How neurons cope with oxidative stress

1.1 Consequences of excessive oxidative stress

The cellular targets of ROS are very diverse. Firstly, the double bonds present in unsaturated fatty acids such as linoleic or arachidonic acid are very vulnerable to attacks by radicals which lead to lipid peroxidation. This initial reaction produces highly reactive lipid peroxy radicals which can cause further damage to nearby lipids. In addition, this chain reaction triggers the generation of breakdown products such as 4-hydroxy-2,3-nonenal (HNE), malondialdehyde and F2-isoprostanes (Barnham et al. 2004). Secondly, proteins can be altered in several ways by oxidative stress. For example, proteins can be modified by breakdown products like HNE, which can crosslink to cysteine, lysine or histidine residues, they can become S-glutathionylated by oxidized glutathione, or they can be carbonylated. Proteins containing labile iron-sulfur clusters such as the mitochondrial aconitase can also lose their iron. Finally, all four bases within DNA can be damaged by oxidative stress via hydroxylation (Gabbita et al. 1998).

The extent of oxidative stress in cells can be measured directly by determining ROS levels via various fluorescent probes (e.g. CellROX (Tian et al. 2021) or CM-H₂DCFDA (Holzerova et al. 2016)), or indirectly either by analyzing the levels of peroxidized lipids (e.g. measured via Liperfluo (Tian et al. 2021) or C11-BODIPY (Tian et al. 2021)), the levels of by-products of lipid peroxidation (e.g. HNE (Smith et al. 1998; Wei et al. 2023)), the level of altered proteins (e.g. measured as protein carbonyl levels (Ferrante et al. 1997)), the amount of damaged DNA (e.g. measured via the deoxyguanosine oxidation product OH⁸dG (Ferrante et al. 1997)) or by analzying the presence of DNA damage response proteins like gH2ax (Junghans et al. 2022; Wei et al. 2023)).

In agreement with the diverse cellular targets of ROS, cellular function and thus cell survival will be compromised at different levels upon oxidative stress exposure. If the damage to lipids cannot be repaired, lipid peroxidation will trigger the ferroptosis cell death program (Stockwell et al. 2017). In addition, breakdown products such as HNE have been reported to initiate apoptosis by activating the kinases c-Jun aminoterminal kinase (JNK) and mitogen activated protein kinase (MAPK) p38 (Tamagno et al. 2003). Damaged DNA can induce the DNA damage response which can lead to apoptosis if the damage exceeds the cell’s repair capabilities. Altered proteins can either lose their functionality or be differentially regulated. For striatal neurons, for example, it has been shown that the S-glutathionylation of the channel TRPC5 causes its activation and thereby an increased Ca²⁺ influx which leads to striatal neuron death (Hong et al. 2015). Therefore, it is essential for all cells to counteract ROS with antioxidant programs.

1.2 Neuronal vulnerability to oxidative stress

If all cells are vulnerable to ROS, why is it of special interest to study how neurons cope with oxidative stress? In fact, the brain is especially vulnerable to oxidative damage, and there are several reasons for this: First of all, neurons have a higher energy demand than most other cells in order to fulfill their various functions such as activating ion transport proteins to maintain the membrane potential, triggering action potentials, releasing neurotransmitters and recycling synaptic vesicles. This high energy demand is reflected by the fact that the brain, despite constituting only about 2 % of our body weight, accounts for 20 % of the overall oxygen consumption. This increased metabolic activity is accompanied by an elevated production of ROS. Secondly, neurons often harbor high levels of redox-active metals, especially iron, which not only supplies iron-sulfur clusters for mitochondrial respiration, but is also involved in myelin synthesis, and neurotransmitter metabolism (Ward et al. 2014). As components of metallo-enzymes redox-active metals are exploited for activating O₂, however, in an unbound state they can also act outside of these metabolic contexts and contribute to ROS production. An example is the Fenton reaction which leads to the decomposition of H₂O₂ generating highly reactive hydroxyl radicals (Jomova et al. 2012). Thirdly, neuronal membranes are rich in poly-unsaturated fatty acids that contain double bonds, which are easily damaged by ROS. Finally, the brain has a very limited potential for regeneration since neurons are very long-lived cell types making oxidative damage more prone to accumulate, and dead cells are not easily replaced. Interestingly, neurons do not only differ from other cell types in their degree of vulnerability to oxidative stress, but also among themselves. For instance, dopaminergic neurons are more sensitive to oxidative stress than cortical neurons (Wei et al. 2023).

1.3 Oxidative stress in neurodegenerative diseases

The importance of a well maintained redox homeostasis for neuronal life is also underlined by the fact that numerous neuronal diseases are associated with an increase of oxidative stress markers, including Parkinson’s disease (Yoritaka 1996), Alzheimer’s disease (Subbarao et al. 1990), Huntington’s disease (Sorolla et al. 2008), Amyotrophic lateral sclerosis (Cunha-Oliveira et al. 2020; Ferrante et al. 1997), epilepsy (Kalita et al. 2019), ataxias (Lupoli et al. 2018) and traumatic brain injury (Varma 2003).

In Alzheimer’s disease, for example, there is increasing evidence suggesting that brain tissue of patients shows elevated ROS levels throughout the disease progression (Gella and Durany 2009). Altered mitochondria are believed to play a significant role in ROS generation in Alzheimer’s disease, and dysregulation of the electron transport chain increases the level of aggregated Aβ and neurofibrillary tangles, which in turn further stimulate ROS production. Subsequently, neuronal function is likely disrupted due to reduced ATP production as a result of damaged mitochondria and damage to DNA, proteins and lipids (Dhapola et al. 2024). In addition, several molecular pathways involving the proteins RCAN, CREB, Nrf2 and PP2A have been reported to contribute to the oxidative stress-induced progression of Alzheimer’s disease (Dhapola et al. 2024). These signaling pathways are altered under stress conditions and affect downstream pathways leading to the progression of the disease. Interestingly, some individuals are able to maintain their cognitive abilities despite the presence of significant Alzheimer’s disease associated neuropathological changes. These resilient individuals showed lower levels of oxidative DNA damage, but the exact cellular and molecular mechanisms are not yet understood (de Vries et al. 2024).

While the relationship between elevated oxidative stress and neurodegenerative diseases remains unclear regarding the question whether ROS are cause or consequence, studies show that overexpression of antioxidant enzymes such as superoxide dismutase (SOD) or catalase can ameliorate symptoms in animal models of neurodegenerative diseases, and antioxidant therapies are currently being tested to combat diverse neurological disorders (Patel 2016). However, in Parkinson’s disease, despite high ROS levels and oxidative stress being considered as major disease factors (Dias et al. 2013), antioxidants have failed to demonstrate symptomatic benefits in clinical trials (Duarte-Jurado et al. 2021, Gogna et al. 2024), leaving ROS’s role as an initial trigger in neurodegenerative diseases unresolved.

1.4 Metabolic adaptation

Since neurons are more vulnerable to oxidative stress, which mechanisms do they employ to avoid excessive ROS production? For a long time, it was assumed that the high energy needs of neurons would necessarily require high levels of mitochondrial respiration throughout the neuron accompanied by a high generation of ROS. However, we know now that neurons exhibit specific metabolic adaptations for mitigating the ROS burden. For maximized energy production, glucose is normally metabolized by glycolysis leading to the generation of pyruvate which is shuttled into the tricarboxylic acid (TCA) cycle to generate reducing equivalents for oxidative phosphorylation. Yet, glucose can also be fed into the pentose phosphate pathway, producing not only precursors for the synthesis of nucleotides, but also maintaining the activity of the antioxidant enzyme glutathione peroxidase (GPx) (Ben-Yoseph et al. 1996) (Figure 1). For the detoxification of H₂O₂ glutathione peroxidase requires reduced glutathione which is oxidized in the process. Reduced glutathione is then regenerated by glutathione reductase, consuming NADPH supplied by enzymes of the pentose phosphate pathway. To keep glycolysis at sub-maximal levels for promoting the pentose phosphate pathway with its antioxidant benefits, neurons downregulate the enzyme 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase-3 (Pfkfb3) which acts as a positive regulator of glycolysis. The importance of this metabolic adaptation becomes apparent upon artificial upregulation of Pfkfb3 in neurons which leads to increased oxidative stress and apoptotic death (Herrero-Mendez et al. 2009).

Figure 1:

Metabolic adaptations of neurons to oxidative stress. (A) Factors causing increased vulnerability of neurons to oxidative stress. (B) Neuronal compartments differ in their preferred energy production pathways. Aerobic glycolysis is favored in neuronal somata keeping the oxidative burden low, while mitochondrial respiration is favored at synapses for efficient energy production at the expense of a higher oxidative burden. (C) Depiction of three metabolic pathways that are differentially regulated across neuronal compartments: PKM2 expression in neuronal somata favors aerobic glycolysis, while PKM1 expression at synapses promotes mitochondrial respiration. At the same time a lower expression of Pfkfb3 in neurons than in astrocytes facilitates the entry of part of the glucose pool into the pentose phosphate pathway with its antioxidant benefits. For more details please refer to the text. F6P, fructose 6-phosphate; F1,6BP, fructose 1,6-bisphosphate; F2,6BP, fructose 2,6-bisphosphate; G6P, glucose 6-phosphate; GPx, glutathione peroxidase; GR, glutathione reductase; GSH, glutathione; OXPHOS, oxidative phosphorylation; PEP, phosphoenolpyruvate; PFK, phosphofructokinase 1; Pfkfb3, 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase; PKM1, pyruvate kinase M1; PKM2, pyruvate kinase M2.

A recent study uncovered another neuron-specific metabolic adaptation for decreasing the ROS burden. By measuring metabolic flux separately in neuronal somata and nerve terminals, Wei et al. realized that somata in comparison to synapses carry out less oxidative phosphorylation and more aerobic glycolysis (Wei et al. 2023). Like many cancer cells neurons can metabolize glucose to lactate despite the availability of oxygen for oxidative phosphorylation, a phenomenon known as Warburg effect. While this entails a much less efficient ATP production (2 ATP instead of 36 ATP per molecule of glucose), it has the advantage of a lower ROS generation. Wei et al. could trace the difference in the utilization of oxidative phosphorylation between somata and terminals back to a higher level of the glycolytic enzyme pyruvate kinase 2 (Pkm2) in somata. Pkm2 is one of the two major pyruvate kinase isoforms in the brain. While Pkm1 favors the entry of pyruvate into the TCA cycle, Pkm2 promotes the conversion of pyruvate to lactate (Christofk et al. 2008; Israelsen and Vander Heiden 2015) (Figure 1). Consequently, elevated Pkm2 levels in somata lead to increased aerobic glycolysis. The relevance of this adaptation could be shown in a Pkm2 knockout (KO) mouse model where the loss of Pkm2 resulted in a switch back to oxidative phosphorylation in neuronal somata resulting in increased oxidative damage and the death of dopaminergic neurons (Wei et al. 2023).

Thus, neuronal compartments differ in their preferred energy production pathway. The use of the more efficient oxidative phosphorylation in synapses likely helps to fulfill the high local energy need for neurotransmission, meanwhile exposing axon terminals to higher oxidative stress levels. However, the increased oxidative damage might be easier to handle by synapses than by cell bodies due to the possibility to exploit synaptic structural plasticity pathways to restore terminals. The increased utilization of aerobic glycolysis in the long-lived somata will protect the bulk of the biosynthetic machinery and the cellular genome from lethal oxidative damage. At the same time, the higher expression of Pkm2 in the somata does not only promote aerobic glycolysis over mitochondrial respiration, but its lower overall pyruvate kinase activity (Board et al. 1990) compared to Pkm1 also allows some of the glucose to enter into the pentose phosphate pathway, thereby further strengthening the antioxidant protection of neuronal somata (Wei et al. 2023) (Figure 1).

Nevertheless, the overall level of aerobic glycolysis has to be well regulated since excessive glycolysis will cause the production of the toxic by-product methylglyoxal. Therefore, the neuronal metabolism seems to be delicately balanced to fulfill the dual neuronal need of high synaptic energy supply and long-time cell survival.

1.5 Antioxidant systems

In addition to metabolic adaptations, neurons have several antioxidant systems in place to cope with the remaining ROS production (Figure 2). A genome-wide CRISPRi screen designed to identify essential neuronal genes emphasized the importance of these systems for neuronal survival (Tian et al. 2021). Among the top 10 hits of this screen, which evaluated the survival of human iPSC-derived neurons upon CRISPRi-mediated knockdown, were the genes for the superoxide dismutases SOD1 and SOD2 (Tian et al. 2021). These enzymes convert free superoxide radicals to molecular oxygen and H₂O₂ which can then be detoxified by glutathione peroxidase or catalase. Interestingly, these genes are not essential for survival in all cell types, supporting the idea that neurons are especially vulnerable to oxidative stress.

Figure 2:

Overview of antioxidant systems. Superoxide dismutases (SOD) convert free superoxide radicals (O₂•-), that are produced by reactions within organelles such as the mitochondrial oxidative phosphorylation (OXPHOS) and by cytoplasmic enzymes such as xanthine oxidase, to hydrogen peroxide (H₂O₂). In the presence of free iron, the Fenton reaction can occur causing the decomposition of H₂O₂ to highly reactive hydroxyl radicals. Catalases can reduce the burden of reactive oxygen species (ROS) by converting H₂O₂ to water and oxygen. In neurons, mitochondria express only low levels of catalases, thus, two alternative detoxification systems are utilized: (1) the glutathione (GSH)/glutathione peroxidase (GPx) system reduces H₂O₂ to H₂O, a process in which two monomeric GSH molecules are oxidized and form one glutathione disulfide (GSSG). GSSG is then converted by the glutathione reductase (GR) back to GSH, consuming NADPH in the process. Thiol groups of proteins (Protein-SH) can also be S-glutathionylated (Protein-SSG) by GSSG. (2) In the thioredoxin (TXN2)/peroxiredoxin (PRDX) system, PRDX reduces H₂O₂ to H₂O accompanied by the oxidation of TXN2, which can by reverted by the thioredoxin reductase 2 (TXNRD2) by using NADPH. PRDX that became overoxidized in this process can be regenerated by reduction through sulfiredoxin and sestrin 2.

To further dissect the factors that protect neurons against oxidative stress, the Kampmann lab designed another genome-wide CRISPRi screen which evaluated neuronal survival under mild oxidative stress (Tian et al. 2021). This mild stress was created by omitting the antioxidants that are commonly contained in neuronal culture medium. Among the identified candidates was the selenoprotein glutathione peroxidase 4 (GPX4) which counteracts oxidative damage to lipids by reducing peroxidized lipids and thereby preventing ferroptosis. Other candidates were factors required for selenocysteine incorporation into selenoproteins such as PSTK, SEPHS2 and SEPSECS. There are 25 human selenoproteins of which many function as antioxidants with the selenocysteine residue catalyzing redox reactions (Zhang et al. 2020). Prominent examples are the already mentioned glutathione peroxidases which detoxify H₂O₂ and repair lipid peroxides, but also thioredoxins discussed below (Zhang et al. 2020). Therefore, defects in the machinery for selenoprotein biosynthesis severely compromise the neuronal ability to cope with oxidative stress. In line with this, a recent preprint demonstrated the importance of a tRNA modifying enzyme, ALKHB8, which promotes selenoprotein synthesis, for redox homeostasis in Drosophila neurons (Madhwani et al. 2024).

Earlier work in mice has also highlighted the importance of antioxidant systems. Consistent with the identification of SOD1 and SOD2 as essential genes in the CRISPRi screen (Tian et al. 2021), SOD2 KO mice showed a severely reduced viability and exhibited neuronal loss (Lebovitz 1996). The murine KO of the small redox protein thioredoxin 2 (TXN2, TRX2) impaired viability even more leading to exencephaly and early embryonic death (Nonn et al. 2003). Investigations of the mutations underlying early-onset neurodegeneration have revealed that thioredoxin 2 is also a crucial component of the neuronal antioxidant defense in humans (Holzerova et al. 2016). Thioredoxin 2, which displays its highest expression levels in the brain (Rybnikova et al. 2000), localizes to mitochondria and constitutes together with peroxiredoxin (PRDX) a part of the mitochondrial H₂O₂ detoxification system. In fact, mitochondria are not only a site of ROS production, but also an important site for ROS removal (Drechsel and Patel 2010). Regarding the employed detoxification systems, brain mitochondria differ from liver or heart mitochondria in expressing little, if any, catalase, an enzyme converting H₂O₂ to water and oxygen (Sinet et al. 1980). Instead, brain mitochondria preferentially utilize two peroxidase systems, the glutathione/glutathione peroxidase system and the thioredoxin/peroxiredoxin system with the latter contributing the larger share (Drechsel and Patel 2010). Peroxiredoxin acts similarly to glutathione peroxidase in reducing H₂O₂. The concomitant oxidation of peroxiredoxin is reverted by an oxidation of thioredoxin 2 which in turn can be reduced via thioredoxin reductase 2 (TXNRD2) which utilizes NADPH for this step such as glutathione reductase. In addition to reducing peroxiredoxin, thioredoxin can also restore thiol groups in other proteins that were damaged by oxidative stress. Holzerova et al. identified a homozygous stop mutation in thioredoxin 2 in a patient suffering from an early-onset neurodegeneration causing severe cerebellar atrophy. Using patient-derived fibroblasts they confirmed the loss of thioredoxin 2 and demonstrated increased ROS levels. Consistent with the notion that the elevated ROS levels are causative for the disease, application of the antioxidant Idebenone (a synthetic analog of coenzyme Q10) ameliorated patient symptoms during a four months follow up (Holzerova et al. 2016).

Moreover, the mitochondrially localized isoform of peroxiredoxin, PRDX3, has been shown to be crucial for the health of the human brain. Recessive mutations in PRDX3 were identified in patients suffering from cerebellar atrophy and ataxia, and PRDX3 knockdown experiments showed increased H₂O₂ levels and decreased cell viability (Rebelo et al. 2021). Another oxidative stress related protein for which loss-of-function mutations have been associated with cerebellar atrophy in mice and humans is OXR1 (Wang et al. 2019). OXR1 stands for oxidation resistance 1, a name reflecting its ability to promote cellular resistance to oxidative stress in E. coli (Volkert and Crowley 2020) as well as in mouse cerebellar granule cells (Oliver et al. 2011). In contrast to thioredoxin 2 and peroxiredoxin 3, the molecular mode of action of OXR1 is still only partially understood. It has been shown that OXR1 can react with H₂O₂ through the oxidation of reactive cysteine residues (Oliver et al. 2011). However, newer results render it more likely that OXR1 exerts its beneficial effects by preserving retromer function, a complex necessary for the recycling of proteins and lipids from endosomes to the trans-Golgi network (TGN) or the cell membrane, and thus maintaining endolysosomal function (Wilson et al. 2024). As we will discuss below, impairments of lysosomal degradation are an important trigger of oxidative stress.

It is intriguing and still an open question why the loss of thioredoxin 2, peroxiredoxin 3 and OXR1 have the most pronounced effects on the cerebellum, while in other circumstances striatal neurons exhibited, at least compared to cortical neurons, a higher vulnerability to oxidative stress (Tian et al. 2021). There is still much to learn about the differences in the antioxidant defenses employed by different types of neurons. In addition to the enzymatic systems that we have discussed, there are also non-catalytic antioxidants present such as vitamin E whose production might also vary in a neuron-type specific manner.

1.6 Lysosomal storage and oxidative stress in neurons

As non-dividing cells, neurons are also more sensitive than other cells to impairments in the breakdown of lysosomal material. Defects in lysosomal hydrolases which catalyze the degradation of diverse molecules that are delivered to lysosomes lead to the accumulation of lysosomal storage material and the further impairment of lysosomal function (Parenti et al. 2021). The loss of the lysosomal protein prosaposin (PSAP) which is associated with lysosomal storage diseases, has recently been shown to cause increased levels of ROS, iron and lipid peroxidation specifically in neurons, thereby compromising neuronal health in line with earlier observations in Drosophila (Hindle et al. 2017). Prosaposin is a precursor of saposins A-D which activate glycosphingolipid degradation by lysosomal hydrolases. In their absence glycosphingolipids accumulate in lysosomes leading to the appearance of aggregated lysosomal storage material, also known as lipofuscin, on electron microscopic images. Iron derived from the lysosomal degradation of iron-containing proteins accumulates within this lipofuscin and is believed to cause ROS generation via the Fenton reaction (Brunk et al. 1992) leading, for example, to lipid peroxidation. The accumulation of oxidized lipids will further impair lysosomal function and in the end might also trigger cell death. In fact, in absence of antioxidants in the neuronal medium, PSAP KO neurons showed high levels of lipid peroxidation which finally induced ferroptosis of the KO neurons (Tian et al. 2021). Thus, efficient lysosomal function is also crucial to protect neurons from oxidative stress.

1.7 Adaptation of antioxidant pathways

The generation of ROS is a normal event in the life of every mammalian cell, and, as we have discussed, there are multiple systems in place to detoxify the highly reactive molecules. Therefore, ROS only turn into dangerous oxidative stress when there is an imbalance between oxidants and antioxidant defenses, leading to a disruption of redox homeostasis and subsequent macromolecular damage. To prevent this, cells generally have stress response pathways which allow their adaptation to adverse changes, and these pathways exist also for oxidative stress in neurons. Only when the adaptive responses are saturated, does ROS becomes a threat to survival. These adaptive responses regulate on the one hand components of the antioxidant systems (e.g. overoxidized peroxiredoxin which can be reduced by sulfiredoxin and sestrin 2) and on the other hand trigger repair processes that can reverse oxidative damage to DNA (e.g. via base excision repair), lipids (e.g. via GPX4) and proteins (e.g. via thiol repair by methionine sulfoxide reductase) (Patel 2016). Even though we do not have a full picture of these neuronal oxidative stress responses yet, some pathways have been elucidated. One of them is the mechanism which tailors the antioxidant system to the level of neuronal activity. It is easy to envision that highly active neurons with their greater energy demand will need elevated antioxidant defenses. In line with this assumption, synaptic activity has been shown to alter the regulation of components of the antioxidant system (Papadia et al. 2008). The neuroprotective effect of synaptic activity was traced back to trans-synaptic NMDAR activation (Papadia et al. 2005, 2008) leading to an upregulation of the thioredoxin-peroxiredoxin system (Papadia et al. 2008). In one branch of this pathway, PI3K and Akt become activated triggering the nuclear export of the transcription factor FOXO (forkhead box O) which positively regulates the transcription of Txnip, a thioredoxin inhibitor. Consequently, less Txnip is transcribed leading to an increased thioredoxin activity and thus better regeneration of oxidized peroxiredoxin. In a second branch the transcription factor AP-1 is activated leading to an upregulation of sulfiredoxin causing a more efficient reduction of overoxidized peroxiredoxin. Thirdly, the transcription factor C/EBPb (CCAAT enhancer binding protein) is transcriptionally upregulated inducing an elevated transcription of sestrin 2, elevating the cellular capacity to reduce overoxidized peroxiredoxin even further (Papadia et al. 2008). Based on these findings Papadia et al. raised the alluring question of whether boosting brain activity might prolong brain health by protecting against neurodegeneration via the upregulation of antioxidant defenses.

However, not only synaptic activity triggers the adaptation of the antioxidant system, also ROS themselves serve as feedback signal indicating that increased antioxidant capacities are needed. A critical ROS sensor is DJ-1 (also known by its gene name PARK7), one of the disease genes underlying familial Parkinson’s disease (Hijioka 2017). DJ-1 defective neurons are more sensitive to oxidative stress (Martinat et al. 2004). DJ-1 contains redox-active cysteines whose oxidation state determines DJ-1 activity thus turning it into a redox sensor. Among its many functions, DJ-1 can activate the Nrf2 (nuclear factor erythroid 2-related factor 2)-dependent antioxidant responsive element (ARE) pathway (Im et al. 2012) which plays a critical role in redox homeostasis. Upon Nrf2 activation, a range of antioxidant genes, including glutathione, thioredoxin, thioredoxin reductase and peroxiredoxin, are upregulated strengthening the antioxidant defenses of the cell. It is important to note, however, that in mammalian forebrain neurons NRF2 is only expressed at low levels (Bell et al. 2011, 2015, Shih et al. 2003). Here, alternative pathways including the transcription factor AP-1, utilizing Ca²⁺ levels as a readout of increased electrical activity, can induce transcription of a similar set of antioxidant response genes as NRF2 (see below; reviewed in detail in (Qiu et al. 2020)).

1.8 ROS as a plasticity signal

As highlighted above, ROS do not only act as cell-damaging agents but can activate cellular signaling systems to upregulate antioxidant capacities of the cell. In addition, ROS signaling can induce structural adaptations in neurons both under pathological and non-pathological conditions. In a Drosophila model of neurodegenerative lysosomal storage disease, it has been demonstrated that elevated levels of ROS lead to an activation of the JNK/AP-1 pathway resulting in excessive synaptic overgrowth at the larval neuromuscular junction (Milton et al. 2011). Interestingly, elevated neuronal activity caused by excitotoxicity results in similar morphological adaptations. These phenotypes can be mimicked by application of oxidants or by increasing the endogenous activity of the JNK/AP-1 pathway. Conversely, the expression of antioxidant enzymes decreases the morphological changes (Milton et al. 2011; Oswald et al. 2018; Peng et al. 2019). Consistently, increasing JNK/AP-1 activity decreased ROS levels and rendered flies resistant to chemically induced oxidative stress while decreased JNK/AP-1 activity increased ROS levels. This pathway is conserved in mammalian cortical neurons with JNK/AP-1, activated by chemically induced ROS, resulting in a compensatory expression of the antioxidant protein SRXN-1 (Ugbode et al. 2020). In contrast, activity-induced elevations in ROS trigger the ERK pathway to achieve a similar upregulation of SRXN-1 (Papadia et al. 2008) which localizes to dendrites and synapses, sites within neurons which are particularly sensitive to ROS (Ugbode et al. 2020).

In non-pathological conditions, ROS has the potential to serve as a synaptic plasticity signal which induces adaptive modifications in synaptic and dendritic structure in response to increased neuronal activity. At the Drosophila presynaptic neuromuscular terminal, the Parkinson’s Disease-linked redox sensor DJ-1b (PARK7) regulates the balance between PTEN, a phosphoinositide phosphatase, and phosphoinositide 3-kinase (PI3K) activity to induce an increase in synaptic arborization and compensatory reductions in the dendritic arbor of motoneurons (Dhawan et al. 2021; Oswald et al. 2018). These studies identified mitochondria as a source of regulatory ROS. However, it has been demonstrated in Drosophila motoneurons that the NADPH oxidases Nox and Duox, located in the plasma membrane of the presynaptic terminal (Nox and Duox) and dendrites (Duox), can generate ROS in the extracellular space. Import of ROS via the two aquaporins Bip and Drip likely contributes to structural plasticity changes and has the potential to integrate ROS signals from neighboring neurons (Dhawan et al. 2021; Sobrido-Camean et al. 2022) (Figure 3).

Figure 3:

ROS as a signaling molecule in structural synaptic plasticity. During neuronal activity, ROS is generated either intracellularly by mitochondria or extracellularly by the NADPH oxidases Duox (postsynaptic) or Duox and Nox (presynaptic). Extracellular ROS, generated by the same neuron or adjacent neurons, can be shuttled into the synaptic terminals by the aquaporins Bib and Drip. DJ-1 acts as a redox sensor and can detect elevated ROS in the neuron. At the postsynapse, oxidation of DJ-1 by ROS was shown to activate downstream pathways resulting in reduced dendritic arbor sizes. At the presynapse, DJ-1 was shown to inhibit the phosphatase PTEN, thereby favoring PIP₃ production by phosphoinositide 3-kinase (PI3K). The c-Jun N-terminal kinase (JNK) responds to increased ROS levels by activating the transcription factor activator protein 1 (AP-1). Both pathways lead to a compensatory increase of presynaptic growth and active zone number. In addition, AP-1 activity regulates the expression of the antioxidant gene sulfiredoxin-1 (SRXN-1), a protein that can reduce overoxidized peroxiredoxin (PRDX), which in turn reduces ROS.