Stepwise activation of SARM1 for cell death and axon degeneration revealed by a biosynthetic NMN mimic

Significance

SARM1 is a NAD⁺ hydrolase that mediates axon degeneration following nerve injury and in neurodegenerative diseases. This study identifies a potent proactivator of SARM1 that is enzymatically converted by nicotinamide phosphoribosyltransferase, a key enzyme in the NAD⁺ biosynthesis pathway, into a direct activator, M1. Structural analyses of SARM1 in complex with M1 reveal two intermediate conformational states, providing a detailed stepwise model of the SARM1 activation process. These findings offer a valuable tool for quantifying SARM1 activation and provide critical mechanistic insights for the development of SARM1 inhibitors as therapeutic agents for neurodegenerative diseases in which SARM1 plays a central role.

Abstract

Axon degeneration, driven by the NAD⁺ hydrolyzing enzyme SARM1, is an early pathological hallmark of numerous neurodegenerative diseases. SARM1 exists in an inactive form and is activated following nerve injury. However, the precise molecular mechanism underlying SARM1 activation remains to be fully elucidated. In this study, we report the identification of a potent proactivator of SARM1, G10, which is converted into a direct activator (M1) by the enzyme nicotinamide phosphoribosyltransferase. Cryoelectron microscopy structures of SARM1 bound to M1, as well as to M1 and a nonhydrolyzable NAD⁺ analog (1AD), captured two intermediate activation states and the fully active state, revealing a stepwise mechanism of SARM1 activation. Further, introducing a disulfide bond to prevent conformational transitions between the two intermediate states mediated by M1 stabilized SARM1 in its inactive form and blocked M1-induced cell death. Together, these findings propose a sequential, stepwise activation model for SARM1 and offer a framework for developing potential SARM1 inhibitors for the treatment of neurodegenerative diseases.

Axon degeneration is a hallmark of many neurodegenerative diseases, where the disintegration of axons disrupts neural communication and connectivity. This process often occurs early in the disease course and precedes the degeneration of the cell body (1–3). A key mediator of axonal degeneration is the sterile alpha and TIR motif-containing 1 (SARM1) enzyme that cleaves NAD⁺ into nicotinamide (NAM) and adenosine diphosphate (ADPR) or cyclic adenosine diphosphate (cADPR) (4–7). SARM1 is kept in an inactive state and becomes activated in response to various neuronal damage signals including metabolic stress of neurons, physical nerve injury, as well as exposure of chemotherapeutical drugs like taxanes, vinca alkaloids, and platinum-based compounds to trigger axonal degeneration (5, 8). Accordingly, deletion or pharmacological inhibition of SARM1 prevents axonal degeneration and neuronal death (5, 9–12). SARM1 is thus a potential therapeutic target for the treatment of many neurodegenerative diseases.

SARM1 is composed of an N-terminal signal peptide for mitochondria targeting and an armadillo repeat (ARM) domain followed by tandem SAM domains and a C-terminal catalytic TIR domain. Before nerve injury, the SAM domains of SARM1 assemble into a central octameric ring, with the ARM and TIR domains forming the peripheral ring, as revealed by recent structural studies (13–18). Interactions between the TIR and the ARM domains prevent the self-oligomerization of the TIR domain, trapping it in an inactive state.

One model of SARM1 activation has proposed that following physical injury, NMN (nicotinamide mononucleotide), a precursor of NAD⁺ in the main NAD⁺ synthesis pathway in most cells, accumulates due to the rapid degradation of the enzyme NMNAT2 (nicotinamide mononucleotide adenylyltransferase 2), the enzyme that catalyzes NMN to NAD⁺, in injured axons, resulting in an increased ratio of NMN to NAD (19, 20). The increased NMN level leads to its binding to the ARM domain of SARM1, disrupting the ARM-TIR interface or creating a steric clash between two adjacent ARM domains, which causes the TIR dislodgement from the ARM domain and subsequent self-association of the TIR domain to form the active site for its NADase activity (18, 21). NAD⁺ competes with NMN for binding to the ARM domain to stabilize the interaction between the ARM and TIR domains, thus keeping SARM1 in the inactive form (15, 17). However, until now SARM1 has only been captured in inactive and fully activated state by structural studies (18), and the absence of intermediate state structures still left the detailed SARM1 activation process to speculation. Moreover, NMN is unable to penetrate the cell membrane, which restricts its usage in investigating SARM1 activation within cells. Although CZ-48, a cell-permeable mimetic of NMN, has been shown to induce SARM1-dependent cell death (13, 22), this molecule fails to trigger axonal degeneration when applied to cultured neurons (14, 23).

Thus, a more potent, cell-permeable SARM1 activator for the study of SARM1 activation as well as the molecular mechanism of axonal degeneration and cell death mediated by SARM1 remains urgently needed. We therefore established a cell-based system and screened a chemical library for such an activator. We have found a small molecule, named G10, that potently induces SARM1-dependent cell death and axonal degeneration. We subsequently uncovered that G10 is metabolized to a direct SARM1 activator, named M1, by NAMPT. The activation mechanism of SARM1 mediated by M1 was further characterized by a detailed cryoelectron microscopic (cryo-EM) analysis of structures of SARM1 bound to M1 as well as M1 plus a nonhydrolyzable NAD⁺ analog.

Results

G10 Induces SARM1-Dependent Cell Death and Neuronal Degeneration.

To investigate the mechanism of SARM1 activation, we conducted a compound screen in HeLa cells ectopically expressing the full-length SARM1 to identify potential SARM1 activators using the known NMN mimics 3AP, CZ-48, and Vacor as positive controls (24, 25). However, Vacor exerted apparent nonspecific cell toxicity, as evidenced by a decline in intracellular ATP in both parental and SARM1-expressing Hela cells (SI Appendix, Fig. S1 A and B). Although the other two NMN mimics, CZ-48 and 3AP showed SARM1-dependent cell toxicity, the effective concentrations were quite high (200 μM and 100 μM, respectively) (SI Appendix, Fig. S1C).

Among 1.36 million compounds, we identified a pyridine-substituted derivative, 7-(4-ethoxyphenyl)-3-((pyridin-3-ylmethyl) thio)-6,7-dihydro-5H-imidazo[2,1-c][1,2,4]triazole, thereafter named G10, as a potent SARM1 activator (Fig. 1A). Treatment of HeLa cells ectopically expressing SARM1 with G10 caused a dose-dependent decline in intracellular ATP levels, with an IC50 at approximately 500 nM. G10 did not show any cell toxicity in parental HeLa cells at this concentration (Fig. 1B). Furthermore, morphological changes in G10-treated SARM1 expressing cells showed similarity to that of CZ-48 treated cells, including cell shrinkage and blistering (SI Appendix, Fig. S1D and Video S1).

Fig. 1.

G10 induces SARM1-dependent cell death and neuronal degeneration. (A) The structure of compound G10. (B) HeLa cells stably expressing SARM1 with Flag and HA tags (HeLa-SARM1-HA-3×Flag) cells and HeLa cells (Control) were treated with G10 at the indicated concentration for 24 h. Cell survival rate was determined by measuring ATP levels. Data are represented as mean ± SD from triplicate wells. (C) NAD⁺ (Left axis) and cADPR (Right axis) levels in HeLa-SARM1-HA-3×Flag cells and HeLa cells were measured at indicated times following treatment with 10 μM G10. Each bar represents mean ± SD, n = 3. (D) Representative images of neurites from WT and Sarm1^−/− DRG neurons treated with indicated concentration of G10. The WT DRG neurons cultured for 7 d in vitro (DIV7) were treated with 0, 25, or 50 μM G10 for 24 h, and Sarm1^−/− DRG neurons were treated with 50 μM G10 for 48 h. (E) Sarm1^−/− DRG neurons were transduced with lentivirus for the expression of human SARM1 at DIV4. Empty vector was used as a control. DRG neurons were treated with 10 μM of G10 4 d after virus transduction. Immunostaining of tubulin in DRG neurons was conducted at 48 h after G10 treatment. (F) Quantification of the axonal degeneration by ImageJ. Each bar represents mean ± SD, n = 3, ****P < 0.0001; n.s., not significant. (G and H) Relative NAD⁺ (G) and cADPR (H) levels were measured at 12 h after treatment with 10 μM of G10, each bar represents mean ± SD, n = 3, ****P < 0.0001; n.s., not significant.

To further confirm that G10-induced cell death is due to SARM1 activation, we measured the levels of the SARM1 substrate, NAD⁺, as well as the SARM1-specific enzymatic product, cADPR, over time in cells following G10 addition (Fig. 1C and SI Appendix, Fig. S1 E and F) (26). Significant NAD⁺ depletion and cADPR production were observed in SARM1-expressing cells, while no such change in NAD⁺ or cADPR was seen in parental cells treated with the same amount of G10. Similarly, treatment with CZ-48 and 3AP also caused similar changes in NAD⁺ and cADPR levels in SARM1-expressing cells (SI Appendix, Fig. S1 E and F).

To test whether G10 induces SARM1-dependent axonal degeneration, we cultured embryonic DRG neurons isolated from wild-type or Sarm1 knockout mice in vitro and treated them with increasing concentrations of G10. Notably, the axons of the wild-type group treated with 25 and 50 μM G10 began to fragment after 24 h, while the axons of the Sarm1 knockout group treated with 50 μM G10 remained intact even after 48 h (Fig. 1D). Furthermore, Sarm1 knockout DRG neurons infected with a lentivirus containing human SARM1 cDNA regained the ability to respond to G10 compared to the vector group (Fig. 1 E and F). Correspondingly, significant NAD⁺ depletion and cADPR production occurred in hSARM1-rescued, but not vector-transfected DRG neurons following the addition of G10 (Fig. 1 G and H). Taken together, G10 is a superior SARM1 activator compared to the previously reported CZ-48 and 3-AP, which either cannot induce axonal degeneration on DRG neurons (CZ-48) or requires much higher concentration to do so (23, 24).

G10 Activation of SARM1 Requires NAMPT.

To investigate the mechanism of G10-induced SARM1 activation, we conducted a gRNA-directed screen to identify genes whose disruption blocked G10-induced SARM1 activation. The gRNA targeting SARM1 was eliminated from the gRNA library beforehand to avoid the inadvertent false-positive results. The screen was conducted after the SARM1-expressing HeLa cells were infected with a lentiviral-encoded gRNA library and then treated with G10. The gRNA sequences from the surviving cells were subsequently decoded through sequencing.

Among the genes against which multiple gRNAs were recovered, we detected NAMPT, the rate-limiting enzyme in the NAD⁺ salvage pathway (Fig. 2A). In this pathway, NAM is first converted into NMN via NAMPT, and NMN is further metabolized into NAM adenine dinucleotide (NAD⁺) by NMNAT. To validate the role of NAMPT in G10-induced SARM1 activation, we used the NAMPT inhibitor FK866 to block NAMPT function. Treatment with both G10 and FK866 effectively prevented G10-induced cell death (Fig. 2B). Although FK866 alone slightly reduced cADPR and NAD⁺ levels compared to the DMSO control, it significantly prevented G10-induced NAD⁺ depletion and cADPR production (Fig. 2 C and D). This inhibitory effect of FK866 on G10-induced SARM1 activation was also observed in DRG neurons, as indicated by axon preservation when FK866 was added together with G10 (Fig. 2 E and F). The reduced NAD⁺ level and cADPR production in DRG neurons following the addition of G10 were also prevented by the presence of FK866 (Fig. 2 G and H).

Fig. 2.

G10 induces cell death via the NAD⁺ biosynthesis salvage pathway. (A) NAMPT appears in the top 100 gRNA hits from a genome-wide CRISPR/Cas9 screen of G10-induced cell death in HeLa-SARM1-HA-3×Flag cells. Fold enrichment-based ranking and the target sequence of each gRNA were listed. (B) Cell survival rate in HeLa-SARM1-HA-3×Flag cells treated with both FK866 at the indicated concentrations and 10 μM G10 for 24 h, determined by measuring ATP levels. Each bar represents mean ± SD from triplicate wells. (C and D) Relative NAD⁺ (C) and cADPR (D) levels in HeLa-SARM1-HA-3×Flag cells measured at 5 h after treatment with 10 μM G10, 100 nM FK866, or both 10 μM G10 and 100 nM FK866. Each bar represents mean ± SD, n = 3, ****P < 0.0001; n.s., not significant. (E and F) Representative images of neurites from WT DRG neurons treated with indicated compounds (E) and quantification of axonal degeneration (F). WT DRG neurons were treated with 10 μM G10, 100 nM FK866, or both 10 μM G10 and 100 nM FK866 for 48 h. Bars represent Mean ± SD, n = 3, ****P < 0.0001; n.s., not significant. (G and H) Relative NAD⁺ (G) and cADPR (H) levels in WT DRG neurons measured at 12 h after treatment with 10 μM G10, 100 nM FK866, or both 10 μM G10 and 100 nM FK866. Each bar represents mean ± SD, n = 3, ****P < 0.0001; n.s., not significant. (I) HeLa-SARM1-HA-3×Flag cells, HeLa-SARM1-HA-3×Flag-NAMPT^−/− cells (referred to as the NAMPT^−/− cells), NAMPT^−/− cells ectopically expressing NAMPT (referred to as the NAMPT-Res cells) were treated with 10 μM G10 for 24 h. SARM1 was immunoblotted with the anti-Flag antibody unless otherwise stated. (J) HeLa cells, HeLa-SARM1-HA-3×Flag cells, and HeLa-SARM1-HA-3×Flag cells ectopically expressing NMNAT2-myc (referred to as the NMNAT2-OE cells) were treated with 10 μM G10 for 24 h. NMNAT2 was immunoblotted using the anti-myc antibody. (K) NMN/NAD⁺ ratio in HeLa-SARM1-HA-3×Flag cells measured at 5 h after treatment with 10 μM G10, 100 nM FK866, or both 10 μM G10 and 100 nM FK866. Each bar represents mean ± SD, n = 3, ****P < 0.0001; n.s., not significant.

The role of NAMPT in G10-induced SARM1 activation was further confirmed by independent NAMPT knockout and rescue experiments. Knockout of NAMPT in HeLa cells expressing SARM1 resulted in a complete blockade of G10-induced cell death. Reintroducing NAMPT cDNA into the NAMPT knockout cells (NAMPT-rescued) successfully restored their sensitivity to G10 (Fig. 2I). Moreover, ectopic expression of NMNAT2 in HeLa cells expressing SARM1, which accelerated NMN conversion to NAD⁺, also successfully blocked G10-induced cell death, suggesting that a G10-based, SARM1-activator was metabolized to an inactive form by NMNAT2 (Fig. 2J). Consistently, the cADPR levels remained at constant low level after G10 treatment in NAMPT knockout cells or NMNAT2-overexpressing cells, while the cADPR level significantly increased in concurrent with NAD⁺ level decrease in the NAMPT-rescued cells (SI Appendix, Fig. S2 A−C). Additionally, neither cellular NMN level nor NMN/NAD⁺ ratio increased in SARM1-expressing HeLa cells treated with G10 (Fig. 2K), further suggesting that the SARM1 activation by G10 was not through indirect manipulation of cellular NMN level but rather the formation of a G10-based SARM1 activator catalyzed by NAMPT.

G10 Is Converted Into a Direct SARM1 Activator by NAMPT.

To further study the G10-based SARM1 activator, we first purified NAMPT from Escherichia coli and immobilized the enzyme onto sensor chips to measure surface plasmon resonance (SPR) changes using a Biacore assay. A series of running buffers with different concentrations of G10 or FK866 (used as a positive control) were flowed over the chip surface. The results indicated that NAMPT bound to both G10 and FK866 in a dose-dependent manner, with their respective dissociation constants in the nanomolar range (SI Appendix, Fig. S2 D and E). We subsequently incubated the purified NAMPT with G10 and phosphoribosyl pyrophosphate (PRPP), followed by HPLC separation of the reaction mixture and mass spectrometry analysis (Fig. 3A). A specific absorbance peak at 6.15 min, in addition to the absorbance peak of G10 at 4.17 min, was observed in the HPLC profile. This product was thereafter referred to as M1. Mass spectrometry showed that M1 has a relative molecular mass of 566.34, consistent with the phosphoribosyl modification of G10 (SI Appendix, Fig. S2F). We subsequently purified the compounds corresponding to the two absorbance peaks and determined their structure using NMR Spectroscopy (Fig. 3B). Some distinctive peaks appeared in the H-NMR resonances of the M1 group compared to the G10 group, with a specific peak at about 6.0 ppm believed to correspond to H¹ of ribose (27). Furthermore, a unique peak appeared in the P-NMR resonances of M1, indicating that a pentaphosphate is covalently attached to G10 (Fig. 3C). Collectively, these results demonstrated that NAMPT catalyzes the conversion of G10 and PRPP to M1 through its phosphoribose transferring activity (Fig. 3D).

Fig. 3.

G10 is converted into NMN mimic (M1) to activate SARM1. (A) The HPLC elution profile of M1 (6.15 min) and G10 (4.17 min) (Top), and mass spectrometry analysis of G10 (Middle) and M1 (Bottom). (B) The ¹H-NMR spectra of M1 and G10. (C) The ³¹P-NMR spectrum of M1. (D) Diagram of SARM1 activation by NAM and G10 metabolism via NAMPT. (E) The amounts of M1 in HeLa-SARM1-HA-3×Flag cells (Left axis), NAMPT^−/− cells (Left axis), NMNAT2-OE cells (Left axis), and NAMPT-Res cells were measured at indicated times following treatment with 10 μM G10. **** highlights statistical significance between HeLa-SARM1-HA-3×Flag cells and NAMPT^−/− cells or NMNAT2-OE cells. Each bar represents mean ± SD, n = 3, ****P < 0.0001. (F) Cell survival rates in HeLa-SARM1-HA-3×Flag cells, NAMPT ^−/− cells, NMNAT2-OE cells, NAMPT-Res cells, treated with 10 μM G10 or 100 μM M1 for 24 h, determined by measuring ATP levels. Each bar represents mean ± SD from triplicate wells. (G) Time course of the base exchange activities of the purified SARM1 incubated with indicated compounds, measured using permeant fluorescent probes (PC6). Each bar represents mean ± SD, n = 3. (H and I) Relative NAD⁺ (H) and cADPR (I) levels in the reaction mixture containing purified SARM1 and indicated compounds that incubated for 45 min. Each bar represents mean ± SD, n = 3, ****P < 0.0001, **P < 0.01.

To further confirm the generation of M1 in cells, we collected metabolites from SARM1-expressing HeLa cells every hour after G10 treatment for UHPLC analysis. We detected M1 production in these cells, whereas no M1 production was observed in NAMPT knockout cells at any point after G10 treatment (Fig. 3E). Furthermore, NAMPT-rescued cells regained the ability to generate M1. Additionally, we observed a significant decrease in M1 production in NMNAT2-overexpressing cells, suggesting that NMNAT2 further converts M1 to an NAD⁺ analog, similar to the conversion of NMN into NAD⁺ (Fig. 3E). Consistently, M1 was detected in the DRG neurons following G10 treatment, while no M1 production was observed after cotreatment with G10 and FK866 (SI Appendix, Fig. S2G).

To confirm that M1 is a true SARM1 activator, we treated SARM1-expressing HeLa cells with increasing concentrations of M1 directly and observed dose-dependent cell death (SI Appendix, Fig. S2H). Moreover, M1 induced cell death in NAMPT-knockout cells, indicating that the need for NAMPT was bypassed (Fig. 3F). Consistently, M1 only caused a 50% intracellular ATP decline in NMNAT2-overexpressing cells compared to cells without NMNAT2 transgene, possibly due to its constant transformation into an NAD⁺ mimic by NMNAT2. In addition to the aforementioned cell-based assay, M1 also directly activated purified SARM1. When purified SARM1 was incubated with M1 and a fluorescent probe PC6 that is able to undergo base exchange catalyzed by SARM1, the addition of M1 significantly increased the PC6 fluorescence, similar to NMN-treated SARM1, and this increase was subdued by a reported SARM1 TIR domain inhibitor, 1AD (Fig. 3G). Consistently, the addition of M1 to purified SARM1 resulted in decreased NAD⁺ levels and increased cADPR levels in the reaction mixture, and these changes were prevented by 1AD (Fig. 3 H and I).

M1 Disrupts the ARM-TIR Secondary Interface and Activates SARM1 by Triggering the Release of the ARM-TIR Domains from the SAM Domain.

To further understand the molecular basis of SARM1 activation by M1, we sought to determine the cryo-EM structures of the full-length SARM1 bound to M1. SARM1 was stably expressed in 293T-SARM1^−/− cells and purified to homogeneity, as determined by size exclusion chromatography (SI Appendix, Fig. S3A). 2D classification and 3D reconstruction revealed that most particles show a central octameric ring of the SAM domains with the ARM and TIR domains missing, whereas a minor portion of particles have the ARM and TIR domains attached to the central ring to form fully assembled octamer (SI Appendix, Fig. S3 B–D). The structures of the fully assembled octamer and the central ring were determined at resolutions of 2.83 and 2.46 Å, and hereafter termed intermediate state 1 (9L2D, EMD-62772) and 2 (9L2E, EMD-62773), respectively (Fig. 4 A–C and SI Appendix, Fig. S3 D–I).

Fig. 4.

Cryo-EM structures of SARM1 in complex with M1. (A and B) Two orthogonal views of the cryo-EM map of M1-bound SARM1 in intermediate state 1 with the model fitted. The SAM, ARM, and TIR domains are colored by green, blue, and magenta, respectively. The EM map of the TIR domain is incomplete due to its high flexibility. (C) Cryo-EM structure of M1-bound SARM1 in intermediate state 2. Only the SAM domains are well resolved. The ARM-TIR domains that are connected to the SAM domain via flexible linkers (dash lines) are dynamic and therefore not visible. (D) The M1 binding site in the ARM domain of SARM1. The EM density for M1 is shown at a contour level of 4σ. Hydrogen bond and salt bridge interactions are represented by dash line. (E) Cell survival rates of HeLa-SARM1^−/− cells ectopically expressing SARM1 WT or mutants treated with 10 μM G10 for 24 h, determined by measuring ATP levels. Each bar represents mean ± SD from triplicate wells. (F) Comparison of the structures of SARM1 in the apo state (colored in gray) and M1-bound SARM1 (SAM in green, ARM in blue, TIR in magenta) in intermediate state 1. (G) Close-up view of the binding interfaces between the TIR and ARM domains. The TIR domain interacts with the two neighboring ARM domains through primary and secondary binding interfaces. (H) Detailed interactions between the ARM and SAM domains in one subunit. (I and J) Effects of mutations in the binding interface of the ARM-TIR domains (I) and ARM-SAM domains (J) on G10-triggered cell death. HeLa-SARM1^−/− cells stably transfected with SARM1 WT or mutants under a Dox-inducible system. Cells were treated with 2.5 μM Dox alone or a combination of 2.5 μM Dox and 10 μM G10 for 24 h. Cell survival rate was determined by measuring ATP levels. Data are represented as mean ± SD from triplicate wells.

As expected, M1 binds to the same position of the ARM domain as NMN. The EM density for the sulfur-linked chemical group (not modeled in the structure) that is attached to the pyridine ring of M1 is invisible, likely due to its high flexibility (Fig. 4D). W103 and two basic residues, R110 and R157, are involved in π–π stacking and salt bridge interactions with the pyridine and phosphate groups of M1, respectively. Consistently, mutations of these residues prevented M1-induced cell death and impaired the host cell NAD⁺ conversion into cADPR in the presence of G10 (Fig. 4E and SI Appendix, Fig. S4 A–E). Notably, the EM density map for the TIR domain in intermediate state 1 is more poorly defined compared to those in the reported inactive states of SARM1 (Fig. 4 A and B and SI Appendix, Fig. S4I). As previously described, the TIR domain in the inactive state interacts with two adjacent ARM domains via two distinct interfaces, referred to as the primary and secondary binding interfaces (Fig. 4 F and G) (25). The primary binding interface contributes most to the binding affinity and mainly involves the ARM4 and ARM5 repeats of the ARM domain that comprises eight ARM repeats (ARM1 to ARM8). In the secondary binding interface, the BB loop that is visible in most inactive structures makes a number of hydrophilic interactions with the ARM2 and ARM3 repeats of the neighboring ARM domain (SI Appendix, Fig. S4J). However, upon M1 binding, the ARM1, ARM2, and ARM3 repeats that are involved in binding M1 move in the opposite direction to the BB loop, thereby disrupting the secondary binding interface, which possibly accounts for the poorer EM density of the TIR domain in intermediate state 1. In contrast, no conformational changes are observed in the ARM4-ARM8 repeats that make no direct contact with M1. Since the TIR domain bridges the two adjacent ARM domains that wrap around the SAM domain and stabilizes their inactive state, disruption of the secondary binding interface would likely cause subsequent dissociation of the ARM-TIR domains from the SAM domains, resulting in intermediate state 2. In this state, the ARM domain remains associated with the TIR domain, but becomes more dynamic due to loss of direct contact with the central SAM domain, which explains why only SAM domain is visible (Fig. 4C).

To further verify that the disruption of the secondary interface and the dissociation of the ARM domain from the SAM domain is critical for SARM1 activation, we introduced mutations that directly disrupted the ARM-SAM interface and assessed their effects on SAMR1 activation. Given the potential constitutive activity of the SARM1 mutants, we performed the cell viability assay using a doxycycline (Dox)-inducible expression system. In the secondary binding interface, K173 and two other residues N170, Q134 in the ARM3 potentially engage in salt-bridging interactions and hydrogen bonding with E604 or D605 and A600 or K602 of the BB loop (SI Appendix, Fig. S4L). The SAM-ARM interface primarily involves hydrophobic interactions among Y380, I405, and F476 as well as salt -bridging interactions between R376 and two acidic residues, E469 and E472 (Fig. 4H). Upon Dox treatment, and in the absence of G10 treatment, the expression of SARM1 mutants with alanine mutations at both K173 and N170 (referred to as the 2AA) as well as Q134 resulted in cell death (Fig. 4I). This observation supports the importance of disrupting the secondary binding interface in SARM1 activation. Consistent with previous findings (16), expression of the Y380A mutant caused immediate cell death without G10 treatment (Fig. 4J). The cell death was associated with a decrease in cellular NAD⁺ and an increase in cADPR, indicating constitutive activation of SARM1 enzyme (SI Appendix, Fig. S4 G–I). In contrast, the F476A or R376A mutation had a minimal effect, suggesting these residues play a minor role in stabilizing the SAM-ARM interface (Fig. 4J and SI Appendix, Fig. S4 J and K). Collectively, these results indicate that SARM1 activation by M1 disrupts the secondary binding interface that is important for the release of the ARM-TIR domains from the SAM domain.

Introducing a Disulfide Bond in the ARM-SAM Interface Inhibits SARM1 Activation by M1.

To further confirm that SARM1 activation by M1 is caused by the release of the ARM domain from the SAM domain but not the TIR domain from ARM (18), we introduced a disulfide bond near Y380 between the ARM and SAM domains to stabilize their interaction while minimizing impact on the ARM-TIR interface. We chose a pair of spatially close residues, S379 and Q423, and mutated them to cysteine for potential formation of a disulfide bond. Mutating either S379 or Q423 to cysteine alone did not influence the G10-induced SARM1 activation (Fig. 5A and SI Appendix, Fig. S5 A and B). However, cysteine mutations at both Q423 and S379 (referred to as the 2Cmut) blocked G10-induced cell death, even with G10 concentration reaching 1 mM, and no change in the levels of NAD⁺ and cADPR were observed following G10 treatment (Fig. 5A and SI Appendix, Fig. S5 C and D). Additionally, we purified the SARM1 2Cmut to analyze its activity in vitro. The 2Cmut exhibited considerably lower base exchange activity compared to the wild type upon addition of M1 or NMN (Fig. 5B and SI Appendix, Fig. S5E). While both M1 and NMN activated SARM1 to convert NAD⁺ into cADPR in a dose-dependent manner, they had minimal effect on the activity of SARM1 2Cmut (Fig. 5 C and D). To further confirm the formation of the disulfide bond and the stabilization of the inactive state by this disulfide bond, we determined the cryo-EM structure of SARM1 2Cmut in the presence of M1 (9L2F, EMD-62774, SI Appendix, Fig. S6 A–F). Indeed, more particles from the 2Cmut display fully assembled octameric rings compared to the WT (SI Appendix, Fig. S6C). Furthermore, the structure also exhibits a clearly defined TIR domain, reminiscent of the inactive state that is characterized by the stabilization of the secondary binding interface (Fig. 5 E–G). The EM density for the disulfide bond is clearly resolved (Fig. 5H), validating the formation of the disulfide bond in the structure. Therefore, the introduced disulfide bond between the SAM and ARM domains blocked M1-mediated SARM1 activation by stabilizing the binding interface between the two domains and preventing conformational transitions from intermediate state 1 to intermediate state 2.

Fig. 5.

Introducing a disulfide bond between the ARM and SAM domains inhibited M1 or NMN-mediated SARM1 activation. (A) Cell survival rates in HeLa-SARM1^−/− cells (Vector) and HeLa-SARM1^−/− cells ectopically expressing SARM1 or SARM1 cysteine mutants, treated with 10 μM G10 for 24 h, determined by measuring ATP levels. Data are represented as mean ± SD from triplicate wells. (B) Time course of the base exchange activities of the purified SARM1 (WT) or SARM1 (2Cmut) incubated with DPBS or M1. Each bar represents mean ± SD, n = 3. ****P < 0.0001, n.s., not significant. (C and D) Relative NAD⁺ (C) and cADPR (D) levels in the reaction mixture containing purified SARM1 WT (Top) or 2Cmut (Bottom) incubated with increasing concentrations of NMN or M1 for 30 min. Each bar represents mean ± SD, n = 3. (E) Cryo-EM map of SARM1 2Cmut in complex with M1. The model is fitted in the map. (F) Comparison of structures of SARM1 WT (colored in yellow) and 2Cmut (colored by domains) in complex with M1. (G) Close-up view of the secondary binding interface. A portion of the BB loop (e.g., D605 and E604) in the structure of 2Cmut is resolved through interactions with the ARM domain (e.g., K173). (H) EM density of the disulfide bond in the 2Cmut at a contour level of 4σ.

The TIR Domains from Multiple SARM1 Octamers That Oligomerize Into a Superhelical Structure.

Previous studies have shown that the TIR domains dissociate from the ARM domains and oligomerize to form multiple functional catalytic sites (18). The reported cryo-EM structure of SARM1 in complex with NMN and a nonhydrolyzable NAD⁺ analog, 1AD revealed that two antiparallel strands of the TIR domains assembled above the octameric SAM domain ring (SI Appendix, Fig. S8A). Both NAD⁺ binding and two-stranded TIR formation are essential for SARM1 activation by M1, as mutations of E642, H685, and Y687, key residues involved in NAD⁺ binding and two-stranded TIR formation, prevented G10-induced cell death (SI Appendix, Fig. S8 B and C). The effects of W662A or Q688A, however, were modest, likely due to their minor roles.

Disruption of the primary binding interface between ARM and TIR domains is likely caused by NAD⁺ binding, since M1 binding has little effect on this interface (Fig. 4G). NAD⁺ binding induces structural arrangement of the BB loop, which possibly weakens the primary binding interface (SI Appendix, Fig. S8D). In addition, the recruitment of the adjacent TIR domain upon NAD⁺ binding causes steric clash with the ARM domain, leading to the dissociation of the TIR domain from the ARM domain. To verify the formation of the two stranded TIR domain during SARM1 activation by M1, we determined the cryo-EM structure of SARM1 in complex with M1 and 1AD (SI Appendix, Fig. S7 A–E). 3D reconstruction and classification yielded two classes in the activated state (SI Appendix, Fig. S7C). One resembles the reported structure of SARM1 in complex with NMN and 1AD with a visible SAM-ARM octameric ring and poorly defined two-stranded TIR domains ( 9L2F, EMD-62774, SI Appendix, Fig. S8 E and F) (18). The other reveals a larger assembly comprising 16 resolved TIR domains, and two poorly defined octameric ARM-SAM rings are attached to the convex surface of the curved TIR domains (Fig. 6A). Larger assemblies with varying number of SARM1 octamers were observed in the negative staining (Fig. 6B). Based on the structural modeling and negative staining, SARM1 could form a superhelical structure with approximately seven SARM1 octamers per turn (Fig. 6 B and C). Superhelical structure has also been observed in a TIR-SAVED effector of a bacterial antiviral defense system, and is required for activation of the TIR domain (28). Therefore, it is possible that SARM1 activation may involve high-order oligomerization that concentrates SARM1 and enhances catalytic efficiency for rapid axon degeneration under certain stimulations.

Fig. 6.

TIR domains from multiple SARM octamers form a superhelical structure in the presence of M1 and 1AD. (A) Cryo-EM map of SARM1 in complex with M1 and 1AD in three different orientations. Two SARM1 octamers are resolved in the map. The curved two-stranded TIR domains and the SAM-ARM ring are docked into the map. (B) Representative negative staining image and 2D class averages of SARM1 in complex with M1 and 1AD. (C) Modeling of the superhelical structure of SARM1 in the presence of M1 and 1AD based on the experimental model in (A). (D) Proposed model for the activation mechanism of SARM1. In the apo state, the TIR domain interacts with the two neighboring ARM domains via the primary and secondary binding interface. M1 binding to the ARM domain disrupts the secondary binding interface (intermediate state 1), leading to subsequent dissociation of the ARM-TIR domains from the SAM domain (intermediate state 2). Upon 1AD or NAD⁺ binding, conformational changes in the TIR domain disrupt the primary binding interface. The TIR domains can interact with one another to form an antiparallel two-stranded structure.

Discussion

In this study, we identified a potent SARM1 proactivator (G10) through a high-throughput screening assay. G10 is converted into the bona fide activator (M1) in cells by the intracellular NAMPT that catalyzes the transfer of the phosphoribosyl group from PRPP to G10. Although NMN is a potent activator of SARM1, its poor membrane permeability limits its use. Molecules like CZ-48 are cell permeable after the phosphate group of NMN is modified, but such a modification significantly reduces its potency. Our studies provided a more potent SARM1 activator that allows the study of the downstream biochemical events leading to axonal degeneration and cell death mediated by SARM1 easier. Moreover, such an SARM1 activator should be a valuable tool for measuring the pharmacodynamics of SARM1 inhibitors in the clinic by applying G10 to the patients’ blood.

The cryo-EM study of SARM1 bound to M1 revealed two intermediate states in SARM1 activation. Our studies, together with previous structural studies, provide a more comprehensive model that can explain the stepwise activation process triggered by NMN or its mimics (Fig. 6D). In the inactive state or apo-state, SARM1 forms a central octameric ring of the SAM domains that are surrounded by the ARM domains. The TIR domain stabilizes the interactions, which consists of both the primary and secondary binding interfaces, between the two neighboring ARM domains. In the presence of M1, we were able to capture two intermediate states of SARM1 activation. The overall structure in intermediate state 1 resembles that in the inactive state with small conformational changes occurring in the ARM domain. M1-induced conformational changes in the ARM domain disrupt the secondary binding interface, leading to the subsequent dissociation of the ARM-TIR domains from the SAM domains (Fig. 4C). The resulting intermediate state 2 has been observed in SARM1 without NMN bound, indicating the low energy barrier between the inactive state and intermediate state 2, which may account for the basal activity of SARM1 in the absence of activators (Fig. 5B) (16, 17). Introducing a disulfide bond between the ARM and SAM domains prevented M1 or NMN-mediated SARM1 activation, emphasizing the crucial role of dissociation of the ARM domain from the SAM domain during SARM1 activation. Upon NAD⁺ (or 1AD) binding, conformational changes in the BB loop and other residues of the TIR domain involved in binding the ARM domain disrupt the primary binding interface, causing the dissociation of the TIR domain from the ARM domain. The TIR domains then self-associate to form two antiparallel strands and catalyze NAD⁺, while the ARM domain reconnects with the SAM domain to form a more compact SAM-ARM ring (SI Appendix, Fig. S8G). The compact SAM-ARM ring lacking the TIR domain was not observed in the M1 bound SARM1 (SI Appendix, Fig. S3D), further indicating that 1AD or NAD⁺ rather than M1 or NMN plays an important role in the dissociation of the TIR domain from the ARM domain.

Furthermore, the TIR domains can potentially form a superhelical structure that concentrates SARM1. The functional significance of the condensation formation of SARM1 is discussed in detail in our companion paper (29). SARM1 inhibitors have garnered significant research attention as a promising therapeutic strategy for various neurodegenerative diseases. The primary focus has been on developing inhibitors that target the NADase of the TIR domain. Our structural analysis of the inhibitory effect of disulfide bond-mutated SARM1 suggests an alternative strategy to develop inhibitors to prevent SARM1 activation by preventing the release of the ARM domains from the SAM domain.

Methods

Plasmids.

pWPI vector (GFP tagged), pLenti-EF1α-teton-TRE3GS vector (with hygromycin selection marker), psPAX2, and pMD2.G were kept in our lab. Full-length cDNA for human SARM1 (a kind gift from Sironax Inc.) was subcloned into the pGEM-T vector. SARM1 mutants were generated using Quickchange Site-Directed Mutagenesis. The full-length cDNAs of human SARM1 (WT and mutants) were inserted into the pWPI vector to generate the pWPI-SARM1 (WT and mutants) HA-3×Flag and pWPI-SARM1-Flag construct. In the inducible system, the cDNA of SARM1 (WT and mutants) were inserted into the pLenti-EF1α-teton-TRE3GS vector to generate pLenti-EF1α-teton-TRE3GS-SARM1 (WT and mutants) HA-3×Flag construct.

Full-length cDNA for human NAMPT and NMNAT2 were PCR-amplified from HeLa cDNA using phanta polymerase (Vazyme) and were subcloned into the pGEM-T vector. Subsequently, the full-length cDNAs for human NAMPT were inserted into pWPI vector (GFP-tagged) and pET28a to generate the pWPI-NAMPT construct and pET28a-NAMPT-His. The full-length cDNAs for human NMNAT2 were inserted into pWPI vector (GFP-tagged) to generate the pWPI-NMNAT2-myc construct.

Two gRNAs targeting SARM1 and three gRNAs targeting NAMPT were subcloned into the pX458 vector to generate the PX458-GFP-SARM1-1/2 and PX458-GFP-NAMPT-1/2/3 constructs.

Virus Packaging.

To prepare the virus, HEK293T cells in 10-cm dishes were transfected with 10 μg of pWPI-SARM1(WT and mutants: W103A, R110A, R157A, E642A, W662A, H685A, Y687A, Q688A, S379C, Q423C, 2Cmut)-HA-3×Flag, pWPI-NAMPT, pWPI-NMNAT2-myc, and pLenti-EF1α-teton-TRE3GS-SARM1(WT and mutants (2AA, Q134A, R376A, Y380A, F476A))-HA-3×Flag together with 6.4 μg of psPAX2 and 3.6 μg of pMD2G using Lipofectamine 3000 (Thermo Fisher) following the manufacturer’s instructions. Eight hours after transfection, the medium was changed to fresh medium. Another 40 h later, the medium was filtered through a 0.22-μm membrane and concentrated by centrifugation at 4,000 rpm for 20 min. The lentiviral particles were resuspended in DMEM or Neurobasal medium before infecting cells or neurons.

Cell Culture and Stable Cell Lines.

Human cell lines HeLa (ATCC® CCL-2) and HEK293T (ATCC® CRL-11268) were cultured in DMEM (Gibco) with 1% Penicillin-Streptomycin Solution (Gibco) and 10% FBS (Gibco) at 37 °C with 5% CO₂. On the first day, cells were seeded in 6cm dishes (about 0.5 × 10⁵ cells). On the second day, HeLa cells expressing Cas9 were infected with SARM1-HA-3×Flag (30). HeLa-SARM1-HA-3×Flag cells were infected with lentivirus containing NMNAT2-myc. HeLa-SARM1-HA-3×Flag-NAMPT^−/− cells were infected with lentivirus containing NAMPT. HeLa-SARM1^−/− cells were infected with lentivirus respectively containing SARM1 (WT or mutants (W103A, R110A, R157A, E642A, W662A, H685A, Y687A, Q688A, 2AA, Q134A, R376A, Y380A, F476A, S379C, Q423C, S379C/Q423C(2Cmut))-HA-3×Flag.

293T- SARM1^−/− cells were infected with lentivirus containing SARM1-Flag or SARM1(2Cmut)-HA-3×Flag. The day after infection, the lentivirus-containing medium was replaced with fresh medium. Four days later, the positive cells were filtered. GFP-positive live cells were sorted using a BD FACSAria II cell sorter to establish HeLa cell lines expressing SARM1 (WT and mutants (W103A, R110A, R157A, E642A, W662A, H685A, Y687A, Q688A, S379C, Q423C, 2Cmut))-HA-3×Flag, HeLa-SARM1-HA-3×Flag-NAMPT^−/− expressing NAMPT, HeLa-SARM1-HA-3×Flag expressing NMNAT2-myc and 293T-SARM1^−/− expressing SARM1-Flag and SARM1(2Cmut)-HA-3×Flag. For dox (Sigma-Aldrich) inducible HeLa cell line expressing SARM1 (WT and mutants (2AA, Q134A, R376A, Y380A, F476A)-HA-3×Flag, positive cells survived under 500 µg/mL hygromycin B (Invivogen) treatment.

Generation of SARM1 and NAMPT Knockout Cells with CRISPR/Cas9.

6 µg of the pX458-GFP-SARM1-1/2 plasmid mix was transfected into 1 × 10⁷ HeLa cells and 293T cells, 6 µg of pX458-GFP-NAMPT-1/2/3 plasmid mix was transfected into 1 × 10⁷ HeLa-SARM1-HA-3×Flag cells using the Invitrogen™ Lipofectamine™ 3000 Transfection Reagent, following the manufacturer’s instructions. Four days posttransfection, GFP-positive live cells were sorted using a BD FACSAria II cell sorter. Two days later, HeLa-SARM1^−/− and 293T- SARM1^−/− cells were dissociated using 0.25% trypsin (Gibco), enumerated, and diluted. Subsequently, the single clones were seeded into 96-well plates and cultured for 14 d or longer, as required by the cell growth kinetics. Notably, GFP-positive live cells from the HeLa-SARM1-HA3×Flag line with NAMPT knockout became fragile and exhibited slower growth. HeLa-SARM1-HA-3×Flag-NAMPT^−/− cells were utilized as a cell pool. The HeLa-SARM1^−/− and 293T-SARM1^−/− cell lines were confirmed by DNA sequencing and western blot analysis using an SARM1 antibody. The HeLa-SARM1-HA-3×Flag-NAMPT^−/− cell line was verified by DNA sequencing and western blot analysis using a NAMPT antibody.

DRG Neuron Culture.

Mouse dorsal root ganglia were dissected at embryonic day 13.5 into L15 medium (Gibco) with 1% Penicillin-Streptomycin Solution (Gibco) and dissociated in 0.25% trypsin at 37 °C for 20 min. The digestion was stopped by adding FBS (Gibco). The DRG neurons were washed three times with DRG culture medium after dissociation. The DRG neuron medium consisted of Neurobasal medium (Thermo Fisher) supplemented with 2% B27 (Thermo Fisher), 50 ng/mL 2.5S NGF (Thermo Fisher), 1% Penicillin-Streptomycin Solution (Gibco), 0.5 mM L-Glutamine (Thermo Fisher), 10 μM 5fluoro-2′-deoxyuridine (Thermo Fisher), and 10 μM uridine (Sigma). Cells were plated as spot cultures in 24-well tissue culture plates (Corning) that were precoated with 100 μg/mL poly-D-lysine (Thermo Fisher) and 5 μg/mL laminin (Corning). The culture medium was replaced with fresh DRG neuron medium every 2 d.

SARM1 Antibody.

The mouse monoclonal antibody against mSARM1 was developed by ABclonal (A23440) and screened by western blot analysis.

Live-Cell Imaging.

HeLa-SARM1-HA-3×Flag cells were cultured in 35 nm glass-bottom culture dishes (MatTek), plated in a 37 °C imaging station chamber, and treated with 10 μM G10. Images were obtained with a Nikon Ti inverted microscope and the PerkinElmer UltraVIEW VoX system. Time-lapse images were captured at 5-min intervals using a Nikon PLAN APO 60× oil objective. The images were processed using Volocity software (Videos S1).

The WT and Sarm1^−/− DRG neurons were plated as spot cultures in 24-well tissue culture plates and subjected to the indicated compound (25 or 50 μM G10) treatments at DIV7. Static bright-field images of WT or Sarm1^−/− DRG neurons were captured at DIV7, DIV8, and DIV9 using NIKON-SIM and processed in NIS-SIM software.

Immunofluorescence.

HeLa-SARM1-HA-3×Flag cells were seeded in 35-nm glass-bottom culture dishes (MatTek) at about 30% confluence. The cells were coimmunostained with mouse anti-SARM1 and rabbit anti-Tom20 (1:1,000 dilution for each antibody) after 8 h of 10 μM G10 treatment. The morphology of SARM1-dependent cell death was captured using a ZEISS 800 60× oil objective. The images were processed in ZEN software.

WT and Sarm1^−/− DRG neurons were plated as spot cultures in 24-well tissue culture plates. Sarm1^−/− DRG neurons were infected with either SARM1 or vector lentivirus at DIV4 and then were treated with 10 μM G10 at DIV8 for 48 h. WT DRG neurons were treated with the indicated compounds (10 μM G10, 100 nM FK866, or a combination of 10 μM G10 and 100 nM FK866) at DIV7 for 48 h. Two days after the initiation of the indicated compound treatments, the DRG neurons were coimmunostained with rabbit anti-α-tubulin antibody, diluted 1:1,000. Images were captured using a ZEISS 800 microscope with a 20× oil objective and processed with ZEN software. The image data presented are representative of at least three randomly selected fields.

The HeLa-SARM1-HA-3×Flag cells or WT and Sarm1^−/− DRG neurons treatment was terminated by washing with DPBS followed by fixation in freshly prepared 4% paraformaldehyde in DPBS for 30 min. The fixed cells or neurons were washed three times with DPBS and then permeabilized in 0.7% Triton X-100 in DPBS for 30 min. The cells or neurons were then blocked for 2 h in blocking buffer (0.7% Triton X-100 and 10% goat serum in DPBS). The primary antibodies were diluted in blocking buffer, and the cells or neurons were immunolabeled at 4 °C overnight. On the second day, the cells or neurons were washed three times with DPBS followed by a 1-h incubation with a fluorescein-conjugated secondary antibody (Donkey anti-Mouse Alexa Fluor 555 for SARM1, Donkey anti-Rabbit Alexa Fluor 647 for Tom20, Donkey anti-Rabbit Alexa Fluor 555 for α-tubulin). DAPI (Invivogen) was diluted in blocking buffer, and HeLa-SARM1-HA-3×Flag cells were immunolabeled for 10 min, followed by three washes with DPBS. Duplicate cultures were examined, and similar results were obtained in at least three independent experiments.

Cell Survival Assay.

Cell survival was assessed using the Cell Titer-Glo Luminescent Cell Viability Assay kit (Vazyme) according to the manufacturer’s instructions. Luminescence was recorded with a Tecan GENios Pro plate reader. 3AP and Doxycycline was purchased from Sigma-Aldrich, Vacor from Greyhound Chromatograph, CZ-48 from MCE, and FK866 from Selleck. Additionally, G10 was a gift from Sironax.

Quantification of Axon Degeneration.

Three imaging fields of each well were randomly selected to calculate axonal degeneration indexes. The axon degeneration index was calculated as the ratio of fragmented axon areas. The particle circularity was limited to between 0.2 and 1 (size: pixel² >20).

Compound Screening.

Each well of a 384-well assay plate was seeded with 3000 Hela-SARM1-HA-3×Flag cells 1 d before compound treatment. Two chemical libraries, containing approximately 65,000 and 71,000 compounds respectively, were delivered into each well at a final concentration of 10 μM. Cell viability was determined by the Cell Titer-Glo assay 24 h later.

Genome-Wide CRISPR/Cas9 Screen.

The gRNA library was purchased from Addgene (#1000000048). For library amplification, 1 μg of library DNA (10 ng/μL) was used to transform 25 mL of electrocompetent E. coli (Takara). Transformed colonies were scraped off for plasmid extraction using the Plasmid PlusMax Kit (QIAGEN). For virus library preparation, HEK293T cells in a 15-cm dish were transfected with 25 μg of library DNA together with 15 μg of psPAX2 and 10 μg of pMD2G. Eight hours after transfection, the media were changed to high-serum DMEM (20% FBS with 25 mM HEPES). Another 40 h later, the media containing the library lentivirus were filtered through a 0.22-μm membrane, and aliquots of 30 mL were stored at 37 °C.

Infection experiments were preconducted on target cells cultured in 12-well plates, with the virus titer adjusted to a multiplicity of infection (MOI) of 0.3, before initiating large-scale screening assays. For the large-scale screen, HeLa-SARM1-HA-3×Flag cells stably expressing the Cas9 protein were seeded in a 15-cm dish (2 × 10⁶ cells per dish) on the first day and infected with the gRNA lentivirus library on the second day. Four days after infection, the culture medium was exchanged with fresh medium containing 10 μg/mL puromycin (MCE) to eliminate noninfected cells. Ten days later, the infected cells were treated with 10 μM G10 to trigger cell death. Surviving cells were collected after growing to near 90% confluence. Genomic DNAs of each group of cells were prepared using QuickExtract (Epicentre).

The gRNAs were amplified by a two-step PCR method using the Titanium Taq DNA polymerase (Clontech Laboratories). For the first PCR, approximately 50 μg of genomic DNA template was used with the forward primer 50 bp-F and the reverse primer 50 bp-R. The PCR program was as follows: 94 °C for 3 min, followed by 16 cycles of 94 °C for 30 s, 60 °C for 10 s, and 68 °C for 25 s, with a final 2-min extension at 68 °C. For the second PCR, 1 µL of the products from the first-step PCR were amplified with the forward primer Index-F and one of the reverse primers (Index-R1 to R9). The PCR program was as follows: 94 °C for 3 min, 18 cycles of 94 °C for 30 s, 54 °C for 10 s, and 68 °C for 18 s, with a final 2-min extension at 68 °C. The final products of the second-step PCRs were purified for sequencing (HiSeq2500, Illumina).

NAMPT Protein Purification in E. coli.

The pET-28a-hNAMPT plasmid was introduced into BL21 competent cells, which were then cultured until the optical density at OD 600 reached 0.5. Protein expression was induced with 1 mM IPTG (Solarbio) at 28 °C overnight. Following induction, cells were harvested by centrifugation at 4,000 rpm for 30 min, and the cell membranes were disrupted via sonication. The lysate was clarified by centrifugation at 20,000×g for 30 min, and the supernatant was filtered through a 0.22-μm membrane.

Protein purification was performed using Ni-NTA agarose (Beyotime). The agarose beads were packed into a gravity flow chromatography column and equilibrated with 50 mM Tris-HCl (pH 8.0), 250 mM NaCl, and 25 mM imidazole, using approximately three column volumes of this buffer. The supernatant was applied to the column twice to ensure adequate protein binding. The beads were then washed three times with the equilibration buffer, and the target protein was eluted using 250 mM imidazole in 50 mM Tris-HCl (pH 8.0) and 250 mM NaCl.

SARM1 Protein Purification from Mammalian Cells.

293T-SARM1^−/− cells stably expressing SARM1 (WT) or SARM1 (2Cmut) proteins were cultured in 15 cm dishes until approximately 90% confluence. Cells were harvested by scraping, followed by centrifugation at 1,000×g for 10 min. The cell pellets were washed three times with DPBS and rapidly frozen in liquid nitrogen for immediate processing or stored at 80 °C. Upon thawing, cells were resuspended in buffer A (DPBS with an additional 150 mM NaCl, 18% glycerol, and 1 mM DTT) and homogenized using a hand-held homogenizer. The lysates were clarified by centrifugation at 18,000×g for 20 min, and the supernatant was filtered through a 0.22μm membrane. The supernatants were incubated overnight at 4 °C with preequilibrated anti-Flag affinity gel (Beyotime). The immunoprecipitates were washed three times with buffer A and eluted with 0.5 mg/mL 3×Flag antigenic peptide in buffer A for 6 h at 4 °C. The eluted fractions were concentrated and further purified by gel filtration chromatography using a Superose-6 10/300 GL column (Cytiva), equilibrated with DPBS containing an additional 150 mM NaCl and 1 mM DTT. The protein peaks corresponding to SARM1 (WT) or SARM1 (2Cmut) were collected, concentrated, and prepared for analysis. For SARM1 (2Cmut) purification, DTT was omitted from all buffers. Proteins were freshly prepared and used throughout the study.

Production of M1 and 1AD.

First, 0.2 µg/µL NAMPT was incubated with 10 mM G10, 1 mM ATP (MCE), 20 mM MgCl_2, and 500 µM PRPP (Sigma-Aldrich) in 50 mM Tris-HCl buffer (pH 7.5) overnight at 37 °C. After incubation, the mixture was centrifuged, and the supernatant was analyzed by high-performance liquid chromatography (HPLC). The lyophilized aqueous phase was subjected to liquid chromatography–mass spectrometry (LC/MS) and NMR spectroscopy for the identification and purification of M1 by the Chemistry Center of the National Institute of Biological Sciences.

Similarly, 0.1 µg/µL SARM1 (WT) was incubated with 500 µM DSARM (MCE) and 500 µM NAD⁺ (MCE) in DPBS buffer containing an additional 150 mM NaCl and rotated at room temperature overnight. The insoluble protein was precipitated by centrifugation at 18,000×g for 20 min. The supernatant was then loaded onto HPLC for 1AD purification, carried out by the Chemistry Center of the National Institute of Biological Sciences.

Samples preparation and purity analysis were conducted on Waters HPLC (Column: XBridge C18, 5 μm, 19 × 150 mm) with 2,998 PDA and 3,100 MS detectors, and Waters UHPLC (Column: BEH C18, 1.7 μm, 2.1 × 50 mm) with PDA and SQD MS detectors, using ESI as ionization. ¹H and ³¹P NMR spectra were recorded on Varian Inova-400 spectrometers. Data for 1H NMR spectra are reported relative to CDCl3 (7.26 ppm), CD3OD (3.31 ppm), or DMSO-d6 (2.50 ppm) as an internal standard and are reported as follows: chemical shift (δppm), multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, sept = septet, m = multiplet, br = broad), coupling constant J (Hz), and integration.

SPR.

Kinetic binding assays for NAMPT and G10 or FK866 (Selleck) were conducted using a Biacore T200. NAMPT protein was prepared in 10 mM sodium acetate buffer (pH 5.5) at a final concentration approximately 50 μg/mL and immobilized onto a CM5 sensor chip (Cytiva, 29104988) using an amine coupling kit (Cytiva, BR1000-50). Approximately 18,000 RU of NAMPT protein was captured on the CM5 sensor chip. A range of G10 and FK866 concentrations (3.12 to 100 nM) was prepared in running buffer (50 mM Tris-HCl, pH 8.0, 250 mM NaCl). These were sequentially injected into both the sample and reference channels for 60 s to allow association, followed by a 180 s dissociation phase with running buffer. The equilibrium dissociation constant (KD) was calculated using the Biacore T200 Evaluation Software.

Base Exchange Activation Assay by PC6 Fluorescence.

All reactions were performed in a DPBS buffer (Dulbecco’s PBS contained an additional 150 mM NaCl, pH 7.4) with 1 μg SARM1 (WT) or SARM1 (2Cmut) protein, 500 μM NAD⁺, and 10 μM PC6 (a kind gift from Sironax) in a total volume of 50 μL per reaction system. The 1AD concentration was at 1 μM and NMN was purchased from MCE Company. The PC6 fluorescence was read by ELISA reader (TECAN) every minute as soon as the protein was added into the reaction.

Metabolite Extraction for UHPLC Analysis.

For the cADPR and NAD⁺ measurement, cell pellets were washed three times with DPBS, and metabolites were extracted by adding 70% acetonitrile (ACN) and incubating on ice for at least 1 h. The supernatants were diluted as necessary and analyzed using ultra-high-performance liquid chromatography (UHPLC). For protein quantification, cell pellets were dissolved in 0.1 M KOH and analyzed using the Bradford assay (Bio-Rad).

Chromatographic separation was performed using a Waters UHPLC system and an Atlantis Premier BEH C18 AX column (2.5 µm, 2.1 × 100 mm). The mobile phases were (a) 5 mM ammonium acetate in water with 0.1% formic acid and 5 µM medronic acid, and (b) 5 mM ammonium acetate in 95% acetonitrile with 0.1% formic acid. The mass spectrometer, an AB Sciex 6500 Plus, was operated in multiple reaction monitoring (MRM) mode at unit resolution in positive ionization mode.

For the NMN/NAD⁺ ratio measurement, the cell pellets were washed three times with DPBS, and metabolites were extracted using a mixture of MeOH:ACN:H₂O (4:4:2) and incubated at −80 °C for at least 1 h. Following the polar metabolite extraction, the cell pellets were centrifuged at 18,000×g for 20 min. The supernatants were then incubated at –80 °C for at least 1 h, followed by a second centrifugation at 18,000×g for 20 min. The supernatants were vacuum-dried, resuspended appropriately, and analyzed using UHPLC. Protein content in the cell pellets was dissolved in 0.1 M KOH and quantified using the Bradford assay (BIO-RAD).

LC−MS/MS analysis was conducted using an Agilent 1290 Infinity UHPLC system coupled with an Agilent 6495 triple quadrupole mass spectrometer (Agilent Technologies, Santa Clara, CA), equipped with an electrospray ionization source. Chromatographic separation was achieved with a Poroshell 120 HILIC-Z column (2.1 mm × 100 mm, 2.7 µm; Agilent Technologies, Santa Clara, CA). The mobile phases employed were (a) 25 mM ammonium acetate and 25 mM ammonia in water, and (b) acetonitrile. The mass spectrometer was operated in MRM mode at unit resolution in positive ionization mode. Data acquisition and processing were performed using Agilent MassHunter Quantitative Analysis B.07.00 software.

The protocol for M1 measurement in stable cell lines is identical to that used for cADPR and NAD⁺ quantification. Additionally, the M1 measurement method in primary DRG neurons is conducted using the same approach as the NMN/NAD⁺ ratio analysis.

NADase Activity Assay of Purified SARM1 Protein.

0.1 µg/µL of SARM1 (WT) or SARM1 (2Cmut) protein was incubated with 250 µM NAD⁺ and either M1 or NMN for a minimum of 30 min. The proteins used to assess SARM1 NADase activity were purified in a buffer devoid of DTT. Reactions were conducted in DPBS buffer (containing an additional 150 mM NaCl, pH 7.4). Upon completion of the reactions, 100% acetonitrile (ACN) was added to achieve a final concentration of 70%, and the mixtures were incubated on ice for at least 5 min. Proteins were precipitated by centrifugation at 18,000×g for 20 min. The supernatants were either stored at −20 °C or directly diluted and analyzed using UHPLC by Sironax (Beijing).

Cryo-EM Specimen Preparation and Data Collection.

SARM1 (WT) protein (about 4 mg/mL) was incubated with 250 µM M1 and a combination of 250 µM M1 and 50 µM 1AD, respectively, for 2 h in DPBS containing an additional 150 mM NaCl and 1 mM DTT at room temperature. SARM1 (2Cmut) protein (about 3 mg/mL) in DPBS containing an additional 150 mM NaCl was incubated with 250 µM M1 for 2 h at room temperature. Prior to Cryo-EM sample preparation, the proteins were centrifuged at 8,000×g for 10 min.

To prepare specimens of all protein complexes, 300-mesh holey carbon grids (Quantifoil R1.2/1.3 Au) were glow-discharged and placed into a Vitrobot Mark IV chamber (Thermo Fisher Scientific) set to 8 °C and 100% humidity. Subsequently, 3.0 μl of the complex sample was applied to the grid, blotted for 5.0 sec, and immediately plunged into liquid ethane. All the grids were stored in liquid nitrogen container before cryo-EM data collection. For the SARM1 (WT)-M1 and SARM1 (2Cmut)-M1 complexes, cryo-EM movies were recorded using EPU on a Titan Krios G4 microscope (Thermal Fisher Scientific) with a Falcon 4 direct electron detector, operating at 300 kV and ×96,000 magnification with a pixel size of 0.808 Å. For the SARM1 (WT)-M1-1AD complex, movies were collected using EPU on a Titan Krios G3 microscope (Thermal Fisher Scientific) equipped with a BioQuantum GIF/K3 (Gatan) direct electron detector at 300 kV in superresolution mode, with a nominal magnification of ×64,000 and a pixel size of 1.080 Å. Detailed Cryo-EM data collection parameters for all samples can be found in Supplementary Table.

Cryo-EM Data Processing.

For the SARM1 (WT)-M1 complex, a total of 6,554 movie stacks were processed using MotionCor2 for motion correction, followed by patch-based CTF estimation in CryoSPARC (31). A total of 2,570,705 particles were picked using Blob Picker and template picker, extracted and subjected to two rounds of 2D classification. Of note, 671,764 particles from 2D classes that displayed typical features of intermediate state 2 were selected for ab initio reconstruction and heterogeneous refinement. A subset of particles, accounting for 36.4%, were then used for homogenous reconstruction and nonuniform (NU) refinement with C8 symmetry, resulting in a final map with a global resolution of 2.46 Å. Concurrently, 357,995 particles from 2D classes exhibiting the features of intermediate state 1 were subjected to Ab initio reconstruction and heterogeneous refinement. And, 31.7% of these particles were selected for homogeneous reconstruction with C1 symmetry. The resulting map was subjected to 3D classification in CryoSPARC to remove particles with partially dissociated ARM-TIR domains. Particles with fully assembled ARM-TIR domains were used for NU refinement with C8 symmetry, yielding a final map with a global resolution of 2.83 Å.

For the SARM1 (2Cmut)-M1 complex, following motion correction, CTF estimation, particle picking, and Topaz Train, a total of 213,363 particles were subjected to 2D classification, followed by ab initio reconstruction and heterogeneous refinement. Of note, 57.4% of the total particles showing the intact SARM1 octameric ring with fully assembled ARM-TIR domains were selected for NU refinement with C8 symmetry. 3D classification was performed to further remove particles with partially dissociated ARM-TIR domains. The resulting particles were selected for NU refinement with C8 symmetry, yielding a final map with a global resolution of 2.41 Å.

For the SARM1 (WT)-M1-1AD complex, the procedures for motion correction, CTF estimation, and particle picking were identical to those described above. A total of 1,083,144 particles were picked from 1,119 movies. Topaz Train was then performed to identify more particles exhibiting features of the active state. In total, 807,946 particles were picked and subjected to 2D classification, resulting in two classes of particles with distinct features. These particles were subsequently extracted with box sizes of 340 and 500 pixels, respectively. The first 2D class, comprising 278,101 particles, was subjected to ab initio reconstruction and heterogeneous refinement. A 3D class showing a well-resolved ARM-SAM ring was further refined by NU refinement, resulting in a final map with a global resolution of 2.89 Å. The second 2D class with the resolved two-stranded TIR domains, comprising 65,732 particles, was subjected to ab initio reconstruction followed by NU refinement, yielding a low-resolution map with a global resolution of 7.63 Å.

Model Building.

The cryo-EM structure of the human SARM1 (PDB: 7CM5) was used as the template for modeling the structures of the SARM1 (WT)-M1 and SARM1 (2Cmut)-M1 complexes.

The model was docked into cryo-EM map using Chimera (32), and then manually adjusted in COOT (33), followed by real-space-refinement in Phenix (34). For the SARM1 (WT)-M1-1AD complex, the cryo-EM structure of the activated human SARM1 (PDB: 7NAL) was docked into the map with a compact SAM-ARM ring and poorly defined two-stranded TIR domains. For the map with 16 resolved TIR domains and two poorly defined SAM-ARM rings, the reported two-stranded SARM1 TIR domains (PDB: 7NAK) and the SAM-ARM ring were docked using Chimera. All models were adjusted in COOT. Figures for the structures and maps were prepared using PyMOL, Chimera, and ChimeraX (35).

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Data, Materials, and Software Availability

The atomic coordinates and cryo-EM density maps were deposited in the PDB and EM Data Bank with codes 9L2D (36) and EMD-62772 (37) (SARM1WT bound to M1, intermediate state 1); 9L2E (38) and EMD-62773 (39) (SARM1WT bound to M1, intermediate state 2); 9L2F (40) and EMD-62774 (41) (SARM1WT bound to M1 and 1AD, octamer); 9L2G (42) and EMD-62775 (43) (SARM1 2Cmut bound to M1).