Research paper

A conformation-specific nanobody targeting the nicotinamide mononucleotide-activated state of SARM1

Yun Nan Hou, Yang Cai, Wan Hua Li, Wei Ming He, Zhi Ying Zhao, Wen Jie Zhu, Qiang Wang, Xinyi Mai, Jun Liu, Hon Cheung Lee, Goran Stjepanovic, Hongmin Zhang, Yong Juan Zhao

Nature Communications 2022 Volume 13, article 7898, 15 pages

Tier 3 · Preclinical or mechanistic with translational relevanceNeurological

Source identifiers

DOI 10.1038/s41467-022-35581-y

PubMed PMID 36550129

Funding This study was supported by grants from the National Science Foundation of China (31871401 to Y.J.Z., 31871403 to J.L. and 31950410540 to S.G.), Ministry of Science and Technology (Synthetic Biology Special Project of National Key R&D Program, 2019YFA0906000 to H.Z.; and Foreign Youth Talent Program, QN2021032004L to S.G.), Shenzhen Science and Technology Innovation Committee (JCYJ20190808163411340 to Y.J.Z., JCYJ20210324125608023 to Y.J.Z.), Shenzhen-Hong Kong Institute of Brain Science-Shenzhen Fundamental Research Institutions (2019SHIBS0004 to Y.J.Z.) and Shenzhen-Hong Kong Cooperation Zone for Technology and Innovation (HZQB-KCZYB-2020056 to Y.J.Z.).

Declared competing interests The authors declare no competing interests.

Study snapshot

Design: Rigorous multi-method structural mechanistic work at a Nature-family journal, with cryo-EM, HDX-MS, XL-MS, and mutational validation converging on the same activation mechanism. Findings independently corroborate concurrent work from the Shi and Figley/DiAntonio groups. Not tier-2 because no in vivo or human data in this paper.
Population: No human subjects and no whole-animal work in this paper. Recombinant human SARM1 (N-terminally truncated at aa 27) expressed in Expi293F cells; HEK293 and HEK293T human cell lines for activity, imaging, and mutant characterisation; one alpaca for nanobody generation (immunisation performed by contract vendor).
Sample: Structural work: 4,605 cryo-EM movies; final particle stack 208,299 particles at 2.7 Å overall; focused class 1 = 298,862 particles at 3.3 Å, class 2 = 133,097 particles at 3.4 Å. Biochemical assays: n ≥ 3 biological replicates throughout; imaging quantification n = 20 cells across 3 biological samples.
Intervention: Nanobody Nb-C6 developed as a conformation-specific probe of NMN-activated SARM1. Activation triggers tested: recombinant NMN; the cell-permeant NMN mimetic CZ-48; low-pH acid activation as a control. Negative controls: irrelevant nanobody Nb-1053 (anti-CD38); NAD-bound SARM1 (autoinhibited state). Twelve SARM1 point mutants generated to probe NMN binding pocket (W103A, R110A, K193M, D317A, D317R) and ARM-domain bending residues (E189Q, L257C, S319F, S319Y, Q320A, Q320Y, F476C).
Endpoints: Three-dimensional structure of NMN-activated SARM1 (deposited as PDB 8GQ5, 8GNI, 8GNJ; EMDB EMD-34198, EMD-34165, EMD-34166); NMN-binding residues and inter-domain interface residues identified; Nb-C6 binding affinity by SPR (KD ≈ 25 nM); SARM1 enzymatic activity (NAD-cleavage rate; cADPR production) in wild-type vs mutants; Deuterium-uptake differences between NAD- and NMN-bound states; XL-MS restraint compatibility with alternative conformations in solution

What the study found, in plain language

This 2022 structural biology paper resolves, at atomic-level detail, how nicotinamide mononucleotide (NMN) switches on the enzyme responsible for axon self-destruction. It provides the mechanistic explanation for the finding first reported by Di Stefano and colleagues in 2015 — that NMN accumulation, when the enzyme NMNAT2 is depleted after nerve injury, drives axons to degenerate.

The protein at the centre of this mechanism is SARM1 (Sterile alpha and Toll/interleukin-1 receptor motif-containing 1). In healthy neurons SARM1 sits in an autoinhibited "donut" shape on the outer mitochondrial membrane. When NMN accumulates and the ratio of NMN to NAD+ tips past a threshold, SARM1 activates and starts cleaving NAD+, collapsing the axon's energy metabolism. Genetic deletion of SARM1 protects axons from Wallerian degeneration in every model organism tested.

The authors, working at Peking University Shenzhen, Southern University of Science and Technology, and the Chinese University of Hong Kong (Shenzhen), tackled a long-standing structural problem: the active form of SARM1 is too floppy to image directly. They solved it by generating a nanobody — a single-domain antibody derived from an alpaca — that binds only the NMN-activated conformation of SARM1 and locks it in place long enough to image. With this tool, they resolved the NMN-activated SARM1 structure by cryo-electron microscopy at approximately 2.7 Å resolution and mapped its solution dynamics with two orthogonal mass-spectrometry techniques (hydrogen-deuterium exchange and crosslinking).

The picture that emerges: NMN binds an allosteric pocket on SARM1's ARM (Armadillo repeat) domain, at the same site NAD+ normally occupies as an autoinhibitor. The two ligands orient differently in the pocket, and NMN binding drives the ARM domain to bend inward, breaking key contacts with the neighbouring SAM domain, releasing the outer ring of the octameric SARM1 assembly. The autoinhibited "donut pie" opens into a "blooming lotus." In solution, the freed TIR (Toll/interleukin-1 receptor) domains transiently dimerise, which is what activates their NADase activity. Site-directed mutagenesis of the NMN-binding pocket and the bending-hinge residues abolishes NMN-triggered activation, confirming the structural model.

For the neurological science surrounding NMN, this paper matters because it identifies — at atomic resolution — how NMN accumulation triggers programmed axon death in the specific pathological context first documented by Di Stefano and colleagues. The mechanism is now sufficiently well-defined that pharmaceutical companies (including several publicly disclosed programmes) are actively developing SARM1 inhibitors as clinical candidates for peripheral neuropathies and other axonopathies.

Key findings

Nb-C6 is a conformation-specific probe. The nanobody binds NMN-activated SARM1 with a dissociation constant of approximately 25 nM but binds inactive NAD-bound SARM1 minimally. Its binding is a linear function of the NMN:NAD ratio in solution, mirroring the physiological trigger for SARM1 activation.
Full-length NMN-activated SARM1 structure resolved. Cryo-EM of the SARM1/NMN/Nb-C6 complex at approximately 2.7 Å overall resolution reveals an octameric ring with ARM and TIR domains swung outward. Two conformational sub-classes are resolved by focused classification.
NMN and NAD occupy the same allosteric pocket but adopt opposite orientations. NMN's nicotinamide ring orients downward; NAD's orients upward. Seven residues form the binding pocket (W103, R110, E149, Q150, R157, H190, K193, S316, G321). Point mutation of any of the direct-contact residues abolishes NMN-triggered activation while preserving acid-triggered activation.
NMN binding drives inward ARM-domain bending. Superimposing NMN- and NAD-bound structures shows the ARM7-8 motifs shift inward and the whole ARM domain rotates, with a root-mean-square deviation of approximately 2.4 Å. Mutants engineered to block this bending (S319F/Y, Q320A/Y, E189Q) are unresponsive to NMN; one mutant (L257C) is constitutively active, consistent with a hydrophobic-interaction requirement for the closed state.
ARM bending breaks the ARM-SAM interface. In NAD-bound SARM1, ARM residues R376 and Y380 anchor to SAM residues E469 and F476. In NMN-bound SARM1, Y380 flips away from F476 and only one interchain salt bridge (R285-D489) survives. F476C mutation renders SARM1 constitutively active, confirming this interface's autoinhibitory role.
Solution dynamics reveal transient TIR dimers. HDX-MS shows that helix α13 in the ARM-TIR interface samples an open state approximately twice as frequently in NMN-bound versus NAD-bound SARM1. XL-MS identifies K636-K694 and K602-K694 crosslinks compatible only with intermolecular TIR-TIR dimers using the BB loop as the interface — the same interface implicated by concurrent independent work.
Three-state activation model. The authors propose: (1) inactive "donut pie" (NAD-bound, autoinhibited); (2) intermediate "blooming lotus" (NMN-bound, ARM released from SAM, TIR still tethered — captured by Nb-C6); (3) active "bloomed lotus" (TIR domains released and dimerised, NADase active).
Convergent evidence across concurrent groups. A parallel structure of NMN-bound SARM1 published during this work's submission (Shi et al., PDB 7NAL) shows the same NMN binding mode with an RMSD of 0.93 Å over ARM residues V61-E400, differing only in the rotation/movement angles between activation intermediates.

What this study can and cannot tell us

Readers should note the following in interpreting this paper's relevance to NMN supplementation:

No animal or human data. All experiments are performed on recombinant human SARM1 expressed in Expi293F cells and on human cell lines (HEK293, HEK293T). The paper establishes the molecular mechanism; it does not itself test whether oral NMN supplementation activates SARM1 in vivo or has clinically meaningful neurological effects.
Mechanism operates via the NMN/NAD ratio, not absolute NMN concentration. Nb-C6 binding depends on the molar ratio of NMN to NAD+, not on NMN alone. In healthy tissue with functional NMNAT2, exogenous NMN is rapidly converted to NAD+, and the ratio does not shift into the activating range. SARM1 activation is expected in contexts where NAD+ synthesis is impaired (injury; NMNAT2 loss-of-function; possibly ageing tissues; possibly during specific drug exposures such as vincristine or paclitaxel that damage NMNAT2). The paper does not resolve whether pharmacological plasma-NMN elevation from oral supplementation can meaningfully perturb the intracellular ratio in intact axons.
Structure captures an intermediate, not the final active state. Nb-C6 stabilises SARM1 in a partially activated conformation in which TIR domains have swung out but remain tethered to ARM domains and do not yet dimerise stably. The final active state — with TIR-TIR dimers formed and NADase activity fully engaged — is inferred from HDX-MS, XL-MS, and orthogonal work by Shi et al. using a different stabilisation strategy, not directly observed in this paper's cryo-EM density.
Nb-C6 itself partially activates SARM1. Because Nb-C6 stabilises the active conformation, its own presence produces modest SARM1 activation and cADPR increase in cells. This means Nb-C6 cannot be used as a purely passive probe in cellular contexts, and the immunofluorescence signal partly reports on a probe-induced state rather than the physiological activation dynamics alone.
Recombinant SARM1 is N-terminally truncated. The recombinant constructs lack the first 27 amino acids that encode the mitochondrial localisation signal, which is necessary for imaging and biochemical work but means the structural conclusions do not directly address possible regulation by mitochondrial-membrane context.
Translational implications remain untested clinically. The paper's therapeutic framing — that SARM1 inhibition may be useful in axonopathies — is prospective. Multiple SARM1-inhibitor programmes are in preclinical or early clinical development at other institutions, but no approved SARM1-targeted therapy exists at the time of this paper's publication or at time of last verification.