Reactive Oxygen Species (ROS) are small molecules that are derived from molecular oxygen. They can either be classed as radicals or non-radicals, depending on whether they have an unpaired electron (1). Superoxide (a radical) is typically restricted to the endosomal compartment, and can be converted into hydrogen peroxide (H₂O₂; a non-radical) at low pH. H₂O₂ can diffuse across membranes to oxidise specific targets, or can be converted to O₂ and H₂O (2).

H₂O₂ is a very useful signalling molecule because it can be rapidly generated and rapidly removed via specific enzymes such as catalase, superoxide dismutase and peroxiredoxin enzymes. It can also be quenched by non-enzymatic means such as glutathione (GSH) (3). As such, by predominantly facilitating cysteine and methionine oxidation, H₂O₂ is integral to regulating several crucial facets of the immune response.

ROS are produced during metabolic reactions within many cellular compartments, including the mitochondria, peroxisome and endoplasmic reticulum (4, 5). This review, however, will focus on ROS specifically produced in the phagosomes and at the cell membrane, by the phagocyte NADPH oxidase NOX2. ROS generation can occur from many sources in cells. These include mitochondria, peroxisomes and the P450 enzyme system. The NADPH oxidase is the first example of an enzyme where generating ROS is the primary function of the system, not a by-product of another process, e.g. the generation of ATP in mitochondria [discussed in (6)].

The physiological production of ROS was first described in 1908, by the German biochemist Otto Warburg, who identified that following the fertilisation of sea urchin eggs, H₂O₂ production succeeded a large and rapid increase in oxygen consumption (7). He suggested the existence of a respiratory enzyme that utilised oxygen to generate ROS, for which he won the Nobel Prize in Physiology and Medicine (8).

The ability of phagocytes to produce ROS was first noted by Baldridge and Gerrard in 1933 who described a marked increase in oxygen uptake by canine neutrophils following phagocytosis (9). Sbarra and Karnovsky extended these findings to show that “this burst of extra respiration” was accompanied by glucose consumption via the hexose monophosphate shunt and lactate production (10). Crucially, inhibitors of mitochondrial respiration have no effect on the oxygen consumption that accompanies phagocytosis. This is because the purpose of the oxygen consumption is independent from aerobic glycolysis, and is instead required to generate ROS.

Further key milestones followed, including (i) the finding that NADPH is the dominant physiological electron donor (although both NADH and NADPH can act in this capacity) that allows the production of ROS (11–13) and (ii) the seminal observation that the process starts with the generation of superoxide (14).

These findings show that neutrophils possess an enzyme that facilitates the donation of electrons to molecular oxygen. The identification of cytochrome b558, which we refer to as NOX2, as the relevant enzyme resulted from insightful biochemistry and the study of the monogenic immunodeficiency X-linked chronic granulomatous disease (X-CGD). This “fatal granulomatous disease of childhood” was first described in the 1950s. It described boys whose neutrophils were unable to kill certain bacteria and did not increase oxygen consumption or produce ROS (15).

In a landmark study for the field, Segal and colleagues showed that neutrophils from patients with CGD lacked both NADPH oxidase activity and a particular unusual b type cytochrome that localised to the plasma membrane (16, 17). The suspected causative genetic region was localised to Xp21 and cloned (18). The cDNA identified from such studies was used to make a translated protein and an anti-serum was raised to it. Elegant studies showed that the anti-sera stained a 91kDa protein found in “purified cytochrome b558” preparations. Crucially, it could not stain neutrophils from patients with X-CGD (19). Thus, the unusual cytochrome identified by Segal was indeed the product of the gene that was disrupted in X-CGD. However, it was clear that the story was not quite that simple. For instance (i) the 91kDa membrane-bound protein transcribed and translated from the X chromosome co-purified with a 22kDa protein (20, 21) and (ii) it transpired that there were autosomal recessive forms of CGD (AR-CGD) associated with a deficiency of other specific proteins (22, 23).

The phagocyte NADPH oxidase (NOX2) is a multi-subunit protein complex that, upon interaction, can form an active enzyme complex capable of producing superoxide. It is comprised of two integral membrane bound components; the 91kDa gp91phox and 22kDa p22phox, which together form cytochrome b558 (16, 21). p22phox binds to and stabilises gp91phox, preventing its degradation and its own in return. The cytosolic components comprise of p40phox (24), p47phox (22), p67phox (22) and Rac1 (25) or Rac2 (26). Following stimulation, p47phox becomes phosphorylated, allowing the complex to translocate to the membrane where it can associate with the gp91phox-p22phox heterodimer, forming the activated complex that transfers electrons from NADPH to molecular oxygen (Figure 1). This process is known as the respiratory burst, which is essential during the innate immune response (27–30). ROS can also be generated by the other NOX family members, NOX1, NOX3, NOX4, NOX5, DUOX1 and DUOX2 (28), however this review will focus on NOX2 derived ROS.

Our understanding of CGD has improved as several large-scale cohort studies have been conducted. X-linked CGD is predominant in Europe (31–33), the United States (34) and Japan (35), accounting for approximately 60% of cases. p47phox deficiency accounts for around 30% of cases and p22phox and p67phox deficiency for the remaining 10%. AR-CGD is predominant in cohorts from countries such as Iran (36) or Turkey (37), where consanguineous marriage is more prevalent.

X-CGD patients have a more severe disease course than patients with the AR-CGD, presumably because they lack all oxidase activity (31). Residual oxidase activity in neutrophils is linked to reduced disease severity and modest production of ROS seems to confer a greater likelihood of long-term survival.

The role of NOX2-derived ROS was first identified in killing microbes. This is well illustrated by the susceptibility of patients with CGD to an extensive, but nonetheless demarcated, range of pathogens. For example, neutrophils from patients with CGD can kill E. coli but not S. Aureus. The focus of this review is the extra-phagosomal role of ROS, but it is worthwhile describing briefly how neutrophil-mediated killing takes place. This can be both direct, where hydrogen peroxide is likely more microbicidal than superoxide, but also indirect through integration with other systems. Here reactive oxygen species collaborate with granule proteins such as myeloperoxidase (MPO). In the presence of H₂O₂, MPO catalyses the production of Hypochlorous acid (HOCL), a very potent anti-microbial agent. HOCL, in turn, can modify multiple proteins (both host and microbe) in the phagosome to generate chloramines and aldehydes.

While we do not cover these processes in detail, we would direct readers interested in this fascinating area to excellent reviews by Nauseef (38) and also, Winterbourn and Kettle (39). These expand more on a few key points:

An important recent addition to the biology of the phagocyte NADPH oxidase and CGD is the discovery of EROS (gene symbol CYBC1) (40). In 2017, we demonstrated that mice deficient in the previously uncharacterised open reading frame, bc017643, were exquisitely susceptible to infection with Salmonella Typhimurium and Listeria monocytogenes. It transpired that they failed to make ROS, owing to almost complete deficiency of gp91phox and p22phox. bc017643 encodes an ER-resident transmembrane protein that co-immunoprecipitates with gp91phox. We also observed that P2X7 receptor expression was downregulated in EROS deficient cells, which was later independently verified by another group (41), demonstrating that EROS mediates the expression of both gp91phox and P2X7. We characterised the protein encoded by bc017643 as EROS (Essential for Reactive Oxygen Species). Recent work in our laboratory suggests that EROS is a highly selective placeholder chaperone, binding to an “unsatisfied surface” on the gp91phox precursor and stabilising it until p22phox can bind (Randzavola, Mortimer et al., submitted). The lack of ROS and NOX2 expression in EROS-deficient mice, akin to that of gp91phox-/- mice, suggested that mutations in the human orthologue C17ORF62 might lead to CGD. In 2018, we and another group reported separate homozygous mutations in C17ORF62, leading to EROS-deficiency, as a novel cause of chronic granulomatous disease (CGD5) (42, 43). C17ORF62 has been re-named CYBC1 (CYtochrome B Chaperone 1).

Mutations in EROS represent the first type of CGD to arise as a result of altered stability or folding of gp91phox. Although they have not yet been implicated in CGD, several other proteins are capable of regulating gp91phox abundance. Negative Regulator of Reactive Oxygen Species (NRROS) was described in 2014 as an ER-resident protein that binds gp91phox and facilitates its degradation (44), its role appearing opposite to that of EROS. Similarly, the heat shock proteins, hsp90 and hsp70, also have opposing effects on gp91phox abundance, stabilising and degrading it respectively (45). The lamin beta receptor (LBR) has an intriguing role in regulating gp91phox and neutrophils that are LBR deficient show lower expression of gp91phox and generation of zymosan-induced ROS (46, 47).

Regulating gp91phox is one method of regulating ROS production. It is essential to tightly control the levels of ROS and subsequent sections of this review examine the tissue damage that can be caused by excess ROS production. However, too little ROS can also lead to inflammation. A key observation from studying CGD is that patients do not only experience opportunistic infections, but also present with autoinflammatory and autoimmune manifestations. These manifestations are characterised by sterile granulomatous inflammation, a hallmark of CGD (48). CGD patients often develop autoimmune diseases in which the pathogenesis is driven by autoantibody production, such as systemic lupus erythematosus and juvenile rheumatoid arthritis (31, 34). By producing H₂O₂, the phagocyte NAPDH oxidase regulates multiple pathways involved in innate anti-microbial defence, often serving to restrain inflammation in the process.

The inflammatory manifestations that affect CGD patients arise as loss of ROS signalling impairs type 1 interferon signalling and autophagy (29). Patients with X-CGD are between 50-90% more likely to experience inflammatory episodes compared to patients with AR-CGD (49, 50), suggesting NOX2, particularly gp91phox, is essential for controlling the balance between a successful immune response and tissue damage.

In normal circumstances, ROS generation occurs concomitantly with phagocytosis, regardless of cell type. Superoxide is a weak base and tends to alkalinise the phagosome, which influences proteolysis. A key study showed that the normal pH of dendritic cells (DCs) is neutral, tending towards slightly alkaline, but that NOX2-deficient DCs have acidic phagosomes which leads to enhanced antigen degradation and impaired cross presentation to CD8+ T cells via MHC Class I (51). The same group reported similar results in human DCs (52) and that Rac2 was key for NADPH oxidase assembly in CD8+ DCs (53). The small GTPase Rab27a was also necessary for NAPDH oxidase assembly (54). However another group, using slightly different conditions, found that while NOX2 did indeed reduce phagosomal proteolysis, this was not associated with significant changes in phagosomal pH. Rather, this group proposed that in DCs and macrophages, NOX2 affects proteolysis through reversible inhibition of the action of cysteine cathepsins via H₂O₂-driven oxidation of cysteine residues. Aspartic cathepsins are unaffected by the presence of NOX2 and thus the phagocyte NADPH oxidase was proposed to alter the activity of only a subset of proteases, skewing the peptide repertoire generated (55, 56).

A recent publication by Reis e Sousa and colleagues provided an intriguing new insight into the role of NOX2-derived ROS in antigen presentation (57). The DNGR1 receptor, expressed on the conventional DC1 (cDC1) subset of DCs, is essential for effective cross-presentation. DNGR1 binds F-actin on dead cell corpses, and has a short hemi-ITAM motif that can recruit and activate Syk. Mice that are deficient in DNGR1 or Syk expression in DCs have impaired cross presentation (58).

This group demonstrated that DNGR1 ligation facilitates Syk kinase activation and this, in turn, leads to NOX2 activation within phagosomes containing internalised antigen. The oxidative stress caused by the resulting free radicals damages the phagosome, causing membrane rupture, thus allowing leak of antigen into the cytosol and its translocation into the MHC class I presentation pathway. Cross presentation is markedly impaired in gp91phox-deficient DCs. How exactly Syk drives NOX2 activation in this context is unknown but the NOX2 activators Vav1 and Rac have previously been shown to be necessary for efficient cross-presentation and they are likely to be involved to some extent.

There is also evidence that NOX2 can influence MHC class II peptide processing and presentation. Extending their previous work on cysteine cathepsins Yates and colleagues demonstrated that NOX2 not only affects the amount of proteolytic processing but affects it qualitatively too (59). They used the myelin oligodendrocyte glycoprotein (MOG) -induced model of experimental autoimmune encephalomyelitis (EAE) and showed that NOX2-derived macrophages (though interestingly not DCs) were defective in their ability to process and present the I-A(b)-immunodominant peptide of MOG. As such, p47phox or gp91phox deficient mice were partially protected from the central nervous system injury and inflammation that characterises EAE. MHC class II presentation is also impaired in human B cells deficient in p40phox (60). This was most marked for cytoplasmic and endogenous antigen but processing of membrane antigen was normal (see Altered Humoral Immune Response in NOX2 Deficiency below).

Oxidative stress is well known to be a contributing factor in the development of neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease and Multiple Sclerosis. Dysregulation or overproduction of ROS leads to oxidative stress which is thought to disrupt immune homeostasis in the central nervous system (CNS) and promote prolonged neuroinflammation (92). NOX proteins are important generators of ROS in the CNS and NOX2 expression has been documented in the CNS in microglia, neurons and endothelial cells (93).

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterised by the progressive loss of motor neurons in the brain, brain stem and spinal cord. Disease progression is rapid, with a prognosis of only 2-5 years after diagnosis for most individuals (94).

Several studies have demonstrated that NOX2 contributes to disease progression in the SOD1^G93A transgenic mouse, a common mouse model of ALS. NOX2 expression and activation was shown to be significantly upregulated in microglia in the spinal cord of SOD1^G93A mice compared to controls (95). NOX2 expression was also found to be increased in the spinal cord of sporadic ALS patients. The authors demonstrated that NOX2 deletion in SOD1^G93A transgenic mice prolonged survival and slowed disease progression, suggesting that NOX2 activity contributes to the degeneration of motor neurons and disease progression in ALS. Another study demonstrated that treatment of SOD1^G93A mice with apocynin, a NOX inhibitor, also increased survival and slowed disease progression (96). The authors also demonstrated that SOD1 regulated RAC1/NOX2 dependent ROS generation in a redox-dependent manner. ALS-associated SOD1 mutants resulted in enhanced activation of RAC1/NOX2 and increased ROS production in both cell lines and the spinal cord of SOD1^G93A mice.

However, in contrast to this, a recent study found that NOX2 inhibition did not extend survival in SOD1^G93A mice (97). Deletion of NOX1 or NOX2 in SOD1^G93A mice did not increase survival or influence microglia activation in this study. Treatment of SOD1^G93A mice with the NOX inhibitors thioridazine and perphenazine did not significantly increase survival or prevent motor neuron degeneration. A study of the oxidative burst in granulocytes in the peripheral blood of ALS patients did not identify any significant difference in NOX2 activity between ALS patients and matched controls. However, patients with lower NOX2 activity were found to have a significant increase in survival (98). Therefore, whilst NOX2 has been shown to be upregulated in both mouse models and patients with ALS, the exact role of NOX2 in ALS pathogenesis remains unclear.

Multiple sclerosis (MS) is a common neurodegenerative disorder characterised by inflammation and demyelination in the CNS. Microglial activation is hypothesised to play an important role in the pathogenesis of MS. NOX2 has also been shown to be upregulated in microglia in active lesions in MS patients (99).

The mouse experimental autoimmune encephalomyelitis (EAE) model is commonly used as a model to study MS. Recent evidence has demonstrated that disease severity is reduced in NOX2^-/- mice after EAE induction. Lymphocyte and microglial infiltration in the CNS was also significantly decreased compared to heterozygous and wild-type controls. Interestingly, the authors showed that only immune infiltration in the CNS is reduced, as immune cell populations in peripheral tissues such as the spleen and cervical lymph node are similar in NOX2^-/- and wild-type mice post EAE induction (100). Microglial activation was also decreased in NOX2^-/- mice, with inflammatory cytokine and chemokine secretion levels in the CNS also decreased.

Another recent paper demonstrated that deletion of NOX2 in conventional DCs (cDCs) reduced disease severity and demyelination in an adoptive transfer model of EAE (101). Interestingly, the authors demonstrated that deletion of NOX2 in cDCs reduced accumulation and activation of autoimmune CD4+ T cells in the CNS in EAE mice, suggesting that NOX2 regulates CD4 infiltration. Deletion of NOX2 also abrogated LC3-associated phagocytosis and CD4+ T cell activation through reduced myelin antigen presentation. This study highlights an important role for NOX2 in promoting inflammation and demyelination in EAE mice. Therefore targeting NOX2-dependent ROS production may slow disease progression and provide therapeutic benefit for patients with MS.

Oxidative stress and damage have also been hypothesised to play a role in Alzheimer’s disease, although the role of NADPH oxidases remains unclear. As mentioned previously, microglia play a vital immune role in CNS homeostasis through clearance of dead cells and debris. However, dysregulation of microglia can lead to prolonged neuroinflammation and the development of neurodegenerative disorders. Microglia have been shown to associate with and promote clearance of amyloid-β (Aβ) deposits in the early stages of Alzheimer’s. However, in aging mice microglia appear to have a reduced ability to clear Aβ deposits and drive inflammation within the CNS (102).

NOX2 expression in microglia has also been hypothesised to play an important role in the pathogenesis of Alzheimer’s disease. A recent study demonstrated high NOX2 expression in microglia and increased microglial infiltration in aged wild-type brains, compared to young mice (103). Interestingly, NOX2^-/- aged mice had significantly less Aβ deposition and plaque formation compared to aged wild-type controls. ROS production was also much lower in NOX2^-/- mice than in wild-type mice, indicating that ROS production in the aged mice was NOX2-dependent. The authors also investigated ROS production in human brain tissue, and found older individuals had higher levels of ROS production when compared to young controls. Stimulation of the BV2 microglial cell line with Aβ₄₂ peptide also resulted in significantly increased NOX2-depdendent ROS production, which could be inhibited using NOX2 inhibitors such as apocynin or Nox2tat. These results indicate that NOX2 may play an important role in the regulation of microglia and the microglial response to Aβ plaques and therefore it may be an important driver of the pathogenesis of Alzheimer’s disease.

Whilst it is clear that oxidative stress is involved in aging and the development of neurodegenerative diseases, the precise mechanisms defining how aberrant NOX2-depdendent ROS production drives neuroinflammation require further investigation. In addition, it remains to be investigated whether targeting of NOX2 through the use of inhibitors would provide therapeutic benefit in neurodegenerative disorders.