Research Article

NAD+ cleavage activity by animal and plant TIR domains in cell death pathways

See allHide authors and affiliations

Science  23 Aug 2019:
Vol. 365, Issue 6455, pp. 793-799
DOI: 10.1126/science.aax1911

NAD depletion as pathogen response

One way that plants respond to pathogen infection is by sacrificing the infected cells. The nucleotide-binding leucine-rich repeat immune receptors responsible for this hypersensitive response carry Toll/interleukin-1 receptor (TIR) domains. In two papers, Horsefield et al. and Wan et al. report that these TIR domains cleave the metabolic cofactor nicotinamide adenine dinucleotide (NAD+) as part of their cell-death signaling in response to pathogens. Similar signaling links mammalian TIR-containing proteins to NAD+ depletion during Wallerian degeneration of neurons.

Science, this issue p. 793, p. 799

Abstract

SARM1 (sterile alpha and TIR motif containing 1) is responsible for depletion of nicotinamide adenine dinucleotide in its oxidized form (NAD+) during Wallerian degeneration associated with neuropathies. Plant nucleotide-binding leucine-rich repeat (NLR) immune receptors recognize pathogen effector proteins and trigger localized cell death to restrict pathogen infection. Both processes depend on closely related Toll/interleukin-1 receptor (TIR) domains in these proteins, which, as we show, feature self-association–dependent NAD+ cleavage activity associated with cell death signaling. We further show that SARM1 SAM (sterile alpha motif) domains form an octamer essential for axon degeneration that contributes to TIR domain enzymatic activity. The crystal structures of ribose and NADP+ (the oxidized form of nicotinamide adenine dinucleotide phosphate) complexes of SARM1 and plant NLR RUN1 TIR domains, respectively, reveal a conserved substrate binding site. NAD+ cleavage by TIR domains is therefore a conserved feature of animal and plant cell death signaling pathways.

Toll/interleukin-1 receptor (TIR) domains are usually found in multidomain proteins involved in innate immunity pathways in animals and plants (1). In mammals, TIR domains are located in the cytoplasmic regions of Toll-like receptors (TLRs) and interleukin-1 receptors (IL-1Rs) and in the cytosolic adaptor proteins involved in inflammatory signaling downstream from these receptors (2). In these molecules, TIR domains function as protein interaction modules; molecular and structural characterizations of TIR domain signaling complexes in the TLR4 pathway suggest a nucleated assembly of open-ended complexes consistent with the SCAF (signaling by cooperative assembly formation) mechanism prevalent in innate immunity pathways (3, 4). The protein SARM1 (sterile alpha and TIR motif containing 1) functions both as a TLR adaptor and as a key executor of axon degeneration (58). In plants, TIR domains are most commonly found as the N-terminal signaling domains of cytoplasmic nucleotide-binding leucine-rich repeat (NLR) resistance proteins, which directly or indirectly recognize effector proteins from pathogens and initiate defense responses (9).

Axon degeneration is a hallmark of many neurological disorders (10), and understanding the molecular basis of SARM1-induced neuronal cell death may offer therapeutic options. Axonal injury is associated with the breakdown of nicotinamide adenine dinucleotide in its oxidized form (NAD+) (11), and SARM1 accelerates NAD+ depletion in nerves postinjury (8). SARM1 is a modular protein with several domains (Fig. 1A). Tandem sterile alpha motif (SAM) domains mediate self-association of SARM1 and are required for SARM1 function (12). Forced self-association of SARM1 TIR domains induces axon degeneration in the absence of injury as a result of rapid NAD+ depletion (8), and the TIR domain has recently been shown to have NAD+ cleavage activity (13).

Fig. 1 hSARM1TIR crystal structure.

(A) Schematic diagram of the SARM1 domain architecture. (B) Structure of hSARM1TIR [cartoon representation; catalytic E642 (orange) and glycerol (green) in stick representation]. (C) Superposition of representative TIR domains from plant NLRs (L6; pink), TLR adaptors (MAL; yellow), TLRs (TLR2; orange) and bacterial proteins (TcpB; green) onto hSARM1TIR (blue). The glutamates equivalent to hSARM1TIR E642 are shown in stick representation. (D) Comparison of the catalytic pockets of hSARM1TIR and MilB. Cartoon representation of the crystal structure of MilB (purple) bound to CMP (cyan) [Protein Data Bank (PDB) ID 4JEM]; below, rotated 180°, stick representation of residues in the catalytic pocket (green) coordinating with the ligand (cyan), and hSARM1TIR structure (blue) bound to glycerol (magenta) with stick representation of residues in the catalytic pocket. (E) hSARM1TIR crystal packing. Three hSARM1TIR molecules are shown in each of the antiparallel strands (colored orange and blue); each strand features a head-to-tail arrangement via the BB-loop (interacting with DE, βE, and αE regions) interface. The association between strands is via the AE interface. (F) Superposition of one strand of the MAL proto-filament (3) (yellow) onto the crystal packing arrangement of hSARM1TIR (orange and blue).

Plant NLRs recognize pathogen effector proteins and trigger the hypersensitive response (HR), a process usually associated with localized cell death, to restrict pathogen infection. Isolated plant TIR domains can trigger cell death when transiently expressed in planta, in the absence of the corresponding pathogen effector proteins (referred to as autoactivity), and this activity is dependent on two self-association interfaces (1416). However, interacting partners or a direct signaling pathway have not yet been defined.

In the current study, we determined the crystal structure of the TIR domain of human SARM1, and the structure revealed close similarity to plant NLR TIR domain structures. In agreement, we show that like the SARM1 TIR domain, the TIR domains from plant immune receptors, including L6 (flax; Linum usitatissimum) and RUN1 (grapevine; Muscadinia rotundifolia) also have self-association–dependent NAD+-cleaving enzyme (NADase) activity. For SARM1, we show that the SAM domains contribute to NADase activity by assembling into an octamer. We demonstrate the mechanistic basis of NAD+ cleavage through structural analysis. Collectively, this work suggests a conserved signaling mechanism involving nucleotide cleavage in cell death pathways and provides a basis for rational drug design in the treatment of axonopathies.

Structure and NAD+ cleavage activity of the SARM1 TIR domain

The crystal structure of the TIR domain from human SARM1 (hSARM1TIR; residues 560 to 700; 1.8-Å resolution) (Fig. 1B and table S1) shows more similarity to plant and TLR TIR domains [Dali (17)] (Fig. 1C and table S2) than to bacterial TIR domains, contrary to conclusions reached from sequence analyses (18). These comparisons also reveal close similarities with enzymes such as the bacterial N-glycosidase MilB, which cleaves the nucleotide hydroxymethyl–cytidine 5′-monophosphate (hydroxymethyl-CMP) (Fig. 1D and table S2). The structure of hSARM1TIR contains a cleft with a bound glycerol molecule (glycerol was used as cryoprotectant). This region, consisting of residues from the βA strand, the AA and BB loops, and the αB and αC helices [the elements of the secondary structure are labeled sequentially (1)], can be superimposed closely with the catalytic sites of MilB as well as human CD38, including a catalytic glutamate residue [residue 642 (E642) in SARM1, E103 in MilB, or E226 in CD38] (Fig. 1D and fig. S1). Structural evidence and the association of hSARM1 with NAD+ depletion suggests hSARM1TIR may cleave NAD+. Based on molecular docking analyses (fig. S2), we show by nuclear magnetic resonance (NMR) and fluorescence-based assays (19) that hSARM1TIR cleaves NAD+ into nicotinamide (Nam) and ADP-ribose (ADPR) (Fig. 2, A and B, figs. S3, S4, and S5, and table S3). The Glu642→Ala (E642A) mutation in hSARM1TIR abolishes this activity, as do alanine mutations at the conserved active site residues tyrosine 568 (Y568), arginine 569 (R569), and R569+R570 (RRAA) (Fig. 2A and fig. S4C). While this work was in progress, similar results were reported by Essuman et al. (13).

Fig. 2 NADase activity of TIR domains.

(A) NADase activity of hSARM1TIR and its mutants, measured by the fluorescence assay using εNAD. (B) NAD+ cleavage reaction time courses of human, Drosophila, and C. elegans SARM1 constructs, monitored by 1H NMR (298 K), using a protein concentration of 20 μM and 1 mM NAD+. (C) NADase activity of RUN1TIR, RUN1TIR(E100A), L6TIR and L6TIR(E135A), measured by the fluorescence assay using εNAD. (D) NAD+ cleavage reaction time courses of RUN1TIR and L6TIR, monitored by 1H NMR (20°C), using an NAD+ concentration of 1 mM. (E) Structure of RUN1TIR in complex with NADP+ and bis-Tris (cartoon representation; ligands and selected residues are shown in a stick representation). (F) Comparison of the binding sites for CMP in MilB (PDB ID 4JEM) and NADP+/bis-Tris in RUN1TIR (stick representation).

Plant TIR domains cleave NAD+

Due to their involvement in cell death and structural similarities with hSARM1TIR, we tested if plant TIR domains can also cleave NAD+. At high protein concentrations, purified TIR domains from the NLRs L6 and RUN1 were capable of cleaving NAD+ into Nam and ADPR (Fig. 2C). The catalytic activities were lower than the activity of hSARM1TIR (Fig. 2D). Like hSARM1TIR, they were also able to cleave NADP+ but not FAD (fig. S6). Activity was not observed for the purified TIR domains from the NLRs RPS4, SNC1, RPP1, RPV1, or ROQ1 (fig. S7). When Escherichia coli lysates (rather than pure proteins) were tested in an enzyme-linked cycling NAD+ cleavage assay (20), NADase activity was observed also for RPS4TIR (fig. S8). Mutations of the residue equivalent to hSARM1TIR E642 in L6TIR and RUN1TIR (E135A and E100A, respectively) abolished NAD+ cleavage activity (Fig. 2C and fig. S9A).

Structural basis of NAD+ cleavage by TIR domains

In NADases, a catalytic glutamate typically interacts with the C-2 and C-3 hydroxyl groups of the Nam ribose in NAD+. In the hSARM1TIR structure, E642 forms hydrogen bonds with the C-2 and C-3 hydroxyl groups of glycerol (Fig. 1D). Attempts to generate crystals of hSARM1TIR or its mutants bound to NAD+-related ligands revealed the structure of hSARM1TIR bound to ribose (1.8-Å resolution) (fig. S10, A to E). The C-2 and C-3 hydroxyl groups interact with the E642 carboxylate, with other hydroxyls interacting with Y568, R570, and aspartic acid residue 594 (D594). A chloride occupies the position of the phosphate group attached to the C-5 atom in the Nam ribose of NAD+. The structure of hSARM1TIR(G601P) (MES bound; 1.7-Å resolution) (table S1) bound to a molecule of 2-(N-morpholino) ethanesulfonic acid (MES) (fig. S10, F to H) also shows interaction of E642, E599, and Y568 with the heterocyclic ring and of R569 and R570 with the sulfonic acid group of MES, which is located in a similar position to the chloride ion in the structure of an hSARM1TIR:ribose complex. MES has been found to mimic ligands in other nucleotide-binding proteins (21).

Attempts to capture NAD+-related ligands in crystals of RUN1TIR led to the structure of the NADP+ complex (Fig. 2, E and F, fig. S10I, and table S1). A molecule of bis-Tris [2,2-bis(hydroxymethyl)-2,2',2"-nitrilotriethanol], a component of the crystallization solution, is in the catalytic site, while NADP+ interacts with the periphery of the active site. The tryptophan at residue 96 (W96) forms a face-to-face aromatic stacking arrangement with the adenosine group of NADP+, with the C-2′ phosphate of the adenosine ribose in NADP+ forming H-bonds with R34, G35, E36, and R39. The analogous C-5′ phosphate protrudes from the binding site, and the nicotinamide mononucleotide (NMN) moiety has no interpretable electron density. The bis-Tris interacts with F33, G35, D60, and the catalytic E100 site. Mutations to residues in the proposed catalytic site affect effector-independent and effector-dependent HR in TIR domains and full-length NLRs, respectively (table S4). Conserved residue mutations R34A, S94A, W96A, and E100A in RUN1TIR also reduce NAD+ cleavage activity (fig. S7D). Double mutation of two arginine residues in the BB loop of RUN1TIR (R64A+R65A) increases NAD+ cleavage activity (figs. S7D and S9, C and D). The bis-Tris inhibits NADase activity (fig. S10J).

NADase activity dependence on SARM1TIR self-association

The disproportional increase of hSARM1TIR activity with concentration (fig. S4B) suggests that self-association may be important for enzymatic activity, consistent with known mechanisms of TIR domain function (3, 15, 22). In agreement with the SAM domains facilitating self-association of the TIR domains in hSARM1, we observed that both the SAM and TIR domains of Caenorhabditis elegans SARM1 (cSARM1tSAM-TIR) were required for NADase activity (Fig. 2B). The lower NADase activity of cSARM1TIR compared to hSARM1TIR is consistent with the delay in neuronal cell death in C. elegans (23).

The tandem SAM domains of human SARM1 (hSARM1tSAM; residues 409 to 561), produced in E. coli, were analyzed using size-exclusion chromatography (SEC) coupled with multiangle light scattering (MALS) (Fig. 3A) and small-angle X-ray scattering (SAXS) (fig. S11). The findings suggested an octameric arrangement. In the crystal structure of hSARM1tSAM (2.7-Å resolution; table S1), each of the two SAM domains (hSARM1SAM1 and hSARM1SAM2) adopts a characteristic five–α-helix bundle (α1 to α5) separated by a 10–amino acid linker (residues 477 to 486) (figs. S12 and S13 and table S5). In the asymmetric unit of the crystal, eight copies of hSARM1tSAM form a ring (Fig. 3B). There are three major protein interaction interfaces in the octamer: an intramolecular hSARM1SAM1:hSARM1SAM2 interface (1034 Å2) and two intermolecular interfaces between the hSARM1SAM1 domains (966 Å2) and between the hSARM1SAM2 domains (685 Å2) (fig. S14 and table S6). Typically, SAM domains form open-ended polymeric structures (24), but the rigid association of the two SAM domains in hSARM1 results in ring formation (fig. S15).

Fig. 3 Octameric structure of hSARM1tSAM is important for function.

(A) Solution properties of hSARM1tSAM (wild type; red) and hSARM1tSAM(5Mut) (blue), analyzed by SEC-MALS. Peaks indicate the traces from the refractive index (RI) detector during SEC; the lines under the peaks correspond to the average molecular mass distributions across the peak. The average molecular mass of hSARM1tSAM is 169.8 kDa (± 0.5%), consistent with an octamer (theoretical molecular mass 161.9 kDa). (B) Cartoon representation of the octameric ring assembly of hSARM1tSAM molecules. (C) Residues mutated in hSARM1tSAM(5Mut). (D) Exogenously expressed mSARM1wt, but not mSARM15Mut, restores the ability of Sarm1−/− neurites to degenerate after cut. Constructs for mouse SARM1 variants were injected along with DsRed vector (40 ng μl−1) to allow visualization of neurites of the injected neurons. The percentage of intact neurites at 24 hours after cut, relative to those present at the time of transection, is plotted (left, individual values and means ± SEM are shown). Representative images of cut neurites at the time of transection (0 hours) and at 24 hours are shown (right). mSARM15Mut does not promote degeneration of cut Sarm1−/− neurites even at the highest concentration used [n.s., not significant (P > 0.05); ***P < 0.001, separate one-way analyses of variance for each concentration with Tukey’s multiple-comparison tests]. mSARM15Mut is expressed at the expected size and at a higher level than mSARM1wt, both in injected SCG neurons and transfected HEK cells (fig. S25).

SARM1TIR behaves as a monomer in solution (based on SEC-MALS) (fig. S16A). Analysis of crystal packing of hSARM1TIR, to demonstrate whether SAM domain-mediated self-association contributes to NADase activity, revealed two major self-association interfaces: an asymmetric BB loop–mediated one (796 Å2) analogous to that observed in the MALTIR filament (3) and a pseudosymmetric one (1452 Å2) similar to the AE interface in plant TIRs (mostly mediated by the αA helices) (15, 22) (Fig. 1, E and F, figs. S17 and S18, and table S7). Mutations in the BB loop (D594A, E596K, and G601P), αA helix (L579A), and EE loop (H685A) reduce the NADase activity of hSARM1TIR (Fig. 2A and fig. S4C), suggesting that both the BB-loop and AE interfaces observed in the hSARM1TIR crystals are functionally important.

In the crystal structure of the impaired BB-loop mutant hSARM1TIR(G601P) (“ligand-free”; 2.1-Å resolution; table S1), the BB loop interaction is modified whereas the AE interface is intact (fig. S19A). The BB loop folds over the catalytic cleft, with the lysine 597 residue (K597) inserted into the active site and interacting with E642 (via water molecules; no glycerol is present) (fig. S19B). This conformation may represent the inactive state of hSARM1TIR, presumably stabilized in this protein by the BB loop mutation. In agreement, the K597E mutant of hSARM1 is active in NAD+ depletion in axons (23). The structure of the inactive AE-interface mutant hSARM1TIR(H685A) (3.0-Å resolution) (table S1) reveals modified interactions between Y568, H685, and R570 in the AE interface (fig. S19C), yet the BB loop interface is intact. Molecular dynamics simulations of monomeric and oligomeric hSARM1TIR reveal that the active site region is less flexible in the oligomeric form (fig. S20 and table S8), suggesting that self-association may stabilize the BB loop and the αB helix region, reversing inhibition of the BB loop due to folding of K597 into the catalytic cleft. Together, our results suggest that both the BB loop and AE interface interactions are required to stabilize the fully active enzyme conformation.

NADase activity dependence on plant TIR self-association

Self-association is important for the NADase activity of hSARM1TIR and cell death activities of plant TIRs. In agreement, addition of Ni–nitrilotriacetic acid beads to 6xHis-tagged L6TIR and RUN1TIR (figs. S7C and S9B) and macromolecular crowding agents, including polyethylene glycol 400 (PEG400) and PEG3350 (simulating a crowded environment inside cells), stimulated the NADase activities of L6TIR and RUN1TIR and led to measurable activities for SNC1TIR, RPP1TIR, ROQ1TIR, RPS4TIR, RPV1TIR, and ROQ1TIR (fig. S7, E and F).

RUN1TIR behaves as a monomer in solution (fig. S16B). In the crystals of RUN1TIR, however, both the AE (1400 Å2) and DE (1140 Å2) interfaces common in plant TIR domains (15, 22) are observed (fig. S21), comparable to the crystals of Arabidopsis SNC1TIR (15). Mutations in the AE and DE interfaces of L6 perturb self-association of L6TIR in solution and its cell death signaling activity in planta (15, 25). Mutation of the catalytic E100 still abrogates NAD+ cleavage activity in RUN1TIR(RRAA) (fig. S7D).

Axon degeneration requires SARM1 SAM domain oligomerization

Based on the hSARM1tSAM crystal structure, we designed point mutations to prevent octamer formation. All the tested single-residue alanine mutations either resulted in insoluble proteins or showed no disruption of the oligomer (fig. S22). Therefore, we designed a mutant [hSARM1tSAM(5Mut)] with five hydrophobic interface residues converted to arginines or aspartates (L442R, I461D, L514D, L531D, and V533D) (Fig. 3C and fig. S14, E to G). This mutant protein is soluble and monomeric, based on SEC-MALS (Fig. 3A), as is its mouse counterpart mSARM1tSAM(5Mut) (fig. S16C). The equivalent mutations in cSARM1SAM-TIR abolish NADase activity (fig. S4D). We then compared the abilities of exogenously expressed full-length, wild-type mSARM1wt and mSARM15Mut to overcome the delay to injury-induced (Wallerian) degeneration of neurites in Sarm1−/− superior cervical ganglion (SCG) neuron cultures. Whereas mSARM1wt restored rapid Wallerian degeneration, mSARM15Mut was expressed but essentially nonfunctional (Fig. 3D and fig. S23). SAM-mediated oligomerization therefore plays a pivotal role in the axon degeneration activity of SARM1.

Implications of NADase activity for plant immunity

The observed NADase activity of plant TIR domains may play a role in the cell death function of plant NLRs, possibly by a mechanism similar to hSARM1. Mutation of the conserved catalytic glutamate in L6TIR-YFP, RUN1TIR-YFP, SNC1TIR-YFP, and RPS4TIR-YFP abrogated effector-independent HR detected by transient expression in N. benthamiana leaves (Fig. 4A and fig. S24A). RUN1TIR(RRAA), which has increased NAD+ cleavage activity, also showed increased HR of YFP and Myc fusion proteins (Fig. 4B and figs. S24B and S25).

Fig. 4 TIR domain functions in plants.

(A) Expression in N. benthamiana (N.b.) of catalytic glutamate mutants of L6TIR, RUN1TIR, SNC1TIR, and RPS4TIR. (B) Expression in N. benthamiana (left) and the eds1-1 mutant (right) of RUN1TIR-YFP and RUN1TIR(RRAA)-YFP, alone or coexpressed with NdEDS1-YFP. (C) Left, Expression in N. benthamiana of hSARM1tSAM-TIR, hSARM1TIR, hSARM1tSAM(5Mut)-TIR, hSARM1tSAM-TIR(E642A), hSARM1TIR(E642A), and hSARM1tSAM. (Middle and right) Expression in N. benthamiana and the eds1-1 of hSARM1tSAM-TIR and hSARM1tSAM-TIR(E642A). (D) Expression in N. benthamiana of fusion proteins of TIR domains from L6, SNC1, RPS4, and RUN1 to hSARM1tSAM and hSARM1tSAM(5Mut). (E) Expression in in N. benthamiana (top) and the eds1-1 mutant (bottom) of TIR domains and TIR-hSARM1tSAM fusion proteins of RUN1 alone or coexpressed with NbEDS1-YFP. (F) Expression of L6-YFP, L6MHV-YFP, L6E135A-YFP, and L6MHV-E135A-YFP in N. tabacum W38 (top) or transgenic W38 expressing AvrL567 (bottom).

Expression in N. benthamiana of hSARM1tSAM-TIR, but not hSARM1TIR or hSARM1tSAM, induced cell death (Fig. 4C and fig. S24C). Disruption of hSARM1tSAM-TIR oligomerization by introduction of the five mutations [hSARM1tSAM(5Mut)-TIR] abrogated cell death, indicating the need for SAM domain–induced hSARM1TIR self-association in the cell death process in planta. The E642A mutation in the hSARM1tSAM-TIR construct also disrupted cell death, implicating NAD+ cleavage in the process.

Fusion of the oligomerizing hSARM1tSAM to L6TIRd28-Myc, SNC1TIR-Myc, RPS4TIR-Myc, and RUN1TIR-Myc enhanced cell death induction of these proteins in N. benthamiana, while fusion of the nonoligomerizing hSARM1tSAM(5Mut) did not (Fig. 4D and fig. S24D). Thus, like hSARM1TIR neuronal degeneration activity, plant TIR-induced HR requires both the catalytic glutamate and the ability to self-associate.

Plant TIR-containing NLR signaling requires the downstream signaling component EDS1 (26). However, hSARM1tSAM-TIR induced HR when it was transiently expressed in N. benthamina eds1-1 knockout mutant lines (27) (Fig. 4C and fig. S24C), implicating a mechanism different to plant TIR-induced HR, or NADase activity involving a separate pathway distinct from the EDS1-mediated signaling pathway. By contrast, hSARM1tSAM fusions of plant TIR domains and RUN1TIR(RRAA) failed to induce HR in the eds1-1 plants (Fig. 4, B and E, and fig. S26).

Mutation of the conserved glutamate (E135A) in the full-length L6 NLR also abolished its ability to induce cell death in the presence of the AvrL567 ligand (Fig. 4F and fig. S24E). Likewise, this mutation abolishes effector-independent signaling by the constitutively active L6D541V variant (28). Consistent with the abrogated HR phenotypes observed for glutamate mutants, our data implicate TIR NADase activity in the cell death process.

Bioinformatic analysis of TIR domains

To analyze the prevalence of NAD+ cleavage activities among TIR domains and the relationships with enzymes such as MilB, we used a recently developed evolutionary model that is based on secondary structure (29) to construct a phylogeny of proteins structurally similar to hSARM1TIR and plant TIR domains [fig. S27; for protein families with limited sequence similarities, such as the TIR domains, phylogenetic analyses using sequence-based evolutionary models are not reliable (30)]. This analysis suggests that TIR domains are part of a large superfamily of enzymes that includes a number of structurally related proteins (which are not usually associated with TIR domains) that bind nucleotides (e.g., glycosyltransferases, nucleoside hydrolases, and flavodoxins) or carbohydrates (e.g., bacterial isomerases) in the analogous region of the protein.

Outlook

We describe the structural basis of self-association–dependent NAD+ cleavage by TIR domains from animal SARM1 and plant NLRs. TIR domains have a flavodoxin fold common to many proteins with diverse functions, and our bioinformatic analyses reveal that they are related to a number of proteins with enzymatic activities such as nucleotide and carbohydrate hydrolysis, consistent with the NAD+ cleavage activity observed for TIR domains. Some branches may have lost enzymatic activity but retained scaffolding protein interaction functions. TLR and adaptor proteins show no NADase activities [fig. S7E; (13)]; although TLRs have glutamate residues in the analogous region of the protein, these are not found in the same spatial locations (Fig. 1). Some bacterial TIR domains have been shown to possess NAD+ cleavage activity (31), and the equivalent glutamate appears to be conserved; however, the analogous regions of the bacterial TIR domains with known structures (1) differ from those in hSARM1TIR and plant TIR domains, with the Cα of the glutamate located >6 Å away (Fig. 1).

In SARM1, the self-association of TIR domains required for NADase activity is at least in part facilitated by the octamer of SAM domains. The ARM domain (by interaction with the SAM and TIR domains) is suggested to hold hSARM1 in an inactive state (12); NMN has been proposed to remove this autoinhibition after axonal injury (32), allowing the TIR domains to self-associate, which activates the NADase activity and leads to axon degeneration (fig. S28A). Full-length SARM1 has been reported to self-associate (8), suggesting it may be constitutively oligomeric.

A structural model of activation may include the following steps: Before activation, the interaction of BB loop K597 with E642 prevents enzymatic activity, and the N-terminal domain of hSARM1 may stabilize this conformation. Upon axon injury, SAM domain assembly facilitates TIR domain association [analogous to animal NLRs causing caspase recruitment domain association (33)] through the BB loop interface, removing K597 from its inhibitory interaction. Further TIR domain association through the AE interface facilitates the interactions of H685 and R570, leading to optimal configuration of the active site, NAD+ cleavage, and axon degeneration.

In plant NLRs, we propose the nucleotide-binding domains (NBDs) (34) facilitate self-association of TIR domains through AE and DE interfaces (15, 16). Downstream signaling by plant NLR TIR domains remains a mystery despite substantial efforts by many research groups, with no direct signaling partners identified. The observed NAD+ cleavage activity of TIR domains could represent the SCAF function (22) responsible for the signaling event. Plant TIR domain cell death signaling is EDS1 dependent (26). However, hSARM1tSAM-TIR induces cell death in eds1-1 N. benthamiana lines; hSARM1TIR hydrolyzes NAD+ much more rapidly than the plant TIR domains examined and thus may cause necrosis through depletion of NAD+, as suggested for axons. Plant TIRs may signal cell death via a more controlled pathway mediated by EDS1 (fig. S28B). Alternatively, the NAD+ hydrolysis catalyzed by plant TIRs may be part of a pathway different from the EDS1-mediated signaling pathway.

Consequently, the products of NAD+ hydrolysis may be involved in cell death signaling. Cyclic ADPR (cADPR), produced by hSARM1TIR (13) and CD38 (35), has been shown to stimulate Ca2+ influx as part of NLR-mediated HR through Ca2+ channels and is also involved in both abscisic acid and salicylic acid signaling pathways (36). NaADP is also involved in Ca2+ signaling in plants, although the enzymes responsible for its synthesis are unknown (37). The NBDs could play a role by not only providing an oligomeric platform for TIR domain self-association but also binding of the ADP group of NAD+ and removing the cleaved ADPR group. Proteins structurally related to TIR domains have nucleotide transfer activities (e.g., ADP ribosylation), and many ADP-ribosylases show low enzymatic activity in the absence of the target of ribosylation (38), which could explain the low NADase activities of plant TIR domains. The low NADase activities we observe may be due to poor self-association abilities of isolated plant TIR domains (supported by the stimulation of activity with affinity beads or macromolecular crowding reagents and in cell lysates, or NAD+ may not be the preferred substrate [many NADases show high substrate promiscuity (39)].

In summary, we show that NAD+ cleavage by TIR domains is a conserved feature of animal and plant cell death signaling pathways, but the differences in structural organization of the relevant proteins and in cellular contexts result in distinctive mechanisms. Our results provide a foundation for future work on the role of NADase activity in plant immunity and development of inhibitors of axon degeneration.

Supplementary Materials

science.sciencemag.org/content/365/6455/793/suppl/DC1

Materials and Methods

Figs. S1 to S28

Table S1 to S8

References (4088)

References and Notes

Acknowledgments: We acknowledge the use of the University of Queensland Remote Operation Crystallization and X-Ray Diffraction (UQ-ROCX) Facility and the macromolecular crystallography (MX) and small/wide-angle X-ray scattering (SAXS/WAXS) beamlines at the Australian Synchrotron, Victoria, Australia. MD simulations were performed on the High Performance Computing cluster “Gowonda” at Griffith University. We thank V. Masic and N. Deerain for technical contributions. Funding: The work was supported by the National Health and Medical Research Council (NHMRC grants 1107804 and 1160570 to B.K. and T.V.. 1071659 to B.K., and 1108859 to T.V.), and the Australian Research Council (ARC grants DP160102244 and DP190102526 to B.K. and P.N.D.). B.K. was an NHMRC Principal Research Fellow (1110971) and ARC Laureate Fellow (FL180100109). T.V. received ARC DECRA (DE170100783) funding and S.J.W. received ARC DECRA DE160100893 funding. J.C. received a Chinese Scholarship Council (CSC) postgraduate scholarship. Y.S. was a Griffith University postdoctoral fellow. J.G. was supported by the UK Medical Research Council and M.P.C. was supported by the John and Lucille van Geest Foundation. M.K.M. was supported by the Australian Government Research Training Program (RTP). Author contributions: S.H., H.B., X.Z., M.K.M., Y.S., J.G., R.O.H., T.B., S.J.W., T.V., P.D., and B.K. designed the research; S.H., H.B., X.Z., M.K.M., Y.S., J.G., J.C., L.W.C., T.Q., J.S.L., W.G., M.X.R., D.J.E., G.F., R.O.H., T.B., and T.V. performed the research; S.H., H.B., X.Z., M.K.M., Y.S., J.G., L.W.C., R.O.H., T.B., J.S.L., M.X.R., D.J.E., G.F., M.v.I., J.P.R., J.D.N., M.B., I.B.D., B.J.S., S.J.W., M.P.C., T.V., P.N.D., and B.K. analyzed the data; S.H., H.B., X.Z., M.K.M., Y.S., T.V., and B.K. wrote the paper; all authors edited and contributed to writing. Competing interests: B.K. is a consultant for Disarm Therapeutics. B.K. and S.H. receive research funding from Disarm Therapeutics. Data and materials availability: Coordinates and structure factors for all crystal structures determined in this study have been deposited in the Protein Data Bank with IDs 6O0S, 6O0T, 6O0R, 6O0Q, 6O0U, 6O1B, 6O0V, and 6O0W.
View Abstract

Stay Connected to Science

Navigate This Article