Permeant fluorescent probes visualize the activation of SARM1 and uncover an anti-neurodegenerative drug candidate

SARM1 regulates axonal degeneration through its NAD-metabolizing activity and is a drug target for neurodegenerative disorders. We designed and synthesized fluorescent conjugates of styryl derivative with pyridine to serve as substrates of SARM1, which exhibited large red shifts after conversion. With the conjugates, SARM1 activation was visualized in live cells following elevation of endogenous NMN or treatment with a cell-permeant NMN-analog. In neurons, imaging documented mouse SARM1 activation preceded vincristine-induced axonal degeneration by hours. Library screening identified a derivative of nisoldipine (NSDP) as a covalent inhibitor of SARM1 that reacted with the cysteines, especially Cys311 in its ARM domain and blocked its NMN-activation, protecting axons from degeneration. The Cryo-EM structure showed that SARM1 was locked into an inactive conformation by the inhibitor, uncovering a potential neuroprotective mechanism of dihydropyridines.

It should be noted that SARM1 is not just a simple NADase activated to deplete the cellular NAD. We have documented that SARM1 is a multifunctional enzyme with properties similar to CD38, a universal signaling enzyme possessing not only NADase activity but also catalyzing both the cyclization of NAD to cyclic ADP-ribose (cADPR) and the exchange of nicotinamide in NADP with nicotinic acid to produce nicotinic acid adenine dinucleotide phosphate (NAADP) . Both cADPR and NAADP are messengers regulating calcium mobilization in the endoplasmic reticulum and the endo-lysosomes, respectively (reviewed in Galione, 1994;Lee, 2012;Lee and Zhao, 2019). The catalytic similarities and its ubiquitous presence in non-neuronal cells suggest that SARM1 may be a calcium signaling enzyme as well.
Since SARM1 is important in axon degeneration and potentially other physiological processes as well, we thus aim to design and synthesize probes for visualizing SARM1 activation in live cells and to screen drug library for potent inhibitors.

Probe design, synthesis, and characterization
We focused on its base-exchange reaction for designing specific probes for SARM1 and had shown that pyridyl derivatives can readily serve as substrates (Graeff et al., 2006;Lee and Aarhus, 1997). We thus conjugated various styryl derivatives to pyridine to produce a series of conjugates (PCs) as fluorescent probes for SARM1 activity ( Figure 1A). We reasoned that conjugating the electron-rich styryl derivative with pyridine should provide a donor-p-acceptor framework (Pawlicki et al., 2009; Figure 1B). The positive charge of the pyridinium moiety of the product should delocalize over the conjugated p-system and lead to fluorescence changes ( Figure 1A). Pyridine conjugates (PC1-9,  Figure 1C, exhibiting the largest fluorescence increase ( Figure 1D).
The time course study for the reaction involving PC6 showed that the UV absorption decreases at 330 nm but increases at 400 nm with an isosbestic point at 350 nm (Figure 1-figure supplement 12; Figure 1E). Corresponding to the absorbance change was the red shift in the fluorescence spectra, from the emission maximum at 430 nm of PC6 to 520 nm of PAD6 ( Figure 1E).
The conversion of PC6 to the exchange product, PAD6, was verified by purifying it using HPLC and characterized by high resolution mass spectrometry (HRMS) ( Figure 1F). The remarkably large spectral changes are anticipated from our design, as the pyridine ring becomes positively charged after its exchange into NAD ( Figure 1F, inlet), a much stronger electron acceptor in the DÀpÀA structure, thereby increasing intramolecular charge transfer and shifting the emission maximum by over 100 nm. The conversion-induced spectral changes were consistent with the spectra of the HPLC-purified products, PAD6 ( Figure 1G).
The observed spectral changes showed a linear dependence on NMN, with as low as 10 mM being effective ( Figure 1H), confirming that SARM1 is an auto-inhibitory enzyme activated by NMN . The fluorescence increase was also proportional to the amount of NMNactivated SARM1 ( Figure 1I), with a detection limit of 48 ng/mL. As an in vitro assay for SARM1, PC6 provides more than 100-fold higher sensitivity over other commonly used probes, such as eNAD, NGD, or NHD ( Figure 1J).
In addition to sensitivity, PC6 also shows exquisite selectivity toward SARM1 versus CD38 and N. crassa NADase (Graeff et al., 1994). All three possess NADase activity as detected by eNAD ( Figure 1K), but only SARM1 could produce large fluorescence increase with PC6.  NMN (10,20,40,60 mM) in the presence of NAD, PC6, and SARM1-dN. Inset: the initial rates plotted to NMN concentrations. (I) Emission spectra with doses of SARM1-dN in the presence of NMN, NAD, and PC6; Inset: the initial rate plotted to SARM1 concentrations. (J) The reaction rates of 10 mM PC6, in the presence of 100 mM NMN and 100 mM NAD, compared with NAD analogss (100 mM) catalyzed by SARM1. (K) The reaction rates of 10 mM PC6 catalyzed by SARM1, NADase, and CD38. PC = pyridine conjugate, NMN = nicotinamide mononucleotide. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Source data in excel for Figure 1D              Imaging SARM1 activation in live cells PC6 was added to HEK293 cells overexpressing either wildtype SARM1 or the enzymatically inactive mutant, E642A (Essuman et al., 2017;Zhao et al., 2019;Figure 2A). Green fluorescence was clearly seen evenly distributed in the whole cells in the wildtype, but not in the mutant cells ( Figure 2B), indicating active SARM1 was required. Intracellular production of PAD6 was confirmed in extracts of wildtype but not the E642A cells ( Figure 2C). CZ-48, a cell-permeant mimetic of NMN and activator of SARM1 , dramatically increased the PAD6 fluorescence ( Figure 2B, right column) and none in E642A-cells. These results indicate that PC6 is cell-permeant and can be exchanged into the cytosolic NAD by the activated SARM1 to produce PAD6 having a large red shift in fluorescence. PAD6 was also cell-impermeant because of its charged ADP-ribose moiety and accumulated in the cytosol, greatly increased its detection sensitivity in live cells.
PC6 also could detect the activity of SARM1 endogenously expressed in HEK293T cells . CZ-48 activated the endogenous SARM1 and produced increase of cytosolic PAD6 signal ( Figure 2D, upper right), but none in the SARM1-knockout cells ( Figure 2D, right lower), confirming the specificity of PC6 for SARM1.
An HEK293 cell line carrying doxycycline (Dox)-inducible SARM1  was used to further substantiate that the PAD6 fluorescence was derived from SARM1 activity. Without induction, only basal SARM1 (    also ( Figure 2E, purple squares), confirming SARM1 is auto-inhibitory. With CZ-48, both the basal and the induced SARM1 were activated, resulting in the largest signal ( Figure 2E, black triangles). In SARM1-knockout cells, no signal was detected ( Figure 2D, SARM1-KO; Figure 2E, blue and red dots).
Consistent with the in vitro results showing that PC6 is highly selective for SARM1 over CD38 in live cells, cells expressing either wildtype or Type III mutant CD38 Zhao et al., 2012) did not show PAD6-signal after 48 hr incubation with PC6 ( Figure 2-figure supplement 1B),

Imaging SARM1 activation during AxD
Vincristine (VCR)-induced AxD in peripheral neuropathy is a common side effect of chemotherapy (Essuman et al., 2017) and is thought to be due to SARM1-activation (Gerdts et al., 2013). Mouse dorsal root ganglion (DRG) neurons were infected with lentivirus carrying TdTomato for visualizing the axons ( Figure 3A and C, orange), and with either a non-targeting ( Figure 3A) or Sarm1-specific shRNA ( Figure 3C). In the non-targeting group, VCR elevated PAD6-fluorescence ( Figure 3A, green), indicating activation of SARM1, by as early as 4-8 hr and reaching a maximum by 16 hr ( Figure 3A and D, blue). AxD started at about 20 hr ( Figure 3F, blue; Figure 3-figure supplement 1A), temporally consistent with a causal role for SARM1. Another measure of SARM1 activation is the elevation of cellular cADPR , which occurred ( Figure 3E, blue) by 12 hr, peaking at 24 hr. Neurons not treated with VCR showed neither SARM1-activation nor AxD ( Figure 3A Reducing endogenous SARM1 using shRNA ( Figure 3B,D and F, KD) reduced the PAD6 fluorescence without altering its peaking at 16 hr ( Figure 3C; 3D, KD +VCR) and reduced AxD ( Figure 3F, KD +VCR; Figure 3-figure supplement 1B), further substantiating a causal role for SARM1. CZ-48 induced SARM1-activation more rapidly ( Figure 3A and D, red) and elevated cADPR higher ( Figure 3E, red), confirming its direct action. Intriguingly, CZ-48 did not induce massive AxD as VCR ( Figure 3F, CZ-48; S5A). These results indicate SARM1-activation is a necessary and causal factor, but not a sufficient one for AxD. Other critical factors and downstream events of microtubular dysfunction might contribute to the degeneration.

Dehydronitrosonisodipine (dHNN) is an inhibitor of SARM1 activation
Another prompt application of PC6 is library screening for inhibitors of SARM1. The feasibility was verified by measuring the IC 50 of a reported inhibitor of SARM1, nicotinamide (Nam) (Essuman et al., 2017). The measured IC 50 value of Nam was around 140 mM, which is consistent with the reported value (Figure 4-figure supplement 1A). Next, we utilized this assay to screen for SARM1 inhibitors. NMN-activated SARM1 was incubated with drugs of the library ( Figure 4A) and its activity measured with PC6 in the presence of NAD ( Figure 1). Out of 2015 drugs, 34 had more than 80% inhibition ( Figure 4B), which were further tested for inhibition of the SARM1-NADase activity using HPLC. Figure 4C shows the plot the IC 50 -values of these drugs measured with both the PC6 and the NADase/HPLC assays. Twenty-four drugs are in the middle sector, indicating they inhibited both reactions similarly. Two inhibited the PC6 activity five fold less than the NADase ( Figure 4C, left sector), and eight in the right sector (seven have IC 50 s higher than 40 mM) inhibited NADase less than the base-exchange. These remarkable differences underscore the importance of using more than one assay for drug screening (see Discussion).
In the middle sector is nisoldipine (NSDP), a calcium channel blocker having beneficial effects on neurodegenerative diseases. Peculiarly, its inhibition of SARM1 varied widely among batches. Investigations indicated the active compound was not NSDP but its derivative. Figure 4D shows fresh NSDP had an IC 50 -value of about 150 mM (squares), but its potency increased 75-fold after exposure to UV ( Figure 4D, triangles, IC 50 = 2.36 ± 0.3 mM). Also, fresh NSDP had an HPLC-elution peak at 12.2 min (Figure 4-figure supplement 1B), but was completely converted by UV to a compound having a peak at 9.8 min that strongly inhibited SARM1 ( Figure 4E, red). HRMS showed that the active compound had a mass of 370.15205 Da ( Figure 4E, inset) identical to a known derivative of NSDP, dehydronitrosonisoldipine (dHNN) (Marinkovic et al., 2003). The HPLC-elution profile of the active compound was also the same as dHNN (Figure 4-figure supplement 1B, purple line and green dash). Indeed, authentic dHNN was active and could not be further activated by UV ( Figure 4F, red line and dash), which also indicates that it is photostable. Another derivative of NSDP, dehydronisoldipine (dHN, elution peak at 8.7 min, Figure 4-figure supplement 1B), showed no inhibition before or after UV ( Figure 4F, black line and dash), indicating that the nitroso group is essential for the effect.

dHNN inhibits SARM1 and AxD by covalently modifying cysteines
The dHNN-induced inhibition of SARM1 was irreversible by washing ( Figure 4-figure supplement 1C, red bars), while that by Nam was reversible. Also, dHNN-inhibition was time-dependent, but not Nam (Figure 4-figure supplement 1D), strongly suggesting dHNN covalently reacted with SARM1.
To determine the mechanism of action of dHNN, we truncated the inhibitory ARM-domain, producing a constitutively active SAM-TIR, which showed a right-shifted inhibition curve compared to SARM1-dN ( Figure 5A), with around 50-fold increase in the IC 50 . The IC 50 of dHNN in the SARM1-dN-expression cells is around 4 mM, close to the IC 50 in vitro. dHNN decreased the cellular cADPR in cells expressing SARM1, but not in those expressing SAM-TIR ( Figure 5B). These results suggest that dHNN is cell-permeant and acts mainly by blocking SARM1 activation but not its enzymatic activities.
The nitroso group of dHNN may covalently modify cysteine residues (Callan et al., 2009)     of other proteins with cysteines were also identified but none showed modification by dHNN (Figure 5-figure supplement 1C), indicating specificity of dHNN. Single mutation of all the cysteines to alanine showed that C311A significantly decreased the response to dHNN ( Figure 5D). However, the IC 50 of C311A was only two fold higher than that of the wildtype, which indicates dHNN might modify other cysteines when Cys311 is mutated and inactivate SARM1.
With Cryo-EM, we found that dHNN stabilized a similar inhibitory conformation of SARM1 as that induced by NAD (PDB: 7cm6) (Jiang et al., 2020). In 2D-classification of the untreated SARM1, most particles presented only the SAM octamer ring ( Figure 5-figure supplement 3A). For the dHNNtreated SARM1, larger octamer ring corresponding to both the SAM and ARM/TIR domains could      , suggesting that dHNN constrains SARM1 in an inactive conformation similar to that induced by NAD. Extra electron density was only observed near residue Cys311, the dHNN-target ( Figure 5C and D), but not other cysteine residues, consistent with it being derived from dHNN (5F, purple). dHNN interacts with Glu264, Leu268, Arg307, Phe308, and Ala315 ( Figure 5F and Figure 5G, green) in the ARM domain, pushing the insertion loop ( Figure 5F, red) toward ARM1 and stabilizes the domain. This is similar to that observed with NAD, which binds at the other side of the insertion loop and stabilizes the ARM domain possibly via ligating ARM1 and the insertion loop ( Figure 5G).
By preventing SARM1 from activation, dHNN also inhibited the VCR-activated cADPR production ( Figure 5H) in neurons and blocked not only the VCR-induced AxD ( Figure 5I, third picture; Figure 5J, red line), as effective as knocking out SARM1 ( Figure 5I, fourth picture), but also AxD after axotomy (Figure 5-figure supplement 4A, third column; B, red line).

Discussion
Visualizing the activity of a signaling enzyme in live cells provides clearer understanding of the spatial and temporal aspects of its mechanism and function, a goal sought by many. The PC probes developed here are particularly advantageous. They are cell-permeant, but the SARM1-catalyzed exchange products are not and accumulate in the cytosol, enhancing their detection. The remarkably large red shift of the product fluorescence provides easy visualization away from the interference of autofluorescence.
Currently, there are several fluorescent probes for SARM1 activity in use. They are all analogs of NAD and are cell-impermeant, such as eNAD. The fact that NAD is now shown to be an inhibitor of SARM1 (Jiang et al., 2020;Sporny et al., 2020) makes the use of these analog probes problematic, as they may affect SARM1 activity as well. PC6 has no such drawback as it is a pyridine, not an NAD analog. Furthermore, the use of PC6 requires neither expression of construct nor cell manipulation, making it suitable for detecting SARM1 activity in any cells. This is documented in this study. Using CZ-48, a cell-permeant activator, to activate endogenous SARM1 produces large increase in PAD6 fluorescence in a variety cell lines as well as primary neurons. With the probes, we provided the first direct evidence in live DRG neurons that SARM1 activation precedes AxD by several hours and that it is a necessary but insufficient factor for AxD.
Screening library to identify compounds of interest is a straightforward strategy widely used. The case for SARM1 is more complicated, as it is not only a multi-domain protein but also an auto-regulated enzyme catalyzing multiple reactions. Compounds may target the regulatory ARM domain as shown here for dHNN, or the catalytic TIR domain as Nam and the inhibitors reported during the preparation of this manuscript (Hughes et al., 2021;Loring et al., 2020). For SARM1, the substrates are different for the base-exchange and the NADase reactions and may thus be differentially affected by the inhibitor-induced conformational changes of the catalytic site. Although the exact reason remains to be determined, the compounds shown here that can selectively block one reaction much more than the other are of interest. Many believe that the NADase activity of SARM1, leading to cellular NAD depletion, is its dominant property for regulating AxD. But the two calcium messengers, cADPR and NAADP produced by its cyclizing and base-exchange reactions may well have functional roles as well. Compounds with differential inhibition can thus be an important tool to resolve the issue.
Much effort is being invested in targeting SARM1-mediated NAD depletion for therapeutic protection from AxD. Chemical blockers may well be an ideal tool for turning off the NAD depletion. dHNN uncovered in this study is the first compound ever described that can block the activation of SARM1, revealing a druggable allosteric site and can thus usher in a new approach for therapeutic drug development. Another point of interest is that dHNN is a derivative of NSDP. Metabolic conversion of NSDP to dHNN, leading to inhibition of SARM1, may well account for the neural protective effects of NSDP (Siddiqi et al., 2019).

Animals
This study was carried out in strict accordance with animal use protocol approved by Peking University Shenzhen Graduate School Animal Care and Use Committee (#AP0015001). All animals (C57BL6/J), purchased from Guangdong Medical Laboratory Animal Center (China), were handled in accordance with the guidelines of the Committee on the Ethic of Animal Experiments. All surgery was performed after euthanasia and efforts were made to minimize suffering.

Cell lines
The HEK293 and HEK293T cells were purchased from the American Type Culture Collection and the identity has been authenticated by STR profiling. They have not been contaminated by mycoplasma. The cells were cultured in DMEM supplemented with 10% fetal calf serum and 1% penicillin-streptomycin solution and maintained in a standard humidified tissue culture incubator with 5% CO 2 .

Synthesis and characterization of pyridine conjugates (PCs)
All air and water-sensitive reactions were carried out with anhydrous solvents in flame-dried flasks under argon atmosphere, unless otherwise specified. All the reagents were obtained commercially and used without further purification, unless otherwise specified. Anhydrous DMF was vacuum distilled from barium oxide, acetonitrile, and dichloromethane was distilled from calcium hydride. Yields refer to isolated yields, unless otherwise specified. Reactions were monitored by thin-layer chromatography (TLC) carried out on 0.25 mm silica gel plates (60 F-254) that were analyzed by UV light as visualizing method and by staining with anisaldehyde (450 mL of 95% EtOH, 25 mL of conc. H 2 SO 4 , 15 mL of acetic acid, and 25 mL of anisaldehyde) or KMnO 4 (200 mL H 2 O of 1.5 g KMnO 4 , 10 g K 2 CO 3 and 1.25 mL of 10% aq. NaOH). Silica gel (200-300 mesh) was used for flash column chromatography. Nuclear magnetic resonance (NMR) spectra were recorded on either a 300 ( 1 H: 300 MHz, 13 C: 75 MHz), 400 ( 1 H: 400 MHz, 13 C: 100 MHz), or 500 ( 1 H: 500 MHz, 13 C: 125 MHz) NMR spectrometer. The following abbreviations were used to explain the multiplicities: s = singlet, d = doublet, t = triplet, q = quartet, dd = doublet of doublets, m = multiplet, br = broad. High resolution mass spectra (HRMS) were obtained from a MALDI-TOF mass spectrometer.

Preparation and quantification of the enzymes
A truncated form of SARM1, SARM1-dN, was prepared as described . In brief, the recombinant protein, SARM1 without the N-terminal mitochondrial signal, was expressed in HEK293T cells and released by 100 mM digitonin in PBS with protease inhibitor cocktail (Roche). The cell lysate of wildtype HEK293T, prepared with the same method, was used as the negative control. To quantify SARM1-dN, the protein was pulled down by BC2 nanobody (Bruce and McNaughton, 2017) conjugated beads which were prepared by conjugating BC2 nanobody to NHS-beads (GE Healthcare). The purified SARM1-dN, named as SARM1-IP, together with the certain amounts of standard protein BSA, was applied to SDS-PAGE, which was stained by Coomassie blue. The protein contents of SARM1-dN were then calculated by the band intensity with BSA as standards.
DtSARM1-dN, with the N-terminal targeting signal removed and tagged with a tandem strep tag II and flag tag for purification, was constructed into Plenti-CMV-puro-Dest (Invitrogene) by LR clonase II enzyme according to the manufacturer's instructions. HEK293F cells overexpressing dtSARM1-dN were constructed by lentivirus infection and selected with 1 mg/mL puromycin. DtSARM1-dN was released by 200 mM digitonin and immunoprecipitated with StrepTactin resin (GE healthcare), washed with buffer W (100 mM Tris-HCl pH8.0, 150 mM NaCl and 1 mM EDTA) for four times and eluted with 2 mM biotin in buffer W. DtSARM1-dN was used in the experiments on the dHNN-modification, cysteine-to-alanine mutants and Cryo-EM structures.

In vitro fluorescence assays
To analyze the activity of SARM1 with PCs in vitro, reactions were started by incubating the enzyme with the reaction mixture, 50 mM PC, 100 mM NAD, and 100 mM NMN in PBS. The absorbance and fluorescence were measured in a quartz cuvette or black 96-well plates (Corning), respectively, in an Infinite M200 PRO microplate reader (Tecan). For the assays with eNAD,NHD, or NGD as the substrate, 100 mM of each probe was incubated with the enzymes and the kinetics of fluorescence production was measured at l ex = 300 nm, l em = 410 nm. The initial rate of the reactions was quantified with the slope of the fluorescence increase in the first several minutes.

HPLC analysis of the base-exchange reaction
The reactions were prepared by mixing SARM1-IP (SARM1 binding on BC2-beads, around 2.5 mg/ mL) with 100 mM NAD, 50 mM PC6, 100 mM NMN, and 0.1 mg/mL BSA in PBS and incubated for 60 min at 37˚C. SARM1-IP was removed by centrifugation at 4,500 rpm for 1 min. The cleaned mixtures were applied to a C-18 reverse phase column equipped on an HPLC (Agilent 1260) with a gradient of 0.1 M KH 2 PO 4 (pH 6.0) and 0.1 M KH 2 PO 4 (pH 6.0) with MeOH (7:3) to elute NMN, cAPPR, ADPR, NAD, and a gradient of ACN from 30 to 70% to elute PAD6 and PC6. The PAD6 fractions were collected and lyophilized for the characterization of absorption and fluorescence spectra.
To analyze PAD6 in cells, the metabolites were extracted from the cells treated with 50 mM PC6 by 0.6 M perchloric acid, followed by the neutralization with Chloroform: Tri-n-octylamine (3:1). The extracts were applied to a C-18 column and PAD6 was eluted with water and acetonitrile by 2% acetonitrile for 8 min, then 30% acetonitrile for 8 min.
Confocal imaging of PAD6 in living cells HEK293 cells, overexpressing wildtype or the enzymatically dead form (E642A) of SARM1 or HEK293T Knocking out NMNAT1 were constructed as before . Cells, grown on 0.05 mg/mL poly-L-lysine coated Chambered coverglass (ThermoFisher, #155411) overnight, were treated with 50 mM PC6 in the presence or absence of 100 mM CZ-48 for 8 hr (for SARM1-OE cells) and 200 mM CZ-48 for 48 hr (for wildtype HEK293T cells), respectively. To demonstrate the edges of the cells, they were stained with 50 mg/mL Concanavalin A, Alexa Fluor 647 Conjugate (Thermo-Fisher) at 4˚C for 10 min before imaging. The fluorescence signals (Ex/Em: 405/525 nm for PAD6; Ex/Em: 561/590 for ConA) were captured under a confocal microscope (Nikon A1).

Analysis of PAD6 signals in live cells by flow cytometry
HEK293 cell line carrying doxycycline (Dox)-inducible SARM1 was constructed as previously described . The cells were treated with 50 mM PC6, 100 mM CZ-48, or 0.5 mg/mL Dox for 4, 8, 12, and 16 hr. The cells were trypsinized and the fluorescence of PAD6 (Ex/Em: 405/ 525 nm) was analyzed by flow cytometry (CytoFlex, Beckman).
On div6, the neurons were infected with lentivirus carrying various expression cassettes of genes or shRNAs. Three days later, the cells were treated with 50 mM PC6 in the absence or presence of 200 mM CZ-48 or 50 nM vincristine. The fluorescence images (Ex/Em: 405/520 nm for PAD6; Ex/Em: 561/590 for TdTomato) were captured under a confocal microscope (Nikon A1) with a 60x object. The mean fluorescence intensity was quantified by NIS-Elements AR analysis (Nikon). Axon degeneration was quantified based on axon morphology using ImageJ. The TdTomato fluorescence images were binarized and measured the total axon area (size = 20 infinity pixels) and the degenerated axon (size = 20-4,000,000 pixels) with particle analyzer module of ImageJ. Axon degeneration index was calculated as the ratio of the degeneration axon over total axon area.
Lentivirus preparation and infection of DRG neurons pLKO.1-shRNA-Sarm1 plasmids were constructed as described previously . Briefly, the shRNA targeting mouse Sarm1 (5'-CCGGCTGGTTTCTTACTCTACGAATCTCGAGATTCGTAGAGTAAGAAACCAGTTTTTG-3') or the scrambled shRNA (5'-CCGGCCTAAGGTTAAGTCGCCCTCGCTCGAGCGAGGGCGACTTAACC TTAGGTTTTTG-3') were inserted to pLKO.1-puro (Addgene, #8453) with EcoRI and AgeI, followed by the replacement of the puromycin resistance gene with a fluorescent protein, TdTomato (Gen-Bank: LC311026.1) with KpnI/BamHI. The lentiviral particles were prepared by transfecting HEK293T cells with the corresponding lentivectors, pMD2.G, and psPAX2  and concentrated with Lenti-Concentin Virus Precipitation Solution (ExCell Bio). The viral particles were finally resuspended in Neurobasal Plus Medium. The virus titer was determined by series infection of HEK293T cells. The virus was added to infect the DRG neurons on div6 with the same MOIs and the experiments were carried out 72 hr after infection.

Imaging and quantification of AxD after axotomy and vincristine treatment
For axotomy, one DRG was seeded into a 24-well plate, and 5 mM 5-fluoro-2'-deoxyuridine and 5 mM uridine were added on the other day. On div5, axons were pre-incubated with the drugs for 0.5 hr and severed near the soma with a 3 mm flat blade under microscope guidance to remove the cell bodies. For vincristine treatment, DRGs were digested with 0.05% Trypsin and seeded into 24-well plates. DRGs on div9-13 were incubated with 50 nM vincristine in the presence or absence of the candidate drugs.
About 9-12 images of the axon were acquired in the bright field with a 20x object for each treatment at the indicated time points using invert optical microscope (Olympus). Axon degeneration was quantified using ImageJ. For each treatment, 60 random grid-squares with 147 Â 147 pixels were cropped, binarized and the total axon area (size = 16 infinity pixels) and the degenerated axon (size = 16-10,000 pixels) were quantified with the particle analyzer module of ImageJ. Axon degeneration index was calculated as the ratio of the degeneration axon over total axon area.

Measurement of the cADPR levels in DRGs
DRG neurons were treated with 50 nM Vincristine or 200 mM CZ-48 for 0, 12, 24, 48 hr on div6. After incubation, DRG was washed with cold PBS and lysed with 0.6 M perchloric acid. The concentration of cADPR was analyzed by the cycling assay, as described previously (Graeff and Lee, 2002).

PC6 assay
For high-throughput screening of the potential inhibitors, 1.5 mg/mL SARM1-dN was pre-incubated with 50 mM drugs (TargetMol, L1000) at room temperature for 10 min and the reaction reagents, including 20 mM PC6, 50 mM NAD, and 50 mM NMN were added to start the reaction. Controls including reactions without the drugs, defined as 0% inhibition, and without both the drugs and SARM1-dN, defined as 100% inhibition. The fluorescence (ex: 390 nm; em: 520 nm) was measured by plate reader (Tecan, M200Pro) and the initial reaction rates were calculated, V x for the reactions with different drugs, V max for the reaction without drugs and V min for the reaction without enzyme. The inhibitory rates were calculated by the equation, (V max -V x )/V max and plotted using GraphPad Prism 8.0. The standard statistics of the screening were calculated as follows: Z' factor=1-(3*SD (V max )+3*SD(V min ))/(Average(V max )-Average(V min )), S/N = (Average(V max )-Average(V min ))/SD(V min ). In the screening of this study, Z' = 0.69 and S/N = 291.96.
For IC 50 measurement, 0.4 mg/mL SARM1-dN was pre-incubated with doses of drugs in vitro for 10 min, and started the reaction by adding 50 mM NAD, 50 mM NMN, and 50 mM PC6. Calculation of IC 50 by plotting the initial rate to dose of compounds.
NADase acitivity of SARM1 analyzed by HPLC 1 mg/mL of SARM1-dN was pre-incubated with drugs for 15 min at room temperature, and the reactions were started by adding 100 mM NAD and 100 mM NMN. They were stopped by removing the enzyme with MultiScreen Filter Plates (Millipore) after 0, 15, and 30 min incubation at 37˚C and the reactants and products were analyzed by a C-18 column (Aligent, 20 RBAX SB-C18) with a gradient of 0.1 M KH 2 PO 4 (pH 6.0) and 0.1 M KH 2 PO 4 (pH 6.0) with MeOH (7:3) to elute NMN, cAPPR, ADPR, NAD, Nam. The amount of ADPR was used to calculate the initial rate. IC 50 was calculated by Graphpad Prism 8.0.

HPLC analysis of NSDP and its derivatives
The NSDP powder was dissolved in DMSO and shined with UV at 254 nm for 30 min and analyzed with a C-18 reverse phase column (ZORBAX SB-C18) equipped on a HLPC (Aligent 1260) and eluted with 50% of 0.1%TFA and 50% of 0.1%TFA in 99% ACN. The products after UV treatment were collected and purified by HPLC, as described above. The inhibitory activity of these fractions was determined by PC6 assay after being neutralized with 100 mM Tris (pH7.5), and the main peak was characterized by HRMS (Thermo, Q Exactive Focus).

The inhibitory activity of dHNN in vitro and in cellulo
To determine whether dHNN inhibits the activation or enzymatic activity of SARM1 in vitro, SARM1-dN, the autoinhibited form, and SAM-TIR, the constitutively active form, were pre-incubated with different concentrations of dHNN at rt for 10 min, after which the activity was measured with PC6 assay and the inhibition rate was calculated.
To test the same effect in cellulo, HEK293 cells overexpressing the inducible SARM1 (iSARM1) or SAM-TIR (iSAM-TIR) were pre-incubated with 20 mM dHNN, or DMSO as controls, for 1.5 hr and then treated with 100 mM CZ-48 or 0.5 mg/mL Dox for the indicated time. The cellular levels of cADPR were measured as described above.

Modification of SARM1 by dHNN
The dtSARM-dN, eluted from the StrepTactin beads, was incubated with 0, 5, or 50 mM dHNN at rt for 40 min and applied to SDS-PAGE. After simplyBlue SafeStain (ThermoFisher), the dtSARM1-dN band was sliced, dehydrated with 100% ACN, and the proteins were alkylated by 22.5 mM IAA for 30 min in dark after the reduction by 10 mM DTT at 55˚C for 30 min. After overnight in-gel digestion with Trypsin, the peptides were extracted and analyzed with HRMS (Thermo, Q Exactive HF-X). The dHNN modifications, determined by Protein Discoverer software (ThermoFisher), were defined as an increase of molecular weight by 370.153 Da,354.158 Da,402.143 Da,or 386.148 Da on the cysteine residues characterized in the MS 2 fragments (Callan et al., 2009;Mö ller et al., 2017). The abundance of each peptide was determined in the MS 1 . Abundance ratio was calculated by dividing the intensity of the dHNN-modified peptides by that of the corresponding peptides.

Data analysis
All experiments contained at least three biological replicates. Data shown in each figure are all means ± SD. The unpaired Student's t-test was used to determine statistical significance of differences between means (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). GraphPad Prism 8.0 was used for data analysis.