A mechanism of allosteric modulation of nicotinamide phosphoribosyltransferase to elevate cellular NAD⁺

Abstract

In aging and disease, cellular NAD⁺ is depleted by catabolism to nicotinamide (NAM). NAD⁺ supplementation is being pursued to enhance human healthspan and lifespan. Activation of nicotinamide phosphoribosyl-transferase (NAMPT), the rate-limiting step in NAD⁺ biosynthesis, has potential to increase salvage of NAM. Novel NAMPT positive allosteric modulators (N-PAMs) were discovered in addition to demonstration of NAMPT activation by biogenic phenols. The mechanism of activation was revealed through synthesis of novel chemical probes, new NAMPT co-crystal structures, and enzyme kinetics. Binding to a rear channel in NAMPT regulates NAM binding and turnover, with biochemical observations being replicated by NAD⁺ measurements in human cells. The mechanism of action of N-PAMs identifies, for the first time, the role of the rear channel in regulation of NAMPT turnover coupled to productive and non-productive NAM binding. The tight regulation of cellular NAMPT via feedback inhibition by NAM, NAD⁺, and ATP is differentially regulated by N-PAMs and other activators, indicating that different classes of pharmacological activators may be engineered to restore or enhance NAD⁺ levels in affected tissues.

Graphical Abstract

INTRODUCTION

Nicotinamide adenine dinucleotide (NAD⁺) and its 2-phosphate derivative (NADP⁺) are formed on electron transfer from NADH and NADPH, respectively. The electron transfer reactions regulated by these enzyme cofactors drive cellular metabolic processes. NAD⁺ serves as a substrate for important enzymes that catabolize NAD⁺ to nicotinamide (NAM), a process that can lead to severe disruption of the cellular NAD⁺ economy. This disruption and depletion of NAD⁺ is closely associated with aging and metabolic disorders; and therefore, replenishment of cellular NAD⁺ by “NAD⁺-enhancing drugs” has evolved as a strategy for therapeutics directed at augmented lifespan and healthspan ^{1, 2}. Administration of NAM or nicotinamide mononucleotide (NMN), was shown to improve aspects of healthspan or to increase lifespan in mice ^{3, 4}. Recently, treatment of prediabetic postmenopausal women for 10 weeks with NMN was shown to restore muscle insulin sensitivity and insulin signaling ⁵. NMN is a biosynthetic precursor of NAD⁺ and is the product of the reaction of NAM with α-d-5-phosphoribosyl-1-pyrophosphate (PRPP) catalyzed by the enzyme nicotinamide phosphoribosyltransferase (NAMPT; EC 2.4.2.12) ¹.

Accumulated DNA damage associated with normal aging leads to elevated poly-ADP-ribose polymerase (PARP) activity: PARP consumes NAD⁺ to ADP-ribosylate proteins at sites of DNA damage ⁶. A second source of NAD⁺ catabolism to NAM is NAD nucleosidase activity associated with CD38, the expression of which increases in multiple tissues with age and is associated with chronic inflammation associated with aging (inflammaging) and cellular senescence ^{2, 7–9}. In response to the cellular stress induced by PARPs and CD38, upregulation of sirtuins (SIRTs) can provide cellular protection; however, SIRTs themselves catabolize NAD⁺ in the process of protein de-acylation. NAD⁺ depletion leads to reduced SIRT1-mediated deacetylation of PGC-1α, which results in defective mitochondria and increased release of reactive oxygen species that cause further DNA damage ^{7, 10}. Other ADP-ribosyltransferases (ARTs) contributing to NAD⁺ depletion include sterile alpha and TIR motif containing 1 (SARM1) ¹¹. Quantitatively, both PARP activity and SIRT activity each account for one third of NAD⁺ consumption ¹².

In mammals, the rate-limiting step in NAD⁺ biosynthesis is the salvage of NAM and conversion to NMN, catalyzed by NAMPT (Figs 1A,B). NAMPT activity is viewed as key in cellular defense mechanisms controlling cell survival and maintaining metabolic homeostasis. ² Female Nampt+/− mice have glucose tolerance and impaired insulin secretion that is ameliorated by administration of NMN.^13–15 NAMPT mediates an adaptive response to inflammatory, oxidative, and genotoxic stress, a response that declines with age, along with a reported significant decline in NAMPT and NAD⁺ levels.^{14, 16, 17} Although the role of NAMPT in physiology and pathophysiology is complex, the possibility of addressing diseases of normal aging, such as Alzheimer’s disease and related dementia (ADRD) and diseases of accelerated aging, such as Type-2 diabetes (T2D), by enhancing NAMPT activity, is compelling.

Figure 1.

(A) NAMPT mechanism: the conversion of NAM to NMN at the NAMPT active site requires ATP to phosphonylate His-247. (B) Superposition of crystal structures illustrating the NAMPT active site (two Mg²⁺ ions shown as green dots and phospho-H247 mimicked by BeF₃⁻ adduct; PDB: 3DKL; 3DHF). (C) Superposition of NAMPT inhibitor FK866 (cyan) and allosteric activator NP-A1 (gold) bound to the rear channel of NAMPT (PDB: 2GVJ; 3DKL; 3DHF; 8DSD).

NAMPT catalyzes the formation of NMN and an ATPase reaction, both at the same active site, in a mechanism delineated in three papers from Schramm, Burgos and co-workers using: 1) kinetic analysis;¹⁸ 2) structural biology;¹⁹ and 3) kinetic isotope effects ²⁰. The chemical equilibrium modulated at the active site of NAMPT (PRPP + NAM ⇋ NMN + PPi) lies to the left in the absence of ATP: i.e. slow conversion of NMN to NAM (Figs 1A,B). The ATPase reaction transferring the γ-phosphate of Mg²⁺-ATP to form N-phosphohistidine (phospho-H247) occurs at the same active site, with nucleobases of ATP/ADP and NAM/NMN sequentially occupying a nucleobase binding pocket. The formation of phospho-H247 leads to a shift in the equilibrium to the right, resulting in a 1,100-fold increased efficiency for NMN formation.²⁰ The increase in k_cat/K_M is driven by a ≤ 170-fold enhancement of PRPP-dependent NAM binding and an increase in k_cat. Importantly, phospho-H247 is required to coordinate two Mg²⁺ ions that stabilize the transition state for phosphoribosyl group transfer from the ribose anomeric carbon of PRPP to the pyridyl-N of NAM (Fig. 1B). Feedback inhibition is observed by both NAM and NAD⁺ and under most conditions, the ATPase reaction is not tightly coupled to turnover of NAM, leading to a seemingly inefficient catabolism of cellular ATP, which is amplified in the presence of NAM and the absence of PRPP. ¹⁸

In HTS assay of small molecule libraries, we identified more than one class of small molecule able to increase enzymic activity of recombinant NAMPT. ²¹ Several phenolic compounds, including quercitrin and other biogenic and/or bioactive phenols were efficacious, with modest potency and binding affinity. These were differentiated from novel NAMPT positive allosteric modulators (N-PAMs) that bind NAMPT to increase activity in biochemical and cell-based assays. The N-PAMs are also differentiated in their ability to couple the turnover of NAM → NMN with the ATPase activity of NAMPT. We were able to obtain multiple co-crystal structures of quercitrin and N-PAMs bound to the rear channel of NAMPT (Fig. 1C). Interestingly, multiple published co-crystal structures also demonstrate the binding of potent NAMPT inhibitors in this rear-channel. The majority of NAMPT inhibitors (e.g. FK866) contain a nitrogenous base that binds in the nucleobase pocket, displacing NAM (Fig. 1C). Structural analysis, combined with enzyme kinetics, support a novel mechanism of allosteric activation of NAMPT by regulation of productive and non-productive binding of NAM and mechanical coupling of the rear channel to the active site.

This work advances understanding of the NAMPT enzyme mechanism, defining a significant role for the rear channel in feedback inhibition by NAM. This mechanism provides the structural and mechanistic basis for the design of small molecules as NAMPT activators of potential therapeutic value. The observations on recombinant NAMPT activity translated directly to observations on NAD⁺ regulation in human cells. We define a role for the rear channel that is unique to NAMPT among phosphoribosyltransferases, raising the question of possible endogenous activators and the contribution of NAMPT activation to the biological activity of biogenic phenols.

RESULTS & DISCUSSION

Activation of NAMPT.

The primary assay of NAMPT enzyme activity measured NAD⁺ production by coupling the NAMPT-catalyzed production of NMN with: (i) the conversion of NMN to NAD⁺ catalyzed by nicotinamide/nicotinic acid mononucleotide adenylyltransferase (NMNAT); and (ii) alcohol dehydrogenase (ADH) and DT-diaphorase (Fig 2A,B). Optimized for HTS, the assay was used to screen: 20,000 compounds from ChemDiv libraries; and 2,000 compounds from the Microsource Spectrum library. The most active validated hit from the ChemDiv library was NP-A1 and validated hits from the Spectrum library included the flavonoid glycoside, quercitrin, onenetin, genestein, and dienestrol (Fig. 2C,F; Fig. S1). The aglycone, quercitin, was devoid of activity at ≤ 40 μM. Further examination of the isomers of NP-A1 revealed that the R-isomer was the most potent compound (EC₅₀ = 37 nM), whereas the S-isomer gave the maximum observed activation of NAMPT relative to vehicle control (Fig. 2C). The trapping of NMN as a fluorescent adduct was the basis of an orthogonal assay for NAMPT activity. ²² This is an endpoint assay in which NMN accumulates, in contrast to the coupled NAMPT assay in which NMN is immediately captured by NMNAT. After completion of our HTS campaign, the 4-pyridyl activator, SBI-0797812, was reported,²³ and it was therefore included in our studies on N-PAMs and quercitrin and confirmed to activate NAMPT (Fig. 2D).

Figure 2.

(A) Coupled primary enzyme assay used for HTS, with (B) raw kinetic traces showing activation, and (C) derived concentration-response curves for HTS hits and NP-A1 isomers. (D) Orthogonal NAMPT assay concentration-response curves for activators and N-PAMs. (E) Binding to NAMPT measured by MST. (F) Structures of N-PAMs and other NAMPT activators.

N-PAM binding to the NAMPT rear channel.

Over 70 crystal structures of NAMPT are represented in the RCS-PDB database, the majority being structures of the human enzyme and with bound inhibitors. As shown in Figure 1C, 3-pyridyl inhibitors, such as FK866, bind in the “rear channel” with the pyridyl group occupying the nucleobase binding site and superimposing on NAM. The nucleobase of both NAM and NAMPT inhibitors is bound in a Phe-193/Tyr-18 pi-pi clamp. The nucleobase pocket must also accommodate the adenine group during the ATPase reaction converting ATP to ADP: adenine binds with the nucleobase displaced and rotated approximately 100° with respect to the orientation of NAM using an Arg-311/Phe-193 cation-pi clamp. Remarkably, quercitrin and N-PAMs were observed to bind to the same rear channel as NAMPT inhibitors, without engagement of the nucleobase pi-clamp (Fig. 1C).

Orthogonal biophysical assays were used to confirm and quantify ligand binding to NAMPT. Microscale thermophoresis (MST) measurements replicated the order of binding affinity indicated from NAMPT kinetic data: FK866 > NP-A1R > quercitrin (Fig. 2E). Given the occupation of the rear channel and nucleobase binding site by NAMPT inhibitors, it was straightforward to utilize an inhibitor scaffold to synthesize and validate a fluorescence polarization (FP) probe (ZN-2-102) to measure binding to NAMPT through FP-probe displacement (Fig. 3). In accord with the MST data, NP-A1R was observed to be a high affinity ligand with K_D = 38 nM, approaching the affinity of the potent NAMPT inhibitor FK866 (K_D = 10 nM); and quercitrin, also mirroring MST data, displayed low affinity for NAMPT (K_D = 16 μM) (Fig. 3A). The S-isomer of NP-A1 displayed similar affinity for NAMPT in the FP-assay to a synthetic analogue, ZN-2-43 (K_D ≈ 400 nM) (Fig. 3B), which was also shown to act as a N-PAM in the enzyme activity assay (Fig. 2D).

Figure 3.

(A) FP binding assay shows similar affinity for inhibitor FK866, and activators NP-A1R and quercitrin, as observed by MST. (B) Binding curves for putative NAMPT activators. (C) Structures of: FP-probe ZN-2-102; 3-pyridyl inhibitor FK866; and putative activator P7C3-A20.

Two compounds reported in the literature to activate NAMPT, SBI-797812 (introduced above)²³ and a phenolic compound, NAT1, demonstrated 2-6 μM affinity for NAMPT (Fig. 3B). To the best of our knowledge, this is the first unambiguous and quantitative evidence that SBI-797812 binds to the rear channel of NAMPT; although, the fact that binding affinity is inferior to potency measured for enzyme activation (Fig. 2D) suggests that rear channel binding may not be the exclusive mechanism of activation of SBI-797812 and related 4-pyridyl activators. Other biogenic and/or bioactive phenols identified in the HTS screen were shown to bind to NAMPT with low affinity, generally matching enzyme activation data. Finally, the carbazole, P7C3-A20, frequently described as a NAMPT ligand and activator, ²⁴ did not show any NAMPT binding activity in this assay (Fig. 3B).

The combined kinetic and biophysical data demonstrated NP-A1R to be a high affinity ligand with modest maximal activation and NP-A1S to be a weaker ligand with superior maximal activation. Quercitrin, as representative of the phenolic activator class, was a weak binding NAMPT ligand with high activation at higher concentrations.

Structural analysis of NAMPT binding.

NAMPT functions as a homodimer in which the dimer interface provides two symmetrically mirrored active sites (Fig. S2). The rear channel, nucleobase pocket, and active site are largely lined by residues of the first monomer; whereas, the pi-pi clamp residue Tyr-18 and hydrophilic residues at the mouth of the active site are supplied by the second monomer. The rear channel of human NAMPT is formed by two small β-strands (β14 and β15) at the dimer interface. Despite strong structural and mechanistic similarities between NAMPT and human nicotinic acid phosphoribosyltransferase (NAPRT), the rear channel of NAPRT, formed by β-strands β13 and β14 is occluded, blocking binding of FK866 and other NAMPT inhibitors. The rear channel of NAMPT is a seemingly unique feature.

NP-A1 structures were obtained with bound NAM and co-crystal structures of quercitrin contained either bound NAM or a stable ADP analogue (AMPCP). The structure of ZN-2-43 did not contain bound NAM, showing that NAM binding was not required for N-PAM binding. An unliganded NAMPT structure, with bound NAM, was also obtained for comparison. A key feature of the rear channel adjacent to the nucleobase binding site is a H-bonding network with three water molecules: two H-bonded to NAM, Ser-241, Ser-275, and Asp-219; and a third water molecule not directly H-bonded to any amino acid residue, which is clearly shown in the NP-A1S structure (Figs 4A,B). In the NAMPT structures without bound ligand, a fourth water molecule was observed, which was displaced by all ligands (Fig. S3). Superposition of FK866 with NAM and NP-A1S bound to the rear channel of NAMPT shows the displacement of two of the three water molecules, with the water interactions being replaced by H-bonding with the amide group of FK866 (Fig. 4C) The phenolic 7-OH of quercitrin displaces one of the three water molecules in the H-bonding network (Fig. 4D). In the two quercitrin co-crystal structures, the 7-hydroxy group sits 3.5 Å and 4.1Å from the amide nitrogen of NAM and from the adenine ring nitrogen of ADP (AMPCP), respectively. In contrast, the third water molecule is not displaced by the NP-A1 isomers and sits only 2.6 Å and 2.2 Å from the methyl groups of NP-A1R and NP-A1S, respectively (Fig. 4D).

Figure 4.

Crystal structures of ligands bound to the NAMPT active site and the rear channel. (A) NP-A1R (gold) and NAM showing key residues and H-bonding network and (B) superimposed with bound FK866 (PDB: 2GVJ). (C) Quercitrin or (D) NP-A1R bound to NAMPT with NAM showing key water molecules. (E) View of bound NP-A1R from mouth of rear channel showing surrounding residues. (F) NP-A1S (gold) superposed with ZN-2-43 (blue).

The replacement of the third water molecule by quercitrin or its apparent compression in the NP-A1 structures might be proposed to “push” NAM towards the anomeric carbon of PRPP, potentially compressing the transition state and lowering the activation energy for phosphoribosyl transfer. However, in all co-crystal structures with activator bound, there was no displacement of NAM relative to the position of NAM in structures without activator. Triangulation of the NAM ring-nitrogen, measuring distances to the α-C of residues surrounding the active site (His-247/Arg-392/Asp-354/Asp-313) demonstrated that the position of NAM is identical in the presence or absence of N-PAMs (Fig. S4). In crystal structures, neither binding of N-PAM nor quercitrin has any apparent effect on the spatial arrangement of NAM with respect to the α-C of catalytic site amino acid residues, demanding an alternative mechanism of allosteric NAMPT activation.

Crystal structures were examined for other pertinent interactions. Superposition of the structures of quercitrin with NP-A1R shows that the rhamnose ring of quercitrin extends further towards the solvent exposed mouth of the rear channel, with the sugar 2-OH and 3-OH interacting with solvent (Fig. S5). The 5-Me group sits in a hydrophobic cleft formed by Pro-307 and Ile-309 and the 4-OH is involved in a H-bond network with the catechol that is itself H-bonded to Lys-189. Comparison with N-PAM co-crystal structures indicates that the H-bonding interaction with Lys-189 is common for these NAMPT activators. Perhaps surprisingly, given its flexibility, the lysine side chain was unperturbed by ligand binding, superimposable in all liganded and unliganded structures (Fig. S6).

Both N-PAM isomers bind in the rear channel cavity formed by Tyr-188/Gly-185/Ala-379 and Pro-272/Pro-307, with the tolyl group enclosed by His-191/Val-242/Ile-351 and the benzyl group by Arg-349/Val-350 (Figs 4E, S6). In addition to the interaction with Lys-189, N-PAMS form a loose H-bonding network at the mouth of the rear channel: for NP-A1R, this network involves multiple water molecules, the amide and pyrimidine nitrogens of the N-PAM and Thr-304. In addition to facilitating this H-bonding for NP-A1R, the major impact of the piperidine stereocenter is to enforce a twist-boat conformation of the piperidine ring in NP-A1S (Fig. 4E,F).

We were not able to obtain a co-crystal structure with the NP-A1 isomers in the absence of bound NAM; however, a structure was obtained with the N-PAM, ZN-2-43, showing that bound NAM is not required for N-PAM binding to the rear channel (Figs 4F, S7). ZN-2-43 binds to NAMPT with similar affinity to NP-A1S (Fig. 2E). In the absence of a nucleobase, the ZN-2-43 structure has the pi-pi clamp residues in the orientation observed for NAM binding (rather than adenine binding). The same pi-pi clamp orientation is seen in all published co-crystal structures of NAMPT inhibitors.

In the three N-PAM co-crystal structures, the ligand adopts an unusual hairpin conformation with intramolecular pi-stacking between the benzyl/benzothiophene and pyrazolo-pyrimidine ring planes (interplane distance ≈ 4.0 Å) (Figs 4E,F). In contrast to quercitrin, water molecules are largely excluded from the rear channel, with the hairpin structure filling the rear channel: the volume of the rear channel cavity was calculated as approximately 490-540 ų (dependent on delineating the mouth of the channel); whereas the volume of the three N-PAMs was calculated to be: 469, 472, and 510 ų; for NP-A1R, NPA1-S, and ZN-2-43, respectively. Intuitively, the hairpin structure suggests an enthalpic cost associated with strain energy; however, DFT molecular orbital calculations show that local minimum energy structures are compatible with the ligand conformations observed in crystal structures (Fig. S8). The R-isomer accommodates a chair conformation of the piperidine ring in the hairpin structure seen in the rear channel, which translates to high affinity and potency (Fig. 2; EC₅₀ ≈ 40 nM; K_d = 38 nM); whereas the S-isomers must adopt a higher energy twist-boat conformation (ΔG = 1.4 kcal/mol) to bind in the rear channel, which translates to lower affinity and potency (Fig. 2; EC₅₀ ≈ 0.7-1.0 μM; K_d = 0.55 μM).

Mechanism of NAMPT activation.

NAMPT uses cellular ATP to drive the salvage of NAM to synthesize NMN. The catalytic efficiency (k_cat/K_M) of NAMPT results from the ATPase reaction, generating phospho-H247. This lowers the K_M for PRPP and consequently lowers K_M(NAM) 170 fold, while increasing k_cat almost 20-fold.¹⁸ In the absence of ATP, the equilibrium established by NAMPT (PRPP + NAM ⇋ NMN + PPi) lies to the left, as demonstrated clearly by the conversion of NMN to NAM seen in the absence of ATP (Fig. 5A). Although NAMPT does not experience zero ATP in cells, this conceptual experiment is useful to emphasize the dependence on ATP. Addition of ATP shifts the equilibrium towards NMN, demonstrating the importance of ATP-mediated formation of phospho-H247 for the forward reaction. In the presence of NP-A1S or SBI-0797812, the equilibrium remains in favor of NAM in the absence of ATP and shifts towards NMN on addition of ATP (2 mM) (Fig. 5A). Similar experiments either without NAM, or without NMN in the incubation mixture, reinforce the importance of ATP (Fig. S10): N-PAM activation of NAMPT is ATP-dependent.

Figure 5.

NAMPT dependence on ATP and NAM. (A) Dependence on ATP of equilibrium shift from NAM to NMN in presence or absence of activators. Dependence of activity on ATP (B), PRPP (C), and NAM (D) for N-PAMs and quercitrin. Dependence of activity on ATP for SBI-797812 (E), NP-A1S (F), and quercitrin (G). (H) Regulation of uncoupled ATPase reaction is ligand dependent.

Activity dependence on NAM, PRPP, and ATP.

In the primary enzyme assay N-PAMs and quercitrin increased rate without significant effect on K_M(ATP); and a similar observation was made for K_M(PRPP) (Figs 5B,C; Table S1). NP-A1S increased catalytic efficiency (V_max/K_M) with respect to ATP tenfold (Table S1). Under the conditions studied, the apparent K_M for NAM (≈145 nM) was not influenced by NP-A1R and modestly increased for NP-A1S; whereas the K_I of NAM was significantly right-shifted by N-PAMs by 10- to 20-fold (Fig. 5D; Table S1). The attenuation by N-PAM of substrate inhibition by NAM is potentially physiologically important, because this opens a window for N-PAM-stimulated NAMPT activity extending to 500 μM NAM, which does not exist for the enzyme in the absence of N-PAM.

The orthogonal enzyme assay was used to compare the ATP dependence of SBI-0797812 activation with that of NP-A1S and quercitrin. Although K_M(ATP) is reported as 7 mM, substrate inhibition by [ATP] > 4 mM, is also reported ¹⁸. The 4-pyridyl activator, SBI-0797812, caused a significant left-shift in K_M(ATP) and also drastically left-shifted the K_I(ATP) to < 1mM (Fig. 5E). As seen in the coupled enzyme assay (Fig. 5B), NP-A1S had no significant effect on ATP dependence (Fig. 5F) and a similar observation was made for quercitrin (Fig. 5C) and a related phenolic activator (Fig. S11). The ATP dependence observed for SBI-0797812 suggests activation will be highly dependent on ATP concentration in a cellular context. Physiological, cellular ATP concentrations vary from 2-7 x 10⁻³ M, depending on cell type²⁵; whereas the concentration of ATP in plasma is 10,000-fold lower.²⁶ The optimal activation driven by SBI-0797812 is therefore predicted to lie between ATP concentrations experienced by cellular NAMPT and extracellular NAMPT (eNAMPT).

The ATPase and NAM salvage reactions catalyzed by NAMPT are uncoupled (or non-stoichiometric), resulting in futile consumption of ATP: i.e. the ratio of ADP/NMN formation is greater than unity under most conditions. ¹⁸ For example, incubating with ATP (2 mM) for 60 min, Burgos et al. reported formation of 30 μM ADP, which was reduced in the presence of NAM+PRPP or NMN+PPi, and increased fourfold in the presence of NAM alone and sevenfold in the presence of PPi alone.¹⁸ Thus, the binding of NAM in the absence of PRPP leads to accelerated non-productive ATP degradation to ADP. Published results with NAM were recapitulated and quercitrin was also observed to increase uncoupled ATP consumption (Fig. 5G). Addition of the NAMPT inhibitor, FK666, also increased ATP turnover to the same high rate observed for NAM (for the inhibitor, the burst of ADP, seen for NAM was not observed). This is consistent with accelerated breakdown of the phospho-enzyme when a ligand is bound to the nucleobase pocket in the absence of PRPP. N-PAMs reduced ATP consumption in the presence of NAM, with the high affinity NP-A1R more effective than the lower affinity S-isomer. The N-PAMs also limited ATP consumption in the absence of NAM. Thus N-PAMs inhibit non-productive binding of NAM and are diametrically differentiated from the phenolic activators in regulation of the ATPase reaction catalyzed by NAMPT.

The importance of the rear channel in NAMPT function.

The detailed enzyme mechanism elegantly delineated by Schramm, Burgos and co-workers does not reference the rear channel.^18–20 An extended mechanism to account for N-PAM activation must take account of the previously detailed mechanism and the role of the rear channel. Applying Occam’s razor, we propose a mechanism that, like Schramm’s, relies primarily on modulation of NAM binding with a potential secondary contribution to k_cat. The proposed reaction profile (Fig. 6A) incorporates: 1) a non-productive pathway (low affinity, high K_M, NAM binding, leading to ATP catabolism); and 2) a productive pathway (high affinity, low K_M, NAM binding leading to turnover to NMN). High affinity NAM binding occurs, reasonably, via the rear channel, since PRPP is required to be bound to the phosphoenzyme and we propose that low affinity NAM binding is also via the rear channel. N-PAM binding inhibits low affinity NAM binding, shifting K_I(NAM) to the right, essentially relieving the effect of higher [NAM] on enzyme activity. In the case of the high affinity N-PAM R-isomer, high affinity NAM binding is also partially inhibited, leading to lower fold-activation.

Figure 6.

N-PAM mechanism of action. (A) Stepwise mechanism starting with (S1) ATP binding and ATPase reaction to give phospho-enzyme, followed by either: 1) Non-productive pathway (orange), (N1) low affinity binding of NAM, leading to (N2) uncoupled, non-productive ATPase reaction; or 2) Productive pathway (magenta), (P1) PRPP binding leads to, (P2) high affinity NAM binding and subsequent turnover to NMN, followed by, (P3) breakdown of the phospho-enzyme by capture of metaphosphate (PO₃⁻) by water, PPi, or ATP, (P4) leaving NAMPT poised for “reloading” by ATP. Red text describes impact of N-PAM binding. (B) Structure of NAMPT active site containing PRPP and phospho-H247, nucleobase pocket containing NAM, and rear channel containing N-PAM or quercitrin. In the proposed mechanism, NAM binding occurs via the rear channel.

The question of N-PAM effects on k_cat is more difficult to assess, because the crystal structures of N-PAM-bound NAMPT do not show any perturbation of NAM nor active site residues. Enzymes have been shown by normal-mode analysis (NMA) ²⁷ to have open and closed conformations that correlate with activity; and, ligand binding of an allosteric modulator would be predicted to stabilize conformers that regulate activity.²⁸ Using NMA, crystal structures obtained in this work were compared to key NAMPT structures from the literature. The NAMPT structures with BeF₃⁻ coordinated to His-247 provide a mimic of the phospho-H247 phosphoenzyme (PDB: 3DKL; 3DHF; Fig. 1B).¹⁹ The structure with an inert analogue of NAM (BzAM), PRPP, and the two Mg²⁺ ions required for catalysis, bound in the active site, can be seen as an ideal model of the enzyme intermediate immediately preceding the transition state for phosphoribosyl transfer to NAM. Comparison of the amplitude (rmsd) of the main-chain residues observed for this structure is indicative of a more “open” protein structure compared to all other structures with bound NAM, which are “closed” (Fig. S9). The lowest frequency mode (Mode 1) was only enriched in structures with open conformations. Visualization of this mode demonstrates coupling of N-PAM binding in the rear channel to residues remote from the rear channel (Fig. S9).

Although NMA observations do not define a specific mechanism of k_cat modulation by N-PAMs, they do support a potential mechanical coupling of the rear channel allosteric site to the active site. We propose a smaller contribution from N-PAM stabilization of the “open” states, containing the phosphoenzyme (E*; Fig. 6A), as suggested by NMA. The phosphoenzyme sits adjacent to the transition states for ribosylation and dephosphorylation; therefore, allosteric stabilization of the phosphoenzyme states will lower the activation barriers for these reactions, translating to an increase in k_cat. The proposed mechanism is driven by modulation of NAM binding and supported by: a) the attenuation of NAM-induced uncoupling of the ATPase reaction (Fig. 5H); b) the differences in potency and activation of turnover by the two NP-A1 isomers (Figs 2C,D); and c) the observed >20-fold right-shift in K_I(NAM) observed for N-PAMs (Fig. 5D).

Regulation of cellular NAD⁺.

Given the dependence of NAMPT activation on the concentration of NAM, ATP, and NAD⁺, the translation of the actions of a biochemical activator to a cellular context is not guaranteed. For example, in addition to NAMPT inhibition by ATP and NAM, the reported K_I for inhibition of NAMPT activity by NAD⁺ is 2 μM;¹⁸ therefore, in a cellular context, inhibition by NAD⁺ might attenuate NAMPT activation. However, NAMPT activity was measured in the presence of N-PAMs at 10 μM NAD⁺, showing retention of NAMPT activity at a level equal or superior to that in the absence of NAD⁺ (Fig. S12). Gratifyingly, biochemical observations on N-PAMs directly translated to elevation of NAD⁺ in human cells (Fig. 7A,B).

Figure 7.

(A) NAD⁺ regulation in THP-1 cells by N-PAMs and activators. (B) Right-shift of FK866 inhibition curve by N-PAMs in THP-1 cells. (C) Biochemical NAMPT activation by N-PAMs compared to other reported activators. Activity relative to vehicle control was measured in the orthogonal enzyme assay: ATP 4 mM; NAM 3 or 300 μM; PRPP 40 μM.

Using FK-866 and NMN as negative and positive controls, respectively, we explored several cell lines to select a highly reproducible model system with high dynamic range. We miniaturized a commercial NADglo assay in the THP-1 human leukemic monocyte cell line, measuring NAD⁺ after incubation with test compounds for 24 h. Cells treated with NP-A1S responded with an increase in cellular NAD⁺ of over twofold, whereas NP-A1R only increased cellular NAD⁺ 1.2-fold (Fig. 7A). The N-PAM, ZN-2-43, was more potent and more efficacious than NP-A1R. Quercitrin was inactive and a recently reported example of a phenolic NAMPT activator, NAT1 (Fig. 1C), increased cellular NAD⁺ 1.25-fold (Fig. 7A). Surprisingly, the NAMPT activator, SBI-0797812, acted as a NAMPT inhibitor reducing cellular levels of NAD⁺ in THP-1 cells in a concentration-dependent manner (Fig. 7A). Since this contrasts with the reported effects of SBI-0797812 in another cell line,²³ activators may have cell or tissue selective phenotypes.

For N-PAMs, the biochemical observations on NAMPT activation fully translated to modulation of NAD⁺ in a cellular context (Fig. 7A). Having demonstrated that N-PAMs bind to the same rear channel as NAMPT inhibitors such as FK-866, we tested the ability of N-PAMs to compete with and displace FK866 in THP-1 cells. Treatment of cells with both NP-A1 isomers in the presence of FK-866 produced a right-shift in the FK-866 response (to IC₅₀ = 0.19 nM and 0.22 nM for S and R isomers, respectively) (Fig. 7B). The effect of the high affinity R-isomer is compatible with its ability to compete with FK866 for binding to NAMPT, but with intrinsically modest activation of NAMPT.

To explore a rationale for the observations on SBI-0797812 in cell cultures, we re-examined the biochemical activation of NAMPT at a higher, physiologically relevant ATP concentration (4 mM) and high and low NAM concentrations (3 μM is at the lower end of plasma NAM levels¹²). Inhibition of NAMPT activity was observed for SBI-079812 (Fig. 7C), compatible with biochemical observations on ATP dependence and cellular concentrations of ATP and NAD⁺.

Biochemical activity does not necessarily translate directly to activity in cells, even with good cell permeability; a phenomenon observed for NAMPT inhibitors²⁹ amongst many other examples. The cellular actions of N-PAMs were as predicted from biochemical observations. N-PAMs increased cellular NAD⁺ levels: the S- isomer increasing cellular NAD⁺ by 250%; and the R-isomer blocking the inhibitory activity of FK866, even though NAMPT activation was modest. Although quercitrin was inactive in cells, the phenol, NAT1, acted as a NAMPT activator in cells and biochemical assays (Figs. 2C,D; 7A). The observation that several biogenic and/or bioactive phenols are NAMPT activators may indicate that this activity contributes to their observed biological actions; although, it is unsafe to ascribe the in vivo phenotype of such phenols entirely to NAMPT activation.

CONCLUSIONS

NAMPT is proposed to play multiple physiological roles in addition to being the rate-limiting enzyme in mammalian NAD⁺ biosynthesis. NAMPT is proposed to act as a cytokine sometimes known as pre-B cell colony-enhancing factor (PBEF),^{30, 31} and as an adipokine known as visfatin.³² In its guise as a cytokine, extracellular NAMPT (eNAMPT) acts as a ligand and agonist at the Toll-Like Receptor 4 (TLR4) activating inflammatory responses.³³ Much remains to be discovered about the interplay of eNAMPT with the cellular NAMPT that controls NAM salvage. These roles are pivotal to cellular bioenergetics, mitochondrial function, metabolic homeostasis, and the Circadian clock.

Small molecules were identified that increase the catalytic efficiency of NAMPT for NAM salvage. Novel NAMPT positive allosteric modulators (N-PAMs) were differentiated from phenolic activators, represented by quercitrin and NAT1, and recently reported 4-pyridyl activators, represented by SBI-0797812. Differentiation was observed in dependence on ATP and NAM concentration and modulation of the uncoupled ATPase reaction of NAMPT. The observations on quercitrin and other biogenic and/or bioactive phenols suggest that, at higher micromolar concentrations, NAMPT activation may contribute to the biological phenotype of these and related compounds and natural products.

Cellular NAD⁺ biosynthesis is tightly regulated; notably via feedback inhibition by ATP, NAM, and NAD⁺. A key aspect of the differentiation of activators is associated with this inhibition by ATP, NAM, and NAD⁺. SBI-0797812 induced a dramatic left-shift in both K_M and K_I for ATP and this “activator” inhibited both NAMPT activity and NAD⁺ formation in THP-1 cells. Although the K_M for ATP and NAM was not perturbed by N-PAMs, the right-shift in K_I(NAM) results in enhanced enzyme activity at higher cellular concentrations of NAM. Given the observed inhibition of NAMPT by NAM, elevated NAM concentrations (>K_I(NAM)), resulting from accelerated NAD⁺ catabolism in stressed cells, would be expected to induce a paradoxical feedback inhibition of NAMPT. However, N-PAM binding to the rear channel of NAMPT was shown to relieve NAM inhibition, right-shifting K_I(NAM), and promoting NAD⁺ biosynthesis. Importantly, biochemical observations on NAMPT activity and affinity translated to regulation of cellular NAD⁺, measured in THP-1 cells. Taken together, variation in concentrations of ATP, NAM, and NAD⁺ in different cells, coupled with differential dependence for activator classes, hint that NAMPT activators can be designed with cell and possibly tissue selectivity.

The mechanism of N-PAM activation was determined by synthesis of novel activators and an FP-probe. Using multiple co-crystal structures and enzyme kinetics, a mechanism is proposed in which N-PAM binding to the rear channel competes with NAM access to the nucleobase pocket via the rear channel. Maximal activation is induced by an N-PAM with affinity for the rear channel that allows productive (low K_M) NAM binding and blocks non-productive (high K_M) binding, representing the primary contribution of N-PAMs to increased catalytic efficiency (Fig. 8). To account for a secondary contribution from an increase in k_cat, N-PAMs are proposed to stabilize the phospho-enzyme through allosteric binding remote (≈ 20Å) from phospho-His-247 through mechanical coupling to the active site. This composite mechanism is compatible with the observed coupling of ATP utilization to NAM turnover by N-PAMs. Central to the mechanism of activation by N-PAMs and phenolic activators is the rear channel, a unique feature of NAMPT, raising the question as to whether endogenous regulators exist that bind to this structural feature.

Figure 8.

NAMPT activation by N-PAMs occurs when Productive NAM turnover to NMN is preceded by high affinity binding of NAM (A). Non-productive breakdown of the phospho-enzyme is preceded by low affinity binding of NAM (B).

Pharmacological activation of NAMPT is a promising therapeutic strategy to elevate levels of cellular NMN and NAD⁺ in metabolic disorders and diseases of aging, ^{13, 14, 16, 17, 31}. The research presented herein provides a mechanism for pharmacological modulation of NAMPT, emphasizing the dependence on cellular concentrations of ATP, NAM, and NAD⁺, which can be exploited therapeutically.

Abbreviations

ADH

alcohol dehydrogenase

ADRD

Alzheimer’s disease and related dementia

ARTs

ADP-ribosyltransferases

eNAMPT

extracellular NAMPT

FP

fluorescence polarization

MST

microscale thermophoresis

N-PAMs

NAMPT positive allosteric modulators

NAD⁺

nicotinamide adenine dinucleotide

NADP⁺

2-phosphate derivative

NAM

nicotinamide

NAMPT

nicotinamide phosphoribosyl-transferase

NAPRT

nicotinic acid phosphoribosyltransferase

NMA

normal-mode analysis

NMN

nicotinamide mononucleotide

NMNAT

nicotinamide/nicotinic acid mononucleotide adenylyltransferase

PARP

poly-ADP-ribose polymerase

phosphor-H247

N-phosphohistidine

PDB

protein data bank

PRPP

α-D-5-phosphoribosyl-1-pyrophosphate

SARM1

sterile alpha and TIR motif containing 1

SIRTs

sirtuins

TLR4

Toll-Like Receptor 4

T2D

type-2 diabetes