Archive for July, 2018

We then reasoned that the sole conversion of Vacor into VMN could not entirely explain its ability to rapidly induce NAD deple- tion and cell death. Indeed, even GMX1778 is phosphoribosy- lated by NAMPT but, although it was among those that most efficiently reduce cellular NAD content (Beauparlant et al., 2009), it appeared much less active than Vacor in depleting NAD contents and triggering cell death (see Figures 2F and S2C). Furthermore, the ability of NMN to fully prevent NAD deple- tion by FK866 or GMX1778 but not by Vacor was a clear hint that NAMPT inhibition is not the sole mechanism responsible for Vacor-dependent dinucleotide shortage and cell killing. We therefore hypothesized that Vacor and/or VMN also target the metabolic step of NAD resynthesis downstream of phosphoribo- sylation, namely the adenylation reaction brought about by NMNAT1–3 (Berger et al., 2005). We found that Vacor up to 400 nM did not affect the activity of NMNAT1–3 (data not shown), whereas a reaction product appeared when NMNAT2 or NMNAT3 was assayed with VMN and ATP as substrates (Fig- ure 4A). To our surprise, LC-MS analysis of the corresponding HPLC peak demonstrated that it corresponded to VAD (Fig- ure 4B). When compared with the adenylation reaction with NMN as substrate, adenylation of VMN by NMNAT2 and NMNAT3 was 0.6% ± 0.08% and 0.04% ± 0.006%, respectively.

Kinetic analysis of the NMNAT2-catalyzed reactions showed that affinity for VMN was 10-fold lower than that for NMN, whereas catalytic efficiency of VAD formation was about 2,000–fold lower than that of NAD (Figure 4C). We also tested VAD against the three recombinant human NMNATs and found that it had no effects on NMNAT1, whereas it inhibited NMNAT2 and NMNAT3 with IC50s of 20 ± 2 mM and 463 ± 99 mM, respec- tively (Figure 4D).

On this basis, to determine the VAD binding site on hNMNAT2 we took advantage of comparative modeling, since no crystal structures are available. By using docking studies, we found that VAD can fit a putative binding pocket (Figure 4E) similar to that reported for NAD cocrystallized in NMNAT1 (PDB: 1KQN [Zhou et al., 2002]), making a net of hydrogen bonds with the side chain of His24, Tyr50, and Ser265 and with the backbone of Ile9, Leu207, and Asn209 (Figure S3E). Of note, His24 is known to be crucial for enzymatic activity (Yalowitz et al., 2004). Structural comparison between our new model of hNMNAT2 and the chain A of crystallized hNMNAT1 (Zhou et al., 2002) and NMNAT3 (Zhang et al., 2003) was then performed by the superimposition of the corresponding 3D structures. As highlighted in Figure S4A, a sequence not aligned with NMNAT2 or NMNAT3 is present at the C terminus of NMNAT1 that bends over the substrate binding site. Possibly, therefore, this sequence of 16 amino acids (DRNAGVILA PLQRNTA) can interact with the catalytic site, thereby regulating substrate entry or enzyme activity. Of note, the last seven amino acids of this terminal sequence can make direct interactions with the side chain of docked VAD. These results taken together may in part explain selective inhibition of NMNAT2 and NMNAT3, but not of NMNAT1, by VAD formation.

We next investigated the effect of VMN on the NAD-synthesiz- ing activity of NMNAT2 at varying NMN concentrations and at a fixed, saturating ATP concentration. The resulting Linewea- ver-Burk plot was consistent with a mixed-type inhibition (Fig- ure S4B). Replots of slopes and intercepts against inhibitor concentrations generated concave-up curves (Figure S4B), indicating a multiple-site inhibition mechanism by VMN charac- terized by a concomitant binding of more than one inhibitor molecule (Segel, 1993). By analyzing the best fitting of the exper- imental initial rates (see equation in STAR Methods), we found a KM forNMNof14±6mM,aKi ofVMNforthefreeenzymeof 203 ± 25 mM, and a slightly higher Ki of VMN for the enzyme-sub- strate (ES) complex (aKi 244 ± 30 mM). The kinetic model of inhibition is shown in Figure S4C.

In principle, given the low catalytic efficiency of NMNAT2 with VMN as substrate, under homeostatic conditions adenyla- tion of Nicotinamide Mononucleotide should outcompete that of VMN. We therefore investigated whether metabolic transformation of Vacor into VAD occurs in intact cells. Remarkably, a time-dependent increase of VMN and VAD was observed in Vacor-sensitive SH-SY5Y cells, whereas low content of VMN and no VAD was observed in Vacor-insensitive HeLa cells (Figure 5A). Inter- estingly, we found that VMN content was higher and VAD con- tent was lower in M26C cells than in SH-SY5Y cells (Figure S5A) in keeping with the lower sensitivity to Vacor of the former compared with the latter (Figure S5B). We also found that Va- cor prompted a rapid drop in NMN that preceded that of NAD in SH-SY5Y cells, whereas it had no effects on NMN and NAD in HeLa cells (Figure 5B). We reason that in Vacor- exposed cells the rapid drop in NMN content due to NAMPT inhibition by VMN allows VMN adenylation and VAD formation, notwithstanding the lower kcat of NMNAT2 with VMN than with NMN as substrate. Together, these findings corroborated the hypothesis that a key determinant of Vacor cytotoxicity is its two-step metabolic conversion into VAD by NAMPT and NMNAT2. Two additional experiments strengthened this assumption. First, sensitivity to Vacor was associated with a high degree of NMNAT2 but not NMNAT1 expression levels (Figures 5C, 5D, S5C, and S5D). Second, VAD displayed iden- tical cytotoxic effects when added to the medium of SH-SY5Y and HeLa cells (Figure 5E), also suggesting intracellular VAD entrance. To rule out a possible extracellular hydrolysis of VAD akin to NAD (Kulikova et al., 2015), we checked by means of LC-MS for the presence of extracellular VMN or NMN in HeLa cell cultures exposed for 1 hr to VAD or NAD (both at 1 mM), respectively. Upon incubation, we found that extracel- lular NMN increased from 2 ± 1.7 to 9.4 ± 1.2 nmol/mL, whereas VMN was not detectable.