. In isotope studies by Liu and colleagues, orally administered NR and NMN substantially replenished the hepatic pool through NAM, whereas intravenous administration made it possible to observe direct delivery of intact forms to several tissues. This distinction rules out a single pharmacokinetic description for “NR in general” or “NMN in general.” A mouse result cannot set a universal fraction of intact absorption in humans or establish that the entire dose inevitably becomes NAM after ingestion.

The microbial nicotinamidase PncA deamidates NAM to nicotinic acid: $\mathrm{NAM}+H_2O\xrightarrow{\mathrm{PncA}}\mathrm{NA}+NH_3$. Subsequent NAAD exposure and NAPRT pathway activity can help identify processing, but a metabolite in blood without isotope resolution does not identify its original atomic source. The contribution of this route changed in models with altered microbiota. Extrapolation to every person is difficult because of microbiome composition, diet, transit time, and whether the dose reaches the intestinal region where the relevant organisms are active.

In the portal system, the liver processes NAM, NA, NR, and their products, and also releases NAM. A change in blood NAD⁺ may therefore reflect hepatic synthesis and exchange rather than superior direct entry of the original molecule into a neuron, muscle, or myocardium. LC–MS/MS is especially informative when it distinguishes NAD⁺, NADH, NAM, NR, NMN, NAAD, methylated products, and isotope-labeled fragments, and specifies the sample matrix and collection time. Whole blood, PBMCs, plasma, muscle biopsy, and extracellular vesicles are different systems. Without tissue data and control for cell composition, a change in whole-blood NAD⁺ cannot be called restoration of the mitochondrial matrix.

Neuronal delivery is a useful example of this boundary. In the Vreones study, after NR, investigators measured NAD⁺ and protein markers in plasma extracellular vesicles enriched for vesicles of presumed neuronal origin, in 22 participants from the original crossover study. They did not collect cerebrospinal fluid or brain parenchyma; still less did they track intact NR crossing the blood–brain barrier. The result is interesting as an indirect biomarker, but claims that “direct NR penetration into the brain has been proven” or “NMN must be hydrolyzed at capillaries” go beyond the experiment. Nucleoside transporters on endothelium offer a candidate mechanism requiring a targeted human test.

The situation in skeletal muscle is analogous: ENT and NRK expression does not guarantee higher NAD⁺ in the muscle matrix or make NAMPT irrelevant. In the Elhassan study, NR altered the NAD-related metabolome and transcriptional signatures in biopsies from older men, but these changes cannot automatically be translated into increased ATP, endurance, or a proven anti-inflammatory treatment. A similar distinction applies in mitophagy research: a change in cellular signaling is not the same as a demonstrated gain in performance.

NAD⁺ effectors: a plausible mechanism does not predetermine an effect

In experimental systems, NAD⁺-dependent SIRT1 deacetylates PGC-1α and FOXO factors, altering transcriptional programs for mitochondrial biogenesis, stress responses, and lipid metabolism. A simplified sequence is SIRT1 → PGC-1α acetylation state → interaction with nuclear respiratory factors → TFAM and respiratory-system gene expression. The sequence also depends on acetyltransferase activity, cellular stress, oxygen, and substrate availability. The claim “a capsule raises NAD⁺, therefore PGC-1α is activated and mitochondria multiply” skips all these measurements. FOXO1/3 regulate, among other things, antioxidant responses; CPT1B participates in delivery of long-chain fatty acids into mitochondrial oxidation. This cross-domain connection is discussed in detail in the article on the CPT1 carnitine shuttle.

In the matrix, SIRT3 deacetylates metabolic proteins, including SOD2 and respiratory-chain components; IDH2 supports the NADPH-dependent antioxidant network. But a rise in total blood NAD⁺ does not measure free NAD⁺ availability for matrix SIRT3. Direct assessment requires tissue analysis, validated markers of target acetylation, and a functional outcome. PARP1 also has a dual role: NAD⁺ is required for repair, while severe DNA damage activates NAD⁺ consumption that can itself cause energetic vulnerability. A precursor cannot be said to “repair DNA” based only on an NAD⁺ concentration.

The source literature contains tables of $K_m$ values for PARP1, CD38, SIRT1, and SIRT3 that are turned into a ranking of “who gets NAD⁺ first.” This is an invalid jump from a purified-enzyme parameter to cellular flux. The relation $v=V_{max}[S]/(K_m+[S])$ applies only under specified conditions and in the absence of additional inhibitors. In cells, the free $[S]$ in a given compartment, active protein abundance, competing substrates, NAM as a product inhibitor, DNA damage, and CD38 topology all matter. Conditional $K_m$ values cannot establish that SIRT3 “shuts off first” or PARP1 “captures” all NAD⁺ during aging. The source’s numerical hierarchy is therefore retained as a subject for critical discussion, not a clinical rule.

Clinical experiments: matrix, population, and primary endpoint

Most interventions test metabolites and tolerability in short studies, not the duration of healthy life. NR and NMN have not been compared in an adequately powered, direct randomized trial with the same population, equivalent dosing, tissue-measurement methods, and clinical endpoint. A “winner” between precursors cannot logically be declared from a separate early study. The precise effects below belong to the named designs; they do not establish an optimal dose for the general population.

The Yoshino result deserves attention precisely because it measured a function in a specific tissue and population; it should not be replaced by the claim “treatment of insulin resistance is proven.” Pencina illustrates a useful counter-boundary: a substantial rise in blood was not accompanied by increased muscle NAD⁺ or improvement in several muscle and metabolic outcomes. Some cardiovascular measures improved in a small, short trial, but clinical reproducibility and causal mechanism are unknown. Huang is methodologically clear: both a biochemical measure and six-minute walk were compared with placebo, and neither had between-group confirmation. Evaluating only the percentage change within the NMN group would produce a false positive conclusion.

There is no evidence that raising NAD⁺ in an accessible sample matrix slows human aging, prevents neurodegeneration, or extends life. Even a positive tissue-biopsy response does not show what fraction was caused by direct NR/NMN entry versus NAM, NA, and resynthesis. Evidence for an advantage of a specific form must demonstrate exposure, mechanism, and a prespecified clinical outcome together, with adequate duration and control for multiple comparisons.

NAM methylation: an established reaction, but an unestablished “methyl deficit”

When NAM formation exceeds its local incorporation through NAMPT, some flux passes through nicotinamide N-methyltransferase (NNMT). The reaction $\mathrm{NAM}+\mathrm{SAM}\xrightarrow{\mathrm{NNMT}}1\text{-}\mathrm{MNA}+\mathrm{SAH}$ transfers a methyl group from S-adenosylmethionine (SAM) to the pyridine-ring nitrogen of NAM. Aldehyde oxidase then forms pyridone derivatives, including 2-PY and 4-PY, which are excreted by the kidneys. Measuring 1-MNA and 2-PY supports the direction of catabolism, but does not by itself demonstrate methylome damage or folate deficiency. The chemistry also needs clarification: the source text incorrectly attributed NNMT attack to the “carbamoylpyridinium nitrogen”; it is the pyridine nitrogen of NAM that is methylated.

SAH can be hydrolyzed to homocysteine. Homocysteine is remethylated through BHMT in the liver using betaine, or through methionine synthase MTR with 5-methyltetrahydrofolate and vitamin B₁₂. These processes are connected to the methionine cycle, but their rates are regulated by nutritional status, liver and kidney function, and reversible reactions. Thus, NNMT stoichiometry cannot be used to calculate a homocysteine increase from the mass of a capsule, nor does it establish a universal hazardous NR/NMN dose threshold. Studies reporting an increase in methylated catabolites should separately assess plasma homocysteine, SAM/SAH in a relevant matrix, and clinical outcomes; without these measures, the hypothesis remains open.

Automatically adding betaine at a 1:1 weight ratio to NR/NMN, as well as fixed high doses of methylfolate and B₁₂, does not follow from the presented clinical trials. Even if a methyl donor participates in the pathway, that does not establish benefit from a preventive “buffer” in the absence of a demonstrated deficiency or elevated homocysteine. Practical assessment should account for the person’s status, medicines, kidney function, and diet; interpretation of laboratory results should consider primary causes of hyperhomocysteinemia. The biochemical role of SAM and the distinction between a substrate scheme and a clinical effect are discussed in detail in the article on S-adenosylmethionine and methylation.

From formulation and timing to a research hypothesis

Doses studied in clinical trials describe protocols; they are not approved “therapeutic ranges for aging.” NR chloride has been used in oral studies with a measured blood response. For NMN, conventional oral and microcrystalline forms have been studied; this evidence base does not establish that sublingual, liposomal, or enteric-coated forms improve human tissue delivery or clinical outcomes. Capsule design alone cannot show that a formulation bypasses first-pass metabolism, CD73, or SLC12A8. If a formulation developer claims such kinetics, comparative studies are needed with the same dose, tracking the parent molecule and metabolites, specifying collection times, and prespecifying the sample matrix.

CLOCK:BMAL1 participates in circadian regulation of NAMPT, while NAD⁺ and SIRT1 feed back on cellular clocks. Studies by Nakahata and Ramsey demonstrate a biological mechanism; they do not establish that NAMPT peaks strictly at 07:00–09:00 in everyone or prove that evening NR or NMN causes insomnia. Circadian phase depends on species, tissue, sleep, and meal timing. Transferring a cellular or mouse phase into a “mandatory morning window” for clinical use is therefore methodologically unsound; trial dosing time should be reported as part of the protocol, not as a therapeutic law.

Flavonoids, including apigenin, have been observed to inhibit CD38 in biochemical and cellular systems. Choosing an oral combination requires data on achievable free concentration in the relevant compartment, specificity, and clinical outcome; an $IC_{50}$ for a purified enzyme does not turn an apigenin dose into treatment for NAD⁺ deficiency. The article on apigenin, CD38, and sleep connects CD38 as a target with the real limits of flavonoid evidence. Likewise, resveratrol/pterostilbene interactions with SIRT1 depend on substrate and assay conditions: “sirtuin activation” is not equivalent to extending human life. Combining urolithin A or PQQ may be a research hypothesis about mitochondrial quality and biogenesis, but it is not a proven “longevity triad” or a basis for a fixed multi-ingredient protocol.

Laboratory monitoring in trials assessed safety and metabolites at specified times, not a universal course every 12 weeks. If NR or NMN is considered in a research or individual clinical context, record baseline status, product composition, concomitant agents, and liver and kidney function; analyze changes using the same method and matrix. Total NAD⁺ in blood is not a validated stand-alone “biological age” test. Homocysteine should be interpreted using local reference ranges and risk factors; a narrow “normal” range from the source text should not be imposed on every laboratory and population.

How this review was assembled and where the evidence ends

This is a narrative editorial synthesis of the supplied literature corpus, not a systematic review or formal GRADE assessment. On October 4, 2026, primary studies underpinning key interpretive boundaries were specifically checked: cellular processing of NMN and NRK1, oral versus intravenous isotope routes, the Vreones sample matrix, and between-group results in the Pencina and Huang clinical trials. The remaining sources are used within the original bibliography; no new exhaustive search, independent peer review, or verification of every citation is claimed. Important uncertainties arise from differences in species, dose, exposure, chemical form, sample preparation, tissue, and duration. Negative findings for some endpoints do not prove that no effect is possible in another population, but they do preclude presenting benefit as already established.

The distinction between NR and NMN primarily creates testable hypotheses about extracellular hydrolysis, transport, and local NAD⁺ synthesis. Both forms may increase accessible markers of NAD metabolism in humans; those markers do not yet identify a universally better precursor or establish prevention of aging. The most informative next studies should measure labeled flux, a specific tissue, and a prespecified clinically meaningful outcome in parallel, and report negative results alongside positive ones.

How this review was prepared

This is a narrative review of NR, NMN, and NAD⁺ metabolism based on a collection of 29 publications. In the October 4, 2026 editorial version, key discrepancies in the interpretation of pharmacokinetic and clinical results were specifically checked. No exhaustive search using a prespecified protocol, meta-analysis, or formal risk-of-bias assessment of each study was conducted.

Cellular precursor-conversion pathways, blood NAD⁺ measurements, levels in individual tissues, and clinical outcomes are considered separately. Data from plasma-isolated vesicles are not a direct measurement of a metabolite in the brain. Results depend on chemical form, route of administration, duration, population, and endpoint; a biomarker change does not establish improved function or long-term benefit.