The SARM1 Axon Degeneration Pathway: Control of the NAD₊ Metabolome Regulates Axon Survival in Health and Disease

Abstract

Axons are essential for nervous system function and axonal pathology is a common hallmark of many neurodegenerative diseases. Over a century and a half after the original description of Wallerian axon degeneration, advances over the past five years have heralded the emergence of a comprehensive, mechanistic model of an endogenous axon degenerative process that can be activated by both injury and disease. Axonal integrity is maintained by the opposing actions of the survival factors NMNAT2 and STMN2 and pro-degenerative molecules DLK and SARM1. The balance between axon survival and self-destruction is intimately tied to axonal NAD⁺ metabolism. These mechanistic insights may enable axon-protective therapies for a variety of human neurodegenerative diseases including peripheral neuropathy, traumatic brain injury and potentially ALS and Parkinson’s.

Wallerian Degeneration: 1850–2012

When an axon is physically separated from its cell body, the portion distal to the injury site undergoes a stereotypical fragmentation known as Wallerian degeneration. This field of research has its origin in experiments performed in the mid-19th century with the observation of the degeneration of the nerve fibers innervating the frog tongue after transection[1]. The stereotypic fragmentation of the nerve was originally believed to be due to a passive wasting of the nerve after its disconnection from its source of nutrients in the neuronal soma.

The first evidence that axon degeneration is an active rather than passive process came in 1989 with the serendipitous discovery of a peculiar mouse strain with drastically slowed Wallerian degeneration, which came to be known as Wallerian degeneration slow (WldS)[2]. In addition to nerve transection, other axonal insults also result in Wallerian-like axon degeneration[3]. Importantly, the WldS mouse provides protection from some of these insults, like the chemotherapy drug vincristine, but also in disease models such as progressive motor neuronopathy, suggesting the existence of a common program of axon degeneration that is activated in a variety of injury and disease conditions[4,5].

WldS encodes an unnatural fusion protein that includes the entire NMNAT1 protein, a nicotinamide mononucleotide adenylyl transferase that converts nicotinamide mononucleotide (NMN) into nicotinamide adenine dinucleotide (NAD⁺)[6]. The axon protection afforded by WldS is explained by the axonal localization and enzymatic activity of overexpressed NMNAT1 protein[7–9], implicating the NAD⁺ metabolic pathway in axonal health and destruction. Although treatment with various NAD⁺ precursors or NAD⁺ itself can result in axon protection and the enzymatic activity of WldS or NMNAT1 is required for protection, their expression does not result in higher NAD⁺ levels within the axon, as was expected from an NAD⁺-synthesizing enzyme[8,10,11].

A molecular understanding for the axoprotection afforded by misexpression of WldS/NMNAT1 awaited another seminal discovery, the identification of SARM1 as an essential pro-degenerative molecule[12]. Loss of SARM1 in fruit flies or mice provides potent cell-autonomous axon protection comparable to WldS expression, demonstrating that SARM1 is required for axon degeneration[12,13] (Fig. 1). SARM1 has N-terminal Armadillo-like repeats, two sterile alpha motifs (SAMs), and a Toll/interleukin-1 receptor (TIR) domain (Fig. 2). TIR domains are the signature signaling domain in innate immune signaling, so prior work focused on SARM1’s potential role in innate immunity[14].

Fig. 1.

A schematic describing the endogenous Wallerian degeneration pathway.

Fig. 2.

A. SARM1 contains an auto-inhibitory N-terminus, two tandem SAM domains and a C-terminal TIR domain. B. The crystal structure of SARM1’s SAM domains, which form an octomer (Image of PDB ID:6QWV[26], created with EzMol[76]). The side-view shows that the tandem SAM domains form two stacking rings in the octomer. C. A schematic of SARM1 at rest, where the N-terminus interacts with the TIR, preventing its dimerization, and after injury-induced activation, where the N-terminus-TIR interaction is disrupted and the TIR’s multimerize, leading to enzymatic degradation of NAD⁺. D. SARM1 cleaves NAD⁺ into Nam and ADPR, and can also cyclize ADPR into cyclic ADPR.

SARM1 is the founding member of a novel class of NAD-consuming enzymes

Our understanding of how SARM1 triggers axon degeneration emerged from detailed structure/function studies that came to the surprising conclusion that SARM1 has a novel enzymatic function that is essential for axon death. In healthy axons, SARM1 has an auto-inhibitory N-terminal region that restrains the degenerative capability of the two SAM domains and a C-terminal TIR domain[13] (Fig. 2). SARM1’s SAM domains mediate its multimerization and mimicking this function with forced dimerization of the TIR domain alone is sufficient to cause axon destruction and neuronal death, as well as a rapid decline in the levels of axonal NAD⁺[13,15], a phenomenon previously described during axon injury[10]. Importantly, a rapid loss of NAD⁺ is sufficient to drive axon degeneration[15], suggesting that activated SARM1’s ability to destroy axons is intimately linked to its regulation of NAD⁺ (Fig. 2). However, TIR domains function as scaffolding proteins in innate immunity signaling pathways[16], and so this known function provided no obvious clues to the mechanism by which SARM1 mediates NAD⁺ destruction.

After fruitless attempts to identify known NAD⁺-consuming enzymes that might cooperate with SARM1, the SARM1 TIR domain was demonstrated to have intrinsic NADase activity, cleaving NAD⁺ and producing nicotinamide, ADPR and cyclic ADPR[17] (Fig. 2). This enzymatic activity relies on a catalytic glutamate residue in the TIR domain and is essential for SARM1 to mediate axon degeneration[17] (Fig. 1). Interestingly, SARM1 is the major producer of neuronal cADPR, which also serves as a biomarker of SARM1 activity[18]. The identification of SARM1 as an enzyme redefined our understanding of the TIR domain, leading to the discovery of enzymatic TIR domains from multiple proteins in species as evolutionarily ancient as bacteria and archaea[19]. Remarkably, SARM1-triggered axon loss is the paradigm for an ancient cell death mechanism that is conserved across biological kingdoms. TIR-containing proteins in plants that mediate leaf cell death in response to pathogen recognition use the same highly conserved catalytic glutamate as SARM1’s TIR domain to degrade NAD⁺ and promote cell death[20,21].

While SARM1’s TIR domain promotes pathological rapid depletion of NAD⁺ and axon self-destruction after injury, SARM1’s TIR domain or other TIR-containing proteins could drive subtler cell signaling through NADase activity. In Drosophila and C. elegans SARM1 regulates signaling pathways[22,23], and yet the TIR enzyme function is conserved in invertebrates[17,24]. It will be important to investigate whether such signaling utilizes this NADase function. Indeed, such a non-degenerative signaling role may occur on the proximal axon side after injury in mammals, where SARM1 transmits a transcriptional immune response[25]. The potential for such a signaling role is supported by the recent finding that there is a low level of basal SARM1 NADase activity in uninjured neurons[18].

The structure of SARM1 has begun to be revealed during the past year. SARM1’s isolated SAM domains form an octameric ring that is necessary for its function[21,26] (Fig. 2). Crystal structures of SARM1’s TIR domain as well as related TIR domains from other organisms revealed a likely binding pocket for NAD⁺ containing the catalytic glutamate and described an interaction between the catalytic pocket and the BB loop of the TIR domain, a conserved region that participates in injury-induced SARM1 activation via an interaction with the auto-inhibitory N-terminus[21,27]. The precise interactions between SARM1’s N-terminus and TIR domain at rest and after injury-induced activation (Fig. 2) remain to be determined and will likely require structural characterization of the full-length protein.

The survival factor NMNAT2 restrains pro-degenerative SARM1 activity

Expression of WldS/NMNAT1 results in axon protection, but until recently it was unknown how its gain-of-function mechanism was connected to the function of the endogenous axon pro-degenerative protein SARM1. Over the past decade, NMNAT2 has emerged as the endogenous NMNAT enzyme present in healthy axons where it functions as an axon survival factor to restrain SARM1 from its destructive activity (Fig. 1).

NMNAT2 is a labile protein in axons and after injury it is rapidly degraded prior to axon fragmentation[28] (Fig. 1). Remarkably, loss of NMNAT2 from axons is sufficient to induce axon degeneration that is entirely dependent on SARM1. Nmnat2^−/− mice are embryonic lethal but they can be rescued by expression of wlds/NMNAT1, a more stable version of NMNAT2, or by genetic loss of sarm1[28–30].

The regulation of NMNAT2 protein levels is a major determinant of axon health. Axonal NMNAT2 can be palmitoylated, a post-translational modification affecting both its trafficking and stability[31]. The non-palmitoylated axonal NMNAT2 is regulated by an atypical E3 ligase complex consisting of PHR1, FBXO45 and SKP1[32] (Fig. 1). Reduction of these genes results in increased levels of NMNAT2 and potent axon protection[33–37]. Palmitoylated axonal NMNAT2 is differentially regulated by a neuronal JNK stress kinase pathway activated by the MAP3K DLK and its paralog leucine-zipper kinase/LZK[32,38,39] (Fig. 1). MAPK signaling promotes the degradation of palmitoylated NMNAT2 and genetic or pharmacologic inhibition of MAPK signaling results in increased NMNAT2 levels and axon protection[32,40]. Axonal stressors that activate DLK also decrease NMNAT2 protein levels, sensitizing axons to SARM1-dependent degeneration[41] (Fig. 1). Interestingly, even a partial decrease in NMNAT2 can lead to peripheral nerve axonopathy in aged mice[42].

The mechanism of SARM1 activation remains an open question. Loss of NMNAT2 is a common effect of many injuries that lead to SARM1 activation, including axotomy, treatment with chemotherapy drugs vincristine and bortezomib, and mitochondrial toxins[28,41,43,44]. Since loss of NMNAT2 results in SARM1 activation and the ability of NMNATs to protect axons is dependent on their enzymatic activity, it was proposed that the NMNAT2 substrate NMN can promote axon degeneration, potentially via SARM1 activation[45,46] (Fig. 3). Increased NMN levels occur in injured axons as NMNAT2 is lost prior to degeneration and inhibition of the NMN-generating enzyme NAMPT provides axonal protection[45].

Fig. 3.

The NAD⁺ metabolic pathway. In the axon, NMN is turned into NAD⁺ by NMNAT2. When activated, SARM1 degrades NAD⁺ into nicotinamide (Nam) and ADPR or cyclic ADPR (cADPR). NMN may activate SARM1 (indicated by dashed line). Nam can be converted back to NMN by NAMPT. Another source of NMN is NR phosphorylated by NRK. The bacterial enzyme NMN deamidase can convert NMN to its deamidated counterpart, nicotinic acid mononucleotide (NaMN). Nicotinic acid riboside (NaR) and nicotinic acid (Na) can be alternate sources of NaMN (via NRK or NAPRT, respectively). NaMN can be also be made into NAD⁺ via conversion to NaAD by NMNAT2 and then by NADsyn into NAD⁺.

A potential pro-degenerative role for NMN is bolstered by experiments showing potent axon protection after injury or genetic loss of Nmnat2 from expression of the bacterial NMN-consuming enzyme NMN deamidase[45,47,48] (Fig. 3). However, manipulations of the NAD⁺ pathway that result in high NMN don’t necessarily promote axon degeneration[49], and can even be compatible with robust axon protection (nicotinamide riboside (NR) treatment, or NAMPT expression)[47]. Limiting NMN accumulation while treating with the NAD⁺ precursor nicotinic acid riboside (NaR) can provide lasting axon protection from vincristine, suggesting that perhaps both lowering NMN and raising NAD⁺ are important for axon protection[49] (Fig. 3). Surprisingly, NMNAT1 and NMN deamidase both functionally protect axons by keeping SARM1 from being activated to degrade NAD⁺ after injury, but not via an obvious common effect on NMN or NAD⁺ levels[47] (Fig. 1).

NMN may promote axon degeneration by activating SARM1 (Fig. 3). Indeed, NMN stimulates SARM1-dependent Ca²⁺ entry into injured axons[46]. A recent study offered the most direct evidence yet that NMN can activate SARM1. A chemically-modified and cell-permeant version of NMN, CZ-48, induced SARM1 NADase activity and cADPR production in HEK cells, potentially through increasing multimerization of SARM1’s TIR domains[50]. CZ-48 or NMN itself could directly activate purified SARM1’s NADase activity[50]. These results await further investigation in neurons. Since CZ-48 is a close analog of NMN, it could also inhibit NMNAT2, which would indirectly alter SARM1 activation. If NMN can stimulate SARM1’s axon self-destruction behavior, it will be of great importance to determine its mechanism of activation and how axons can survive or even maintain protection in the presence of elevated levels of NMN[47].

The Wallerian axon degeneration pathway is activated in many neurodegenerative conditions and can be therapeutically inhibited

While Wallerian degeneration is a specific pathological response to injury, the SARM1 pathway mediating Wallerian degeneration also promotes axon loss in various neurodegenerative conditions. The WldS mouse provides functional improvement in models of glaucoma, ischemia, Parkinson’s and Charcot-Marie-Tooth neuropathy[51]. These phenotypes likely reflect inhibition of SARM1 by WldS. Importantly, loss of SARM1 is much more potent than expression of WldS in rescuing Nmnat2^−/− mice[52], implying that in mouse models of neurodegeneration that have been helped by WldS expression, the effect of SARM1 inhibition will be stronger.

Indeed, recent studies demonstrate that the absence of SARM1 is profoundly neuroprotective in a number of models of neurodegeneration including peripheral neuropathy. Peripheral neuropathy is the most common neurodegenerative disease, and involves a dying-back axon loss without cell body death, and so is a prime candidate to be mediated by SARM1. A series of studies demonstrate that SARM1 is essential for the development of chemotherapy-induced peripheral neuropathy in response to mechanistically distinct chemotherapeutics[43,53,54]. Loss of SARM1 is also protective in models of diabetic neuropathy[55]. These initial results suggest that SARM1-mediated axon destruction is a common mechanism in a variety of peripheral neuropathies and that its inhibition is a promising therapeutic strategy.

SARM1 also plays a role in central nervous system axon degeneration. After traumatic brain injury (TBI), mice lacking SARM1 have preserved neurological function and improved long-term axon integrity[56–59]. The axon degeneration pathway is also implicated in adult-onset neurodegeneration conditions such as amyotrophic lateral sclerosis (ALS). Activation of the pro-degenerative DLK MAP3K pathway is observed in a number of neurodegenerative mouse models and in Alzheimer’s and ALS patient samples[60], and activated DLK can sensitize axons to SARM1-destruction via lowering levels of the survival factors NMNAT2 and STMN2/SCG10, another labile axonal protein[41,61] (Fig. 1). Excitingly, direct links have recently emerged between two neurodegenerative diseases and the Wallerian degeneration pathway. Loss of STMN2 protein is a key downstream factor in TDP-43-associated ALS[62,63] and Parkinson’s disease[64] (Fig. 1), suggesting that the Wallerian degeneration pathway may be worth pursuing as a therapeutic approach in these neurodegenerative conditions. Thus far, mouse models of TDP-43- and SOD1-ALS have seen some or no benefit, respectively, from loss of SARM1[65,66], but the human data on STMN2 encourages further investigation.

The first mutations in endogenous Wallerian degeneration genes causing human disease were described in 2019. Missense mutations in NMNAT2 were found in two siblings with childhood onset polyneuropathy and accompanying erythromelalgia[67] and two stillborn siblings with fetal akinesia deformation sequence[68]. These NMNAT2 mutations range from partial to complete loss-of-function, with correlated severity of disease phenotypes[67,68]. The discovery of these mutations underscores the relevance of the Wallerian degeneration pathway to human disease and implies that even mild loss-of-function alleles in NMNAT2 or mutations in other genes in the pathway like SARM1 may directly cause or sensitize people to various neuropathies or neurodegenerative conditions.

Finally, if the Wallerian degeneration pathway is playing a pro-degenerative role in these various neurodegenerative conditions, it will be of great therapeutic interest to inhibit it. Manipulations that result in increased NMNAT2 should be axon protective and efforts are ongoing to create pharmacological DLK inhibitors or to find small molecules that alter NMNAT2 levels[32,69,70]. The discovery of enzymatic NADase activity in SARM1’s TIR domain also makes it an attractive candidate for the development of small molecule inhibitors. Recent insights into the SARM1 catalytic mechanism may facilitate the development of such an inhibitor[24]. In addition, AAV-mediated delivery of a potent dominant-negative SARM1 transgene provides profound in vivo axon protection[71], raising hopes for multiple lines of therapeutic modalities targeting the Wallerian degeneration pathway.

Conclusions

A comprehensive mechanistic understanding of the axon degeneration pathway emerged over the past five years. Essential axon survival factors like NMNAT2 and STMN2 are continuously supplied to axons and when this supply is lost or their levels are otherwise reduced, the pro-degenerative protein SARM1 is activated to destroy NAD⁺ and cause irreversible axon fragmentation (Fig. 1). The levels of these axon survival factors are tuned by key regulators, including an E3 ligase complex consisting of PHR1, FBXO45 and SKP1, a neuronal stress kinase pathway featuring dual MAP3Ks DLK and LZK, and now a central neurodegenerative disease protein, TDP-43 (Fig. 1). This mechanistic understanding is leading to the development of therapeutic targets for neurodegenerative disease.

Despite this rapid progress, several important questions remain. How are the axonal levels of the survival factors NMNAT2 and STMN2 regulated in vivo? How do DLK and the MAPK pathway promote NMNAT2 degradation? How does disruption of mitochondrial function result in lowered NMNAT2 and STMN2 and will other mitochondrial diseases result in SARM1 activation[44,72]? How is NMNAT2 palmitoylation regulated and what is the function of the different pools of palmitoylated vs. non-palmitoylated NMNAT2 in axons?

The details of SARM1 activation are also not yet fully understood. If NMN can activate SARM1, how does this occur? How do NMNAT1 and NMN deamidase keep SARM1 from being activated by injury? Are there other triggers for SARM1 activation? Is SARM1 multimerization regulated in a similar manner to other large signaling complexes like the inflammasome[73]? How does loss of NAD⁺ lead to axon fragmentation and are there meaningful steps downstream of SARM1, such as the yet unknown function of the Drosophila gene Axed[74]? Is axonal ATP loss in the axon a direct consequence of NAD⁺ decline and, if so, how does the resulting local energy deficit promote axon fragmentation[39,75]? Surprisingly, although there is a substantial increase in cADPR in injured axons that precedes morphological fragmentation, manipulations that raise or lower cADPR levels in axons do not change the time-course of degeneration, demonstrating that cADPR is likely not a pro-degenerative byproduct of SARM1’s cleavage of NAD⁺[18]. However, a potential pro-degenerative role for nicotinamide and ADPR, SARM1’s other byproducts, remains to be investigated.

There is great promise of a meaningful role for the SARM1 axon degeneration pathway in human disease. A major question in the field is which diseases show activation of the pathway and which approach will be the most fruitful for therapeutic intervention to promote axonal health and nervous system function. The identification of cADPR as a gene-sensitive biomarker of SARM1 activity in nerves and plasma neurofilament light chain (NfL) as a biomarker of axon fragmentation provides the first molecular assays of SARM1-dependent axon degeneration in vivo[18]. These biomarkers should aid in determining the status of SARM1 activation and the efficacy of therapies designed to inhibit the pathway. With the current pace of progress and the recent elucidation of the axon degeneration pathway, we can expect answers to many of these questions in the near future.

Highlights.

• A mechanistic understanding of the endogenous axon degeneration pathway has emerged over the past few years.

• Axonal health is maintained by a balance of axon survival factors NMNAT2 and STMN2 and pro-degenerative molecules DLK and SARM1.

• The pro-degenerative protein SARM1 is the founding member of the TIR-domain family of NAD⁺-consuming enzymes and this enzymatic activity of SARM1 is essential for axon degeneration.

• The endogenous axon degeneration pathway is activated in many neurodegenerative conditions and the pathway can be therapeutically targeted in several ways that are likely to be beneficial to treating human diseases.