Why Some Neurons Break While Others Bend: Redox Thresholds Hold the Answer

For decades, neuroscientists have wrestled with one of the most puzzling contrasts in biology: why neurons in the central nervous system (CNS) shatter under stress while their counterparts in the peripheral nervous system (PNS) often bend, survive, and regrow. Now, in a provocative Essay published in PLOS Biology, Shuji Wakatsuki and Toshiyuki Araki of the National Center of Neurology and Psychiatry in Tokyo propose a framework that may finally explain this asymmetry—not as a simple difference in antioxidant strength, but as a difference in multilayered signaling thresholds that decide when reactive oxygen species (ROS) become irreversible degenerative signals.

The idea arrives at a critical moment. Oxidative stress has been implicated in virtually every major neurodegenerative disorder, from Alzheimer’s and Parkinson’s diseases to amyotrophic lateral sclerosis (ALS) and peripheral neuropathies. Damaged proteins, oxidized lipids, and injured nucleic acids accumulate in diseased neural tissue, and more than 30 causative genes have been identified for ALS alone. Yet mutations in any single gene explain only a small fraction of cases, and clinical trials of broad-spectrum antioxidants have repeatedly disappointed. Wakatsuki and Araki argue that this failure reveals a fundamental truth: the biological consequences of ROS are determined not merely by their abundance, but by the molecular architecture that interprets and transduces redox signals within neurons.

At the heart of the new framework is a distinction that has often been blurred in the literature. ROS themselves are not simply toxic byproducts; they are physiological signaling molecules that regulate cell proliferation, immunity, autophagy, metabolism, and tissue repair. Oxidative stress, by contrast, denotes a condition in which ROS production exceeds the cellular capacity for detoxification. The presence of ROS therefore does not by itself specify whether a neuron will adapt, degenerate, or die. What matters is the threshold—an emergent property of a signaling network shaped by signal amplitude, duration, subcellular location, prior priming, feedback, and the simultaneous engagement of independent pathways.

The authors organize this threshold logic into four partially overlapping layers. The first is ROS generation and oxidative input, driven by the intense mitochondrial activity required to meet neurons’ high energy demands and by NADPH oxidase (NOX) activity. The second is antioxidant buffering and adaptation, mediated by pyridine nucleotide systems (NAD/NADH and NADP/NADPH) and thiol-based redox couples such as glutathione (GSH/GSSG) and cysteine/cystine. The third is redox-sensitive signal detection and molecular gating, in which oxidative modifications of proteins alter enzymatic activity, interactions, localization, and stability. The fourth comprises cellular outcomes—degeneration, survival and adaptation, or repair and regeneration. Crucially, downstream commitment mechanisms need not themselves be directly redox-sensitive; rather, they help set the effective threshold by determining whether an oxidative perturbation is tolerated, amplified, converted into structural degeneration, or followed by successful clearance and repair.

Threshold logic of this kind is well established in other biological systems. In the DNA-damage response, the duration and temporal pattern of p53 activity can direct recovery, arrest, senescence, or apoptosis even when cumulative signal exposure is similar. Individual p53 target promoters decode amplitude, duration, and frequency using different threshold and filtering properties. In T cells, distinct pathways downstream of the same antigen receptor show different activation thresholds, allowing signal strength to be translated into graded and digital outputs. The parallels do not imply that neuronal oxidative signaling uses the same molecular circuitry, but they illustrate a general design principle: biological specificity can reside in the architecture and dynamics of signal processing rather than solely in the identity or amount of the initiating stimulus. Neurons are especially suitable for such regulation because signals must be integrated across long distances, highly specialized compartments, and long lifetimes—meaning that local redox events in an axon, synapse, mitochondrion, or glial contact may be interpreted differently even within the same cell.

The buffering layer reveals why antioxidant capacity cannot be reduced to a single variable. In the CNS, mature forebrain neurons can exhibit epigenetic repression of the transcription factor NRF2, whereas astrocytes retain strong NRF2-dependent transcriptional responses, and neuronal activity can induce NRF2-dependent glutathione metabolism genes in astrocytes—illustrating that antioxidant protection in the brain is dynamically distributed across cell types. Genome-wide CRISPR interference and activation screens in human induced pluripotent stem cell-derived neurons have identified glutathione peroxidase 4, selenocysteine metabolism, lysosomal function, iron handling, and autophagy as major determinants of survival under chronic oxidative stress. Prosaposin, a lysosomal precursor protein required for sphingolipid degradation, proved particularly important: its deficiency produced a neuron-selective pathway leading from lysosomal failure and iron trapping to lipid peroxidation and ferroptosis. Antioxidant capacity, in other words, is a coupled network linking thiol metabolism, organelle quality control, iron homeostasis, and lipid composition.

The PNS, meanwhile, is not simply biased toward antioxidant protection. Schwann cells can support axons and adopt repair-promoting states after injury, but they can also generate spatially restricted oxidative signals that amplify neuroinflammation—a Schwann cell TRPA1–NOX1 circuit sustains macrophage-dependent oxidative signaling after peripheral nerve injury. NRF2 activation can be protective in experimental peripheral neuropathy, yet these observations do not establish that basal NRF2 activity is globally stronger in every PNS neuron or glial state. CNS and PNS tissues may instead utilize different combinations of neuronal and glial redox-buffering mechanisms that vary according to the type and temporal stage of injury.

The gating layer provides some of the most mechanistically detailed examples. Oxidation of catalytic cysteine residues in protein phosphatases can reversibly suppress their activity, prolonging receptor- and MAPK-dependent signaling; in neurons, reversible oxidation of ERK-directed phosphatases can convert oxidative stress into sustained ERK activity and toxicity. Redox modifications can also activate, rather than inhibit, signaling enzymes: glutathione S-transferase pi 1-mediated S-glutathionylation of Aurora-A at Cys290 promotes its dimerization and autophosphorylation, coupling localized redox fluctuations to MEK–ERK signaling and postsynaptic maturation. The ubiquitin system is similarly regulated by redox-mediated modification—nitrosative or oxidative modification of the cysteine-rich E3 ligase parkin can alter and subsequently impair its activity, linking redox stress to defective protein quality control in models of Parkinson’s disease. These gates filter out weak, transient, or spatially inappropriate signals through priming phosphorylation, scaffold assembly, compartmental recruitment, relief of autoinhibition, or loss of an inhibitory metabolite. The number, order, and subcellular localization of required events may be as important as the absolute abundance of ROS.

Among the pathways linking oxidative stress to neuroaxonal degeneration, the E3 ubiquitin ligase ZNRF1 provides a particularly clear worked example. ZNRF1 is expressed in neurons in both the CNS and PNS and participates in axonal degeneration in response to oxidative stress, ischemic injury, and axon severance. ROS-dependent activation of EGFR promotes phosphorylation of ZNRF1, inducing its ubiquitin ligase activity. Activated ZNRF1 ubiquitinates AKT and promotes its proteasomal degradation, releasing GSK3β from AKT-mediated inhibition and activating GSK3β-dependent pathways that destabilize microtubules and drive neurite fragmentation. The events upstream of ZNRF1, however, are not identical in the experimental CNS and PNS paradigms examined. In cultured central neurons exposed to oxidative stress, ROS-dependent EGFR activation seems directly coupled to ZNRF1 phosphorylation. In injured peripheral axons, by contrast, ZNRF1 activation additionally requires p38 MAPK-dependent phosphorylation of p47phox at Ser345, which promotes NOX-dependent ROS signaling. The authors interpret this additional requirement as a two-step gate: injury-associated p38 signaling supplies a permissive state, whereas NOX-derived ROS supplies the activating redox signal. This two-step architecture may reduce inappropriate pathway activation by requiring coincident signals before Wallerian degeneration proceeds.

Downstream of these gates lie commitment mechanisms that determine whether an upstream perturbation becomes irreversible. The dual leucine zipper kinase (DLK) pathway can serve as a signaling mediator for both protective and degenerative responses—supporting regenerative transcription after axonal injury, but also activating JNK–JUN-dependent degeneration and neuronal death. Stress-induced recruitment of DLK to palmitoylated, sphingomyelin-rich vesicular assemblies is required for local kinase activation, illustrating how subcellular organization creates a regulated signaling checkpoint. The NMNAT2–SARM1 system provides an even clearer commitment threshold: mitochondrial impairment can deplete axonal NMNAT2 and increase the NMN/NAD+ ratio, which SARM1 senses through its autoinhibitory armadillo repeat domain; once activated, its NAD+ hydrolase activity triggers a feed-forward metabolic catastrophe that commits the axon to degeneration. These pathways also demonstrate that degeneration and regeneration are not simply opposite ends of a single linear scale—a neuron may have a high threshold for somatic death but a low threshold for local axon degeneration, or it may undergo rapid Wallerian degeneration while retaining robust axonal regenerative capacity.

The therapeutic implications are profound. The limited clinical success of broad antioxidant therapies suggests that simply lowering ROS levels may be insufficient—and may also interfere with physiological redox signaling. The threshold framework instead identifies several possible intervention levels: increasing buffering in a vulnerable cell type, preventing pathological priming or scaffold assembly, inhibiting conversion of a redox signal into a degenerative program, or blocking the execution step after upstream stress has occurred. The DLK and SARM1 fields illustrate both the promise and the risk of this strategy. Pharmacological SARM1 inhibition protects axon structure and function in preclinical models, including chemotherapy-induced peripheral neuropathy. By contrast, a phase 1 study of the DLK inhibitor GDC-0134 in ALS identified dose-limiting sensory and ocular adverse events, emphasizing that chronic cell-wide inhibition of a multifunctional stress pathway may have unintended consequences. More selective interference with stimulus-dependent axonal DLK palmitoylation and localization has recently been shown to suppress pro-degenerative signaling while avoiding cytoskeletal effects associated with direct kinase inhibition—supporting the targeting of a disease-relevant gating event rather than eliminating an entire signaling system.

Wakatsuki and Araki are careful to bound their claims. The available ZNRF1 studies used different neuronal preparations and stress paradigms and did not generate matched ROS dose–response curves in CNS and PNS neurons; they support a model of distinct activation requirements but do not yet establish a quantitatively higher ZNRF1 activation threshold throughout the PNS. The relevant unit of comparison, the authors insist, is not the CNS neuron versus the PNS neuron, but a defined neuronal subtype in a defined compartment and injury context, together with its associated glia and immune cells. Single-cell profiling of dorsal root ganglia after peripheral and central axon injury has shown that neuronal and non-neuronal populations engage distinct transcriptional and intercellular programs, while spatial transcriptomic studies of human neurodegenerative tissue reveal region- and cell-state-specific molecular programs associated with selective vulnerability.

The next step for the field will be quantitative comparison. Experiments that measure input-output relationships at single-cell and subcellular resolution, under matched CNS and PNS conditions and in the presence of relevant glia, will help determine whether proposed gates truly alter the threshold for degeneration. If manipulation of p38 MAPK or p47phox Ser345 alters ROS production but does not change the relationship between local ROS levels and ZNRF1 activation, the pathway would be better regarded primarily as a mechanism controlling ROS production; if the pathway status alters ZNRF1 activation at comparable local ROS levels, this would support a genuine gating mechanism. Separating ROS generation from the molecular decisions that interpret ROS may ultimately reveal why apparently similar oxidative burdens produce divergent outcomes—and may help identify therapeutic targets that preserve physiological redox signaling while preventing pathological commitment to neurodegeneration.

Subject of Research: Oxidative signaling thresholds shape vulnerability in the central and peripheral nervous systems

Article Title: Oxidative signaling thresholds shape vulnerability in the central and peripheral nervous systems

Article References: Wakatsuki, S., & Araki, T. (2026). Oxidative signaling thresholds shape vulnerability in the central and peripheral nervous systems. PLOS Biology, 24(10), e3004032. https://doi.org/10.1371/journal.pbio.3004032

Image Credits: AI Generated

DOI: 10.1371/journal.pbio.3004032

Keywords: Oxidative, signaling, thresholds, shape, vulnerability, central, peripheral, nervous, systems, scientific research