NADPH Oxidases: Redox Regulation of Cell Homeostasis & Disease

Abstract

The redox signaling network in mammals has garnered enormous interest and taken on major biological significance in recent years as the scope of NADPH oxidases (NOXs) as regulators of physiological signaling and cellular degeneration has grown exponentially. All NOX subtypes have in common the capacity to generate reactive oxygen species (ROS) superoxide anion (O₂^·-) and/or hydrogen peroxide (H₂O₂). A baseline, normal level of ROS formation supports a wide range of processes under physiological conditions. A disruption in redox balance caused by either the suppression or “super” induction of NOX off balance with antioxidant systems is associated with myriad diseases and cell/tissue damage. Over the past two to three decades sour understanding of NOXs has progressed from almost entirely a phagocyte-, antimicrobial-centered perspective to that of a family of enzymes that is vital to broad cellular function and organismal homeostasis. It is becoming increasingly evident that highly regulated, targeted oxidative protein modifications are elicited in a spatiotemporal manner and initiated at cell membranes in humans by seven NOX isoforms (NOXs 1, 2, 3, 4, 5 and DUOXs 1 & 2). In a sense, this renders NOX-ROS signaling akin to that of other second messenger systems involving localized Ca²⁺ dynamics and tyrosine kinase transactivation. Accordingly, the study of ROS compartmentalization in subcellular organelles has been shown to be crucial to elucidating their role in cell phenotype modulation under physiological and pathophysiological conditions. The NOXs are as distinct in their distribution and activation as they are in their cellular functions, ranging from host defense, second messenger PTMs to transcriptional, epigenetic and (de)differentiating effects. The review integrates past knowledge in the field with new focus areas on the leading-edge of NOX-centered ROS signaling including how a new wave of structural information provides insights for NOX biology and targeted therapies.

Graphical Abstract

1. INTRODUCTION

NADPH oxidases comprise a family of oxidoreductase enzymes that expressly catalyze the formation of reactive oxygen species (ROS), i.e. their only known function to date is the sole, evolved, and deliberate generation of ROS. NADPH oxidases contain a distinctive hemoprotein core responsible for electron transfer from NADPH to molecular O₂ resulting in the formation of superoxide anion (O₂^·−) as their primary metabolite.

Over the past 30 years, the study of NOXs has undergone a rebirth from its original focus on their antimicrobial function or as “perpetrator” of oxidative stress to that of a family of enzymes vital to cell signaling and organ function. It is becoming increasingly clear that this collection of variably complex monomeric and multimeric proteins constitutes a family of highly specialized and localized enzymes which subserve innumerable roles in biology and disease.

1.1. About this review

This review is decidedly focused on NADPH Oxidase (NOX) biology and the wealth of information surrounding each of its mammalian NOX isoforms in signaling. In Chapter 1, we provide perspective to its storied beginnings from the discovery of the respiratory burst to its unparalleled, systematic inquiry into the biochemistry and its implications for NOX activity in phagosomal antimicrobial action. Approximately three decades ago our knowledge of NOX biology entered a period of rapid expansion coinciding with the discovery of their non-phagocytic ROS-generating activity. In Chapter 2, we discuss the biochemistry, unique structural characteristics, tissue distribution and function of individual NOX isotypes, their cytosolic regulatory modulatory factors (aka subunits or components), signal transducing agents, and organelle distribution. In Chapter 3, the review delves into current knowledge with respect to the NOXs’ increasingly acknowledged biological roles in physiology. Accordingly, Chapter 4 provides a comprehensive look at the array of cell types and organ systems with initial emphasis, wherever possible, on the physiology of NOXs. The review takes a dive into the onset and progression of disease for which much investigation has been prioritized. We survey compelling evidence of the role of ROS derived from NOXs broadly ranging in scope from inflammation, cancer, cardiovascular diseases and diabetes to neurological disorders and discuss single nucleotide variations and polymorphisms that have the potential to influence disease. The status of NOX inhibitors, drug development and relevant clinical trials finishes out the discussion in Chapter 5. In the final chapter, we offer conclusions and provide a brief prospectus on the future of NOX biology, its prospects, and challenges.

It has become quite evident that referring to ROS exclusively as injurious or necessary evils (1, 2) is no longer valid. Indeed, NOX-derived ROS are, by nature, physiological and genuine signaling agents and all discussions of the NOXs/ROS ought to begin with that premise. As such, we have chosen to begin each subchapter’s discussion with this in mind wherever possible and whenever the literature allowed. In this review, we refer to the seven mammalian NADPH oxidase enzymes as NOX oxidases, NOX isoforms since the products of the reaction can differ. For brevity, we may simply refer to these enzymes as NOXs 1 through 5, and DUOXs 1 and 2. Indeed, in many instances, NOX, per se, refers to the transmembrane hemoprotein of an entire NADPH oxidase complex to which it gives its name [e.g. NOX2 in the multi-subunit NOX2 oxidase complex]. An exhaustive database and literature review was conducted in the writing of this manuscript but due to the prohibitive scope of information available, we have limited our discussion almost entirely to mammalian NOXs in an effort to include as many landmark papers by leaders in this field as well by those who are not normally, or less-, acknowledged. Inevitably, as was frequently the case, we found it necessary to limit consideration to an abridged subset of information in a particular focus area. Along with this comes regret and an apology for unintended omission of deserving references. Still, as evidenced by the bibliography, the number of citations discussed was exceedingly lengthy and extraordinary. Finally, to appropriately give credit where credit is due, we strove to prioritize original research findings wherever possible. In fact, on all counts, we have strived in this review to be as unbiased as possible and to take the “road less traveled” in some areas by citing the less popular perspectives. With this comes the likelihood that the prevailing Zeitgeist in the field will be unintentionally spurned at times in the interest of challenging dogma and encouraging new investigation.

1.2. Reactive oxygen species

Reactive oxygen species (ROS) are molecular oxygen-derived small molecules which are divided into two groups: free radical and non-radical species. Free radical species possess an unpaired electron in their outermost atomic shell and include superoxide radical (O₂^·−) as well as hydroxyl (HO^·), alkoxyl (RO^·) and peroxyl radical (RO₂^·), and other less commonly encountered radicals. Non-radical ROS that are oxidizing include: singlet oxygen (¹O₂), ozone (O₃) and hydrogen peroxide (H₂O₂), hypochlorous acid (HOCl), and peroxynitrite (ONOO⁻). In this context, acquisition of an electron by molecular oxygen (O₂) facilitated by the NOX leads to the formation of O₂^·− – popularly referred to as the “mother of all free radicals” as it is considered the primary ROS generated by the majority of the NOX and multiple other enzyme systems in oxidase biology. O₂^·− is converted in a step-wise manner to other ROS beginning with its spontaneous dismutation or through metabolism by superoxide dismutases (SODs), and catalyzed further by the Haber-Weiss and Fenton reactions (3, 4), or by reacting with nitric oxide radical (^·NO) to form peroxynitrite anion (5). Importantly, O₂^·− has an intrinsic chemical half-life of 10⁻⁹ to 10⁻¹¹ s whereas in the presence of SOD, its half-life is decreased to 10⁻¹⁵ s (5). The reaction catalyzed by SOD chemically reduces O₂^·− to form oxygen and H₂O₂ which, in turn, is reduced to water (H₂O) and O₂ via peroxidases and catalase, but also by the redoxin family of proteins (6). H₂O₂ can evade degradation, cross membranes and traverse biologically relevant (long) distances in vivo permitting it to have autocrine and paracrine signaling effects in cells and tissues (7). O₂^·− reacts with transition metals such as the reduction of Fe⁺³ to Fe²⁺ yielding O₂ in Haber-Weiss reaction (4). Furthermore, Fe²⁺ reacts with H₂O₂ forming Fe³⁺, ^·OH and OH⁻ in Fenton reaction (4). The reaction of O₂^·− with ^·NO, rate-limited by diffusion of both radicals, forms the potent oxidant peroxynitrite (ONOO⁻) (8). ONOO⁻ is subsequently oxidized or reacts with a hydrogen radical (H^·) to form the stable HOONO (9). The latter decomposes rapidly into ^·OH and reactive nitrogen species (9, 10). Thus, concentrations of hydroxyl radical increase by means of H₂O₂ conversion (metal-dependent) and HOONO dismutation (metal-independent pathway). All told, a variety of ROS singularly, or in combination, react exceedingly well in an often highly controlled and directed manner with receptive moieties on cellular targets. With that said, among all others, H₂O₂ has emerged as the foremost and credible biological transducer of NOX signaling. With respect to the extensively studied role and importance of reactive nitrogen species stemming from ^·NO metabolism and its interaction with O₂^·−, we refer the reader to multiple reviews on the topic (11–14). In the interest of scope, this review only discusses ROS derived from the NOX family. However, it would be woefully inadequate not to acknowledge that the intricacies of NOX signaling involve an admixture of the effects of reactive oxygen and nitrogen species on downstream targets (15–17).

1.3. Historical perspective: from the sea urchin to clinical trials

The early advancement of our understanding of the NOXs is a tale of multiple pioneering observations that unfolded progressively, however disjointedly, over three quarters of a century and through the study of various organisms and cell types. The first observation, to our knowledge, is attributed to the discovery by Otto Warburg over a century ago of a rapid and sharp rise in oxygen consumption in sea urchin eggs upon fertilization, which he attributed to respiration (18). In 1932, the phagocyte NOX field was ignited by the discovery of the respiratory burst by Baldridge and Gerard (19). Using Warburg manometers, the pair detected increasing O₂ consumption by canine leukocytes fed gram-positive bacteria. Sbarra and Karnovsky in 1959 demonstrated reliance of this consumption on glucose metabolism (20). Prescient in its implications, they observed that O₂ consumption accelerated at a rapid pace during phagocytosis and thus proposed the involvement of the hexose monophosphate shunt and lactate production (20). However, key to its discernment from mitochondrial oxidases, oxygen consumption in this process was not inhibited by potassium cyanide or antimycin which suggested it did not depend on mitochondrial oxidative phosphorylation (21). In the interim, Iyer and coworkers had reported that the guinea pig phagocyte respiratory burst involves the generation of H₂O₂ (22) and later studies demonstrated the pivotal oxidation of NADPH in this generative process (23).

On the heels of these discoveries, in 1964 Rossi and Zatti proposed that an oxidase was responsible for this respiratory burst and that oxidation of NADPH by intact granules drives induction of the hexose phosphate shunt (24). Additional studies using subcellular fractions of stimulated neutrophils conclusively demonstrated a >100-fold preference of the enzyme for NADPH over NADH (24). About a decade later, Babior et al. identified the initial metabolite of the respiratory burst oxidase as O₂^·− based on an observed SOD-inhibitable reduction of ferricytochrome c (25) and a couple of years after that, in 1975, Clark & Klebanoff and colleagues (26) demonstrated the importance of myeloperoxidase in the antimicrobial activity of neutrophils (see (27) for more detail). Indeed, while NOX is crucial to this process, myeloperoxidase (present in neutrophils but not in macrophages) boosts the antimicrobial action of NOX-derived oxidants by at least three log orders and is especially important in combatting certain virulent bacteria like S. aureus. It is important to note at this juncture that NOX is undeniably essential as the source of H₂O₂ (acting directly or via its conversion by myeloperoxidase to the more oxidizing HOCl) and serves as the primary oxidant-generating antimicrobial enzyme in human neutrophils (27). Underscoring the clinical relevance of these findings, the essence of a catalytic component NOX in human neutrophils was identified when scientists and physicians reported on a cohort of patients exhibiting symptoms of a fatal granulomatous disease of unknown etiology (28). Indeed, it was learned that afflicted children in this cohort suffer from frequent and severe infections as their neutrophils are deficient in oxidase and microbicidal activity (29). Over a period of approximately three decades, physicians and scientists observed that neutrophils from patients with the disease were compromised in their respiratory burst activity and often incapable of killing phagocytized staphylococci - thus rendering them susceptible to frequent and often fatal infections (30–34).

Early on, Hattori as well as Shinagawa et al. identified cytochrome b in membranes of horse and rabbit neutrophils (35, 36) and Segal and Jones later described the same in human neutrophils and suggested that recurrent infections were due to the lack of this cytochrome b in chronic granulomatous disease (CGD) patients (37). Soon after, a soluble FAD- and NADPH-dependent O₂^·--generating activity isolated from human neutrophils and sedimenting with cytochrome b was identified and characterized – subsequently named glycoprotein 91 (91 kDa MW) of phagocyte oxidase, gp91^phox, and later officially renamed NOX2 (38–41). Moreover, its essential clinical function is observed in X-linked CGD patients, an immunodeficiency syndrome caused by defective phagocytic NOX2 activity. Patients with X-linked CGD, caused by a defect in either the gene encoding NOX2, lack cytochrome b₅₅₈ in their phagocytes (39) as well as stable detectable levels of co-stabilizing p22^phox (22 kDa) as a part of the same cytochrome (38, 39). Taken together, these data revealed that a heterodimer was required for stable expression of both gp91^phox and p22^phox. Thereafter, the story of a more complex system arose with the discovery of other key proteins in the NOX2 complex: p47^phox, p67^phox and small GTP-binding proteins Rac1 or Rac2 (42–47). Later on, came the discovery of an additional cytosolic regulatory subunit, p40^phox (48).

In the early to mid-1990s, the often-overlooked discoveries of a series of non-phagocytic oxidases were watershed advances for the field. Indeed, beginning with a rigorous study published by Quinn and coworkers in 1993 (49), a structurally and genetically distinct cytochrome b₅₅₈ was found in human fibroblasts; yet this finding has been largely overlooked. While these fibroblasts appeared to contain cytochrome-b₅₅₈ (detected spectrophotometrically), multiple polyclonal and monoclonal antibodies directed at various epitopes of gp91^phox and p22^phox were incapable of detecting the subunits. Moreover, while polymorphonuclear neutrophils from a patient suffering from X-linked CGD showed O₂^·−-generating activity 90% lower than normal patients, their fibroblasts showed no difference in activity compared to controls, respectively (49). These findings were confusing at the time as to whether or not a NOX protein complex had indeed been detected in fibroblasts. Nevertheless, the discovery of cytosolic NOX subunits in dermal fibroblasts and simultaneous discoveries reported in the mid-1990s by the laboratories of Griendling in vascular smooth muscle (50), Wolin in endothelial cells (ECs) (51) and Pagano in adventitial fibroblasts (52, 53) piqued awareness of a non-phagocytic signaling role for NOX and broadened conventional thinking beyond that of the phagocytic NOX (50, 53–55). Indeed, findings of a bona fide vascular NOX activity induced by the pro-constrictor agent angiotensin II (AngII) and inhibited by diphenylene iodonium (DPI), but not by KCN, were paradigm-shifting. Moreover, immunological detection of NOX2, p22^phox, p47^phox and p67^phox in aortas, and p67^phox in plasma membranes of cultured fibroblasts plus the first demonstration that immuno-sedimentation of a NOX subunit outside of the phagocyte (p67^phox) was capable of abrogating NOX-derived O₂^·− provided compelling evidence of functional NOX2 in a non-phagocyte (50, 53, 54). And, in the wake of these seminal discoveries, a series of important studies corroborated a role for NOX2 components in vitro and in vivo in vascular cell ROS and phenotype (56–60). These novel findings and what was to follow in the cardiovascular field and beyond were transformative and prefaced the discovery of a unique NOX dubbed Mox-1 (later NOX1) followed in suit by seminal findings of NOX3, NOX4, NOX5 and the DUOXs 1 & 2.

Beginning with the original discovery of the phagocytic respiratory burst, the field has appreciated NOXs’ roles in vital physiological processes involving NOX-generated ROS, e.g. host defense, innate and adaptive immunity, signal transduction, cell proliferation, self-renewal (61), cellular defense (62), gene expression, angiogenesis, hearing, platelet recruitment, ion transport, wound repair and memory (63–67). As such, Noxologists are increasingly aware that, as in many other biological processes, NOXs have evolved to maintain important fundamentally protective functions, e.g. cell renewal and cell and tissue homeostasis. Indeed, the revelation that NOXs are present in prokaryotes has solidified this premise (68). Like nitric oxide (^·NO from nitric oxide synthase (NOS) whose role is broadly accepted as salubrious in limited amounts, so must ROS derived from NOXs be viewed. Thus, the objective in disease treatment might be more wisely viewed as returning organ systems and patients to their normal state of basal NOX-ROS signaling.

From a disease perspective, an unrecognized inflection point for the field came also with the realization of NOXs’ roles in hypertension and related end organ damage (53, 57, 69). In the period since the mid to late 1990s, the field has geometrically expanded to include pathologies spanning cancer, renovascular hypertension, atherosclerosis, ischemia-reperfusion, and CNS disorders including Alzheimer’s disease and Parkinson’s disease. Admittedly, NOX-linked pathology, i.e. in diabetes, heart and neurodegenerative diseases, pulmonary disease and cancer among others (65, 70–91) is still foremost on the minds of many translational, clinical and even basic scientists. Indeed, the widely held, unyielding perception of NOX activity as synonymous with oxidative stress and disease is slow to fade.

2. THE NADPH OXIDASE (NOX) FAMILY OF ENZYMES

In higher mammals including humans, the NOX family is comprised of a group of seven enzyme isoforms (NOXs 1 – 5 and DUOXs 1 & 2) that act singularly or in combination with one or more auxiliary subunits. The catalytic cores of these subtypes serve to transfer electrons from NADPH to O₂ across the enzyme systems: (a) highly homologous binding sites for NADPH and flavin cofactor (FAD) on exceedingly well-conserved dehydrogenase domains (DHs); and (b) six structurally conserved transmembrane (TM) catalytic core helices which, at their center, coordinate the arrangement of two sequential prosthetic heme groups, one proximal and one distal to the cytoplasm. Together, these functional groups are sequentially aligned for an efficient and unidirectional electron relay from NADPH to O₂ on opposite sides of membranes. The DUOXs, by exception, are comprised of seven transmembrane domains with a characteristic N-terminal extracellular peroxidase homology domain tethered to its extra TM helix (Figures 1 & 2). Considering how rapidly the field has grown in recent decades, this chapter (and all others for that matter) is limited by the depth we can reasonably provide in several areas. Thus, for detailed information on biochemistry and structure, the authors refer the reader to the following comprehensive reviews (92–98).

Figure 1. NADPH oxidase (NOX) family members.

NOXs comprise a family of transmembrane proteins with 6 or 7 transmembrane (TM) domains. NOX1–3 require assembly of cytosolic regulatory subunits (NOXA1, NOXO1, p47^phox, p67^phox and Rac1/2) for activation. NOX4 is constitutively active though its catalytic activity is increased by Poldip2 and tyrosine kinase substrates with 4 or 5 Src (SH3) homology domains (Tks 4/5). Tks 4/5 can also activate NOX1. NOXs 1–4 catalytic subunits are associated with p22^phox which stabilizes the complex. NOX5 and DUOX1 & 2 do not bind to p22^phox but have 4 or 2 Ca²⁺ binding sites, EF-hands, respectively. NOX5 possesses a calmodulin binding-domain, and its stability is regulated by heat shock protein 90 (HSP90). DUOX1 & 2 have an extra TM from which a peroxidase-like domain extends extracellularly at the N-terminus of the isoform.

Figure 2. Identifying domain maps of NOX membrane and cytosolic regulatory subunits.

Structural domains within membrane and cytosolic subunits of the NOX isoforms. NOXs 1 – 5 contain requisite 6 transmembrane domains (TM) and cytosolic NADPH and FAD binding domains. NOX5 and DUOXs 1 & 2 possess Ca²⁺ binding EF hand domains toward their N-terminus; wherein the DUOXs 1 & 2 also contain an extra TM domain with a tethered peroxidase-like domain. A hallmark proline rich domain (PRR) in p22^phox, p47^phox and NOXO1 is key for its interaction with Src homology domains (SH3) within p47^phox, p67^phox and NOXA1, respectively. Tetratricopeptide repeats (TPRs) in p67^phox permits its binding to Rac1/2 and a Phox and Bem1 (PB1) domain is required for its interaction with p40^phox. PX domains in p47^phox, p40^phox and NOXO1 facilitate binding to phosphoinositides on the plasma membrane. Different domains in Rac1 include the nucleotide binding sites (NBS), switch 1, switch 2, polybasic region (PRB), insert region (IR) and the CAAX box.

2.1. NOX Subunits

2.1.1. Membrane subunits

2.1.1.1. Prototypical NOX (NOX2) and catalytic electron transfer

NOX2 (also known as gp91^phox) is the classical hemoprotein catalytic core of the NOX2 oxidase complex to which other NOX hemoproteins are compared and differentially homologous. Originally defined in macrophages and neutrophils, it is among the most widely detected of the NOX subunits (Figure 4). Additionally, the NOX2 gene is located on the X chromosome’s positive strand and has an mRNA sequence of 4276 bp that is translated to a 570-amino acids protein with a molecular mass of 65 kDa (Table 1). NOX2 TM domains comprise the core of a ferric reductase (FDR) to which it has been compared phylogenetically (99, 100) (Figures 1 & 2). By comparison to other FDRs like STEAP (six-transmembrane epithelial antigen of the prostate enzymes), a six-helical transmembrane motif common for all NOXs takes on an hourglass shape, narrow in the middle and broad on each side of the membranes wherein the electron accepting substrates and heme cofactors bind. As such, this structural arrangement facilitates electron passage across the membrane by means of an elaborate electron-transfer cascade, or "electron hopping" mechanism (100).

Figure 4. Tissue distribution of the NOX in human tissues.

Graphs correspond to the consensus dataset of expression levels/tissue expressed as RPKM (reads per kilobase of transcript per million mapped reads) created by HPA and GTEx transcriptomics datasets. Y axis in each graph is adjusted to the maximum expression for the given NOX. Anatomical display of the expression levels shown in males. Adapted from https://www.proteinatlas.org/. More details and information can be found under HELP: Assays & Annotations on the Protein atlas website. Note that the website is constantly been updated with new information, new tissues, etc.

Table 1.

Nomenclature, chromosomal location, and molecular dimensions of NOX genes and gene products

NOX	Other Names	NCBI Gene ID	# of transcriptsa	RefSeq Accession Number: mRNA,Proteinb	Chromosome Locationa,c	Gene Length	mRNA in bp	# Amino Acids	Molecular Mass	Coding mutationsd	Non-Coding mutationsd
NOX1	GP91–2, MOX1, NOH-1, NOH-1L, NOH1	27035	4	NM_007052, NP_008983	[-] Xq22.1; chrX:100,098,313–100,129,348	31036 bp	2543 bp	564	64871 Da	12	0
NOX2	AMCBX2, CGD, CGDX, GP91–1, GP91-PHOX, GP91PHOX, IMD34, NOX2, p91-PHOX	1536	1	NM_000397, NP_000388	[+] Xp21.1-p11.4; chrX:37,639,312–37,672,714	33403 bp	4276 bp	570	65336 Da	768	160
NOX3	GP91–3, MOX-2	50508	1	NM_015718, NP_056533	[-] 6q25.3; chr6:155,395,368–155,455,839	60472 bp	1980 bp	568	64935 Da	6	0
NOX4	KOX, KOX-1, RENOX	50507	7	NM_001291927, NP_001278856.1	[-]11q14.3; chr11:89,324,353–89,589,557	265205 bp	4209 bp	578	66932 Da	4	1
NOX5		79400	3*	NM_024505, NP_078781	[+] 15q23; chr15:68,930,525–69,062,762	132238 bp	8405 bp	765	86439 Da	7	0
DUOX1	LNOX1, NOXEF1, THOX1	53905	2	NM_175940, NP_787954	[+] 15q21.1; chr15:45,129,933–45,165,576	35,644 bp	5483 bp	1551	177235 Da	44	3
DUOX2	LNOX2, P138(TOX), P138-TOX, THOX2	50506	2	NM_014080, NP_054799.4	[-]15q21.1; chr15:45,092,650–45,114,172	21523 bp	6361 bp	1548	175364 Da	450	52
p22phox	CybA, Cytochrome B-245 Alpha Polypeptide; Cytochrome B(558) Alpha Chain	1535	1	NM_000101, NP_000092	[-] 16q24.2; chr16:88,643,275–88,651,083	7,809 bp	692 bp	195	21013 Da	89	20
p47phox	Ncf1	653361	1	NM_000265, NP_000256	[+] 7q11.23 chr7:74,774,011–74,789,315	15,305 bp	1349 bp	390	44682 Da	55	8
NOXO1	SNX28; P41NOX; P41NOXA; P41NOXB; P41NOXC; SH3PXD5	124056	4	NM_172168, NP_751908	[-] 16p13.3; chr16:(1978917– 1981469)	5,276 bp	1556 bp	376	41253 Da	7	0
p67phox	Ncf2; Neutrophil Cytosolic Factor 2	4688	5	NM_000433, NP_000424	[-] 1q25.3 chr1:183,554,461–183,601,849	47,389 bp	2267 bp	526	59762 Da	92	17
NOXA1	p67phox-Like Factor; NCF2-Like Protein	10811	3	NM_006647, NP_006638	[+] 9q34.3 chr9:137,423,350–137,434,406	11,057 bp	1678 bp	476	50933 Da	7	0
p40phox	Ncf4; Neutrophil Cytosol Factor 4	4689	2	NM_013416, NP_038202	[+] 22q12.3 chr22:36,860,988–36,878,017	17,030 bp	1378 bp	339	39032 Da	12	5
DUOXA1	NIP; mol; NUMBIP	90527	7	NM_144565, NP_653166	[-] 15q21.1; chr15:45,117,366–45,129,938	12573 bp	1996 bp	343	37815 Da	5	1
DUOXA2	TDH5; SIMNIPHOM	405753	1	NM_207581.4, NP_997464	[+] 15q21.1; chr15:45,114,326–45,118,421	4096 bp	1755 bp	320	34787 Da	73	13

All information was collected from CBI-Gene card (https://www.genecards.org/) (1) unless otherwise indicated.

^a:Isoform numbers correspond to # REFSEQ mRNAs as reported in GeneCards, numbers different from what is reported in the literature are marked with *.

^b:RefSeq Accession Numbers for mRNA and protein of isoform selected by HGMD database.

^c:Gene orientation on chromosome indicated in square parenthesis as positive or negative.

^d:Number of mutations in coding and non-coding regions per HGMD.

In NOX2 five inter-helix loops (A to E) connect six transmembrane spanning helices, followed in sequence by a C-terminal DH domain that extends into the cytosol (101, 102). Early structural analyses showed that the enzyme’s A, C and E loops are on the extracellular side of the membrane (and in the case of organelles e.g. phagosome/endosome they extend into the lumen), while the B and D loops are on the intracellular side of the membrane or on the cytosolic side of organelles (101, 102) (Figure 1). Recent analyses shed new light on the structure and function of the three extracellular loops (A,C & E) detailing that they form a cap over the outer heme in the TM which suggests their role in limiting access to the oxygen reducing site (103). More specifically, Loop A appears to serve as a stabilizing buttress underneath Loop C and Loop E thus supporting their configuration (104). An intra-loop disulfide bond (between cysteines 244 and 257; C244-C257) stabilizes Loop E in a compact structure (104). Historically, mutations causing substitution at C244 linked to CGD were identified as preventing maturation of Loop E (105–107). Corroborated by recent structural data, on the opposite side of the membrane, the B and D loops are in accessible proximity of the protein’s pliable C-terminal DH domain constituted by the FAD-binding (FBD) and NADPH-binding (NBD) sub-domains (101, 102, 108–110). As such, the B and D loops are critical for proper orientation of cytosolic subunits and their interaction with, and thus electron transfer from, the DH domain (102, 111–114). Two pairs of strictly conserved heme binding motifs are present on the third and the fifth TM helices (104, 115) and their imidazole rings position the two B-type hemes orthogonal to the plane of the membrane for fluid transfer of electrons to O₂ (116). For more detail into structure function relationships uncovered biochemically, we direct the readers’ attention to a series of comprehensive reports (117–119).

With respect to stability of the protein, human NOX2 is variably glycosylated on its second and third extracellular loops (C and E). As is commonly the case with glycosylated proteins, it is visualized on an immunoblot as “laddered” or as a “smear” from its predicted 65 kDa up to 91 kDa molecular mass proportional to the degree of glycosylation (120). Loop E sits atop the NOX and contains an acknowledged glycosylation modification at N240. Additionally, two N-linked glycosylation moieties reside on amino acids 149 and 132 on Loop C (104).

A common motif that was discovered in NOX2 (104) presents itself as an ordered non-protein density region enveloped by the highly conserved R54, H119, and the outer heme appears in structural analysis to be occupied by a water molecule bound precisely at the NOX-emblematic, oxygen-reducing center (104, 108, 110). In addition, there is evidence for a characteristically hydrophilic tunnel that connects the extracellular environment to the oxygen-reducing center with a radius sufficiently large to allow passage of O₂ and the release of O₂^·- (104). The outer and inner heme (and F215 between them) constitute a conduit for electron transfer (104, 108, 110).

From the early days of protein sequence information and biochemical characterization prior to crystallography and cryo-EM analyses, it was known that most of the NOXs possess six transmembrane helices and a DH containing NADPH and FAD sites at their C-terminus (Figure 2). Along this common backbone structure, five redox centers central to their function are distributed in sequential order: the NADPH, FAD, a proximal (internal or inner) heme, distal (external or outer) heme and O₂ (terminal acceptor) sites. Intriguingly, these redox centers have an incrementally less-negative redox potential up to the second-to-last step in the chain. That is, the distal heme has a lower redox potential than the proximal heme, thus creating a redox “pit” that must be energetically overcome for the transfer of electrons to O₂. Intrinsic to their physical composition, NOX protein(s) contain three out of these five redox centers: the FAD site and two hemes, whereas the other two are extrinsic to the molecule – those being NADPH and molecular O₂. In accordance with this, NOXs’ transmembrane electron transfer is initiated with the movement of an electron from cytosolic NADPH at the NADPH-binding site to the FAD cofactor giving rise to FADH₂ (121, 122). Following this is electron transfer from reduced FAD (FADH₂, E^0’ = −304 mV) to the first heme (E^0’ = −225 mV) which leaves FAD in its semiquinone radical form. The electron proceeds from the proximal to the distal heme (E^0’= −265 mV), and terminally to the final electron acceptor - O₂ (E^0’ = −160 mV) that generates O₂^·−. However, the transfer of the electron from the first to the second heme runs counter to electromotive forces; and as such, their redox potential differential [more negative at the outer heme (E^0’= −265mV) renders electron transfer between them unfavorable (119). Nevertheless, O₂ binding to the outer heme creates an overall energetically favorable state for the outer heme to essentially “drag” the electron across this proverbial electron “pit” (117, 119, 123). This “trough” of sorts prevents an excessively rapid electron transfer and ensures a more gradual and smooth electron flow across the channel without impediment or accumulation (124). Once the cycle is complete, the proximal heme can accept another electron from NADPH. This time around, however, the donor initiating the electron transfer is distinctly the semiquinone FAD (E^0’= −256 mV).

The majority of knowledge available on the NOX family comes from extensive studies of NOX2, which, along with its smaller co-stabilizing subunit p22^phox, comprises the cytochrome b₅₅₈ core of NOX2 oxidase (125–127); and, like other NOX isoforms, the anchoring catalytic NOX2 hemoprotein gives the entire oxidase complex its name. Incidentally, the cytochrome was initially called flavocytochrome b₂₄₅ and later renamed cytochrome b₅₅₈ due to its confirmed peak absorbance at 558 nm obtained by calculating the reduced minus oxidized spectrum (128, 129). It has long been appreciated that the heterodimer NOX2-p22^phox flavocytochrome is modulated by association with its organizing subunit p47^phox (130, 131), activating subunit p67^phox (132–135), p40^phox (48, 136–138) and the GTPase Rac (134, 137, 139) (Figure 1). Importantly, the activation domain of p67^phox is arguably central to and facilitates electron transfer from NADPH to FAD (135). However, for this to happen, p67^phox must arrive at the membrane to bind NOX2 (Figure 3). In that regard, the cell’s detection of a pathogen triggers phosphorylation of cytosolic factors at the center of which is p47^phox. This, in turn, induces translocation of the tripartite complex of p40^phox-p47^phox-p67^phox to membrane-bound components of NOX2 and facilitates catalysis of O₂^·− production.

Figure 3. Activation of NOX2 is driven by p47^phox S-379 phosphorylation leading to the interaction between regulatory and membrane components.

In the resting state, NOX2 oxidase is disassociated into its subunits and cofactors, while NOX2 and p22^phox are in the membrane, p47^phox–p67^phox – p40^phox are in an associated, yet dephosphorylated, state in a soluble trimeric complex, in which the p67^phox and p40^phox PB1 domains bind and a SH3 domain at the C-terminus of p67^phox binds to the PRR of p47^phox. At this time, the SH3 super-groove and PX domain of p47^phox are masked by the polybasic auto-inhibitory region (AIR) which restricts p47^phox to its folded and inactive state. Upon activation by diverse stimuli, protein kinase C (PKC) gets activated and phosphorylates p47^phox on critical serines, this in turn disrupts the hydrogen bonds linking the C-terminal AIR and the tandem SH3 domains and exposes the supergroove pocket. This allows for the secondary phosphorylation of serine residues (S379) (see insert) permitting translocation and binding of p47^phox to p22^phox’s PRR domain. p47^phox and p40^phox acting through hallmark binding domains outlined in Figure 2 function to engage p67^phox and Rac1 and scaffold the entire complex in place for NOX2 activation.

While NOX2 is activated posttranslationally, at the gene level its promoter region accommodates a host of transcription factors that upregulate its transcript and, in turn, protein levels. This, however, is outside the scope of this review and is thoroughly reviewed elsewhere (140, 141). All the same, many of the known transcription factors for each of the NOX core proteins are presented in Table 2.

Table 2.

Transcriptional regulators of NOXs.

Subunit/Isoform	Transcription Factor	Cell Type	Transcriptional Effect	Reference
NOX1
	STAT1	Aortic Smooth Muscle	Upregulation	(1304)
		Colon Epithelial Cells	Upregulation	(747)
	STAT3	Aortic Smooth Muscle	Upregulation	(1304)
		Aortic Tissue	Upregulation	(861)
	GATA4	Colon Epithelial Cells	Upregulation	(1305)
	GATA6	Colon Epithelial Cells	Upregulation	(1305)
	NFκB	Aortic Smooth Muscle	Upregulation	(1306)
	AP-1	Aortic Smooth Muscle	Upregulation	(1307)
		Vascular Smooth Muscle	Upregulation	(1308)
	MEF-2B	Vascular Smooth Muscle	Upregulation	(1308)
		Vascular Smooth Muscle	Upregulation	(1309)
	C/EBPα	Aortic Smooth Muscle	Upregulation	(186)
	C/EBPβ	Aortic Smooth Muscle	Upregulation	(186)
		BMDM	Upregulation	(1310)
	C/EBPδ	Aortic Smooth Muscle	Upregulation	(186)
		BMDM	Upregulation	(1310)
	ATF-1	Vascular Smooth Muscle	Upregulation	(1311)
NOX2
	NFκB	Monocytes, Microglia	Upregulation	(1312)
		Monocytes	Upregulation	(1313)
	STAT3	Aortic Tissue	Upregulation	(861)
	PPAR-α	Macrophages	Upregulation	(1314)
	HIF-1α	Microvascular Endothelial Cells	Upregulation	(447)
	CDP	HL60, HEL, PLB985	Downregulation	(1315)
	CP1	HL60, HEL, PLB985		(1315)
	YY1	HeLa, K562, PLB985	Upregulation	(1316)
	Elf-1	PLB985, U937, Jurkat, K562	Upregulation	(1317)
	PU.1	PLB985, U937, Jurkat, K562	Upregulation	(1317)
		Neutrophils, Mononuclear Leukocytes	Upregulation	(1318)
		HeLa, U937	Upregulation	(1319)
	ICSBP	HeLa, U937	Upregulation	(1319)
	IRF-1	HL60, HeLa, PLB985		(1320)
		HeLa, U937	Upregulation	(1319)
	IRF-2	HL60, HeLa, PLB985		(1320)
	HOXA9	U937	Upregulation	(1321)
	PBX1	U937	Upregulation	(1321)
	Meis1	U937	Downregulation	(1321)
	HOX10A	U937	Downregulation	(1321)
p47 phox
	NF-κB	Monocytes	Upregulation	(1313)
	AP-1	Aortic Smooth Muscle	Upregulation	(1307)
	Ets-1	Aortic Smooth Muscle	Upregulation	(1322)
	PPAR-α	Macrophages	Upregulation	(1314)
p67 phox
	NFκB	Monocytes	Upregulation	(1313)
	AP-1	Aortic Smooth Muscle	Upregulation	(1307)
		HL-60, PLB985	Upregulation	(1323)
	PPAR-α	Macrophages	Upregulation	(1314)
		Aortic Smooth Muscle	Upregulation	(1324)
	PPAR-β/δ	Aortic Smooth Muscle	Upregulation	(1324)
	PPAR-γ	Aortic Smooth Muscle	Upregulation	(1324)
	PU.1	HeLa, U937	Upregulation	(1319)
		HL-60, PLB985	Upregulation	(1323)
	IRF1	HeLa, U937	Upregulation	(1319)
	ICSBP	HeLa, U937	Upregulation	(1319)
p22 phox
	NFκB	Aortic Smooth Muscle	Upregulation	(1325)
	C/EBPβ	Vascular Smooth Muscle	Upregulation	(1217)
NOX4
	HIF1α	Pulmonary Artery Smooth Muscle	Induced	(1326)
	NFκB	Aortic Smooth Muscle	Induced	(1306)
	AP-1	Aortic Smooth Muscle	Upregulation	(1307)
	STAT1	Aortic Smooth Muscle	Upregulation	(1304)
	STAT3	Aortic Smooth Muscle	Upregulation	(1304)
		Aortic Tissue	Upregulation	(861)
	C/EBPα	Aortic Smooth Muscle	Upregulation	(186)
	C/EBPβ	Aortic Smooth Muscle	Upregulation	(186)
	C/EBPδ	Aortic Smooth Muscle	Upregulation	(186)
	E2F1	Vascular Smooth Muscle	Upregulation	(1327)
	PPAR-α	Aortic Smooth Muscle	Upregulation	(1324)
	PPAR-β/δ	Aortic Smooth Muscle	Upregulation	(1324)
	PPAR-γ	Aortic Smooth Muscle	Upregulation	(1324)
		HUVEC	Downregulation	(1328)
NOX5
	NFκB	Aortic Smooth Muscle	Upregulation	(1329)
	STAT1	Aortic Smooth Muscle	Upregulation	(1329)
	STAT3	Aortic Smooth Muscle	Upregulation	(1329)
	AP-1	Aortic Smooth Muscle	Upregulation	(1329)
	C/EBPα	Aortic Smooth Muscle	Upregulation	(186)
	C/EBPβ	Aortic Smooth Muscle	Upregulation	(186)
	C/EBPδ	Aortic Smooth Muscle	Upregulation	(186)
	PPAR-α	Aortic Smooth Muscle	Upregulation	(1324)
	PPAR-β/δ	Aortic Smooth Muscle	Upregulation	(1324)
	PPAR-γ	Aortic Smooth Muscle	Upregulation	(1324)

Data obtained from the literature and discussed in (140, 141, 150)

2.1.1.2. NOX1

NOX1, the second protein in the NOX family to be discovered, by way of sequence tag (EST) screening for homologs of human gp91^phox (142), has, notably, three splice variants reported. NOX1 is also X-linked and sits on the reverse DNA strand. Its mRNA transcript of 2543 bp, which is considerably shorter than that of NOX2 (4276 bp), encodes a 564 amino acid (aa) protein that is only six aa shorter (circa 60% exact aa homology) than NOX2 (Table 1) and possesses a very high degree of functional and structural homology with it. From a historical perspective, it was originally termed mitogenic oxidase 1 (Mox1), on account of its supposed involvement in proliferation (142, 143). However, the described increase in the mitogenic rate was later challenged as to whether this might be the consequence of Ras-mediated transformation (142, 143). In any event, an EST screen based on the gp91^phox third transmembrane domain revealed a sequence homologue initially named NOH1 and found to map on the human chromosome Xq22 (144). To our reading, NOX1 has been detected in virtually every cell albeit at far lower levels than in the colon (Figure 4). Its cellular localization is primarily assigned to the plasma membrane (more accurately caveolae) and endosomes. Nothwithstanding the perennial challenges faced with antibody specificity, we have chosen to include the most comprehensive tissue localization information available for NOX family members in Figure 4.

Like NOX2 oxidase, the NOX1 oxidase is comprised of the NOX1 catalytic subunit along with its stabilizing partner p22^phox (Figure 1). Its canonical regulatory subunits, homologues of p47^phox and p67^phox, NOXO1 (NOX organizer 1) and NOXA1 (NOX activator 1), respectively (145), together with the GTPase Rac1, are required for its catalytic competency (Figure 1). Intriguingly, one salient characteristic of NOX1 is that it is normally perceived as a constitutively expressed enzyme in its canonical form by virtue of deficiency of an autoinhibitory domain in its organizing subunit NOXO1 (Figure 2). That said, the complex is induced by a host of factors including, among others, PDGF, EGF, AngII, prostaglandin F_2α (PGF_2α), interferon-ϒ, PKC-δ and TLR2 (142, 146–150). Non-canonical yet enzymatically competent (hybrid) NOX1 oxidase can have its canonical cytosolic regulatory subunits NOXO1 and NOXA1 replaced in function by one or both NOX2 oxidase cytosolic subunits (p47^phox and/or p67^phox). Additionally, tyrosine kinase substrate with four (Tks4) and five SH3 domains (Tks5), may support localized O₂^·− production and, in turn, H₂O₂ from NOX1 (151). Concretely, Tks4 and Tks5 bind directly to NOXA1, and not p67^phox (152). Transcription factors promoting NOX1 expression are thoroughly reviewed elsewhere (140, 141) and presented in Table 2.

2.1.1.1. NOX3

NOX3 was first revealed by a number of seminal reports in and around the discovery of NOX1 (63, 67, 140, 153, 154). Its gene sequence contains an exon transcript of 1980 bp which is located on chromosome 6 on the reverse strand (Table 1). Despite the large difference in its mRNA length compared to NOX2, its translated protein is remarkably close in length and function to that of NOX2, i.e. 568 aa in length and 65 kDa MW. Although it exhibits ∼56% amino acid identity to NOX2, a considerably higher degree of structural and functional homology is preserved owing to conserved similarity of substituted amino acids (Figure 2). Concordantly, hydropathy plot analysis and sequence alignment predict structure-function of NOX3 to closely align with that of NOX1 and NOX2 (154, 155). Like NOX1, NOX2 and NOX4, NOX3 requires p22^phox as its co-membranal stabilizing partner; but rather uniquely among the first three NOXs, it displays varying degrees of activity in the presence vs. absence of one or more of a combination of NOX1 and NOX2 cytosolic subunits (63, 156, 157) and may require a stimulus to be detectable. Indeed, in the presence of NOX2 cytosolic subunits (p47^phox and p67^phox), NOX3 displays relatively low constitutive activity but is sharply increased by phorbol myristic acetate (155, 156). In fact, it does not require a stimulus for appreciable activity when expressed in the presence of NOX1 cytosolic subunits (NOXO1 and NOXA1) (63). Interestingly, p67^phox and NOXO1 appear sufficient to sustain robust constitutive NOX3 oxidase activity. However, when p47^phox and NOXA1 are added to a cell-free system containing NOX3, activity is stimulus-dependent (63). At least one report suggests that Rac may be dispensable for NOX3 activity (156).

Quite distinctly among the NOX isoforms, the majority of reports consigns NOX3 to the inner ear, including the cochlear and vestibular sensory epithelia and the spiral ganglion (63) and it has been linked to hearing and hearing loss (158). Moreover, in the vestibular system, it functionally participates in the maintenance of corporal equilibrium and gravity perception (159). As such, due to its largely benign and physiological roles, scientists have seemingly shied away from developing inhibitors for NOX3. Furthermore, in a publication by Bánfi and colleagues largely credited for debuting and characterizing the biochemistry and distribution of NOX3, it was also found in the skull, brain and embryonic kidney although its physiological role was called into question by virtue of its diminishingly low expression there (63, 155) (Figure 4). Intriguingly, however, other studies over the years reveal that NOX3 has a wider “repertoire” as it is detected in hepatoma cells and in placenta, testes, the pancreas, lung, the cardiovascular system, adipose tissue, and adrenal glands (140, 141, 154, 155) and Figure 4. Yet, the verdict is still out on its definitive physiological and pathophysiological roles in those cells and tissues with some noteworthy exceptions in the lung, pancreas and cardiovascular system (160). One of the most salient of these, in the lung, is that NOX3 is definitively tied to TLR4 signaling and emphysema (161). Individual NOX roles in processes and disease are delineated in detail in the remainder of this review (see below).

2.1.1.2. NOX4

NOX4 has arguably been the most investigated NOX isoform (behind NOX2) in the twenty plus years since it was discovered. It was also among the first identified to significantly deviate from the canon of NOX 1 – 3 oxidase complexity and organization. Located on chromosome 11 on the reverse strand, its primary transcript of 4209 bp encodes a 578 aa protein rendering it of similar length and structure to that of NOX2 (Table 1; Figure 2). Thus, despite large variations in the open reading frame for NOXs 1 – 4, NOX4 is remarkably similar in structure and length (Figure 2). In further detail, despite possessing ~39% precise amino acid homology to NOX2, NOX4 weighs in at a similar 67 kDa MW and has been identified as having seven splice variants (Table 1). Unlike the other NOXs, NOX4 does not appear to require organizer or activator subunits (i.e. p47^phox, NOXO1, p67^phox, NOXA1), and seemingly requires only p22^phox as a membranal stabilizing subunit. Moreover, there is a general perception that, unlike NOXs 1 – 3, its elevated activity depends on transcriptional induction by a wide variety of transcription factors in distinct cell types (Table 2). Whereas there is evidence that polymerase delta-interacting protein 2 (Poldip2) amplifies NOX4 function (162), in addition to other proteins such as protein disulfide isomerase (PDI) and tyrosine kinase substrate 4/5 (Tks4/5) that influence it (163), these are not regarded as obligatory for NOX4 activity. As such, NOX4 is generally presumed to be constitutively active under normal conditions and upregulated by its expression in response to NFκB, E2F1, PPARs and the C/EBP series among others (Table 2) (164), save for select evidence that posttranslational modifications alter NOX4 activity (165). Parenthetically, many studies refer to H₂O₂ as the primary metabolite generated by NOX4 (166, 167); however, multiple studies have also detected O₂^·- (168, 169). These discrepancies might be attributed, in part, to the capacity of some probes to more easily detect H₂O₂. However, one very compelling finding is of the ability of NOX4’s E-loop to internally dismutate O₂^·- to H₂O₂ and thereby not permit release and detection of O₂^·- by the isoform (167). Moreover, from one intriguing physiological perspective, NOX4 was initially ascribed a role of cell O₂ sensor. To that argument, its exceptionally high K_m for O₂ (∼18%), similar to the values of known oxygen-sensing enzymes, contrasts with NOX2’s K_m of 2–3% (166). However, this concept has largely been abandoned since NOX4 expression itself can respond linearly and progressively to changing pO₂ levels (166).

2.1.1.3. NOX5

NOX5 was introduced in 2001 as a homologue of NOX2/gp91^phox independently by the Lambeth (155) and Krause (170) laboratories. Besides shared characteristics with other of NOX isoforms, NOX5 boasts a couple of unique features, of which the most striking is that it appears self-sufficient, uniquely requiring no other subunit for its activation, not even p22^phox (171). On an extended cytosolic N-terminus, it contains four Ca²⁺‐binding helix–loop–helix structural domains (EF hands) (Figures 1 & 2) (172). These domains afford it highly sensitive and scalable activity depending on Ca²⁺ concentration (170, 171). Predictably, NOX5’s mRNA transcript is substantially longer than NOXs 1 – 4 at 8405 bp coding a protein 765 aa long with a mass of 86 kDa (Table 1). NOX5 is most commonly expressed in the spleen, testis, uterus, ovary, endothelium and smooth muscle cells where it has been implicated in widespread roles from immunity to reproduction and cardiovascular disease (155, 173, 174). However, recent rigorous inquiries by Geiszt and colleagues into the presence of NOX5 protein, challenge the notion that NOX5 is involved in gametogenesis or immune cell function in the spleen (173). In fact, transcript detection and cross-reactivity in these tissues might rather be attributed to ECs or auxiliary cell types in their midst (173). With respect to fertility, it appears that NOX5’s involvement with fertilization is conflated by NOX’s broader and inexorable association with sea urchin fertilization. On the other, cross-tissue detection in ECs, could provide greater proof for its role in cardiovascular system (174). Still, with the advent of new and better tools to specifically detect and manipulate NOX expression will inevitably come greater scrutiny into their localization and function particularly with respect to this isoform that is not found in rats and mice.

The human NOX5 gene encodes six identified isoforms [α, β, γ, δ, ε (also called short NOX5, NOX5S)] and ζ (175) and is located on the positive strand of chromosome 15. NOX5 isoforms α, β, γ and δ were identified by Bánfi et al. (170), and possess N-terminal EF-hand domains, conferring Ca²⁺-activation. While isoforms α, β and γ are functionally active and generate ROS, the δ, ε and ζ isoforms seem to be relatively inactive when it comes to O₂^·– generation and it is unclear whether they have functional significance. NOX5α and NOX5β are reportedly the major isoforms expressed in human cells and negatively regulated by NOX5ε (176). The NOX5ε and NOX5-S isoforms, identified by Cheng et al. lack EF-hand domains, and, as such, are structurally similar to NOXs 1 – 4 (155). To that extent, NOX5ε, incapable of Ca²⁺ binding, depends on cAMP-response element-binding protein for activity (177); and even though p22^phox can interact with NOX5, it is not required for NOX5 function (178).

Similar to other NOXs, NOX5 is regulated post-translationally by a variety of mechanisms including phosphorylation (179–181), S-nitrosylation (182), SUMOylation (183), palmitoylation (184), and oxidation (185), but unlike NOX2 is not glycosylated. Moreover, NOX5 gene promoter activity in vascular smooth muscle cells is driven by transcription factors such as STATs 1 & 3, AP-1, and NFκB (Table 2) and epigenetic factors including overexpression of histone deacetylase 2 (186).

Major advancements in our understanding of NOX structure came with much-anticipated findings from X-ray crystallography of Cylindrospermum stagnale NOX5 (csNOX5) published in a revealing landmark paper by Magnani and coworkers in 2017 (108). CsNOX5 bears rather low (40%) homology to human NOX5 but more importantly appears to rather faithfully replicate key structural aspects of the human NOX (108). Intriguingly, the transmembrane domains of csNOX5 take on a pyramidal shape with a triangular base (cytosolic side) and a narrower apex toward what the authors define as “the outer membrane face” of the subunit (108). From that perspective, the arrangement of NOX resembles an upside-down “funnel” through which electrons flow and are directed. Moreover, csNOX5 TM domains span six helices (h1–h6) and a previously undisclosed, to our knowledge, N-terminal α-helix which runs at the surface of and parallel to the inner side of the membrane and for whom, on closer look, no function appears to have been ascribed to date (108). Further, csNOX5 TM core is aligned by four lipid ligands to a few cross-membrane helices and two hemes of the TM portion of NOX are positioned with their planes perpendicular to the lipid bilayer; and the vector connecting iron atoms in the hemes is almost exactly perpendicular to the plane of the bilayer (108). It seems reasonable to propose then that these physical characteristics in conjunction with the identified lipids in the membrane provide stability and optimal orientation of the prosthetic groups and facilitates electron passage across the membrane by means of an elaborate electron-transfer cascade, or "electron hopping" mechanism from the inner to the outer side of the membrane. Moreover, it is tempting to speculate that the disclosed extra α-helix that spans a portion of the inner side of the membrane could provide greater stability to the TM domains or even recognition and engagement by other yet-unidentified cytosolic partners (187).

As previously predicted, csNOX5 topology attests to one of the hemes being proximal to the cytosolic (inner) side of the domain, while the second heme is located toward the outer side (108). Magnani et al. assert a plausible electron transfer path to be from the inner heme traversing Trp378 to the second heme (108). Structural analysis remarkably also put forth a mechanism by which O₂ positions itself for reduction. That is, a small cavity was detected directly above the outer heme and occupied by what appeared to be a “highly ordered water molecule”, and lined by conserved residues Arg256, His317, and iron-affixed His313. The H-bonding environment of the cavity expectedly promotes O₂ binding and retention, while the positive charge of Arg256 electrostatically promotes the catalytic production of O₂^·– (108). Incidentally, the inner side of TM domains interacts reciprocally to the bi-lobal DH surface, wherein the flavin ring is exposed, while the C terminal csTM structure (residue 412) must necessarily be close in space to the N terminus of csDH (residue 413) (108).

2.1.1.4. p22^phox

p22^phox maintains the stability of the catalytic subunits of NOXs 1 through 4 by preventing its counterpart NOX from proteasomal degradation (188). As mentioned, it is a membrane-integrated stabilizing counterpart to NOX1–4 (see below) with an affinity domain that anchors p47^phox, the SH3 domains of which bind to p22^phox (Figures 2 & 3). Biochemical characterization and subsequent partial crystallization revealed a proline rich region (PRR) in the cytosolic domain C-terminus that is necessary for pivotal interaction with organizing subunit p47^phox and is required for NOX activation in NOXs 1 through 3. In retrospect, experimental evidence over the years including epitope mapping (189), peptide walking (190) and analyses of CGD patient variant proteins (191, 192) has given rise to exceptional models with two or four membrane-spanning helices, placing both the N-terminus and C-terminus on the intracellular side of the membrane (193). Recently obtained cryo-EM structures settled that debate with confirmation of four TM helices (104, 194), of which p22 TMs 1 & 4 interact with TMs 3, 4 and 5 of NOX2 to form the heterodimer (104). A naturally occurring mutation P156Q in p22^phox identified in CGD patients, prevents p47^phox translocation to the membrane after activation (193). In fact, p22^phox contains three conserved prolines, which are essential for its contact with p47^phox: P152 and P156 that bind one (N-terminal) SH3 motif of p47^phox, and P158, which interacts with the other SH3 domain (C-terminal) of p47^phox, respectively (195). These prolines undergird recruitment of p47^phox; subsequent NOX assembly is described below (193) (Figure 3, inset).

2.1.1.5. DUOX1 and DUOX2

Dual Oxidases or DUOXs diverge from the classical NOX structure and composition significantly. Formerly termed thyroid oxidases (THOX1 and 2) (Table 1), DUOXs contain seven TM domains from N- to C-terminus, a peroxidase homology domain (PHD), a pleckstrin homology-like domain (PHLD), cytosolic Ca²⁺-binding EF hand domains, a double heme catalytic core and a DH domain (Figures 1 & 2) (109, 110, 140, 196–198). Similar to all of the NOXs, the C-terminal DH domain includes requisite FAD-binding and NADPH-binding domains, while TM domains comprises the catalytic module with ~50% homology to NOX2 (199).

Distinctly, both DUOXs possess a unique N-terminus that extends into the extracellular space rather than the cytosol and contains a peroxidase-like domain possessing 43% similarity to thyroid peroxidase (198). DUOXs were first purified as flavoproteins from pig thyroid plasma membrane (200), and subsequently cloned as two human cDNAs clones encoding proteins of 1551 (DUOX1) and 1548 aa (DUOX2) (Table 1) (196). The encoded polypeptides showed 83% similarity to each other and 53 and 47%, respectively, to NOX2.

Both murine and human DUOX1 possess a characteristically large extracellular N-terminal PHD domain (109, 110) whose function is to this day a topic of consternation by Noxologists as to whether the PHD is, indeed, a peroxidase. Unfortunately, while the domain has been described to have the “architecture” of a peroxidase, cryo-EM did not detect the presence of heme or histidines required for heme orientation in the domain deemed necessary for peroxidase activity (109, 110). These findings are, indeed, largely in agreement with previous biochemical analyses (201). Nevertheless, in addition to a highly conserved TM domains, DUOX1 possesses emblematic cytosolic domains including an FBD, an NBD and a Ca²⁺-binding EF domain (109, 110). Furthermore, two cation-dense PHD regions, CBS1 and CBS2, are thought to be crucial for folding (109, 110). Importantly, mutations in both these cationic zones disrupt the protein’s ability to bind to its maturation factor, DUOXA (109, 110) and, thus, they can be expected to play an essential role in the complex’s stability. The sphere-shaped PHD is perched above the TM domains reportedly held in place by at least one covalent interaction, i.e. a disulfide bond between C118 on PHD and C1165 on loop C of the TM and multiple other non-covalent interactions (109, 110). Until the PHD domain is ruled in or out as playing a substantive role in DUOX catalytic activity, strategies at disrupting this interaction could be a potential opportunity for therapeutic intervention in cases in which DUOXs’ roles are deleterious. Moreover, an agent which blocks proper rotation and orientation of the EF and DH domains (109, 110) could present another prospect for competitive blockade.

Like in the other NOXs, NADPH binding involves proper positioning of the TM domains and NBD of DUOX1. On the underside of the TM domains and plasma membrane, a lipid molecule modulates the interplay between the TM domains and NADPH (109). A conserved ‘GXGXG’ motif, lying under the NADPH diphosphate group, is “sighted” in the tight loop linking the first β-strand and α-helix of TM (109). Collectively, cryo-EM studies corroborate the direction of electron flow and home in on a key phenylalanine in the process: NADPH : FAD : HEME : Phe1097 : HEME (109). As such, Phe1097 mutation diminished H₂O₂ production in DUOX1 (109).

It has been known for some time that DUOXs 1 & 2 do not require classical NOX regulatory subunits. Even so, they do rely on particular maturation factors DUOXA1 or DUOXA2, respectively, for proper plasma membrane translocation and enzymatic function (198). The genes for DUOX1 and DUOX2 can be found on chromosome 15, forward and reverse strands, respectively (199), and are organized in opposite transcriptional directions, but with distinct promoters (199).

DUOX maturation factors DUOXA1 and DUOXA2 are glycoproteins that regulate DUOXs 1 and 2, respectively by facilitating their ER to Golgi transition, glycosylation, maturation, membrane fate and protein stability (201, 202). Remarkably, DUOX/DUOXA genes are situated on opposite strands and transcribed outwardly and bidirectionally from a shared promoter region (203, 204). This ordered and elegant expression is fascinating and suggests an absolute co-dependence of each for the biological function of the other. In keeping with this, forced cellular expression of DUOX or DUOXA alone cause each to remain in the ER, stunted in their ability to mature as a plasma membrane-integrated and functional protein. Not surprisingly, since the DUOX1 & DUOXA1 and DUOX2 & DUOXA2 combinations are Ca²⁺-dependent, they produce the largest amount of H₂O₂ in those configurations, respectively (205, 206). Exchanging the combinations, i.e. DUOX2-DUOXA1 and DUOX1-DUOXA2 results in considerably less H₂O₂, putatively in that order (199). Interestingly, the DUOX2-DUOXA1 has the capacity to produce O₂^·- which presumably offsets its ability to produce as much H₂O₂ as otherwise might be expected and is determined by identified sequences in DUOXA1 (207). Further to their function, DUOX-DUOXA complexes are modulated posttranslationally, e.g. N-glycosylation of DUOX1 and DUOXA1 promotes maturation and DUOX1 phosphorylation amplifies H₂O₂ production (199).

DUOX1-DUOXA1 exist in configurations alternating between inactive “dimer-of-dimers” and active heterodimer configurations (199). Structurally, DUOX1-DUOXA1 in the “dimer-of-dimers” inactive state has two structural hallmark characteristics compared to its active form: a more mobile cytoplasmic domain complex (containing the PHLD, EF, and DH subdomains) and displacement away from the TMs, which likely disrupts the proper orientation of the complex with TMs and makes electron transfer across the TM unfavorable (110). In other words, in the inactive state, the cytoplasmic domains are flexible and not optimally positioned for electron transfer (110). Furthermore, the DUOX1–DUOXA1 dimer docks with another just like it into the “dimer-of-dimers” structure causing steric hindrance between the NBD and FBD. Finally, formation of this complex causes the solvent cavity to be sealed and inaccessible to O₂ (109).

Conversely, in the active state, the FBD and NBD of DUOX1 latch onto the TMs, and the O₂ entry site is fully exposed (110) making for a favorable electron flow from NADPH to O₂ and hence the production of H₂O₂ (110). Indeed, it is highly likely that DUOX1–DUOXA1 and DUOX2–DUOXA2 dismute O₂^·- within the complexes themselves (110); and appears to reject the long-held notion that the extracellular PHD is involved in H₂O₂ generation. While the mechanism governing the transition from inactive to active DUOX1–DUOXA1 complex remains elusive, the data suggest that FAD and NADPH binding might facilitate the docking of DUOX cytosolic domains onto the TMs and ease the transition to a heterodimeric state or heterotetrameric state (110). Due to high structural homology of DUOX1–DUOXA1 and DUOX2–DUOXA2, it is likely that these complexes are similarly regulated (110).

2.1.2. Cytosolic regulatory subunits of the assembled NOX complex

In phagocytic cells, catalytically competent and prototypical NOX2 oxidase, containing NOX2 and p22^phox at its core, is regulated and maximally active with the aid of four cytoplasmic components: p47^phox, p67^phox and p40^phox and Rac1 or 2 (118, 208) (Table 1 & Figures 1 & 2).

2.1.2.1. p47^phox

p47^phox, also termed neutrophil cytosolic factor 1 (Ncf1), is at the heart of most characterizations of the NOX2, and, as such, has been termed the organizing subunit of the oxidase complex. The human p47 gene is found on chromosome 7 (7q11.23) and encodes a protein 390 aa in length with a molecular mass of approximately 45 kDa (Table 1). At its N-terminus, p47^phox has one PX domain, which interacts with phosphoinositides (chiefly phosphatidylinositol 3,4-bisphosphate and phosphatidic acid) and promotes p47^phox’s translocation and binding in membranes (209). The central portion of p47^phox harbors two critical SH3 domains (SH3A and SH3B; SH3 supergroove) and a polybasic auto-inhibitory region (AIR) (210, 211) (Figure 2). In the resting state, the SH3 super-groove and PX domain is masked by the AIR which restricts p47^phox to its folded and inactive state. This region is interspersed with crucial serines whose phosphorylation facilitates unfolding and exposure of the SH3 supergroove to p22^phox’s PRR domain (210). C-terminal to AIR is a PRR domain that allows docking with p67^phox and NOXA1 (118, 210, 212–214) (Figure 3).

Over the past two decades, multiple teams of investigations have illustrated that critical structure-modifying phosphorylation on p47^phox occurs at its C-terminus (131). Essentially, manipulations of variable combinations of serine mutations in p47^phox showed that serine phosphorylation is indispensable for NOX2 activation (215). To iterate some key examples, the replacement of serine 379 with alanine, which avoids phosphorylation, abolishes NOX2 activity (215), while individual mutation of select serines between 303 (S303) and S370 to alanine inhibited NOX2 activity by only about 50% (215). On the other hand, in silico and docking studies showed that phosphorylation of S379 (beyond the PRR domain) (216) disrupts linkage of the C-terminal AIR and the tandem SH3 domains and exposes their intervening supergroove. This, in turn, allows for secondary phosphorylation of serines in the supergroove pocket permitting translocation and binding to p22^phox PRR domain (Figure 3). Incidentally, a tryptophan (W193) appears to play a central role in the N-terminal SH3 of p47^phox that interacts with p22^phox upon NOX2 activation and assembly (130, 131, 217, 218). From a host of follow-up studies including X-ray crystallography, a clearer picture has emerged that phosphorylation of both serine-rich regions are cumulative in their effect on p47^phox binding to p22^phox and full activation of the NOX2 complex.

2.1.2.2. NOXO1

Human NOXO1, a 376 aa protein, is encoded by exons totaling 1556 bp in length on chromosome 16 (219) (Table 1) and is the canonical NOX1 organizing subunit (145, 220, 221). Like NOX1, it is enriched in colon epithelia (145, 220); and unlike its counterpart, p47^phox, NOXO1 lacks an AIR domain which is consonant with conventional wisdom that it is avidly bound to p22^phox and is permissive of constitutive NOX1 oxidase activity (221, 222). A tantalizing alternate theory is that the absence of an AIR domain within NOXO1 allows its exposed and promiscuous SH3 domains to interact intramolecularly with NOXO1’s own PRR region (and potentially other proteins), and make it, at times, less avid in its binding to p22^phox (223).

2.1.2.3. p67^phox

The NOX2 activator p67^phox, originally named Ncf2, is a 526 aa protein encoded on chromosome 1. p67^phox is an essential component of a latent ternary complex with p47^phox and p40^phox with 1:1:1 stoichiometry in the cytoplasm of quiescent phagocytes (Table 1; Figure 3). The human p67^phox ortholog contains four protein–protein interactions that comprise a Rac-binding domain with four tetratricopeptide repeat (TPR) motifs at its N-terminus; an N-terminal SH3 domain of unknown function; a PB1 (Phox and Bem1) domain which stabilizes its association with p40^phox; and a cardinal SH3 domain at its C-terminus that interacts with p47^phox (224). Structural flexibility of p67^phox allows simultaneous association and stable complexation with p47^phox and p40^phox (134, 225, 226) (Figures 2 & 3).

Employing small-angle X-ray scattering analysis, p67^phox was confirmed as a monomer that adopts an elongated conformation with few to no significant associations among its internal domains (227), in contrast to p47^phox (228) and p40^phox (224). Unlike the relatively singular ability of excess p67^phox (plus Rac1) to sustain NOX2 activity in a reconstituted system (229), in cells and in vivo p47^phox and p40^phox function as carrier proteins for p67^phox positioning it for NOX2 activation (see below) (230) (Figure 3).

2.1.2.4. NOXA1

The NOXA1 gene is located on chromosome 9 (9q34.3) and is responsible for a 476 aa soluble protein that, in comparison with p67^phox, lacks a central SH3 but contains a C-terminal SH3 of high homology to its counterpart (Table 1, Figures 2 & 3). NOXA1 triggers NOX1 more efficiently than does p67^phox and is a poor NOX2 activator compared with p67^phox (221). Similar to a role for NOXO1 and p47^phox in NOX3 function, NOXA1 is also able to activate NOX3 (63, 157).

NOX1 oxidase-organizing and activating subunits are also modulated by way of phosphorylation. Similar to its homologue p47^phox, NOXO1 phosphorylation enhances its binding to p22^phox. Plus, NOXO1’s interaction with NOXA1 is mediated by NOXO1 phosphorylation at Thr341 by protein kinase C (PKC) (231). Counter to conventional wisdom that phosphorylation results in complex formation and NOX activation, PKC-mediated NOXA1 phosphorylation has been suggested to prevent NOX1 hyperactivation (232). Paralleling this effect, PKA phosphorylates NOXA1 at Ser172 and Ser461, which permits NOXA1 interaction with 14–3-3ζ and dissociation of NOXA1 from the NOX1 complex - thereby negatively regulating NOX1 activity. On the contrary, PKA can positively influence NOX1 through phosphorylation of NOXO1 at Ser154, whose constitutive phosphorylation even under non-stimulated conditions may be responsible for baseline NOX1 oxidase activity (231, 233). Additionally, NOX1 phosphorylation at Thr429 by PKCβ1 (234) allows for the binding of the NOXA1 activation domain with NOX1. Thus, it appears that the location and balance between phosphatases as well as PKC and PKA isoforms in the vicinity of the canonical NOX1 oxidase might finesse its regulation.

2.1.2.5. p40^phox

p40^phox, alternatively termed neutrophil cytosolic factor 4 (Ncf4), is a 339 aa, 39 kDa protein that was first reported as constitutively associated with p67^phox in the cytosol of dormant phagocytes (48). The Ncf4 gene is located on chromosome 22 (22q12.3) (Table 1). Initial experiments performed in mature neutrophils of CGD patients lacking p67^phox indicated a need for p67^phox in the stable expression of p40^phox and association with the former via its carboxyl terminus. It is logical, therefore, that in CGD patients lacking p67^phox, p40^phox is present in reduced amounts (235, 236). From a structural standpoint, p40^phox contains domains in common with other cytosolic subunits and one that is unique. To that extent, from its N to C-terminus it is comprised of a PX domain, an SH3 domain and a PC (phox and cdc24p) domain, also termed the PB1 domain. As depicted in Figure 3, the subunit docks to the membrane via its PX domain and its key interactions with phosphoinositides, and to p67^phox by way of its PB1 domain (237) (Figures 2 & 3).

Interestingly, in response to p40^phox deletion, the degree of defect of NOX2 oxidase in innate defense depends considerably on the stimulus involved (238). With that said, the degree to which p40^phox deficiency compromises S. aureus killing may be as severe as with neutrophils lacking p47^phox (238–240). In fact, a variety of reconstituted systems (241–243) suggest p40^phox is a positive regulator of the NOX2 system (244), and to that end, p40^phox serves as an “adaptor” that recruits p67^phox and p47^phox to membranes (245). p47^phox interactions with membrane phospholipids appear crucial for p40^phox function and potentially vice versa (246, 247) in phagosomes (248) and early endosomes (247, 249) (Figure 3). Moreover, p47^phox and p40^phox PX domains bind to phosphatidylinositol-3,4-bisphosphate and phosphatidylinositol-3-phosphate, respectively (247). In that regard, PIP3s are necessary, but not sufficient, for binding of domains to the plasma membrane (250). As such, p47^phox and p40^phox cellular localization involves phosphoinositide-3-OH kinase PI(3)K activity (247, 250) and in this capacity, PI(3)K inhibition can disrupt p40^phox PX domain localization to PIP3-enriched early endosomes (247).

Thus, unique PI(3)K products in the vicinity of PI(3)K bind each p47^phox and p40^phox and juxtapose their membrane location (250). Moreover, p40^phox and p47^phox function as “adaptor” or “carrier” proteins of p67^phox when arachidonic acid is used as a stimulus in a cell free system (251), the physiological relevance of which is supported by the functional roles of phospholipase A₂ (PLA₂), which releases arachidonic acid from phospholipids (252). Finally, p40^phox can take the place of p47^phox as a p67^phox adaptor protein in late stages of sealed phagosome formation and prolong retention times of p67^phox on phagosomes (244). By no means, however, is p40^phox limited to its role as a positive modulator of NOX. Some reports describe it as not required for or eliciting a negative impact on NOX2 oxidase activity (253, 254).

2.1.2.6. Rac GTPases

Rac GTPases are indispensable for the catalytic competency of NOXs 1 – 3 but reportedly not required for NOX4, NOX5 and the DUOXs (164). Significantly, three highly homologous 21–22 kDa Rac proteins are present in mammals. Whereas Rac1 has been found to be ubiquitously expressed (42), Rac2 is the abundant and recognized NOX-related GTPase in neutrophils which is substantiated by its primary expression in myeloid cells and isolated from neutrophil cytosolic fraction (45). Uniquely, Rac3 appears to be most abundant in the developing brain (255) but is also found in some cancer cells (256).

Although Rac1 and Rac2 are reportedly 92% sequence-identical (257) their distinctions are related to their effector domain sequence diversity (residues 26–45), the so-named insert region (amino acids 124–135) and the most variable C-terminal polybasic region (258) - the latter being the domain that is reportedly critical for activation of the NOX (259, 260). Further supporting their distinction is the notion that Rac1 appears to be more uniquely required for the activation of NOX1 (259). Residues A27 and G30 in the effector-specific domain in the N-terminus of Rac have been characterized as indispensable for the support of NOX oxidase activity (260, 261). Moreover, Rac interacts directly with p67^phox in a GTP-dependent manner. Alterations in Rac by mutations in the effector site did not stimulate oxidase activity or bind to p67^phox, leading, therefore, to the notion that this cofactor is the Rac effector protein in the NOX complex (262).

2.1.2.7. p47^phox–p67^phox – p40^phox complex formation: electron transfer facilitation

In its latent state, NOX2 oxidase is not assembled with its cytosolic cofactors. Activation of the NOX2 complex requires the assembly of a multimolecular complex at the plasma membrane consisting of two integral membrane proteins, NOX2 and p22^phox, as well as two cytosolic factors, p67^phox and p47^phox. On the one hand, flavocytochrome b₅₅₈ (NOX2/p22^phox) is located in the membranes of intracellular granules and secretary vesicles (129). On the other hand, Rac exists in a GDP-bound cytoplasmic complex with Rho-GDI. In more detail, p67^phox and p40^phox are in an associated, yet dephosphorylated, state in a soluble trimeric complex, in which the p67^phox and p40^phox PB1 domains bind (263), and a SH3 domain at the C-terminus of p67^phox binds to the PRR of p47^phox (244, 264) (Figure 3).

NOX2 activation requires the en bloc assembly of the multi-component complex in the cytosol to the flavocytochrome in phagosome membranes by way of a sophisticated sequence of inter-protein, protein-lipid and intra-protein interactions. Unmasking of p47^phox’s autoinhibited SH3 domain enables p47^phox translocation and binding to p22^phox, inducing conformational changes and exposure of its PX domain (265). The exposure of PX domains on p47^phox and p40^phox is achieved by phosphorylation and results in the anchoring of the p47^phox–p67^phox–p40^phox complex to membranal phosphoinositides. Thus, when joined by Rac, the assembled complex enables efficient electron transfer and O₂^·− production (212). Upon phagocyte stimulation, subcellular granules containing flavocytochrome b₅₅₈ fuse with maturing phagosomes, and the p67^phox – p47^phox –p40^phox triad translocates to the phagosomal membrane (224, 244, 251).

2.1.2.8. X-ray crystallography and Cryo-EM

Until the past decade, structural data available for the NOX had been scarce on account of the complexity of the NOX isoforms and their subunits and the challenge of determining the structure of insoluble transmembrane proteins. For context, Magnani and coworkers, reporting on the crystal structure of the TM and DH domains of csNOX5 made highly illuminating comparisons with human NOX2 which paved the way for multiple structural examinations of the NOX (108). Indeed, at the time of the writing of this review, a full crystal structure of Streptococcus pneumoniae NOX2 (F397W mutant) and its subsequent dehydrogenase domain was delineated (103). While not the human NOX2 (25% sequence identity), this report has allowed for deeper kinetic and mechanistic insight into NOX2 since hallmark motifs are highly conserved across organisms. More importantly, this report advances the field in a very substantive way as the first X-ray crystal structure of a full-length NOX2 (103). In conjunction with a successive wave of other reports using cryo-EM to elucidate active and inactive structural dynamics in NOX2 and DUOX (86, 104, 109, 110, 194), the field has been transformed with respect to a confirmed structural elegance and the fine-tuned internal interactions of the NOX. Together, these papers have been trail-blazing and are expected to forge a new inflection point in the field with exponential discovery for years to come. For new revelations and corroborations made by these reports, the reader is referred to respective sections above on NOX2, p22^phox, DUOXs 1 & 2 and DUOXA1 & A2.

Unfortunately, for the remaining NOX core proteins and subunits, full structures do not generally exist to our knowledge. That is, cytosolic subunit structures, though often incomplete, include one with respect to p47^phox (PDB: 1NG2) (131) and p40^phox (PDB: 2DYB) (224), and a couple for p67^phox (e.g. 1HH8 (266) and 1K4U (226)). Similarly, there are crystal structures of various interactions between essential subunits, such as the interaction between p67^phox and Rac (PDB:1E96) (134) and two of structures of p47^phox: p22^phox interaction: one unveiled with X-ray crystallography (PDB: 1OV3) (131) and another with solution NMR (PDB: 1WLP) (267).

2.2. Modulation of NOX activity

This section is intended to give a high-level perspective on major modulators of the NOX and provide a subset of salient influencers of the NOX. Admittedly, there is a far greater number of enzymes, chaperones, scaffolding proteins and factors that merit discussion and toward that end, we refer you to an excellent recent comprehensive survey given by Laurindo and colleagues on the topic (268).

2.2.1. Activation of the oxidase by phosphorylation of regulatory subunits

Upon activation by a series of stimuli including pathogen-associated molecular pathways (PAMPs) via toll-like receptors (TLRs) and multiple cascades stimulating PKC activity, phosphorylation of critical serines on p47^phox lead to its unfolding, exposure of the SH3 supergroove and anchoring to p22^phox’s PRR domain (269, 270) (Figure 3). Studies corroborate that phosphorylation of p47^phox reinforces the binding of its latent cytosolic partners p40^phox and p67^phox to cytochrome b₅₅₈ (271). And, a major role of PKC notwithstanding, interleukin-1 receptor-associated kinase 4 (IRAK4), for one, has been shown to directly phosphorylate p47^phox leading to NOX2 activation (214, 272, 273). However, the extent to which it does this and influences p47^phox activation is unclear. Indeed, it is likely that a variety of previously undiscovered kinases modulate p47^phox phosphorylation and activation to varying degrees. Incidentally, the potent influence of phosphorylation extends to NOXO1 as does the interaction between NOXO1 and NOXA1 (231). Following on that premise, inactivation of phosphatases by way of oxidation augments NOX signaling via unabated phosphorylation of myriad kinases (274, 275). All told, phosphorylation of NOX components is the major means by which these enzyme systems are tightly regulated. In turn, oxidation and phosphorylation/activation of signaling kinases downstream of the NOX is a common mechanism by which this signaling is transduced (see other sections for greater detail on mechanisms).

2.2.2. Activation by lipids and arachidonic acid

Early discovery of the arachidonate-activated cell free NOX2 system set the stage of the discovery of cytosolic components of the NOX. Indeed, the discovery that long-chain fatty acids and anionic amphiphiles including SDS and lithium dodecyl sulfate could supplant fatty acids was a major paradigm shift allowing the study of the NOX in cell-free systems (276, 277). Moreover, phosphoinositide 3-kinase (PI3-K) generates phosphatidyl inositol (PI) derivatives that bind PX domains in p47^phox and p40^phox (247); these include those described in 2.1.2.1, and PI3P for p40^phox (248). PIP2 and PIP3, like diacylglycerol, stimulate PKC activity utilizing phosphatidyl serine as a cofactor, reduce Ca²⁺ levels and phosphatidyl serine concentrations, and are complementary in their stimulatory effects (278). Ca²⁺ ion release from the endoplasmic reticulum (ER) evoked by IP3 binding permits activation of cytosolic PLA₂ (279), which, in turn, liberates arachidonic acid from its lipid-bound source (e.g. phosphatidylcholine) in membranes. Additionally, free-form arachidonic acid stimulates PKC and elicits NOX activation (280). Accompanying this action, arachidonic acid, and other anionic amphiphiles, are suspected to induce conformational change in p47^phox and p67^phox supporting their association with p22^phox and NOX2 (281). Moreover, arachidonic acid can release the p47^phox SH3 tandem domain from its AIR, thus promoting p47^phox binding to p22^phox (282, 283). Intriguingly, exogenous addition of arachidonic acid in Rac2 knock-out neutrophils rescued NOX2 activity (284).

2.2.3. The “troposphere” of influence: proteins modulating NOX activity

2.2.3.1. Protein disulfide isomerases

Protein disulfide isomerases (PDIs) are redox thiol chaperones of the thioredoxin class of proteins that are principally found in the ER where they serve a canonical function in aiding the appropriate folding of nascent proteins. The PDI family prototype PDIA1 or simply PDI is a 55-KDa protein that has four thioredoxin-like domains named a-b-b’-a'. The two “a” domains contain the redox catalytic Trp-Cys-Gly-His-Cys (WCGHC) motifs, while the "b” domains participate in recognition and binding of the substrate, with the b´ domain being the main substrate binding site (285, 286). PDI is a flexible protein, depending on its oxidized or reduced state. Oxidized PDI shows an open conformation, which allows for interactions with unfolded proteins (287). A regulatory role of PDI on NOX2 was demonstrated by coimmunoprecipitation of PDI with NOX subunits and by inhibition of AngII-induced ROS generation in vascular smooth muscle cells (VSMCs) upon inhibition or silencing of PDI (288). Notably, PDI’s effect has been established in a cell-free assay to test for NOX assembly, the results showed oxidized PDI promoting O₂^·- generation while reduced PDI inhibiting it (289). In macrophages, PDI has played a role in facilitating NOX2 assembly and phagocytic respiratory burst activity (290). These results indicated a novel function of PDI in the regulations of NOXs (288). The mechanisms involved are still being investigated but seem to involve a two-staged process with one at the membrane being a redox-mediated activation of NOX assembly and the other, in the cytosol, involving a redox-dependent association with p47^phox in a long term regulatory loop consistent with NOX complex recycling (289). This is especially evident for NOX1 hybrid oxidase assembly wherein the molecular mechanisms involve the formation of a disulfide bond between PDI’s Cys400 with Cys196 located in p47^phox SH3 binding domain, facilitating NOX1 assembly and contributing to atherosclerosis (291). Along those lines, increased expression of PDI in mesenteric resistance arteries of spontaneous hypertensive rats contributes to an increased NOX1-mediated ERK1/2 MAPK signaling in response to AngII (292). Additionally, PDI silencing led to a disruption in Rac1 and RhoA activation and cytoskeletal disorganization and, additionally, silencing or overexpression of an inactive PDI decreased migration of VSMCs (291, 293). A recent study identified the role of PDI as a mediator of NOX1 upregulation and vascular dysfunction in hypertension. In that vein, overexpression of PDI in VSMCs increased the translocation of ATF1 to the nucleus and ATF1 binding to NOX1 gene regulatory regions upregulated NOX1 transcription (294). Altogether, PDI’s role in NOX1 organization and its association with p47^phox and Rho GTPases which are at the juncture between NOX1 activation and cytoskeletal organization, could explain the role of this chaperone in sustaining vascular patency during vascular remodeling. Additionally, PDI-induced upregulation of NOX1 expression appears to sustain vascular contractility and contribute to vascular dysfunction in hypertension (294–296).

2.2.3.2. Heat shock proteins

As with other proteins, heat shock proteins (Hsp) and molecular chaperones 70 and 90 facilitate NOX function by ensuring their proper folding and stability. The balance between the two appears to be the primary determinant of the fate of NOX proteins vis-à-vis ubiquitination and degradation through proteasomal degradation. Loss-of-function of Hsp90 brings about a decrease in steady-state levels of NOX1, NOX2, NOX3 and NOX5 consistent with Hsp90 binding stabilizing the isoforms and redirecting them from degradation (297). Interestingly, Hsp70, whose levels are upregulated by a decrease in Hsp90 and factors that induce NOX activity like mitogens and PMA, has the opposite effect (297). That is, overexpression of Hsp70 decreases NOXs1, 2, 3 and 5 levels and activity, i.e. Hsp70 binds to the same NOX hemoproteins and appears to commit them to degradation (297, 298).

2.2.3.3. Peroxiredoxins

Peroxiredoxins (Prdxs) comprise a family of proteins with peroxidase-like activity whose primary function appears to be the scavenging of H₂O₂ and other cellular hydroperoxides. Prdx6 is unique from other mammalian Prdxs in its capacity to reduce phospholipid hydroperoxides and in its PLA₂-like activity (299). These properties justify unique interest in Prdx6 in the release of arachidonic acid and conformational activation of the NOX. Indeed, studies have shown that Prdx6 participates in NOX2 activation in neutrophils, macrophages and ECs (300) and in an unbiased search for NOX binding partners employing a yeast two-hybrid system using NOXA1’s C-terminal SH3 as bait, Prdx6 was proposed as a binding partner. Further experiments showed that Prdx6 stabilizes NOXA1 through its SH3 domain (300). Furthermore, mutation of this SH3 domain disrupts binding, and depletion of Prdx6 inhibits NOX1 activity, while its overexpression enhances it as well as the presence of both NOXA1 and NOXO1 in a variety of cell lines (300). Paralleling this role in in ECs, Prdx6 also binds NOXA1 homolog p67^phox (301).

2.2.3.4. EBP50

Ezrin-radixin-moesin (ERM) binding phosphoprotein 50 (EBP50) is an adaptor protein classically discovered in the apical region of epithelial cells (302). It contains two postsynaptic density 95/disc-large/zonula occludens (PDZ) domains conferring it the capacity to scaffold other proteins to integral membranes and the actin cytoskeleton by virtue of its binding to ERM proteins. EBP50 facilitates assembly and regulation of multiprotein complexes that tether transporters to the cytoskeleton (303). Our group revealed that EBP50 mediates NOX1-dependent biological effects, i.e. VSMC proliferation, hypertrophy and vasoconstriction through its binding to p47^phox (and not NOXO1) and activation of the non-canonical or hybrid NOX1 oxidase (304). Evidence for direct p47^phox binding to EBP50 was proved using confocal microscopy, proximity ligation assay, co-immunoprecipitation and FRET. Deeper inquiry revealed that a unique four amino acid C-terminal domain in p47^phox, but not NOXO1, conferred selectivity to the interaction and insertion of the four-amino acid A-S-A-V PDZ motif from p47^phox into NOXO1 bestowed on the latter the ability to bind EBP50 and had no effect on the NOX2 or NOX1 canonical systems. In accordance with p47^phox’s known binding to and translocation along the actin cytoskeleton, it stands to reason that EBP50 uniquely scaffolds and chaperones p47^phox into a hybrid NOX1 system. These experiments appear to pave the way for the discovery of an assortment of to-date undiscovered specific protein-protein interactions as facilitators in the delicately orchestrated assembly of NOX subunits. Moreover, it is tempting to speculate that EBP50’s historical role in epithelial cell apical membranes indicates a possible previously undefined role in colonic NOX1 action.

2.2.3.5. Essential for Reactive Oxygen Species

Another intriguing chaperone for the NOX was identified in mice lacking a previously uncharacterized open reading frame bc017643. These mice displayed almost complete insufficiency of gp91^phox and p22^phox and were susceptible to Salmonella typhimurium and Listeria monocytogenes infection (305). bc017643 encodes an ER transmembrane protein that co-sediments with gp91^phox named EROS (Essential for Reactive Oxygen Species) (305, 306). EROS is described as a selective “place holder” chaperone for NOX2, binding to the NOX2 precursor and maintaining its stability and proper conformation until p22^phox can bind to it. Human EROS mutant C17ORF62 that was re-named to CYBC1 (CYtochrome B Chaperone 1) has been asserted as a novel cause of CGD (307).

2.2.3.6. DNA Polymerase Delta Interacting protein 2

DNA Polymerase Delta Interacting Protein 2 (Poldip2) was originally established as associated with DNA-Polymerase delta (δ) and proliferating cell nuclear antigen (PCNA) and, thus, was implicated in gene expression regulation and DNA repair (308). Poldip2 has been localized to the nucleus, cytosol, plasma membrane and mitochondria and performs multiple functions independent of NOXs, which have been detailed elsewhere (308). Germane to this discussion, Poldip2 reportedly associates with smooth muscle NOX1 and NOX4, but only with NOX4 stimulates ROS in a manner involving p22^phox expression (162). Loss-of or gain-of Poldip2 decreases or increases ROS production, by NOX4/p22, respectively, and Poldip2 silencing promotes a phenotype much like that elicited by NOX4 suppression, uncoupling NOX4/p22^phox from focal adhesions. Somewhat unexpectedly, under or overexpression of Poldip2 impairs migration of VSMCs. This seemingly conflicting observation is proposed to be a consequence of a disrupted balance and coordination of regional (front vs. back) adhesions and loss of cell force polarization (309).

2.2.3.7. Negative regulator of reactive oxygen species

Negative regulator of reactive oxygen species (NRROS aka LRRC33) is an ER-localized transmembrane protein with several leucine rich domains that limits ROS generation in phagocytes by desensitizing TLR signaling during inflammatory responses (310, 311). Further, NRROS limits ROS generation by directly interacting with the nascent NOX2 (Cybb) monomer and facilitates its proteasomal degradation. Indeed, it is localized to the ER where it can associate with and degrade NOX2 via ER-associated degradation and NRROS-deficient phagocytes generate higher levels of ROS upon an inflammatory stimulation along with improved anti-bacterial function. Therefore, modulation of NNROS levels in the ER is a means by which phagocytes can regulate appropriate levels of microbicidal ROS, and also maintain a level that avoids unintended collateral damage. In that sense, it is logical that NRROS-induced degradation of NOX2 may curb hyperinflammation and autoimmune diseases. Indeed, NRROS-null phagocytes produce excessive ROS in response to inflammatory challenges leading to severe autoimmune disorders like encephalomyelitis (310). For an unabridged discussion of NOX modulatory factors please see (268).

2.3. NADPH Oxidase-mediated signal transduction

2.3.1. Redox-sensitive signaling proteins

In redox signaling, target proteins or lipids are transiently oxidized or reduced in a graded manner by an oxidizing or reducing species, respectively. This commonly involves NOX-derived H₂O₂-induced oxidation of cysteine residues in signaling proteins (but can also include methionines or selenocysteines), i.e. progressive oxidation of their accessible and susceptible thiolate anion to sulfenic, sulfinic and sulfonic acids (reversible to irreversible in that order) which then by influence on catalytic sites or by allostery stimulates (or inhibits) protein activity. This effect, in turn, modulates a proximal signal. Multiple other consequential alterations include, among others, the formation of mixed disulfides and internal disulfide bridges. For far more detail on these processes and the factors influencing them, the authors refer you to (312–315). Primary oxidant targets of NOX include proteins with a broad spectrum of functions of which only a subset is described below.

2.3.2. Ion channels/Ca²⁺ signaling

One of the first studies to report on NOXs’ role influencing channel activity was that of volume-sensitive chloride (Cl^-) currents in osmotic swelling of cardiomyocytes. In that report, the authors identified a causal role for NOX2 in the process and demonstrated its abrogation by the specific NOX2 inhibitor gp91ds-tat (316). Expanding on this theme, Earley and coworkers showed that an interplay between NOX2 and TRPA1 causes spikes in calcium entry in cerebral arteries by H₂O₂- and lipid peroxide-triggered single channel opening, i.e. activation of NOX2 induced TRPA1 sparklets within milliseconds and vasodilatation immediately after channel opening (317). The investigators also showed that NOX2 colocalizes with TRPA1 and that its signal can be abolished by NOX2i (57, 317). More recently, TRPC3 and TRPC6 activated in vascular SMCs required NOX1 for calcium influx (318). In connection with this, Park et al, showed that AngII-induced elevations in calcium related to contraction were dependent on NOX1 mobilization of extracellular calcium, and suggested that calcium mobilization from intracellular stores (such as store-operated calcium channels) is not dependent on NOX1 (318).

Plasma membrane Ca²⁺ ATPase is transiently inhibited by H₂O₂ (319). On the other hand, the inositol 1,4,5-triphosphate (IP3) receptor (IP3R)-mediated Ca²⁺ release is stimulated by ROS (319) i.e. cysteine residue modification in the IP3R sensitizes it to IP3 (320). Substantiating this finding, generation of IP3 and addition of inhibitor of sarcoplasmic reticulum Ca²⁺ ATPase (SERCA), thapsigargin, did not alter calcium influx in VSMCs from NOX1-deficient mice (321). Likewise, NOX1 participates in thrombin-stimulated smooth muscle cell intracellular calcium mobilization via activation of L-type voltage dependent calcium channels; this dynamic and its pro-migratory effects are abolished in NOX1-deficient mice (321). As a final point, insofar as NOX1 and NOX4 have been localized to the mitochondria, it is consistent with their roles to contain susceptible (low pKa) cysteine and methionine residues that can mediate this effect. Indeed, multiple studies demonstrated the ability of ROS to stimulate entry via both L-type (322–324) and T-type voltage-gated channels in VSMCs (325).

Intriguingly, NOX4 has been implicated in the activity of K⁺ channels involved in the response to hypoxia through a direct protein-protein interaction with TASK-1, a two-pore K channel (326) which appears to sense O₂ level declines by way of NOX4. As for the ryanodine receptor (RyR), it possesses redox-active cysteines that are oxidized by NOX4, increasing calcium release and muscle contraction (327).

2.3.3. Inhibition of phosphatases

One of the best characterized effects of NOX in signaling is the inhibition of protein tyrosine phosphatases (PTPs) by NOX-derived ROS. Key reactive thiols on PTP1B exhibit low pKa values and therefore are reactive at physiological pH of 7.0 (328); this may more logically occur when receptor tyrosine kinases like EGFR inhibit peroxiredoxins in the vicinity of the PTP (329) (Figure 5). Oxidative inhibition of PTP1B has a permissive effect - allowing kinase signaling to proceed unabated (330). EGF, PDGF and insulin act through tyrosine kinase receptors, induce NOX-derived oxidants and inhibit a select number of PTPs (331). For example, NOX4-derived H₂O₂-mediated inhibition of PTP1B accelerates cell proliferation and migration of glioblastoma cells by way of sustained coronin-1C phosphorylation, a mediator of cell motility (332). Moreover, GPCR activation depends on NOX-regulated PTP inactivation and AngII induces the activation of NOX1 leading to oxidative phosphorylation of SHP2, another PTP (275). Furthermore, NOX4-induced inhibition of focal adhesion kinase (FAK) tyrosine phosphatases contributes to the maintenance of FAK phosphorylation and integrin signaling required for cell migration and survival (333, 334).

Figure 5. Model of compartmentalized NOX-ROS downstream signaling. Constellation of intracellular mechanisms channeling NOX-derived H₂O₂ to kinase targets and various phenotype shifts.

Diagrammatic representation of how oxidative and phosphorylative mechanisms in the vicinity of the NOX2 source in lipid rafts can theoretically channel NOX-derived H₂O₂ for redox signaling. In this scenario, growth factor receptor (GFR) transactivates and stimulates NOX2 assembly and H₂O₂ production which, in turn, activates a target kinase. GFR’s phosphorylation and inhibition of peroxiredoxin I (Prx I) prevent its degradation of NOX2-derived H₂O₂. Simultaneously, NOX2 H₂O₂ oxidatively inactivates Prx II further augmenting steady-state H₂O₂ in the vicinity of NOX2 and its target kinase. Additionally, H₂O₂ via cysteine oxidation of phosphatases inhibits phosphatase activity which is permissive of target kinase activity and feedforward signaling. Away from NOX2, Prx I and II remain active and scavenge H₂O₂ from other potentially interfering sources. Collectively, these effects protect and allow localized H₂O₂ to rise close to target kinase(s) and propagate downstream signaling leading to an array of phenotypic shifts. In sum, these reactions restrict and direct NOX- H₂O₂ while, in other areas of the cell, active Prxs degrade H₂O₂ preventing interference with signaling (see section 2.4).

2.3.4. Kinases

While often triggered by NOX/ROS-mediated signaling, c-Src non-receptor tyrosine kinase regulates multiple downstream signaling targets including MAPK, PI3K and PKC (71). Interestingly, c-Src can be also upstream of NOX, since NOX1 and 2 are regulated through c-Src-dependent processes (71, 335). Molecular mechanisms linking EGFR to downstream signalling involve NOX and c-Src. c-Src activation is achieved via cysteine oxidation; i.e. NOX-induced oxidation of Cys185 and Cys277, resulting in their sulfenylation, serves as a molecular “switch” that promotes its activity (336). Moreover, c-Src can activate NOX1-mediated ADAM17 and cause EGF shedding (337) and more recent studies suggest a role for c-Src in vascular contraction mediated by NOX5 (71).

2.3.5. Tyrosine kinase substrates

Tyrosine kinase substrate (Tks) proteins, discovered as Src tyrosine-kinase substrates, are scaffolding proteins that promote cell migration, proliferation and differentiation primarily by regulating the EGFR signaling (268). Interestingly, Tks4 and Tks5 were found to be homologous to the p47^phox organizer superfamily which clued investigators into the discovery of their NOX activating role (102, 268). On their N-terminus are successive SH3 domains and a PX domain, multiple PRRs and several Src phosphorylation sites (268). Interestingly, Tks4 and TKs5 expression are instrumental for invadopodia and podosome formation and function (338–340) and by binding to NOXA1, Tks4 promotes NOX1 recruitment to invadopodia (152). Lastly, Tks4 and Tks5 were shown to support NOXs1 – 4-dependent ROS generation in a variety of cell types (152, 341).

2.4. NOX subcellular compartmentalization

Broadly speaking, the scope and intensity of a particular ROS’s effect has long been proposed to depend on the location of its source, its reactivity, quantity, and sphere of influence. The latter naturally depends heavily on the subcellular organelle and its localized environment which includes antioxidant enzyme systems and nearby amino acids, e.g. active cysteines and methionines, on proteins in response to ROS derived from NOX. Indeed, subcellular compartments within cells vary widely as to their redox potential and there is skepticism as to the functionality of NOXs in a number of organelles (342). Subcellular localization of NOX subunits enables their spatio-temporal action and specific and diverse biological functions. A thorough survey of the literature on NOXs subcellular localization, however, comes with caveat of their precision due to a limited reliability of various antibodies.

That said, groundbreaking work by Rhee and colleagues in 1999 set the stage for the notion that NOXs in lipid rafts can effect localized oxidation and activation of signal transduction pathways (6). That is, cellular “machinery” is in place to facilitate compartmentalization and protection of localized ROS production by the NOXs for discrete action upon downstream targets (343, 344). In that regard, the peroxiredoxins (Prxs) that are localized to the cytosol, the mitochondria and the ER are capable of delimiting NOX-mediated signaling (343, 344) (see Figure 5). Indeed, NOX-derived oxidants are met by a sentinel of peroxidases that localize their signaling effects while guarding against random and widespread oxidation. In the case of Prdx1 (aka Prx I), it is phosphorylated and inactivated by receptor tyrosine kinases (e.g., PDGFR, EGFR) in the region close to the production of ROS (by NOX), thus allowing a sustained localized H₂O₂ to exert its effect on signaling (Figure 5). Whereas Prdx2 (aka Prx II) is not inhibited by phosphorylation, it is oxidatively inhibited by sustained levels of H₂O₂ in the vicinity of its source. Additionally, localized H₂O₂-mediated PTP inactivation promotes the action of downstream kinases to exert sustained activation and thereby allows the signal to fan out and propagate (275, 345). On the other hand, away from the receptor kinases, Prxs remain active and attenuate H₂O₂ (6); thereby avoiding toxicity to the cell from errant H₂O₂ as well as keeping interference by H₂O₂ from other sources at bay (Figure 5). These fascinating revelations provide the foundation on which we may appreciate how localized NOXs act on their targets.

In fact, over the years, there has been a growing appreciation for the notion that NOXs differentially assemble within discrete organelles like endosomes and lipid rafts and exert their effects as agents of localized signaling. For instance, cytokine-induced NOX1 or NOX2-derived H₂O₂ putatively activates NFκB in the vicinity of endosomes and AngII receptor in caveolin-associated lipid rafts transduces a signal to transmodulate EGF receptor and Akt via NOX1 (346–353). At the time of the seminal findings by Miller and coworkers in 2007 with respect to NOX1 in endosomes activating NFκB, we postulated that endosomal NOX1 might be viewed as mobile generators and instigators (personifying them as “moving marksmen”) of differential signaling across the cell (354). As of now, there is no evidence to our knowledge that definitively affirms this viewpoint. However, we predict that with the development of new tools for the dynamic detection of NOX motility, ROS production and bio-panning techniques, major breakthroughs will elucidate the array of oxidases and their alternating targets across subcellular organelles. Indeed, recent studies have supported this general contention in the etiology of disease (355).

2.4.1. Plasma membrane and specific granules

The subcellular localization of NOXs in the cell membrane ensures the release of ROS into the extracellular space, which is undeniably important for intercellular ROS signaling, thyroid function, gut-microbiota symbiosis and pathogen elimination. Membranal NOX1 and DUOX2 in epithelial cells of the gut are at the vanguard between the colon and the microbiome, where they have been ascribed a role in the control of microbiome colonization (356, 357) and in host defense (357–359). In resting neutrophils, on the other hand, NOX2 is located at the plasma membrane and in specialized granules. Upon activation by a variety of receptor-mediated events converging on PKC activation and phosphorylation, NOX2 becomes more enriched in specific submembranal granules that fuse at first with the plasma membrane and later with phagosomes in conjunction with azurophilic granules containing myeloperoxidase in the lead up to the generation of H₂O₂ and HOCl during phagocytosis (95). For greater detail on this captivating activation process a pair of in-depth reviews are recommended (95, 96). In non-phagocytes, there have been multiple reports suggesting that NOX2 on the plasma membrane supplies H₂O₂ in crosstalk with other nearby cells as an agent of paracrine signaling (360–366). A similar situation is described for NOX1 and NOX3, which are described to regulate migration of cancer cells by ROS-dependent formation of invadopodia (151). Moreover, NOX5 is demonstrably present at the plasma membrane, confined to specific membrane domains that appear to allow release of ROS into the extracellular space, although the verdict is still out on its function there (367). Integrating form and function, it is tantalizing to propose that, in addition to possible paracrine signaling roles in non-phagocytic cells, NOXs 1, 2, 3 and 5 play some role in the stability and allosteric activation of nearby cell surface proteins not the least of which are inducible metalloproteases involved in migration (368).

2.4.2. Endosomes/redoxosomes

The plasma membrane can also transduce signals via endosomes/redoxosomes starting with invaginations which become agents of signaling as endosomes in the cytosol, not unlike that which occurs in phagocytes. Thus, it is not unreasonable to propose that redoxosomes might serve a role as depots in which NOX can travel across the cell (369) and transduce their signals to targets in the cytosol or organelles where they are required (see above). In fact, one might envision an endosome/redoxosome as a mobile “oxidizing command center” which delivers the NOX to its site of action (354). Assuredly, NOXs in redoxosomes can interact with lipid rafts on juxtaposed membranes and coordinate modulatory signals through the activation of ion channels, transporters, and receptors (370).

2.4.3. Mitochondria

For quite some time we have known that there is crosstalk between cellular NOXs and mitochondria (371). One prime example of this is that mild mitochondrial dysfunction induced by mtDNA mutations can abolish NOX activation by AngII, in part, due to a failure to upregulate NOX1 (372). Moreover, mitochondrial perturbations can cause an increase in NOX4 mRNA suggesting induction of cellular NOXs by even minor stresses to the DNA or mitochondrial electron transport (373). Indeed, an earlier report of mitochondrion-localized NOX4 having a functional role in disease reinforces these findings (374). On the other hand, the impact of mitochondrial stress on NOX appears cell and condition dependent (322–324).

Bedside their crosstalk in the cell, NOXs are described to localize to different compartments of mitochondria. Indeed, multiple lines of evidence indicate that NOX4 is endogenously localized to the mitochondrion via a 73-aa mitochondrial targeting sequence on its N-terminus directing it to the inner mitochondrial membrane (375, 376). To that point, NOX4 has a putative mitochondrial localization signal and was reported to localize in mitochondria in NIH3T3 cells (377) and is mitochondrially detected in kidney cortex, cardiomyocytes (378) and in cancer cells (379). Similarly, NOX4 was found in the mitochondria of the rat kidney cortex and glomerular mesangial cells as detected by immunoblotting (374). Incidentally, AngII-stimulated NOX is associated with opening of mito-KATP channels, depolarization of membrane potential and mt-ROS production (379) and VEGF-induced angiogenesis involves sequential activation of NOX2, NOX4 and mtROS (380).

2.4.4. Endoplasmic reticulum

The question has been raised by some as to whether NOX-derived ROS in the endoplasmic reticulum might be involved in protein maturation and quality control. To that point, however, a major generator of ROS and player in protein folding in the ER is endoplasmic reticulum oxidase 1 (ERO1) (381). By multiple accounts, however, an array of NOXs including NOX2, 4 & 5 may be detected in the organelle (381) that have raised the possibility of a similar role for them. Nevertheless, with the possible exception of NOX4, NOXs do not appear capable to produce appreciable ROS in the ER (268) and by comparison the whole of ER NOX-derived ROS likely pales in comparison to ERO1 (381). Indeed, ERO1 and PDI are primary to proper protein folding in the ER as PDI facilitates the formation of disulfide bonds and ERO1 rejuvenates PDI by accepting electrons from it and generating ROS (redox cycling) (381). Taking these findings together, there remains considerable doubt as to whether ER NOXs contribute substantively to steady-state ER ROS, and by extension protein folding (97, 382). On the other hand, PDI can both (a) facilitate NOX1 assembly and activity; and (b) upregulate NOX1 mRNA & protein expression (but not NOX4 level) ultimately leading to increased NOX1 activity (291, 383). Somewhat oddly, this appears to suggest that PDI can promote transcription of NOX1 but the mechanism for this is still open for debate (291, 383).

Moreover, in spontaneously hypersensitive- (SHR) vs. normotensive Wistar Kyoto (WKY) rats, Camargo et al. show increased NOX4 expression and ROS in the ER of vascular SMCs that correlated with ER stress, hyperoxidation and phosphorylation of unfolded protein response (UPR) mediators PERK and IRE1α. These findings hint at a localized effect of ER NOX4 on ER stress associated with pro-apoptotic events which contribute to vascular dysfunction in hypertension (384, 385). In contrast, previous studies by Santos et al. had shown that, compartmentalized ER NOX4 in the heart, per se, is induced during UPR and ensures cell survival (385). In aggregate, these studies suggest an interplay between NOX4 and ER stress activating either pro-survival or pro-apoptotic UPR signaling pathways in a context-dependent manner.

2.4.5. The nucleus

NOX2 and/or NOX4 have been localized to the nucleus of epithelial cells, ECs, smooth muscle cells, hepatocytes, renal cells, cardiomyoblasts, myelodysplastic cells, and hemangioendothelioma cells (386–393). In rat vascular SMCs, NOX4 immunostaining is detected in nuclei, even more specifically in the nucleolus (388). Additional studies revealed perinuclear or nuclear distribution of NOX4 in COS7, HEK293, and ECs (387, 394).

While some data suggest an effect of NOXs in the nucleus, it is unclear at this time, what biological effects they confer there. On the one hand, data appear to suggest an influence on nuclear gene expression via the activation of transcription factors in the cytosol. These include factors like AP-1, nuclear factor (NFkB), p53, CREB, HIF-1alpha and HIF-like factor. Activated and oxidized forms of these factors in the cytosol are reduced by redox factor-1 (Ref-1) in the nucleus whose redox state, under some circumstances, can be dependent on thioredoxin (395). In contrast, a pair of reports suggest NOXs might influence gene expression and splicing and/or trigger the DNA damage response at elevated levels (387, 396). To the last point, NOX4 was found to stimulate gene transcription as well as track with nuclear domains involved in pre-mRNA splicing and an indicator of DNA damage H2AX (396).

3. BIOLOGICAL FUNCTIONS

3.1. Proliferation

ROS derived from NOXs are implicated in modulating the proliferation of myriad cell types, in vitro, in vivo and ex vivo (91, 394, 397–410) among which NOX1 has arguably been the most thoroughly investigated. In fact, dating to its original discovery, NOX1 was referred to as Mox1 (mitogenic oxidase 1) on account of its perceived ability to mediate normal serum-induced growth in smooth muscle cells and to transform fibroblasts to a neoplastic phenotype (142, 143). As alluded to above, this deduction was later called into question by the presence of transformed Ras in the preparation (411). Salient examples of this NOX-modulated pro-proliferative cellular effect follow.

A large number of publications on proliferation implicate or deem NOX1 as causal (91, 401, 412–414) in many cell types and primarily the colon where it is enriched (see also sections on the colon and cancer). Interestingly, multiple cell types in lung as well have received a great deal of attention in this regard. One cell, type in particular, is lung epithelium (413). In brief, NOX1-ROS stimulated cell proliferation and delayed cell cycle withdrawal in actively cycling cells by reducing the threshold for growth factors to induce cyclin D1 expression, whereas during cell cycle re-entry, NOX1 is required for transcriptional activation of c-Fos and Fra-1 pivotal in immediate early gene response (413). Consistent with those findings, inhibition of NOX1 expression evokes a G₁/S block in colonic epithelial cell cycle progression associated with a decrease in cyclin D1 levels and profound MAPK signaling suppression that is not accompanied by increased apoptosis (414). In vascular cells, a series of reports have implicated a role for NOX1 in proliferation which invoke the role of a variety of factors including AngII, thyroid hormone and oxidized LDL (415–419).

With respect to the vasculature in the lungs, in pulmonary hypertension (PH) lung arterial SMCs proliferate in a NOX1-H₂O₂-dependent manner following stimulation with an estrogen metabolite, 16α-hydroxyestrone (420). This is consistent with findings that NOX1 promotes: (a) vascular hyperplasia in human pulmonary arteries from patients with pulmonary arterial hypertension (PAH); and (b) proliferation of human PAECs in vitro (415, 417). In the above-mentioned study examining the effects of estrogen, NOX1-mediated ROS triggered oxidation (and inhibition) of protein tyrosine phosphatases, increased phosphorylated p38 MAPK, and decreased Nrf2 bolstering the pro-proliferative effects of the estrogen metabolite (420). Some other notable NOX1 downstream mediators in PH include bone morphogenetic protein receptor 2, sonic hedgehog signaling and specificity protein 1 in the lung endothelium (415, 416). In greater detail, our group has identified a two pronged feed-forward loop in endothelial proliferation involving on the one hand, PKA, CREB, bone morphogenetic protein receptor 2 and sonic hedgehog signaling, and, on the other hand, specificity protein 1 and serum-derived factor-1 as drivers of pulmonary hypertension (415, 416).

Although less prevalent in the literature, multiple reports invoke a role for NOX2 and NOX4 in proliferation through activation of a variety of signaling pathways (332, 394, 403, 421, 422). For example, NOX2 and NOX4 have been shown to converge on induction of MAPK/ERK, Akt, and SMAD pathways in EC proliferation but also have the capacity to include one or more of NFκB, PDGF and eNOS in their signaling repertoire (423). Substantiating a role for NOX4 is the demonstration that Poldip2 promotes proliferation in ECs (424). Additionally, NOX4 was noted to interact with phosphorylated VEGF receptor 2 and elicit proliferation of VEGF-induced human retinal microvascular ECs (425). In support of a feed-forward interaction in this pathway, NOX4-derived ROS can activate NOX2, and, in turn, induce VEGFR increasing human and mouse EC proliferation (423). Indeed, in PH, mitochondrial H₂O₂ (plausibly derived from both mitochrondrial electron transport and mitochondrial NOX4) plays a crucial role in a propagatory response (426). Thus, the complexity of these pathways is only heightened by a bifurcating role of mitochondrial ROS and additional studies showing that NOX4 can set-in-motion amplified ROS generation from multiple subcellular organelles (394, 426) involving an array of pro-proliferative growth factors and transcription factors (332, 403, 422).

In PASMCs, hypoxia-induced NOX4 upregulation results in increased H₂O₂ and a reduction in PPARγ expression involving activation of a proline-rich tyrosine kinase 2 (Pyk2)/ERK1/2/NFκB, a process which expectedly leads to greater NOX4 expression. In synchrony, this appears to propagate NOX4’s role in proliferative signaling (427–429); and evidence in a separate model suggests a similar pattern. That is, hypoxia-induced PPARγ downregulation is expected to spur the expression of thrombospondin 1 (TSP1) which, in turn, induces NOX4 expression and PASMC proliferation further contributing to vascular changes in the development of PH (430).

Additionally, stimulation of human microvascular ECs (HMEC) with low dose glucocorticoid increased their proliferative response, with implication that p22^phox and NOX4 contribute to ensuing systemic hypertension, right and left ventricular dysfunction and PH-related vascular remodeling (84). Indeed, Kračun and colleagues argue that the role of these NOX components in systemic hypertension might be explained by increased NOX-derived O₂^·- in peripheral arterioles and augmented vascular tone. Mechanistically speaking, ROS from NOX2 and NOX4 were shown to be responsible for the activation of HIF1 by glucocorticoids that expectedly is a prime driver of the endothelial proliferation (84). Further evidence can be found in human hybridoma ECs wherein NOX2 and NOX4 overexpression increase ROS production and proliferation, while their silencing inhibited ROS and basal EC proliferation (394). In addition, overexpression of NOX5S and NOX5β in HMEC1 results in significantly augmented cell proliferation and angiogenesis compared to control vector-transfected cells (178). By way of knockdown, the authors were able to substantiate EC NOX5’s role in thrombin’s pro-proliferative effect and capillary-like structure formation (178).

On the whole, we have observed that during the earlier days of tissue NOX research, the role of NOXs in the mechanistic regulation of cell proliferation was more deeply explored than it is today. More recent studies have demonstrated an unfortunate trend. That is, they appear to almost entirely center on the pathology in question identifying only whether or not a particular NOX is involved. From our viewpoint, more recent studies either describe the proliferation as the governing mechanism of a certain type of cancer or as a main mediator of pulmonary vascular remodeling (loss-of and gain-of-function). In that respect, a “reboot” of deeper mechanistic interrogations of biochemical and signaling pathways affected by NOXs seem highly warranted.

3.2. Migration

Cell migration is one of the basic cellular processes underpinning morphological changes, i.e., angiogenesis, extravasation of innate and adaptive immunity-driving cells, SMC migration into neointima, and invasion and metastatic cancer cells. Importantly, NOX-derived ROS actively participate in the precision signaling and tight regulation of migration that involves cell polarization and actin polymerization and depolymerization in various physiological and pathological settings.

Further to the point, multiple cell types migrate in response to a variety of tissue factors ranging from cytokines, chemokines, growth factors to hormones. For example, primary agents of proliferative and angiogenic processes, growth factors are stimulated by oxidants derived from NOXs both directly and, indirectly via HIF1 (371, 431, 432). Much has been reported on the role of NOX in cellular migration with a major focus on vascular smooth muscle cells (94). In fact, a multitude of factors and downstream kinases including Src, CAMK-II, PKC, and JAK-STAT have been shown to act on or in conjunction with MAPK pathways to effect proliferation and migration (94, 433). As a prime illustration of this, CAMK-II and 14–3-3 oxidation induce MMP9 and de-phosphorylation and activation of cofilin, respectively, and collectively contribute to migration at the cells’ leading edge (434, 435).

With respect to the role of individual NOXs, SMCs derived from NOX1-deficient murine thoracic aorta showed decreased oxyethidium fluorescence, EGF transactivation and migration in response to thrombin compared to WT suggestive of a role for O₂^·--induced EGF in these effects (436). Indeed, NOX1 deficiency prevented thrombin-induced phosphorylation of Src, transactivation of EGFR, and the activation of MMP9 by ERK1/2 (436). Moreover, thrombin’s trademark N-cadherin shedding in this process proved to be NOX1-dependent since it was inhibited by NOX1 knockout and an EGFR or metalloproteinases inhibitor (436). Likewise, the migratory phenotype that was impaired in NOX1 KO SMCs, was rescued by NOX1 replenishment (436). The study in question was closed out with the demonstration that NOX1-deleted ECs did not migrate, nor coalesced to form tubules on a fibrin gel (437). Moreover, NOX1 deletion in smooth muscle prevented 14–3-3 oxidation and activation of cofilin further establishing a causal role for NOX1 in the migratory response (435). Astonishingly, in aortic ECs from old mice, NOX1 inhibition ameliorated age-impaired migration and angiogenesis (438, 439).

In response to hyperoxia, ECs exhibit heightened migration, which is dampened by NOX4 deletion as well as N-acetylcysteine (NAC) pretreatment (440). ER-localized NOX4 reportedly interacts and stabilizes VEGFR2 in the ER, thus maintaining continuous transport of VEGFR2 to the cell surface via the Golgi apparatus – a phenomenon that enables directed migration of ECs (441). Similarly, hyperoxia-induced ROS and EC migration are, in part, attributed to NOX2 and NOX4 in human lung arterial ECs; and NOX2 & NOX4 demonstrably compensate for each other in this migratory response (440). Angiopoietins, such as Ang-1 and Ang-2, play a key role in the proliferative and angiogenic responses of ECs. Accordingly, adenoviral depletion of NOX2 decreased Ang-1-mediated ROS and angiogenesis by HUVECs (442). Likewise, Ang-1-induced migration and vessel outgrowth were blocked in ECs from p47^phox-deficient mice, to which the authors assigned a critical role of ROS-mediated activation of AKT and p42/44 MAPK (443). Yet another important mechanism of NOX-regulated migration is remodeling of the cytoskeleton, and one of the best-known studies on this process involved NOX’s role in focal adhesion and the linkage of Poldip2 and RhoGEF to cytoskeletal remodeling (162, 309). In that regard, in SMCs following PDGF stimulation, Poldip2 elevated NOX4 activity, activated RhoA, and reinforced focal adhesions and stress fiber formation triggering migration (162). A number of excellent reviews on the topic over the past decade may be additionally consulted for an extensive exploration on the subject (94, 444, 445).

3.3. Angiogenesis

Angiogenesis is a finely orchestrated physiological process of capillary formation by ECs and in that regard, NOXs 1, 2, 4 and 5 and their compartmentalization and control of unique signaling pathways dominates this discussion (371). For instance, NOX1’s localization to caveolae appears to position it for activation of PTP1B in lipid rafts and augmentation of VEGFR2 signaling and angiogenesis (446). On a larger scale, angiogenic stimulation of primary mouse and human ECs increase NOX1 expression and activity (437); and studies report impaired angiogenesis in mice lacking NOX1, but not NOX2 or NOX4 (437). As well, NOX1 silencing inhibited PPARα-mediated activation of NFκB that sequentially decreased EC migration and tube-like structure formation. Likewise, administration of NOX1/4 inhibitor GKT136901 reduced angiogenesis and tumor growth in vivo in a PPARα-dependent manner from which a role for NOX1, and not NOX4, was interpreted in new vessel formation (437). As alluded to above, angiogenesis might be promoted in microvascular ECs by low dose glucocorticoids, shown to be mediated by p22^phox, NOX2 and NOX4 - similar to the situation observed with microvascular EC proliferation (84). Furthermore, vascular sprouting of vena cava explants induced by urotensin-II (U-II) was significantly reduced in NOX2^−/− mice as compared to controls (447); and, in agreement with this, NOX2^−/− mice displayed reduced U-II-induced invasion of new vessels into matrigel plugs (447). On a subcellular level, NOX2 and NOX4’s differential distribution in plasma membranes, ER, mitochondria and nuclei are almost certain to explain their divergent signaling roles. For example, NOX2’s relocation with p47^phox in lamellipodia and membrane ruffles and cysteine oxidation in scaffold proteins at the cell’s leading edge appear to reinforce a causal link between NOX2 oxidase and migration (347, 448).

In still another study, NOX4 promoted angiogenic responses, at least partly, via enhanced activation of receptor tyrosine kinases (449) which would suggest localization of the transduced signal in this scenario to be proximal to the plasma membrane. That is, either NOX4 siRNA or expression of a dominant negative NOX4 blocked VEGF-mediated angiogenic responses by way of suppressed receptor tyrosine kinase and ERK activity, respectively (449). By the same token, NOX4 overexpression enhanced VEGF-mediated angiogenic responses and reduced serum-deprived EC apoptosis (449). Further corroborating those findings, overexpression of NOX4 (NOX4-Tg) in ECs was sufficient to promote proliferation, tube formation and blood flow in mice (450) and NOX4-Tg favored accelerated recovery from hindlimb ischemia, and enhanced aortic capillary sprouting (450). Even more compelling is the notion that EC NOX4 was sufficient to promote angiogenesis in an eNOS-dependent manner, since enhanced sprouting seen in aortas from NOX4-Tg mice was ablated in aortas from NOX4-Tg eNOS^−/− mice. Notwithstanding its suggested role at or near the plasma membrane, a number of studies have pinpointed NOX4 to the perinuclear and nuclear compartments (451). Thus, much more work needs to be done with respect to the dynamics of specific NOX compartmentalization in the cell and the array of signaling pathways a particular NOX can prompt. One might envisage, in that sense, a particular NOX having more than one signaling target depending on the state of the cell and the stimulus at any particular point in time. As a final note, capillary-like structures arising from HMEC1 in response to thrombin can be ablated by NOX5 siRNA (178) and one study infers that NOX5 may mediate this phenotype via calcium-activated potassium channels (452).

For greater depth on the association of NOX with angiogenesis, the reader is referred to some noteworthy reviews (371, 453). Expectedly, this topic will be covered in additional depth in the section on cancer.

3.4. Cell death

Cell death can arise either as a direct cytocidal effect of excessive oxidation and necrosis or subtly regulated by ROS signaling via a programmed cell process such as apoptosis. Namely, major induction of NOX/ROS in response to inflammation or invading pathogens that is not controlled or contained, for example, can cause irreparable damage to lipids, proteins and DNA that directly disrupt the integrity of the cell. Conversely, in a more coordinated manner, NOXs play a role in systematic attenuation of growth or via “measured” processes of autophagy and apoptosis. The following comprises a list of notable examples of NOXs’ role in the latter.

By way of example, mouse models of dextran sulfate sodium (DSS)-induced colitis and azoxymethane/DSS-induced colon cancer showed a protective role of NOXO1 by virtue of its role in natural killer cell differentiation and infiltration, colon epithelium apoptosis, and prevention of tumor formation (454). Interestingly, this function seems to be limited to the canonical NOX1 oxidase since knockout of NOXO1 reduced the generation of O₂^·- in colonic crypts, an effect which could not be restored by p47^phox expression (454). Likewise, in pheochromocytoma cells of the adrenal medulla, nerve growth factor (NGF)-induced neurite outgrowth is suppressed by increased NOX1-derived O₂^·- (455). In distinct fashion, neuronal apoptosis is triggered in mild traumatic brain injury through increased NOX1 - ROS (proxy-detected by lipid peroxidation) and the generation of neurotoxic product SDF1 (456). Moreover, in favor of a broader role of NOX1 in apoptosis, in murine models of sepsis induced either by intraperitoneal administration of LPS or cecal ligation and puncture surgery, cardiomyocyte apoptosis and caspase-3 activation is suppressed in NOX1 knockouts as compared to WT systems (457).

With respect to NOX4, upregulated NOX4 in pheochromocytomas by amyloid beta (Aβ) instigates autophagy and apoptosis (458). In neurons, this process is mediated via glutamate activation of N-methyl-D-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptors, Ca²⁺ influx and H₂O₂ generation by NOX4 (459). Furthermore, whole organ degeneration and neuronal apoptosis in response to cerebral artery occlusion and reperfusion was explored in rats pretreated with either vehicle, or NOX2 inhibitor (NOX2ds-tat aka gp91ds-tat) or NOX4 inhibitor (GKT137831) (460). Perhaps not surprisingly, both treatments mitigated ROS levels, neuronal apoptosis/degeneration and blood brain barrier damage due to cell death (460). Thus, considerable evidence exists for NOXs’ role in apoptosis and homeostasis. In sharp contrast, a recent study illustrated that genetic and pharmacological suppression of NOX4 induced apoptotic clearing of myofibroblasts and reduced fibrosis in dystrophic respiratory and limb muscles (461). Thus, it is not yet clear what mechanisms ultimately govern pro-apoptotic vs. anti-apoptotic signaling in response to NOX across cells and tissue, but it is safe to say that they are cell- and context-specific and likely relate to where in the cell or in the life cycle of the cell the particular NOX is perturbed.

Finally, a positive feedback parallel mechanism could be involved in microglia and other cell types wherein elevated Ca²⁺-induced activation of protein tyrosine kinase PYK2/MEK/ERK signaling amplifies PARP1 activation, TRPM2-mediated Ca²⁺ overload and cell death (462) much like that stimulated by NOX2 and NOX4 in response to Aβ₄₂ (463). As such, a large body of data appear to suggest that NOXs plays a central role in a propagative cycle of cell death arising from brain injury and neuroinflammation as well as in the regulation of tissue fibrosis.

3.5. Differentiation

Differentiation processes are heterogeneous since there is a clear difference between the processes in which: (a) pluripotent stem cells differentiate into one of many cell types; and, for example, how (b) monocytes differentiate towards their terminal phenotype. Hence, between proper differentiation, which is a term usually ascribed to stem cell phenotypes, and terminal differentiation (e.g. macrophages and cardiomyocytes) there is a broad spectrum of processes that are referred to as differentiation with respect to the degree of cellular potency.

Modulated by ROS derived from NOXs, differentiation of cells to their mature states requires a coordinated and synchronized biological and metabolic process. Prime examples of the role for NOXs in differentiation (reviewed in depth in other sections of this review and elsewhere in (405, 464–466) include the following. In the case of M-CSF-treated monocytes, simultaneous deletion of NOX1 and NOX2 but not by single gene deletion of either of the isoforms significantly blocked ROS generation (467). The resulting ROS deficiency prevented activation of ERK and JNK and impaired differentiation and polarization toward M2 macrophages. This impaired phenotypic switch has been corroborated in another study using preventative effects of the antioxidant butylated hydroxyanisole (468). In agreement with this, in NOX1/2 double knockout mice, polarization toward M1 macrophages did not occur. Indeed, wound healing in NOX1/2 double knockout mice was delayed and M2-type macrophages infiltrated to a lesser degree to the wound edge compared with WT mice (467). Moreover, in early stage differentiating ECs derived from murine inducible pluripotent stem cells (iPSCs), the absence of NOX2 reduced the expression of endothelial markers CD31, CD144 and eNOS (469). Furthermore, in mouse neurons, NOX2 expression is elevated and NOX2 deficiency decreases expression of neural differentiation markers (470). Continuing on this theme, expression of NOX2 can vary depending on the stage of differentiation since it has been shown to be low in undifferentiated iPSC, rise upon neuronal cell induction, and vanish during neuronal cell differentiation. Notably in this regard, in iPSCs from CGD patients, the lack of NOX2 caused an abnormal neural induction in vitro, as revealed by a decreased generation of mature neurons (470).

Intriguingly, VSMCs in vitro display markers of differentiation that are causally linked with relatively high NOX4 at early stages of primary culture (466). On the contrary, NOX1 upregulation correlates with dedifferentiation upon passaging (466). Consistent with this, NOX4 co-localized with αSMA-based stress fibers in differentiated SMCs and translocated to focal adhesions in de-differentiated, proliferating SMCs. Importantly, NOX4 siRNA was shown to reduce O₂^·− production in serum-deprived VSMCs and to downregulate SM-alpha actin, SM-MHC, and calponin, as well as SM-alpha actin stress fibers (466). In contrast, NOX1 depletion elicited no change in these differentiation markers (466).

To date, it should not be surprising that NOX4 has been most broadly implicated in cell differentiation including osteoclasts (471), adipocytes (405) and fibroblasts (472, 473). As such, NOX4 contributes to trans-differentiation even in cells in which NOX4 is not highly expressed (465). Indeed, NOX4 has been shown to play a role in the differentiation of pre-adipocytes to adipocytes, with the expression of both NOX1 and NOX4 being increased during insulin-induced differentiation (405). By comparison, silencing of NOX4, but not NOX1, inhibited insulin-induced differentiation and H₂O₂ production and predictably NOX4 overexpression increased MAP kinase phosphatase1 (MKP1), attenuated MAPK signaling and promoted a mature adipocyte response to insulin (405). In stark contrast, some reports reveal the opposite role for NOX4; namely its ability to cause epithelial dedifferentiation and mesenchymal transition (474). On balance, however, the consensus in the field is that NOX4 is pro-differentiating in cell types in which its role has been closely interrogated.

3.6. Regeneration and self-renewal

Tightly regulated and cell-specific NOXs represent one of the major sources of oxidant signaling molecules participating in tissue development, regeneration, and stem cell self-renewal (73, 475–477). A pivotal role for NOXs in the regenerative processes were highlighted in a very recent study in zebrafish, in which duox, nox5 and cyba (p22^phox) nulls were crossed with a transgenic line ubiquitously expressing H₂O₂ reporter HyPer (74). Homozygous duox deletion engendered the greatest effect to reduce H₂O₂ levels and the rate of fin regeneration compared to the other singular mutants (74). Not surprisingly, therefore, the duox:cyba double mutant showed an even greater effect on fin regeneration, suggesting that one or more of Noxs 1, 2 and 4 that co-depend on p22 ^phox could play a role in regeneration (74). Notably, zebrafish lack Nox3.

As for NOX1’s role in mammalian regeneration, the isoform is knowingly most abundant in the colon, where is has been shown to participate in physiological functions including tissue repair following colitis (478). That is, in a murine model of experimental colitis in mice, following dextran sulfate-induced colitis NOX1-deficient mice were not able to repair mucosal wounds compared to control littermates. Mucin-secreting goblet cells in NOX1-deficient mice exhibited decreased cell survival, migration, and terminal differentiation (478). Extending these observations to the colon epithelium, during the repair process NOX1 expression and ROS production in colonic crypts is increased and NOX1 deficiency suppresses ROS and the proliferation and migration of crypt progenitor cells (478). Collectively, the results suggest that NOX1 aids in epithelial recovery after colitis by promoting bioactivity of crypt progenitor cells and thus mucosal wound repair (478).

With respect to NOX2, tissue specific knockouts demonstrated that myeloid, but not EC, NOX2 is required for EC regeneration (479); and after carotid artery injury, vitamin D3-induced NOX2 improved vascular regeneration. Experimentally, mice were treated with daily injections of vitamin D3 for 5 days and carotid injury was induced (479). Regeneration of the EC layer was lost in NOX2 knockout, but not in NOX1 or NOX4 knockout mice. To the point of cellular contribution to this regeneration, vitamin D in a NOX2-dependent manner activated MAPKs and induced vascular cytokine stem cell-derived factor-1 (SDF-1) which, in turn, mobilized pro-angiogenic myeloid cells to the site of injury and promoted endothelial growth (479). Accordingly, SDF1 induction was lost after global deletion of NOX2 (479); and importantly, EC-specific knockout of NOX2 revealed that EC NOX2 had no bearing on this recovery and affirmed the role of myeloid cells in the regenerative response.

When it comes to NOX4, its role in regeneration has largely been investigated in skeleto-muscular system, and the liver. Indeed, NOX4 is implicated in some very basic processes of myoblast fusion which are essential for its role as a modulator of skeletal muscle development and regeneration (480). The cross-sectional area of myofibers in regenerating muscle of NOX4 ^-/- mice was lower than that in WT mice. Similarly, myotubes differentiated from NOX4^-/- primary myoblasts show significantly lower diameters and fusion index values than those from WT myoblasts (480). Taken together, these findings are consistent with NOX4 contributing to myoblast fusion, and putatively through the regulation of myomaker Tmem8c expression (480). Regeneration and growth of skeletal muscle are stimulated by mitochondrial ROS (mtROS) and ROS generated by NOX4 (480). Similarly, NOX4 has been implicated in bone regeneration, i.e. osteoblast development, stromal cell self-renewal, and proliferation. As such, ROS production in osteoblasts from global NOX4^-/- mice as well as conditional mouse NOX4 knockout in the limb bud mesenchyme was significantly lower compared to control mice (73). At 3-weeks old, mice lacking NOX4 showed significantly lower bone volume, bone mineral density and trabecular number compared with controls independently of gender (73). Thus, these data suggest that in bone and osteoblastic cells, NOX4 regulates osteoblast differentiation, proliferation, and maturation (73). Finally on the level of pathophysiology and clinical relevance, the skeletal muscle of dystrophic mice as well as of Duchenne muscular dystrophy patients have higher levels of expression of NOX4 that seem to localize to interstitial cells found between muscle fibers (461). Targeting of NOX4, both pharmacologically or by genetic manipulation significantly decreased the level of fibrosis in respiratory and limb muscles of dystrophic subjects. That is, the lack of NOX4 rejuvenated muscle regeneration by decreasing the number of collagen-generating myofibroblasts and restoring muscle-specific stem cells to their physiological niche suggesting NOX4 as a therapeutic target to promote the beneficial tissue remodeling in Duchenne muscular dystrophy and, potentially, other muscular dystrophies and muscle pathologies (461).

Along similar lines, but in this case potentially lethal, NOX4 was shown to regulate proliferation and stemness of glioblastoma stem cells which are responsible for spurring gliobastoma (481). In fact, NOX4 was found to be more highly expressed in glioblastoma stem cells than in their more differentiated neural counterparts. To elaborate a bit further, NOX4 was induced by TGFβ and its overexpression drove further expression of glioblastoma stem cell markers (481). Additionally, these results point to the possible utility of NOX4 as a valuable stratification marker for this pernicious and deadly disease.

On the topic of liver regeneration however, data from the Fabregat group have convincingly shown that both global NOX4^-/- and NOX4 hepatic cell-specific null mice subjected to 2/3 partial hepatectomy possess a capacity for quicker recovery of liver-to-body weight ratio and an increased survival rate compared to WT mice (81). The liver is indeed a unique organ in the reparative and regenerative response whereby all cell types must proliferate to re-establish functional liver mass (81). Further to that theory, the regenerative hepatocellular fat accumulation and parenchymal reorganization recovered faster in NOX4^-/- livers. Indeed, hepatocyte proliferation was accelerated and increased in NOX4-deleted mice concomitant with increased Myc and cyclin expression (81). Consistent with this last finding and supportive of physiological relevance, the proliferative activity of non-transformed human hepatocytes in vitro is increased in cells lacking NOX4 consistent with increased cyclin D1 and cell cycle progression, while in vivo NOX4 expression is downregulated after partial hepatectomy and pathological proliferative conditions such as diethylnitrosamine-induced hepatocarcinogenesis (482).

4. NADPH OXIDASES IN HEALTH AND DISEASE

4.1. Stem cells

ROS signaling from NOXs mediates stem cell renewal vs. differentiation via a host of redox-sensitive proteins including FoxOs, p38 MAPK, Nrf2, HIF1, ATM and p53 (464, 483–487). For example, HIF1 has been described as a master regulator in the maintenance of cell cycle stasis and stemness and p38 MAPK is but one among many kinases that subserves a pivotal role in preserving a pool of stem cells (464, 484) and where NOX-ROS signaling logically acts and is promoted. In general, ROS levels are observed at low levels in stem cells undergoing self-renewal whereas they increase in differentiated successors of these cells (488). On the other hand, excessive ROS can cause irreparable DNA damage, cell cycle arrest, senescence and impaired reprogramming (489). This rather non-descript, non-delineated and generalized association of the relative level of ROS involved in stemness vs. differentiation is unfortunate but will assuredly become more defined by necessarily rigorous future studies that quantify levels of NOX activity and ROS generation and interpolate them on a spectrum from stem cell to highly differentiated cell types.

Regarding their experimental value and therapeutic potential, the best studied of these cell types are iPSCs that are redirected from adult somatic cells to pluripotency through forced expression of key genes and factors to an embryonic stem (ESCs) cell-like state (490). As such, iPSCs and ESCs share many of the same cardinal features required for preserving their self-renewal and pluripotency including low steady-state ROS (491) maintained by FoxO-targeted genes like SOD2 and catalase (464, 492, 493). ESCs reportedly require low levels of NOX-derived ROS to maintain its homeostasis and self-renewal potency and appear to have a “fail-safe” mechanism when faced with elevations in ROS. Accordingly, in response to exogenous H₂O₂, ESCs experience a shortened G₁ cell cycle phase wherein they are homeostatic, relying metabolically on glycolysis and pentose phosphate pathways (494). Moreover, in agreement with a fine-tuned homeostasis, p38α kinase acts in a negative feedback loop in ESCs to suppress NOX2 and O₂^·− and H₂O₂ in ESCs (495). On the other hand, ESCs display increased NOX2 protein and ROS levels that trigger differentiation (464, 495). To that point, but somewhat inferential in interpretation, NOX2 seems to be transiently activated and triggered early in murine ESC differentiation consistent with p67^phox expression being elevated in 2–3-day- vs. 11–12-day old embryoid bodies (464, 477).

In perinatal amniotic stem cells, NOX4 expression in the nucleus varies depending on the cell donor and transcription factors which participate in stemness: e.g., Oct4, SSEA-4, and SOX2 (464, 489). During transcriptional regulation of stem cell differentiation potential, nuclear NOX4 appears to couple with the redox-sensitive transcription factors Nrf2 and NFκB (464, 489). Indeed, NOX4 was shown to predominantly localize in promyelocytic leukemia protein nuclear bodies (PML-NB) where it purportedly interacts with prelamin A, consistent with stem cell fate regulation through transcription factor recruitment and DNA damage modulation (464, 489, 496, 497). Additionally, NOX4 post-translational modification (i.e., sumoylation) is involved in its sequestration, proteasomal degradation and tempered ROS generation which curb DNA damage and premature aging (489).

Retrogression of somatic stem cells to stemness, aka nuclear reprograming, requires ROS signaling and at least one study examining iPSCs attempted to test the role of multiple NOXs in this process (464, 498). In this context, examining mRNA for various NOXs and their subunits, an increase in mRNA limited to DUOXs 1 & 2 and p67^phox was demonstrated. Despite these findings, the genetic knockdown of p22^phox compromised reprogramming efficiency of iPSCs, from which, on its own, one might invoke the role of any one or more of NOXs 1 through 4. However, no apparent attempt was made to test causality of any one NOX in this report (498). Thus, in that regard, the report leaves unanswered questions as to which NOX oxidase or NOX oxidases definitively participate(s). Later in the report, the authors do show that NOX2 along with p22^phox and p67^phox are upregulated concomitantly at the early stage of reprogramming and by association deduce that NOX2 oxidase is the culprit. However, to our reading, they do not attempt to knockdown NOX2 or p67^phox in this or other subsequent reports. On the other hand, in an apparent effort to examine the effect of excess ROS on reprogramming, the authors proceed to show that exogenous overexpression using NOX2 plasmids and increased NOX2 mRNA (no ROS measurement) along with antioxidant depletion cause a marked decrease in stem cell population (498). Thus, by inference and coincidence, one might conclude that modulation of NOX2 oxidase to produce low vs. high levels of ROS is pivotal to nuclear reprogramming (stemness) vs. differentiation. More rigorous testing employing NOX isoformchallenge the notion that the isotype is involved in loss-of-function and gain-of-function is in view of this shortcoming warranted before clear conclusions can be drawn.

In other studies on neural cells, in addition to a declared role of low constitutive NOX-derived ROS generation in progenitor cell sustenance and replenishment, differentiation of neuronal cells appears NOX2- and NOX4-dependent (464, 499–503). In line with this observation, neurogenesis in response to AngII and injury in mice has been shown to involve NOX4 (421, 504).

Overall, the findings bring into focus a fine-tuning of reprogramming by NOX wherein cells strike a balance between ROS and antioxidant activity that determines stemness vs. differentiation (498). Importantly, each kind of stem cell possesses a distinct amalgamation of NOX subtypes suggesting that the NOXs in unison, and in combination, play varied roles in the differentiation of various tissue types (464).

In summary, NOX research in the context of stem cells is clearly an under-explored area. Extensive studies are needed so that we can arrive at the juncture of meta-analyzing robust data on the topic, which would help in defining the future directions of the field. Understanding the role of NOXs in processes of differentiation of stem cells would unveil enormous potential of the modulation of their programming by NOX inhibitors. In this respect, one can envisage new therapeutic strategies given the growing field of stem cell therapy.

4.2. Cells of hematopoietic lineage

4.2.1. Erythrocytes

NOX1, NOX2, NOX4, and NOX5 are detected in red blood cells (RBCs) from control subjects and patients suffering from sickle cell anemia (505). NOX2 and perceptibly NOX1 appear more abundant in sickle cell anemia samples. Intriguingly, both a pan-NOX inhibitor as well as NOX2-selective inhibitor NOX2ds-tat potently and to a similar extent inhibit the 3-fold increased ROS generation between sickle cell vs. normal erythrocytes. In contrast, neither a xanthine oxidase inhibitor nor a mitochondrial complex I inhibitor was capable of blocking ROS production (505). Unfortunately, a later study published in 2019 was the only other study we could find that referred loosely to NOXs’ involvement in red blood cells physiology and pathobiology employed broad-spectrum, non-specific inhibitors that compromise multiple oxidoreductases (506). Thus, a host of definitive studies targeting one or more of the multiple NOXs in RBCs under normal and diseased conditions appears ripe for consideration.

4.2.2. Eosinophils and basophils

Like neutrophils and macrophages, human eosinophils display robust expression of the major NOX2 oxidase subunits (507–509). In fact, in an early report, eosinophils were shown to express NOX2, p47^phox, and p67^phox in higher amounts than neutrophils and displayed O₂^·--generating activity that was at least twice that of neutrophils in response to N-formylmethionine-leucyl-phenylalanine (fMLP) and four times as great in response to PMA (510). In another report, leukotriene B4 robustly upregulated phosphorylation p47^phox and NOX2 expression, highlighting the role of NOX2 in granule exocytosis in human eosinophils stimulated with LTB4 (511). In human peripheral blood eosinophils, mRNA for NOX2 and DUOX1 and DUOX2 were detected in appreciable amounts but only NOX2 increased with disease severity in asthmatic patients (512). Basophils, on the other hand, were shown to express NOX2 (along with p22^phox) but not its p47^phox and p67^phox cytosolic subunits and, compatible with these findings, a functional NOX activity could not be detected. Thus, basophils reveal no evidence of a respiratory burst found in other white cells to varying degrees (513, 514).

4.2.3. Monocytes

Aside from their characteristic NOX2/phagosomal activity, in CD11b+ monocytes/macrophages and CD11c⁺ dendritic cells, LPS-induced NOX1 was demonstrated and implicated in collagen-induced arthritis (515). Moreover, both NOX1 and NOX2 appear critical for monocyte to macrophage differentiation, M2-type macrophage polarization, and emergence of tumor-associated macrophages (467, 516). Additionally, NOX5α and NOX5β transcription variants together with NOX5 polypeptides are constitutively expressed in the THP1 monocytic cell line and in human primary CD14⁺monocytes and potentially play a role in atherosclerosis (517).

4.2.4. Macrophages

Found in virtually every organ, macrophages function as a part of both innate and adaptive immunity due to their role in antigen presentation and ability to surveil for and digest potential pathogens. Consequently, macrophages contribute to tissue inflammation and related tissue phenotypic changes and the progression of these changes depend on the balance of M1/M2 macrophages regulated by NOXs (518–520). In addition to the classical NOX2 involved in phagocytosis, colonic and peritoneal macrophages express NOX1 (521) and NOX1-selective deletion in macrophages abrogates hepatic inflammation and tumorigenesis (522). Interestingly, NOX1/NOX2 double knockout, but not NOX1 nor NOX2 single knockouts, induced monocyte to macrophage differentiation and diminished M2 polarization. Mechanistically speaking, NOX1 and NOX2 ablation prevented JNK and ERK activation that is responsible for this differentiation response (467). Moreover, loss of both NOX1 and NOX2 did not affect the M1 population of macrophages which retained their normal function (467). Intriguingly in another study, a CGD model of p47^phox-deficient but not NOX2-deficient, mice displayed skewed differentiation towards an M2 macrophage phenotype (523). Although at first these findings may be confounding, they appear to invoke a role for NOX1 hybrid oxidase (utilizing p47^phox) in the shift between the M2 and M1 phenotype. On the contrary, the absence of NOX2, per se, in still another study, caused attenuation of macrophage STAT1 signaling in favor of STAT3, which led to expression of M2 anti-inflammatory markers (524). In further contrast, macrophage NOX4 deletion resulted in M1 polarization as a result of reduced STAT6 activation, increased NFκB activity and ramped up NOX2 levels (525). Thus, the contribution of any one or more of the NOX1 hybrid oxidase, NOX2 and NOX4 oxidases could positively or negatively be involved in macrophage differentiation.

Lastly, DUOX1 is reportedly expressed in macrophages (526, 527) and appears fundamental in alveolar macrophages to the attenuation of phagocytic activity and cytokine secretion (528). Similar to the effect of NOX4 deficiency, DUOX1^−/− macrophages were biased towards a proinflammatory M1 phenotype characterized by IFNγ, CXCL9, CCL3, and CCL5 secretion and demonstrate anti-tumorigenic properties and promotion of CD8⁺ T cells (528, 529). Collectively, therefore, these studies tentatively ascribe an anti-inflammatory role to both NOX4 and DUOX1.

4.2.5. Chronic granulomatous disease

CGD, an inherited primary immunodeficiency caused by functional impairment of the NOX2 complex in neutrophilic granulocytes and monocytes, is characterized by recurrent and severe infections, dysregulated inflammation, and autoimmunity (530). Loss-of-function mutations in any of the five major structural subunits of the NOX2 complex in neutrophilic granulocytes and monocytes result in defective ROS generation. NOX2 CYBB gene on the X chromosome accounts for approximately two-thirds of cases of CGD. Autosomal recessive mutations in NCF1 (p47^phox) account for about 20–25% of cases, and mutations in CYBA (p22^phox) and NCF2 (p67^phox) each account for about 5% of cases (530–536). In addition, there has been one reported case, to our knowledge, of an NCF4 (p40^phox) mutation resulting in CGD (537). For a much more in-depth discussion of CGD, the reader is referred to Chapter 4.16 on NOX polymorphisms with reference to Table 3.

Table 3.

Single Nucleotide Variations and Single Nucleotide Polymorphisms for NOX Isoforms and their Regulatory Subunits.

Gene	RS number	Amino Acid	Nuc. Change	Disease	Supporting Publications
NOX1 [-]	rs142303829	p.R241C	c.721C>T	Inflammatory Bowel Disease (IBD), very early onset	(1167)
	rs34688635	p.D360N	c.1078G>A	IBD, very early-onset	(777)
NOX2 [+]	rs137854588	p.R73Term	c.217C>T	CGD	(105)
	rs137854591	p.R91Term	c.271C>T	CGD	(1172)
	rs139670417	p.R229H	c.686G>A	Crohn's Disease	(1169)
	rs137854585	p.P415H	c.1244C>A	CGD	(41)
	rs13306300	p.G472S	c.1414G>A	CGD	(1170, 1171)
	rs137854593	p.D500G	c.1499A>G	CGD	(1173, 1174)
	rs151344490	p.L505R	c.1514T>G	CGD	(1175)
NOX3 [-]	rs757425327	p.R74Q	c.221G>A	Autism spectrum disorder	(1176, 1177)
	No rs # listed	p.A318=	c.954G>A	Developmental disorder	(1177, 1178)
NOX4 [-]	rs34495256	p.(E3dup)	c.7_9dupGAG	Functional Phenotype: Reduced Protein Expression	(1181)
	rs1836882	n/a	c.-8021A>G *mut. in promoter at −8021	*non-coding exception for NOX4 Increased ROS in peripheral blood mononuclear cell	(1182)
NOX5 [+]	no rs # listed rs144275394	p.H55R p.V446M	c.164A>G c.1336G>A	Congenital Heart Disease? Development Disorder	(1183) (1184)
p22phox [-]	rs4673	p.H72Y	c.C214T	Coronary Artery Disease (CAD), reduced susceptibility	(1186, 1188, 1330)
			aka c.C242T	Lower Risk of Metabolic Syndrome	(1331)
				Cerebrovascular Disease & Stroke, enhanced risk	(1205, 1206)
				Atherosclerotic Stroke	(1196–1199, 1212)
				CGD	(126)
				Cardiovascular events	(1192, 1195)
				ESKD, Increased Hypertension Risk	(1200)
				Diabetes	(1201–1204, 1212)
				Diabetic Nephropathy	(822)
				Myocardial Infarction	(1191)
				Ulcerative Colitis	(1332)
	rs1049255	3'UTR Δ	c.A640G	Atherosclerosis, subclinical	(1197)
			*complementary, reverse strand	Coronary artery disease	(1195)
				Coronary artery disease, younger patients	(1187)
				Non-Hodgkin Lymphoma; lower p22 levels & NOX activity	(1216)
	rs9932581	n/a	c.-932A>G	Diabetic Kidney disease	(1201)
			promoter substitution −932 from start codon	Diabetic Atherosclerosis	(1217)
				Organ Rejection	(1208)
				Hypertension & Blood Pressure	(1194)
				CAD	(1333)
p47phox [+]	rs139225348	p.G83R	c.247G>A	Dermatophytosis (Fungal Infection); Functional Defect in neutrophils	(1219, 1220)
				Crohn’s disease, pediatric IB	(1169)
	rs201802880	p.R90H	c.269G>A	Systemic Lupus Erythematosus	(1221)
				Sjögren's Syndrome	(1221)
				Rheumatoid Arthritis	(1221)
				CGD	(1222)
				Spontaneous Abortion	(1334)
p67phox [-]	rs17849501	p.A202A	c.606G>A	Systemic Lupus Erythematosus	(1223)
	rs13306581	p.T279M	c.836C>T	CGD	(1174)
	rs35937854	p.V297A	c.890T>C	Systemic Lupus Erythematosus	(1223)
NOXO1 [-]	rs200352693	p.(His198Profs*38)	c.593delA	Cardiovascular Traits	(1224)
NOXA1 [+]	rs564363346	p.V318M	c.952G>A	Autism spectrum disorder	(1176)
	rs953881310	p.S373L	c.1118C>T	Developmental disorder	(1178)
DUOX1 [+]	rs371937236	p.R50Q	c.149G>A	Hypothyroidism	(1225)
	rs747868839	p.R746W	c.2236C>T	Hypothyroidism	(1225)
	rs145668427	p.R925W	c.2773C>T	Central serous chorioretinopathy, in females	(1226)
DUOX2 [-]	rs199957468	p.L171P	c.512T>C	Inflammatory Bowel Disease, increased risk of	(79)
	rs201197899	p.T423I	c.1268C>T	Hypothyroidism	(79), (1225)
	rs753591292	p.Y1450H	c.4348T>C	Hypothyroidism	(1227, 1228)
				IBD, increased risk; disturbed microbiota-immune homeostasis	(79)
	rs376623263	p.E1469K	c.4405G>A	Hypothyroidism	(1225, 1229–1231) (79, 1227)
DUOXA1[-]	rs149960164	p.T213M	c.638C>T	Hypothyroidism	(1232)
	rs16977686	p.S313G	c.937A>G	Hypothyroidism	(1232)
DUOXA2[+]	rs200789957	p.L204P	c.611T>C	Hypothyroidism	(1233)
				IBD, increased risk	(79)
	rs4774518	p.Y246X	c.738C>G	Hypothyroidism	(1234) (1235)

Initial Source Database - HGMD (Qiagen). SNVs and SNPs in the exonic regions of the gene are listed with minor exceptions.

Notes: Primary entries in bold are SNPs in one or more of the population cohorts; RS numbers are SNV/SNP identifiers that, when available, are searchable in NIH dbSNP database and in a few cases the gnomAD database.

The [+] or [-] symbol following the gene name indicates which strand the gene is on and whether it is forward- or reverse-expressed. Thus, some mutations are indicated as they would appear on the [+] strand even if they are represented by the more relevant complementary mutation on the [-] strand for that gene.

In the case wherein there were few or no SNPs that could be found for a particular component, SNVs are listed and indicated by shaded rs numbers. MAFs obtained from a combination of dbSNP and gnoMAD databases. Supportive publications curated by HGMD and via PubMed search. HGMD listings that did not have dbSNP or gnomAD links were not deemed SNPs.

This deep inquiry into SNV/SNPs was conducted in September 2023.

4.3. Immune system and immunity

4.3.1. Innate immunity and inflammation

In a broad sense, and as a first line of defense, immune system cells recognize pathogens by the so-called pattern recognition receptors (PRRs) in two ways: directly, by recognizing pathogen-associated molecular patterns (PAMPs), and, indirectly, by detection of the host’s own components released during cell damage or death manifested as damage-associated molecular patterns (DAMPs) or “alarmins”. Toll-like receptors (TLRs) are a prime example of PRRs and core mediators of inflammatory pathways that modulate the immune response to pathogens. They are expressed on cell membranes of macrophages, natural killer (NK) cells, dendritic cells (DC), T- and B cells and PRR receptor-mediated ROS is coupled chiefly with the NOX2 oxidase (538). TLRs recruit hallmark adaptor signaling molecules like Myd88, TRIF and TRAM whose function it is to engage the inflammasome and activate, among other transcription factors, NFκB which promotes the expression of pro-inflammatory cytokines and chemokines (Figure 6) (539, 540).

Figure 6. NOX2 regulation of NFκB is at the intersection between inflammatory and autoimmune disease.

In macrophages, NOX2-derived oxidants prevent thioredoxin-1 (TRX1) reduction by thioredoxin reductase (TR-1) keeping TRX1 from entering the nucleus, thereby preventing NFκB-mediated pro-inflammatory cytokine production (TNFα, IL1 and IL6) and inflammatory disease (A). In contrast, regulatory T-cells (Tregs) play an essential role in protection against autoimmunity by suppressing T-cell responses. In NOX2-deficient Tregs, TRX1 levels accumulate in the nucleus and augment NFκB-mediated transcription of immunosuppressive cytokines (CD25, CD39 and CD73), which in turn, cause an increase in Treg-suppressive activity and the infiltration/proliferation of effector T cells, thus, inhibiting a hyperimmune/autoimmune response, and transplant rejection (B).

NOX2 is crucial not only for microbial killing but also for suppressing the immune response by macrophages. This effect is corroborated in studies in Cybb-deficient CGD mice wherein NOX2 suppresses macrophage-centered inflammatory responses (541). With respect to process, NOX2 tempers LPS-induced lung inflammation in mice through ROS-mediated suppression of thioredoxin (TRX) reduction, thwarting its nuclear translocation and its ability to induce NFκB promoter activity (542) (Figure 6A). This, in turn, reduces gene expression of the pro-inflammatory cytokines (e.g., TNFα, IL1, IL6), chemokines (MCP1, IL8, MIP1a), adhesion molecules (ICAM1, VCAM1), enzymes (COX2, iNOS), and immune receptors (MHC, IL2 receptor, IFNγ receptor). When NOX2 activity is compromised as in CGD, TRX levels mount in the nucleus and augment inflammatory disease and a hyperimmune response (543).

In contrast to NOX2’s mechanism of action in innate immunity, a direct interaction of NOX4 C-terminus with TLR4 can trigger an NFκB-activated immune response in a variety of cells including kidney epithelial HEK293T cells, U937 monocytic cells and ECs (544). The TLR4 Toll/IL-1R region physically interacts with the C-terminus of NOX4, which is critical for LPS-induced NFκB activation and cytokine and adhesion molecule expression (544). Likewise, TLR5 and DUOX2 interactions provoked by bacterial flagellin appear to generate ROS and IL8 expression in airways. The Toll/IL-1R region of TLR5 also activates DUOX2 in a Ca²⁺-dependent manner (545–547). By comparison, NOX1 was reportedly activated by NOXO1 and NOXA1 in human colonic T84 cells in response to recombinant flagellin from Salmonella enteritidis (548). Besides TLR and related receptors, inflammasomes are an essential processing unit contributing to PAMP and DAMP sensing and subsequent signaling. Findings are consistent with NOX2 activating the NLRP1 inflammasome (549, 550) and with NOX4 activating the NLRP3 inflammasome (551–553).

4.3.2. Antigen presenting cells

NOX2 is essential for promotion of antigen cross-presentation by DCs to CD8⁺ T lymphocytes (554). A special class of phagocytic cells, DCs serve to process antigens to a correct size for their loading onto major histocompatibility complex (MHC) class II in mixed ER-phagosome compartments. In this scenario, NOX2 is essential for antigen presentation to CD8⁺ cells, it produces low levels of sustained ROS which concomitantly causes alkalinization of the phagosomal milieu (554). Additionally, NOX2 activity supported by Rac2 rather than Rac1 causes endosomal membrane lipid peroxidation that enables endosomal antigen release into the cytosol for cross-presentation (555).

4.3.3. Adaptive immunity

The adaptive immune response comprises the expansion of cell-mediated and humoral immunity in which three major subtypes of T cells are involved: CD8⁺ (cytotoxic or killer), CD4⁺ (helper), and regulatory T cells (Tregs). With respect to their main function, CD8⁺ cells directly attack and kill bacteria and viruses and surveil cancer cells. Additionally, through the secretion of cytokines, CD8⁺ cells recruit other cell types that participate in the immune response. To that extent, CD8⁺ T cells are critical to elimination of intracellular pathogens for which NOX2 plays a major role. For instance, in a model of Trypanosoma cruzi infection, NOX2 deletion augments the susceptibility of mice to infection (556). Moreover, p47^phox-/- mouse CD8⁺ T cell count was lower even under non-stimulated conditions and did not proliferate in response to T. cruzi infection (557). Yet, in a model of murine lymphocytic choriomenigitis virus infection, CD8⁺ T cells from p47^phox-/- mice displayed improved survival with reduced choriomenigitis viral titers (558). In addition, with respect to influenza, NOX2^-/- mice show reduced inflammation and viral titers in vivo (559). Clearly, in this way, NOX2’s role in adaptive immunity appears pathogen specific.

Alternatively, the roles of CD4⁺ cells in adaptive immunity include memory B cell and cytotoxic T cell activation. Depending on their cytokine-secreting profile, CD4⁺ T cells are further divided into subclassifications of Th1, Th2, Th17 and Tregs. As for the latter, Tregs can be viewed as a separate T cell subtype since their primary role is distinct and important for providing a mechanism of self-tolerance and prevention of autoimmune responses. Importantly, T cell activation depends on three signals: [1] antigenic peptide: MHC complex binding; [2] the interaction between costimulatory molecules on T cells and antigen presenting cells; and [3] innate immune-derived ROS and proinflammatory cytokine production. On top of that, T cells express a functional, phagocyte-like NOX2 involved in the immune response and promote AngII-induced hypertension and cardiac remodeling (560, 561). Interestingly, NOX2-deficient CD4⁺ cells display enhanced and prolonged ERK activation, an increase in T helper type 1 cytokine secretion and a preferential Th1 response (562). Furthermore, CD8⁺ T cells from these mice display augmented IFNγ secretion (typical of Th1 response) and lesser Th2 cytokines IL4 and IL5 secretion (562). On the contrary, in asthmatic mice, deletion of NOX2 causes enhanced IL2 and IL4 secretion from CD4⁺ cells in response to anti-CD3 and anti-CD28 stimulation, indicative of a Th2 response (563). And, in this context, NOX2 inhibitor gp91ds-tat was shown to be capable of instigating proliferation of CD4⁺ T cells that produced greater IL-4 (563). Alternatively, Tse and coworkers observed a decrease in IFNγ and IL2 secretion and sharply higher IL17 and TGFβ (typical of the Th17 response) secretion in p47^phox-/- CD4⁺ T cells (564). Notably, the authors attributed these discrepancies to use of the non-obese diabetic mouse strain known for its autoimmune phenotype (564). Overall, it would appear that a Th1/Th17 response is primarily elicited by NOX2 insufficiency, with increased IFNγ, IL17 and associated transcription factors T-bet and RORγt (565–567).

In Tregs, recent studies indicate that NOX2 suppresses TRX1 translocation and downstream expression of pro-inflammatory genes like CD25, CD39, and CD73 (560) (Figure 6B). When NOX2 is present, NFκB’s activation and ability to promote expression of its target hallmark genes is disrupted (543, 568). Conversely, when NOX2 is inhibited, Tregs unleash their ability to immunosuppress through increased TRX1 translocation to the nucleus, NFκB’s activation, higher expression of CD25, CD73 and CD39, and decreased cytotoxic CD4⁺ T-cell activity, which averts autoimmune disease and tissue rejection (568) (Figure 6B). In line with this, the Shah group intriguingly demonstrated that NOX2 deletion gives way to a more suppressive Treg phenotype which attenuates CD4⁺ T-cell function, decreasing autoimmune responses and fibrosis (560).

Furthermore, in autoimmune diseases, when suppression of the immune response by regulatory T cells is defective, there is an expansion of proinflammatory T cell populations, such as Th1 and Th17, and reduction in the number of Tregs (569). Numerous murine models (570, 571) and patient cohorts (572, 573) exhibit an association between genetic NOX2 subunit deletion, leading to ROS insufficiency, and autoimmune disorders. Complicating matters further, insufficient NOX2-derived ROS increase the number of reduced thiol groups on T cells, and lower the T cell activity and proliferation threshold, as well as provoke T cell-dependent autoimmune responses (570, 574). As far as treatment of autoimmune diseases, Treg mouse models (575) are expected to provide greater insight into the mechanism of Treg NOX2 inhibition which would not only attenuate the immune response but, in so doing, facilitate would healing (576) and organ transplant tolerance (577, 578) (Figure 6B).

Cytokine withdrawal-induced apoptosis, which mimics various characteristics of T cell death in vivo, demonstrates that NOX2 is indispensable for antigen-induced T cell apoptosis and resolution of an immune response (579). Indeed, NOX2^-/- T cells show significantly higher resistance to apoptosis following cytokine withdrawal compared to WT T cells (579).

4.3.4. Humoral immunity

B cell NOX2 bolsters the humoral adaptive immune response. Interestingly, reports show that NOX2 is primarily responsible for rapid ROS generation and antimicrobial activity in B cells, while later stage ROS production seems to be NOX2-independent (305). NOX2 knockout mice do not produce ROS upon B cell receptor (BCR) activation, even though B cell receptor’s proximal and downstream signaling remains intact (580, 581). However, NOX2^-/- B cells reenter the cell cycle and proliferate more readily following B cell receptor stimulation (402, 581). Inferentially, NOX2 appears to satisfy a negative feedback role on B cell receptor-mediated proliferation of B cells.

4.4. Skeletomuscular system

4.4.1. Skeletal muscle cells

NOX1, 2, 4 and the DUOXs have all been reported on in skeletal muscle and ATP, glucose and fatty acids play a generalized role in NOX expression and activity (582). As for NOX1, little is known other than its capacity to be upregulated by myostatin (583). Importantly, NOX expression distributes in a fiber-type-specific manner in response to exertion: more specifically, NOX2 increases in type-2a and NOX4 in type-1 fibers (584).

In skeletal muscle cells, the greatest focus appears related to the NOX2’s role. That is, NOX2 and p22^phox are found along with p47^phox, p67^phox and p40^phox in membrane-enriched fractions, co-localized with membrane proteins of the sarcolemma and transverse tubules (82, 585–587). Interestingly, reports indicate that NOX2 is not present in the sarcoplasmic reticulum (585, 586, 588), but rather in transverse tubules and linked to increased frequency of muscle contraction (585). Mechanistically, activation of the ryanodine R1 receptor by NOX2/ROS causes Ca²⁺ release from the sarcoplasmic reticulum in this process (585); and mitochondrial ROS are likely to weigh heavily in this response either directly as a byproduct of electron transport or indirectly via NOXs (589). That is, ROS leakage from mitochondria under relatively normal conditions or an amplification of ROS by a feed-forward activation by NOX-ROS ensures a participatory role for the organelle (590). Likewise, localization of p47^phox to the sarcolemma and transverse tubules in limb, diaphragm and flexor digitorum brevis muscle assembly with NOX2 (activation) and its known association with cytoskeletal proteins and contractile elements is consistent with conserved NOX2 signaling to actionable Ca²⁺ release and contraction across skeletal muscle systems (82, 587, 591–594). Along similar lines, a pair of studies illustrated a role for NOX2 in protection against ethanol-mediated growth plate disruption and obesity-related impairment in skeletal muscle adaptation (595, 596). Taken together, numerous findings point to NOX2-derived ROS as central to the normal functioning of the contractile machinery in skeletal muscle.

Distinctly from NOX2, NOX4 might serve as an O₂ sensor in skeletal muscle sarcoplasmic reticulum, transducing a signal from molecular oxygen elevations to vicinal ryanodine R1 receptor, effecting intracellular calcium release and contraction (327). On the contrary, one study at least suggests a deleterious role for NOX2 and 4 in skeletal muscle function (597) and another hints at their role in sarcopenia (598).

In the context of physical exercise, myotubes exposed to rises in IGF-I exhibit increased ROS and hypertrophy that are inhibited by NOX4 knockdown (599). The findings of that study also suggested that NOX4-derived ROS sensitize the response to IGF1, in addition to providing negative feedback for the IGF-I signaling (599). Also contradictory to a biological role, Rac1 activity is increased in skeletal muscle of Duchenne’s muscular dystrophy mice (mdx) and seemingly contributes to the pathophysiology of mdx via its effects on NOX2 and mTOR (600, 601). In C2C12 myotubes from dystrophic mice, knockdown of the deacetylase SIRT1 increases NOX4 protein levels, suggesting that SIRT1 is a negative regulator of NOX4 expression and a means by which dietary antioxidants may exert their protective action against excessive NOX-derived ROS in the disease (602). DUOXs1 & 2 have also been found in C2C12 muscle cells although little is reported on their biological or pathophysiological importance except for one study, to our knowledge, of their potential role in myogenesis (603). For further detail on the regulation of NOXs in skeletal muscle, we direct your attention to some of the most recent reviews we could find on the topic (582, 604).

4.4.2. The skeletal system

4.4.2.1. Osteoclasts and Osteoblasts

ROS were initially identified at the osteoclast-bone interface, where they were credited with a key role in osteoclast differentiation and bone resorption/degradation (605, 606) wherein NOX1, 2 and 4 have been implicated (607–609). Adherent to the bone surface, osteoclasts, large multinucleated cells that originate from monocyte lineage, participate in resorption of bone components in a dynamic ongoing process of bone remodeling. Delving into its possible causality, a study in NOX1 knockout mice indicated that NOX1 is not involved in the differentiation of osteoclasts (471). On the other hand, in one study NOX1 deletion blocked RANKL-mediated ROS production pivotal to osteoclast differentiation (607). As further evidence of a role for NOX1, reduction in NOX1 translation and activation is ascribed to reduced osteoclastogenesis of murine macrophage-like (RAW264.7) cells in response to the phytoestrogen genistein (610, 611). Furthermore, silencing of NOX1 in bone marrow macrophages is reported to significantly decrease ROS and inhibit osteoclast differentiation (607). In aggregate, these data suggest that osteoclast differentiation and activity depend on NOX1 in a condition- and cell type-specific fashion. Incidentally, NOX1, NOXA1, and NOXO1 expression are increased in the process of endochondral ossification in murine femurs and have been implicated in chondrocyte maturation and bone matrix formation (608).

Initially, p47^phox and NOX2 were described in rat tibial osteoclasts (612) followed by the discovery of NOX2 in murine osteoclasts (613). NOX2 knockout mice showed reduced NFATc1 expression during RANKL stimulation, thus highlighting the importance of NOX2-derived ROS in osteoclast biology (614). In early experiments, vitamin D-mediated differentiation of osteoclasts was described as H₂O₂-dependent, which is consistent with more recent suggestions of NOX4’s role in osteoclast proliferation and decreased trabecular thickness (471, 615, 616). RANKL and M-CSF induce NOX4 expression during the process of monocyte differentiation into osteoclasts (608). This suggests that even small differences in NOX activity (as the authors argue by way of low-abundance NOX4) can alter the phenotype of bone (608). Furthermore, NOX4 deletion ablates both increased H₂O₂ generation and intracellular calcium signaling in response to RANKL during osteoclastogenesis and is accompanied by impaired activation of NFATc1 and c-Jun (471, 608). Accordingly, NOX4^-/- mice display higher bone density and a reduction in OSCAR, a characteristic receptor signature for osteoclasts (471). In line with this, in a model murine of osteoporosis induced by ovariectomy, both acute genetic knockdown of NOX4 or pharmacological inhibition attenuate trabecular bone loss (608), underscoring the potential for NOX4i as a treatment for osteoporosis. Importantly, human bone obtained from patients with increased osteoclast activity reveals increased NOX4 expression; and, interestingly, women with a single nucleotide polymorphism (SNP) in an intronic region of the NOX4 gene that results in increased NOX4 expression (but curiously not NOX4 activity), display elevated circulating markers of bone turnover and decreased bone density (608). The notion that increased NOX4 expression, and not activity, associates with negative outcomes, while peculiar, might suggest a ROS-independent role for NOX4 in osteoclastogenesis (608). Finally, a handful of studies corroborate NOX4’s role in both osteoblastogenesis and osteoclastogenesis (73, 471, 617, 618). In aggregate, an overwhelming number of reports point to extensive involvement of NOX4 in bone loss, making its targeting a potential therapy or adjunct therapy for the treatment of osteoporosis and other bone disorders (471).

4.4.2.2. Osteoblasts

Bone morphogenetic protein 2 (BMP2) induces an osteoblast phenotype in murine osteoblastic (2T3) cells, coinciding with increased alkaline phosphatase expression and mineralized bone through a NOX-dependent pathway (619). Osteoblasts, terminally differentiated from mesenchymal stem cells, are immature bone cells that give rise to bones in vertebrates by secreting the proteins comprising the organic substance of bone matrix (the osteoid) and controlling its mineralization (620). Although necessary for osteoblast-mediated mineralization of bone, phosphate can increase O₂^·- production and suppress BMP2-induced differentiation of MC3T3-E1 cells into osteoblasts. This effect was shown to be consistent with a role for NOX1 or NOX4 (621). This, at first, may appear contradictory to phosphate’s role in bone formation and the knowledge that phosphate promotes trans-differentiation of VSMCs into osteoblasts in vascular wall calcification by way of ROS signaling (622). The discrepancy, however, might be explained by the amounts of ROS produced and the state of differentiation of the cells in discrete milieu (621) and is fascinating considering the complexity of pathways effected by BMP2 including its ability to promote osteoblast differentiation (623). In the latter, BMP-induced osteoblast differentiation was blocked by both pan-NOX inhibitor DPI and antioxidant NAC, as well as by a dominant-negative NOX4. In contrast, the glucocorticoid dexamethasone induces apoptosis of osteoblasts concomitant with an increase in NOX4 expression (624). Indeed, NOX4 was found to be responsible for the activation of the intrinsic mitochondrial apoptosis pathway and NOX4 suppression prevented dexamethasone-induced osteoblast apoptosis (624). Further in support of NOX4’s role in bone ossification, Ambe and colleagues showed that NOX4 rather than NOX1 is likely involved in chondrocyte maturation and the process of bone matrix formation during endochondral ossification of cartilage in murine femurs (608, 625). Additionally, TGFβ-induced NOX4 and the TGF-NOX4-FAK axis stimulate migration of primary human osteoblasts during fracture healing (609, 626).

4.4.2.3. Osteoarthritis

NOXs’ potential role in osteoarthritis has been postulated in a mouse model of acute arthritis, wherein generalized NOX inhibition partially suppressed proteoglycan synthesis in arthritic joints (627). By blocking peroxynitrite generation, purportedly as a consequence of NOX inhibition, an anti-arthritic strategy was proposed (627). Paralleling these findings, another report asserted that selective NOX2 inhibition by CYR5099 compound and attenuation of neutrophil O₂^·− reduced the severity of arthritis (628). However, challenging this notion, p47^phox knockout mice exhibited more severe antigen-induced arthritis compared to wild type, accompanied by increased MMP levels and inflammation (629, 630). Similarly, NOX2^-/- mice develop arthritis spontaneously, an age-related effect that sees 60% of mice being arthritic at 15 to 18 weeks (566); and the animals exhibit decreased bone mineral density, increased inflammatory markers, RANKL, and IgG against collagen in knee joints (566, 609). On balance, these results support baseline NOX2 activity as required for maintenance of healthy joints and are corroborated by in vitro studies wherein knockdown of NOX2 blocked quinolone-induced apoptosis of potentially pro-inflammatory chondrocytes (609, 631–633). Contrary to these findings, a study in synovial fibroblasts demonstrated that pro-inflammatory VEGF expression was blocked by broad-spectrum NOX inhibitors (634). Similarly, these agents blocked the production of inflammatory markers (TNFα, IL1β, MMP3, MMP13, and VEGF) in response to stimulation of synoviocytes with advanced oxidation protein products (635).

Furthermore, in a mouse model of post-traumatic osteoarthritis induced by anterior cruciate ligament rupture, NOX4 activity was acutely decreased within 24 h of injury, followed by a sustained low-level rise in activity (636) and inhibition of NOX1/NOX4 with GKT137831 protected against structural changes in the knee joint after injury (609, 636). In chondrocytes, IL1β and post-traumatic oxidative stress upregulate NOX2 and NOX4 (637) and pan-NOX inhibition with APX-115 significantly attenuates IL1β-induced ROS and protected mice against osteoarthritis by modulating the oxidative stress, MMP13 and Adamts5 expression (637). On a more mechanistic level, pan-NOX inhibition and combined NOX2 and NOX4 inhibition suppressed Rac1, p38- and JNK MAPK in addition to restoring mitochondrial oxidative phosphorylation in IL1β-treated chondrocytes (637), highlighting a possible ameliorative role for NOX2- and/or NOX4-targeted inhibition in the pathogenesis of post-traumatic osteoarthritis. A recent review on the dynamics of osteoblast vs. osteoclast NOXs in dynamic bone growth and loss provides additional detail on the subject (87).

4.5. Sensory system

4.5.1. The Eye

In retinal vascular inflammatory responses, both endotoxemia- and diabetes-induced increases in ICAM1 and leukostasis are sharply diminished in NOX2^-/y, identifying NOX2 oxidase as a critical mediator in both pathologies (638). NOX2 knockout effectively prevented diabetic increases in ICAM1, leukostasis, and breakdown of the blood-retinal barrier, underlining a fundamental role of NOX2 in the onset of diabetic retinopathy (638). Moreover, NOX2 has been implicated in VEGF-A and elaborated NOX2 expression which appear to be causal in multiple manifestations of neovascularization of the retina (639). Along similar lines, NOX1, 4 and 5 appear to contribute to oxidative stress and damage to the blood retinal barrier in the context of hypertensive diabetic retinopathy (640). In murine corneas, NOX2 and NOX4 transcription and expression were significantly increased following alkali burns (641) and largely co-localized with CD11b-positive inflammatory cells in the corneal stroma (641). Topical administration of non-specific inhibitors apocynin and DPI abolished infiltration consistent with a potential, however not definitive, role for NOX2/4 in corneal damage (641). Collectively, these data appear to assign a role to NOX2 and NOX4 in inflammatory cell infiltration and tissue degeneration in response to injury. Finally, a handful of recent studies implicate the NOX including NOXs 1, 2 and 4 in glaucoma and at least one study invokes the role of iNOS and Nrf2 in the etiology of the disease (642–644).

4.5.2. The Ear

A number of empirical studies have described NOX’s importance in ear physiology. Primarily, these experiments have centered on the inner ear and NOX3, wherein NOX3 is commonly discussed as rather uniquely delimited in scope compared to other NOXs. In this way, NOX3 mRNA was reported to be at least 50-fold higher in the inner ear than in other tissues, and most specifically in spiral ganglion neurons, the organ of Corti and the vestibular system (63, 67, 645). The common embryological ancestry of these structures also suggests that NOX3-expressing cells share a common role in development (67, 150).

Intriguingly, NOX3 in vestibular and cochlear epithelia has an important role to play with respect to otoconia formation - small crystals of the vestibular apparatus that serve to detect linear acceleration and gravity. The significance of NOX3 to the balance maintenance is unquestionable in mice that harbor a mutation responsible for a loss of function in NOX3, which due to their inability to form otoconia, suffer from the so-called “head-tilt” phenotype (67). Another murine model with a loss of function mutation in the subunit NOXO1, serving as ‘organizer’ of NOX3 in this setting, displayed a similar phenotype (“head-slant”) (646, 647). Particularly, NOX3 in conjunction with lactoperoxidase appears to be responsible for lipid peroxidation, which, as a consequence, produces alterations in the membrane characteristics of the Ca²⁺-containing vesicles thus facilitating reaction of calcium-carbonate with otoconin-90/95 (OC-90/95) to form otoconia crystals (646). Therefore, NOX3 loss-of-function disrupts key processes participating in gravity perception. With that said, head-tilt mice displaying vestibular dysfunction appear not to suffer from hearing impairment (67). NOX3, nevertheless, has been localized to regions of the inner ear associated with hearing and hearing loss as a consequence of drug toxicity (e.g. cisplatin), noise and aging by at least one account via the activation of TRPV1 channels (67, 150, 645, 648, 649). Looking ahead to potential drug therapies, in addition to the difficulty of acquiring selectivity for the NOX3 complex comes the added challenge of penetrability into the inner ear and discerning the contribution of NOX3 to hearing, hearing loss and balance at any particular stage of life and disease.

4.6. Reproductive system

4.6.1. Spermatogenesis

ROS signaling and hypoxia are implicated in the self-renewal of germline spermatogonial stem cells (SSCs) that ensure the continuity of spermatogenesis. As is usually the case, ROS effects are biphasic or multiphasic; that is, there are often diminishing returns of beneficial effects with cell death following the high ROS levels. Although low levels of ROS from NOX1 had been shown to participate in SSC self-renewal, deprivation of essential amino acid glutamine sharply upregulated NOX1 activity and caused apoptosis of SSCs (88). In accordance with this, NOX1 deletion attenuated p53-mediated apoptosis (88). Thus, from a physiological standpoint, it would appear that glutamine ensures baseline ROS-dependent SSC-self-renewal by defending against excessive NOX1-derived ROS while providing sustenance to the cells in the way of restrained NOX1 activity and low-level generation of ROS (88). Although not entirely clear, glutamine’s protective effects have in part been ascribed to the generation of cellular glutathione (88). Additionally, crosstalk between ROS and hypoxia seems to even more-tightly regulate the renewal process since NOX1 knockout mice exhibited reduced HIF1α levels (650), which, in turn, can unleash mitochondrial ROS excess (651).

In murine SSCs, NOX3 expression correlates with self-renewal by way of mechanisms that recruit various NOX proteins differentially via the PIK3-Akt and MAP2K1 pathways (464, 652). Indeed, MAP2K1 influences NOX3 but not NOX1 or NOX2 in these cells. In support of a multi-NOX-dependent amplification process are observations that NOX3-derived ROS drive SSC self-renewal through feed-forward ROS from other NOXs (650). Notably, in this regard, histone methyltransferase SETDB1-mediated control of NOX4 in the modulation of spermatogonial stem/progenitor cell survival was recently discovered (653). And, in support of a rather unique role for NOX4 in the process, suppression of mitochondrial ROS, which is commonly held to be influenced by NOX4, did not influence SSC fate (650).

Parenthetically, ROS are physiologically important for processes involving sperm function such as motility, capacitation, acrosome reaction, and sperm-oocyte fusion. High ROS levels, however, may cause changes in membrane permeability affecting flagellar movement and thus, decreasing fertility (654). Under physiological conditions, NOX5 has been described as the main source of ROS in spermatozoa and to play a role in human sperm motility (655). Not surprisingly, treatment of sperm with the NOX inhibitor GKT136901, which purportedly inhibits NOX5, showed a significant decrease in the percentage of acrosome-reacted spermatozoa (656). On the other hand, in teratozoospermic semen samples, sperm displayed upregulated NOX5 expression compared to normal sperm (657, 658). As usual, positive and negative effects of NOX-generated ROS are in balance, as sperm exhibit low antioxidant enzyme levels they are more vulnerable to changes in this balance. However, germane to this subsection's discussion on SSCs, as mentioned earlier in this review recent inquiries by Geiszt and colleagues into the presence of NOX5 protein, challenge the notion that the isoform is involved in gametogenesis (173).

4.6.2. Oogenesis

Endogenous ROS play key roles as signaling molecules that modulate ovulation (659) but little hard data as to causality of NOXs in mammalian oogenesis has stood the test of time. With that said, the literature supports PKC and NOX2 as linearly linked to murine oocyte maturation (660) and in Drosophila NOX has been proved critical in follicle maturation, rupture and ovulation (661, 662). NOX1 has not, to our knowledge, been reported in ovaries and, for that matter, reports of the role of NOX2 (660) might be attributed to infiltrating leukocytes, especially since it has been suggested that neutrophil NOX2 participates in corpus luteolysis and oocyte degradation (663). Indeed, NOX2 is linearly linked to murine oocyte maturation even though the precise cell type in which NOX2 is functional is not entirely clear but could play a supportive role by way of its effects in cumulus cells and cumulus-oocyte complexes (660). Moreover, NOXs 4 and 5 have been detected in human follicles, with both purportedly being expressed in human granulosa cells (664). However, only mRNA levels were reported and causality was not tested, to our knowledge, in either of references (664) and (665). In contrast, MATER (maternal antigen required by embryos) protein interacts with PKC in cumulus cells under physiological conditions; in both younger and elder female subjects MATER and NOX4 proteins are highly correlated and levels of both are much higher in young infertile patients as compared to older (666). These results are consistent with a key role of both MATER and NOX4 for correct follicular development and establishment of a pregnancy, however, again causality of NOX4 was not tested (666). Moreover, NOX4 was detected in isolated differentiated, in vitro fertilization-derived human granulosa-lutein cells, in proliferating human granulosa tumor cells, as well as in situ in cells of growing ovarian follicles (667). H₂O₂ levels in these cells were reduced 50% by GKT137831, consistent with a considerable NOX4-H₂O₂ contribution to these processes (667). Although NOX5 is expressed in follicles, its role in oogenesis has not been conclusively determined to our knowledge as of the writing of this review.

4.6.3. Embryonic development

mRNA transcripts for p22^phox have been detected as early as mouse embryonic day E5.5, for p67^phox at E7.0 and for p47^phox at E7.5 before yolk sac hematopoiesis (E8.0) (668). In the developing kidney, NOX1 is primarily expressed in ureteric bud cells and condensed mesenchymal cells at E16 (669). At E18 and P1, NOX1 was also identified in immature glomerular endothelial and mesangial cells and tubular epithelial cells in nephrogenic zones, whereas its expression was detected as weak in the adult kidney cortex (P20) (669). Contrary to this finding, NOX1 was robustly expressed in renal tubules in the medulla of adult kidney (P20). The levels and distribution of other NOXs and subunits (NOX2, NOX3, NOX4, p47^phox, and p67^phox) were significantly elevated postnatally and comparable to those of NOX1 (669) which might suggest a spatiotemporal and orchestrated modulation of NOXs during embryonic kidney development.

In an early study of tissue NOX in embryonic and fetal hematopoiesis, NOX2 mRNA was explored and identified as first expressed at E9.0; however, among all NOX2 oxidase components, only p22^phox was expressed in circulating hemocytoblasts by E9.0 (668). By comparison, no other embryonic tissue displayed NOX2, p22^phox, p47^phox or p67^phox mRNA, either before or after the onset of hemocytoblastic circulation (668). However, the four transcripts and corresponding proteins are expressed in clusters of developing granulocytes in the liver by E14, with expression continuing at E16 and E19. While spleen and limb bone marrow showed inconsistent results, cord blood neutrophils contained all of the above-mentioned subunits.

4.6.4. Placenta

NOX1 is identified as a major source of increased ROS in syncytiotrophoblasts, extravillous trophoblasts and vascular endothelial cells in preeclamptic human placentas (670). NOX1 and NOX5 have been detected in human cytotrophoblasts (671). In prior studies, NOX2 was found in normal placental villi, though its expression may have come from placental macrophages (671). In placentas of pigs under a high-energy vs. low-energy diet regimen, NOX2 mRNA and protein were sharply increased as compared to those from dietary controls concomitant with decreased placental vessel density and markers of angiogenesis (672). Another study from the same group showed that impaired placental angiogenesis correlates with the higher expression of NOX2 in low birth weight pig fetuses, and, in fact, demonstrated that NOX2 deletion significantly increases fetal angiogenesis in these fetuses (673). Additionally, higher NOX2 and NOX4 protein levels were demonstrated in preeclamptic human placentas (674). A more detailed and rigorous analysis of NOXs expressed in the placenta showed that mRNA for all members of the family can be detected in chorionic villi between 7–14 weeks of gestation, while at term all but NOX3 mRNA could be detected (675). In terms of protein, all NOXs could be detected in chorionic villi through pregnancy except for DUOX2, which was only detected at term in villous cytotrophoblasts (675). Immunohistochemical analysis showed temporal and spatial differences in NOX isotype expression during pregnancy. Interestingly, NOX4 was detected throughout villous cytotrophoblasts during early pregnancy but it seemed to translocate to the nucleus at term. Furthermore, spatio-temporal differences are also affected in early- and late onset of preeclampsia (675). In aggregate, ROS from a variety of NOX isoforms appear to be distributed in a cell-specific and temporal manner in a broad range of stages of placental development, with implications for pregnancy and preeclamptic hypertension.

4.7. Skin

Unstimulated normal human epidermal keratinocytes under basal conditions express rather robust amounts of DUOX1 mRNA and threshold amounts of NOX1, NOX2, NOX4, NOX5 and DUOX2 but no NOX3 (676). Skin contains many histological features of connective and adipose tissue with its distinguishing cell types being keratinocytes and melanocytes. In response to UV light, a biphasic pattern of keratinocyte NOX1 expression appears to be activated that is causally involved with nucleotide excision repair and reduced apoptosis (677). In another report, genomic instability induces NOX1-derived and mitochondrial O₂^·−, dispersion of mitochondria and decreased ATP production which is ablated upon impairment of NOX1 (but not NOX2) (678). That said, what makes the latter study remarkable is that the protective effects of NOX1 suppression could not be replicated by overexpression of catalase or cytosolic or mitochondrial SOD (678). These data could suggest either (a) the antioxidant enzymes were not sufficiently expressed or able to access NOX1-enriched organelles; or (b) that NOX1 is eliciting its negative effects independent of ROS (677, 678). Additionally, correlative data imply a varied participation for NOX isoforms in immortalized keratinocyte cell lines (e.g., HeLa, HaCaT) which express one or more NOXs (NOX1, 2, 4 and/or 5) (679, 680). Moreover, other studies imply a causal role for NOX2 and NOX4 in TLR4-instigated keratinocyte autophagy coincident with characteristic inflammatory signaling (680, 681).

In contrast to keratinocytes, normal human epidermal melanocytes appear to express NOX1, NOX2, NOX4 and p22^phox (682, 683). In a study by Brar and coworkers in 2002 comparing the contribution by NOX2 and NOX4 oxidase, only NOX4 antisense oligonucleotides prevented melanoma cell proliferation (682). Those findings have been substantiated by a recent report correlating NOX4 with melanoma in BRAF-mutated patients, indicating that NOX4 together with the increased c-met might serve as a possible marker for severity of melanoma (70). In contrast, Liu et al. in 2012 detected elevated NOX1 oxidase in human melanoma vs. control cells, and NOX1 overexpression and inhibition augmented and attenuated their proliferation, respectively (684). Moreover, active assembly of non-canonical NOX1 oxidase (utilizing p47^phox) was suggested to drive oxidase activity involved in migration, accumulation and growth of A375 melanoma cell in vivo (683). On the other hand, other studies suggested that NOX4 is involved in the cellular differentiation rather than dedifferentiation (617, 685–687). Indeed, a key study links NOX4 to the antineoplastic property of proopiomelanocortin (688). In aggregate, NOX4 inhibitors must be applied with caution of possible anti-differentiation and pro-tumorigenic effects in skin cancer and more broadly in other tissues wherein NOX4 has been described as foundational to differentiation.

Lastly, proinflammatory agents such as sphingosylphosphorylcholine were shown to induce NOX5 expression in keratinocytes where it is responsible for increased ROS in a calcium-dependent ROS manner (679, 689). Moreover, perinuclear NOX5 is detected in the non-epithelial metastatic human melanoma cell line UACC-257 (690) but its role there has yet to be defined.

Other Skin-Related Diseases

NOXs in melanocytes and perilesional keratinocytes appear to contribute to the ROS-mediated destruction of pigment-producing cells in vitiligo patients (691, 692). NOX1 was shown to be the culprit for ROS generation by keratinocytes in response to ultra violet B light (UVB) and when NOX1 was knocked down pivotal p38 MAPK activation was impaired, followed by reduced IL6 levels and attenuated cell cytotoxicity in vitro (693). Interestingly, when lysates of UVB-irradiated NOX1^-/- vs. wild type cells were injected intradermally in mouse ears in vivo, those receiving NOX1 null lysates displayed abrogated swelling and abolished IL6 mRNA levels convincingly demonstrating a crucial for NOX1 in the inflammatory process (693). In another in vitro model, keratinocytes treated with disease-specific pro-inflammatory cytokines exhibited upregulated NOX1 and NOX4, increased indication of ROS production and DNA damage consistent with acute dermatitis (AD) (76). ML171, a reportedly NOX1-selective NOX1 inhibitor, abrogated oxidative stress and p38 activation as effectively as a dual NOX1/4 inhibitor in AD, indicating a greater role of NOX1 (>NOX4) (76). In a murine bleomycin-induced scleroderma model, the dietary antioxidant alpha-lipoic acid exerted protective effects against dermal thickness, inflammation, and TNFα expression (685). Moreover, transcripts of NOX4 and p22^phox along with αSMA, collagen type I and fibronectin, TGFβ1 and p-Smad3 were significantly induced in the bleomycin group consistent with myofibroblast differentiation. Importantly, all of these salient changes were significantly diminished by the antioxidant and radical scavenger alpha-lipoic acid signifying a feed-forward role of oxidants in this model (685). Overall, the findings appear to support NOX1 as causal for keratinocyte dysfunction and raise the possibility of a role for NOX4 that requires further examination.

As for skin wound healing, differentiation of crucial myofibroblasts involved in closure of the wound and healing does not appear to depend on NOX4 in vivo or in vitro despite its robust upregulation (694). This runs contrary to previous studies implicating NOX4 in differentiation of vascular and cardiac fibroblasts (687, 695, 696). The discrepant findings are likely attributed to distinct downstream signaling roles of NOX4 across cells and tissues with respect to phenotype. Additionally, distinct and varied distributions of NOX isoform in different cell types could confound a clear phenotypic role for NOX4.

4.8. The Excretory system

The original site of NOX4 detection (aka Renox for Renal Oxidase), the kidney, is one of the most complex organs and subserves a fundamental role in corporal homeostasis through essential functions like glomerular filtration of solutes, tubular reabsorption, and the countercurrent multiplication system. Among myriad other functions including urea excretion, the synthesis and release of renin that converts angiotensin I to vasoconstrictor hormone AngII, it plays multiple regulatory roles in blood pressure maintenance and hypertension. Kidney pathologies that relate to the NOX and diabetic nephropathy and hypertension are addressed in subchapters 4.10 and 4.11, respectively.

Although the role of ROS has been vastly investigated in the context of electrolyte transport, fluid reabsorption and ion channel activity, many of the early studies were conducted with either ROS scavengers or nonspecific NOX inhibitors. Hence little could be deduced from those studies when it came to a concrete isoform role in urine formation and processes contributing to it. For instance, in the proximal tubules, apocynin and siRNA against p22^phox normalized reduced reabsorption in SHR compared to WKY rats, suggesting that increased ROS from NOXs 1 – 4 in SHR reduce salt and water reabsorption via a Na⁺/H⁺ exchanger (697). In opossum kidney cells of high in vitro passage, increased Na⁺/K⁺-ATPase activity is consistent with excessive reabsorption of fluid at the thick ascending limb of the loop of Henle (TALH) and was detected alongside increased NOX1, SOD1, SOD2, SOD3, and H₂O₂ accumulation (698). Pretreatment with apocynin reduced H₂O₂ levels and Na⁺/K⁺-ATPase activity hinting at a role for one or more NOXs in this age-related response as NOX1, or no other NOX for that matter, was definitively examined for causality (699). Moreover, increased flow rate in the TALH was shown to stimulate PKC and p47^phox-dependent NOX activity, resulting in increased O₂^·− and Na⁺ reabsorption in the thick limb (700). Demonstrating more on the potential role of NOXs in electrolyte transport are studies implicating ROS as potential regulators of epithelial sodium channels (701, 702). To that point, losartan (AngII type 1 receptor blocker), DPI and PKC inhibition uncoupled NOX activity from AngII-stimulated epithelial sodium channel function. In support of these findings was an increased epithelial sodium channel activity following PKC activation and addition of exogenous ROS (701, 702).

NOXs also appear to account for the regulation of potassium intake as suggested by a pair of studies. More explicitly, low K⁺ intake induced O₂^·− levels in the renal cortex and was associated with a reduction in potassium channel (ROMK) activity and curbed K⁺ excretion (703, 704). In support of this notion, Tempol reduced cortical O₂^·−, increased ROMK activity and restored K⁺ excretion in low K⁺ diet rats (703). Overall, the studies hinted that NOXs might play a role in the regulation of electrolyte transport and fluid absorption and have consequences for the regulation of blood pressure (discussed in subchapter 4.11 of this review), especially in models using AngII as the pressor agent. On the whole, many of the above findings, while implicating NOXs as culprits, required follow up studies to provide more direct and definitive confirmation of NOXs’ participation. For example, with respect to NOX involvement, NOX2 deletion was found to alleviate declines in ROMK channel activity in low K⁺ diet rats (705, 706).

4.8.1. NOX expression & Functionality in the Kidney

Over the past decade or so, it has become evident that numerous NOX isoforms including NOX1, NOX2, NOX4 and NOX5 are expressed in the kidney (707) and are causal in numerous of scenarios of biology and disease in the kidney. For example, renal vessels primarily express NOX1, NOX4 & NOX5 and in both ECs and SMCs (707, 708) including NOXA1 which, along with NOX1, contributes to renal artery constriction (709). One study, on the other hand, illustrates the aberrant expression of NOX2, in addition to NOX1 and NOX4, in kidney vascular dysfunction (89). Another implicates NOX2 in the maintenance of tone of afferent arteriole in response to AngII (710). Moreover, pharmacological NOX1 blockade ameliorated MAPK signaling and renal vascular ischemia reperfusion injury (711).

The glomerulus, comprised of mesangial cells and podocytes, exhibits upregulated NOX1, NOX2, NOX4, and NOX5 (along with canonical NOX2 cytosolic subunits) which reportedly contribute to mesangial hypertrophy, ECM accumulation, tissue expansion and apoptosis of podocytes common in kidney disease (707, 712). To that point, NOX2 can activate podocyte TRPC6 channels which contribute to glomerular and kidney disease (713) yet NOX4 and NOX5 dominate the discussion in this regard. In its own right, NOX4 has been known for some time to align with glomerulosclerosis in human diabetic kidney disease and targeting of NOX4 in this context in animal models proved renoprotective (714, 715). However, at least one study challenged that assertion showing no effect of NOX4 deficiency on nephropathy (716). More recently, studies have emerged linking increased NOX5 expression in a variety of kidney cells with renal inflammation and nephropathy in humans and rodent models, respectively (717, 718). Indeed, by some accounts, NOX5’s role in kidney disease may compound and supersede that of NOX4 in humans. In mice overexpressing human NOX5 in mouse kidney mesangial cells on a NOX4 null background exacerbated renal injury in response to diabetes (717, 718). Moreover, NOX5 increased redox-sensitive factors like EGFR1, ERK1/2 and TXNIP in this model and NOX5 silencing in mesangial cells suppressed these factors and prevented fibrosis and inflammation (719). Additionally, human NOX5 overexpression in mouse podocytes can drive inflammation and glomerular degradation (181, 719, 720).

Proximal tubules express all of the same isoforms and NOX5 has definitively been shown to participate in oxidative stress and inflammation therein (721–723). On the other hand, NOX2 participates in functions of renal tubules such as glucose handling and electrolyte homeostasis (724). Furthermore, NOX4 in mouse proximal tubular cells has been implicated in diabetic nephropathy in response to high glucose and, appropriately, GKT136901, NOX1/NOX4 inhibitor, ameliorated the symptoms (725). As for the macula densa, it expresses NOX2 along with its cytosolic subunits, and NOX4 (723, 726). In fact, recent studies have proved the definitive regulation of ENaC by NOX4 in salt-sensitive rats (727). The authors direct the reader’s attention to a pair of excellent reviews for more detail on the subject (174, 728).

4.9. Digestive system

NOX1 and DUOX2 are the predominant NOX subtypes found in the gastro-intestinal tract. Unlike DUOX2 that is evenly distributed in the intestinal tract, NOX1 is predominantly expressed in the ileum, cecum, and colon where it expectedly plays a role in host defense (454, 729, 730). Indeed, NOX1 is expressed in the colon in a graded manner, increasing from proximal to distal segments (731), and governs homeostasis and injury repair of the colon via Wnt-β-catenin signaling (732, 733) as well as orchestrates symbiosis between gut microbiota and the intestine by way of a variety of redox-sensitive signaling pathways including ERK, FAK and NFκB (356). DUOX2’s physiological role seems in large part to be the regulation of interactions between the intestinal microbiota and the mucosa important for the immune homeostasis (734). Indeed, missense, loss-of-function defects in NOX1 and DUOX1 display abnormal ROS production and intestinal secretory cell (Paneth) metaplasia, dysfunctional defense to C. jejuni and very early onset inflammatory bowel disease (735). In contrast to these homeostatic roles, both NOX1 and DUOX2 are shown to trigger the development and progression of ulcerative colitis (454, 735, 736).

NOXO1 in human colon cancer cells is a target of proteasomal degradation which reduces NOXO1 expression and stability and, as a result, NOX1-dependent ROS (454). On the other hand, NOXO1 is markedly increased in human colon cancer compared to normal colon which is consistent with a role of the NOX1 oxidase in colon cancer progression (737). On its own merits, the role of NOXs and DUOXs in colon cancer is described in greater detail in the Cancer subchapter (4.15) of this review.

4.9.1. Esophagus

A pro-survival role for NOX1 is described in esophageal epithelium and reportedly involves MAPKs, JNK and NFκB pathways (738). On the flip side, the esophagus is malignantly eroded by gastroesophageal acid reflux which can cause disease evolution from Barretťs esophagus to esophageal adenocarcinoma (739). To that point, silencing of NOX1 and NOX2 by RNAi was shown to suppress DNA damage in esophageal cells treated with bile acids as compared to controls; and in support of this observation, NOX2 inhibitory peptide gp91ds was also effective (57, 739, 740). More mechanistically, esophageal cell exposure to acidic bile salts induces phosphorylation of the p47^phox and its migration to the plasma membrane (739).

Overexpression of NOX5-S and p50 subunit of NFκB significantly increased silencer-of-death domain (SODD) protein expression consequently exerting an anti-apoptotic effect in esophageal adenocarcinoma cells and, in turn, proliferation (741). Concordantly, knockdown of NOX5-S and NFκB’s p50 prevented acid-mediated increase in SODD protein (741) and averted proliferation (742). Moreover, consistent with a deleterious role for NOX5, under conditions or transient acid loading, chronic elevations in NOX5 were shown to cause proliferation of Barrett’s cells and esophageal carcinoma cells (742). In the human Barretťs cell, NOX5 mRNA (but not DUOX1 or DUOX2) and p16 promoter methylation were elevated following pulsed acid treatment in a time-dependent manner (743); and these increases appear to reverse themselves after short-term acid exposure (4 and 8 wks) upon return to normal culture medium. In contrast, after long-term acid exposure (12 wks) restoration is only partial (743). In the same study, palliative effects of proton pump inhibition in Barretťs esophagus patient mucosa were indicated by demonstration that a one-month proton pump inhibitor treatment could prevent elevations in NOX5 (743).

4.9.2. The Colon

NOX1 expression is relatively low in the proximal vs. distal colon (150, 731, 744). Moreover, in one key study, NOX1 transcripts appeared evenly distributed between luminal surface and crypts within the colon wall (150, 731), and in another study were deemed highly localized in the lower portion of the crypts where the epithelium proliferates and differentiates (358). Yet another study asserts that the highest NOX1 protein expression is in the surface mucosa of guinea pig colon and is acutely activated by TLR4 and TLR5 signaling (548). Consistent with that discovery, an array of reports revealed the expression of NOXO1 and NOXA1 in guinea pig, mouse, and human colon epithelium (145, 220, 221, 548). Histochemically, NOX1 is found at the brush border of colon epithelial cells and in the cytoplasm in multiple studies, wherein its expression in the colon epithelial cell mucosa is heightened in patients with prolonged constipation (464, 745).

Additionally, NOX1 and O₂^·− are reportedly increased upon differentiation and growth arrest in colonocytes (358) and the highest levels of NOX1 were found at subconfluence (150, 746). The latter appears to have led to speculation that NOX1 might act in a growth phase-dependent manner; however, NOX1 predominant luminal surface location appears to disfavor this hypothesis, at least under physiological conditions (548, 731). Indeed, it is worth emphasizing that NOX1 is indispensable for gut microbiota and intestinal symbiosis (356).

Continuing on that line of thought, NOX1 expression is higher in the distal colon, and coincident with increased presence of bacteria (731). Furthermore, NOX1 expression is enhanced by inflammatory cytokines such as IFNγ (358, 747) or activated by LPS (748) or bacterial flagellin (548). Disturbed rhythmicity of NOX1-derived intestinal H₂O₂ is attributed to altered microbial oscillations (gut dysbiosis) leading to significant disruptions in gut microbiota, nutrient and energy extraction and growth retardation that has been associated with obesity, type II diabetes and atherosclerosis. As such, typical NOX1-derived H₂O₂ generation contributes to normal rhythmicity which governs the growth and survival of intestinal bacteria and preserves gut homeostasis (749, 750). Meanwhile, in the small intestine, NOX activity is primarily localized in enterocytes and attributed to high levels of inflammatory stimuli-induced NOX4 and markedly lower NOX2 (751). Intriguingly, it was discovered earlier on that NOX4 is essential for cell-cell communication of enterocytes and proper coordination of the innate host response (752).

Moreover, DUOX2 is a significant source of H₂O₂ in the small and large intestine, in particular the sigmoidal colon, the cecum, salivary glands and rectal glands (359, 753, 754). Intestinal epithelial cells express DUOX2 protein mainly at the tip of the epithelium where it is more strongly expressed in conventionally raised vs. germ-free mice (357). Duox2 gene expression is induced by colonization of germ-free mice with microorganisms, but interestingly not when they are colonized with diverse commensal bacteria (357). These microbiota not only induce Duox2 expression but achieve its activation by exploiting two distinctive signaling pathways; fast induction as seen in colon involves MyD88 and p38 MAPK while a slower induction in ileum is controlled by adaptor proteins including interferon-β (TRIF) and canonical NFκB signaling (357, 540). Indeed, distal GI tract DUOX2 plays a role in host defense, providing the source of H₂O₂ for lactoperoxidase-catalyzed generation of antimicrobial hypothiocyanite ions, and, similar to NOX1 in the proximal colon, provides protection from and modulates interactions between intestinal microbiota and the mucosa (358, 359, 730). On the other hand, with respect to dysfunction, DUOX2, like NOX1, protein is also elevated in patients with constipation compared with controls (540, 745).

4.9.3. Liver

Liver resident cell types, such as hepatocytes, hepatic stellate cells, macrophages, ECs and Kupffer cells express varied combinations of NOX isoforms ranging from NOX1 to the DUOXs with NOX1, NOX2 and NOX4 being the most studied (755). ROS derived from NOXs are implicated in all stages of liver disease, from non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) to liver fibrosis and cirrhosis (756–762). In fact, NOX1 is significantly upregulated in livers of NASH patients as well as in mice fed a high fat diet (HFD) (763); and NOX1 deficiency can attenuate HFD-induced liver injury and suppress the accumulation of nitrotyrosine-protein adducts in liver sinusoids (763). In primary cultured liver sinusoidal endothelial cells, palmitic acid has been shown to upregulate NOX1, but not NOX2 or NOX4, mRNA and reduced endothelial cell relaxation (763). Moreover, NOX1 deletion attenuated serum aminotransferase and liver caspase 3 suggesting a key role of NOX1 in NAFLD progression (763). In the context of liver cancer, injection of the hepatocarcinogen diethylnitrosamine induced elevations in liver to body mass and serum alanine aminotransferase concentrations were significantly lower in NOX1^−/−, but not NOX4^−/−, mice (522). In addition, diethylnitrosamine challenged NOX1^−/− mice exhibited lower hepatocyte proliferation and 80% fewer tumors together with decreased cyclinD1 levels compared to wildtype mice (522). Moreover, hepatic inflammatory cytokine and chemokine expression, i.e. IL6, TNF, CCL1 & CCL21α, were strongly induced in control mice after diethylnitrosamine injection, while no change was detected in NOX1^-/- mice, coinciding with a key role of macrophage-specific NOX1 in regulating their production from macrophages and their paracrine role in hepatocarcinoma (522). As such, NOX1 appears to play a fundamental role in the pathology of the liver with varied roles in distinct cell types ranging from pro-inflammatory to pro-proliferative as the basis for fatty liver disease and hepatic carcinoma.

Furthermore, a handful of findings support a role for NOX2 oxidase in steatosis and insulin resistance. That is, NOX2-deficient animals are protected against HFD-induced steatosis and insulin resistance (764). Complementary to this, progression of NAFLD has been attributed to NOX2 (756) and NOX2 deficiency normalized mitochondrial oxidative phosphorylation dysfunction, oxidative DNA damage and NASH caused by HFD (757). Intriguingly, palmitate-stimulated hepatic infiltrating macrophages (wherein NOX2 activation predominates), but not Kupffer cells, are critical mediators of the hepatic response to liver injury and inflammation in NAFLD (764). Overall, the reports collectively corroborate involvement of oxidative stress in patients with NAFLD, which is aggravated when progression to NASH occurs. Thus, NOX2-ROS and downstream signaling contribute to the genesis of NAFLD/NASH even though the proposed pathways inciting NOX2 differ widely (765–767).

As well, elevated NOX4 levels were found in patients with NASH compared to healthy controls (760). And, mice with hepatocyte-specific NOX4-deficiency exhibited attenuated oxidative stress and liver fibrosis under HFD-induced steatohepatitis (760). Liver fibrosis resulting from chronic liver injury, is another process wherein NOX isoforms participate in the activation of hepatic stellate cells, the predominant collagen-producing cells in the liver (768, 769). The role of NOX1 and/or NOX4 in liver fibrosis was demonstrated by reduced levels of carbon tetrachloride-induced hepatic collagen deposition and other fibrogenic markers after treatment with GKT137831 (770). Similarly, when liver fibrosis was induced by bile duct ligation, NOX4 pharmacological downregulation by GKT137831 or genetic deletion (NOX4^-/-) attenuated the fibrotic phenotype (771). On the other hand, NOX4 induction emerges as necessary for a pro-apoptotic effect in liver cells (766) and NOX4 deletion appears to augment liver regeneration (81). NOX2 also participates in hepatic fibrosis (761, 772, 773). With respect to that assertion, liver fibrogenesis induced both by carbon tetrachloride or bile duct ligation were attenuated in NOX2^-/- (as well as in NOX1^-/- mice) (761) and in p47^{phox -/-} mice (774) suggesting participation by the NOX2 canonical or NOX1 hybrid oxidase. While the absence of NOX5 in mice hinders the study of its role in animal models of liver fibrosis, in human hepatic stellate cells, various NOX5 splice variants are detected (their expression upregulated by TGF-β and AngII) which cause hepatic stellate cells to proliferate and increase their collagen synthesis via p38 MAPK (397). These results support the likelihood that NOX1, 2, 4 and 5 all play some role in liver fibrosis and other liver pathologies. Even so, the majority of attention with respect to therapeutic targeting has been placed on NOX1/NOX4 inhibition (770, 771, 775, 776). For a comprehensive review of NOXs in liver biology and disease see (755).

4.9.4. Inflammatory bowel disease

NOX1 as well as DUOX2–DUOXA2 are upregulated in Crohn’s disease, irritable bowel syndrome, pouchitis and ulcerative colitis which one might theorize are attempts by the bowel to augment host defense and healing (540). Supporting clinical significance in that regard, inactivating missense NOX1 and DUOX2 mutants are associated with very early onset inflammatory bowel disease (VEOIBD, ≤ 6 years) patients (735). In fact, a pair of studies of IBD patients harboring NOX1 mutations showed that the catalytic activity of multiple NOX1 variants negatively correlates with the disease though in a context-specific manner (540, 777). In addition, in VEOIBD patients one of the Duox2 alleles harbors a mutation in the vicinity of the first transmembrane domain of DUOX2 that impedes epithelial cell surface expression in cell-based studies (540, 778). Likewise, in adult IBD, a monoallelic Duox2 mutation was found to be associated with Crohn’s disease in some individuals in the Ashkenazi population (779). Not only was reduced epithelial NOX function associated with VEOIBD, mutations in all major NOX2 complex subunits (CYBB, CYBA, NCF1, NCF2 and NCF4) (presumably causing the inactivation of leukocyte NOX function) (780) that determined susceptibility to infection were also identified (540, 781). Thus, a popular notion that oxidative stress or ROS overproduction elicited by the NOXs is culpable for IBD is predictably erroneous. For more detailed information with respect to polymorphisms in NOX and inflammatory bowel disease, please see Chapter 4.16.

4.10. Endocrine system and Obesity

4.10.1. Endocrine – the pancreas

In the human pancreatic β-cell, Oliveira and colleagues in 2003 revealed the presence of p47^phox, NOX2 and p22^phox transcripts and, notwithstanding widely recognized limitations of antibodies, p47^phox and p67^phox proteins were detected by Western blot in the central islet (782). Moreover, colocalization of p47^phox, p67^phox and NOX2 protein in the islet along with glucose-elicited p47^phox translocation from the cytosol to the plasma membrane suggest an active NOX2 oxidase (782). Concordant with those findings, DPI impaired glucose metabolism and glucose-mediated insulin secretion (782) and these data were recently corroborated as potentially involved in circadian clock protein-driven insulin secretion (783). Moreover, those findings have been supported by and extended to other NOX isoforms in rat pancreatic islet cells wherein NOX1, NOX4, p22^phox as well as p47^phox, NOXO1, NOXA1 and p40^phox were detected even though the singular or cumulative causality of one or more NOX oxidase system to pancreatic function was not delineated (784). In other studies, glucose metabolism-derived or generalized NOX-derived H₂O₂ was proposed as having a positive effect on insulin secretion whereas chronic oxidative stress conferred a negative influence on secretion (785, 786). Additionally, NOX2 activity has been linked with mitochondrial “fuel”-induced insulin secretion as well as dysregulation in islets (787, 788). However, no in-depth studies could be found contrasting the mechanistic impact of NOX2 oxidase vs. other NOX isoforms individually or in conjunction and/or with the mitochondrion, and thus the need for deeper inquiry is evident.

4.10.2. Obesity and adipocytes

A shift in adipocyte redox balance is a hallmark of metabolic disorders and amplified NOX expression and ROS production in adipose tissue of obese mice go hand in hand with obesity-related metabolic disorders (582, 789). In that regard, NOX1, NOX2 and NOX4 have been localized to adipocytes and their expression differs widely across tissue and cell types (582, 789). Along those lines, NOXs are broadly and variably upregulated, for example, across skeletal, vascular, adipose and renal tissues in diet-induced obesity (582, 789, 790).

Related to adipocytes, per se, the most studied and ostensibly abundant NOX has reportedly been NOX4 (405, 791) with one report indicating measurable changes in NOX2 and NOX3 expression along with p47^phox, p67^phox and NOXO1 (405, 792). In healthy adipocytes, NOX4-mediated H₂O₂ production is induced by insulin, which enhances its signal and drives preadipocyte differentiation to adipocytes. Moreover, NOX4-mediated H₂O₂ amplifies its signaling through oxidative inhibition of the protein tyrosine phosphatase 1b (PTP1b), which, in turn, promotes insulin receptor activation and glucose uptake (582, 791). Still, both NOX1 and NOX4 are increased coincident with ROS production during insulin-induced differentiation of preadipocytes to adipocytes. However, rather uniquely it appears that NOX4 influences various mediators of the insulin receptor pathway, i.e. MAPK phosphatase-1, insulin receptor substrate-1, and, in this way, serves to modify the balance between proliferative and differentiated phenotypes (405). To that point, silencing of NOX4, but not NOX1, by RNAi inhibited insulin-induced differentiation and ROS production and promoted proliferation (405). Furthermore, as an early occurrence in obesity, nutrient excess, and increased glycolysis was shown to enhance NOX4 levels and its translocation from the cytoplasm to lipid rafts where its efficient activation is proposed (793). The related ROS amplification, though not delineated as to the specific pathways involved, appears to elicit activation of NFκB and chemotactic factor expression and, in turn, local inflammation inducing insulin resistance in differentiated adipocytes (793). In the same report, it was further established that NOX4 is the source of ROS in response to high glucose and palmitate treatment excluding mitochondrial oxidative phosphorylation as such (793); and that of all the NOX and DUOX isoforms examined, only NOX4 was present in appreciable amounts and sharply induced by high glucose at the mRNA level (582, 793). Similarly, albeit to a far lesser degree, NOX4 protein levels are induced, concomitant with an outsized rise in NOX4 activity and NOX4-derived ROS. This perhaps suggests that NOX4 is principally controlled by post-translational modification under glycolytic conditions (793) and invites speculation that Poldip2, in this scenario, might augment elevated ROS similar to the situation in vascular injury models (794). In support of this notion is the observation that Poldip2 associates with p22^phox to activate NOX4 (162). Relatedly, treatment of 3T3–L1 adipocytes with high glucose and palmitate for 7 days resulted in increased glycolytic capabilities, NADPH content, and NOX4-derived ROS generation (793). And, in accordance with the latter, blockade of the pentose phosphate pathway as well as NOX4 silencing resulted in diminished H₂O₂ production, suggesting alignment of the pentose phosphate shunt, NADPH generation and NOX4 activity (793). An additional study is consistent with many of the abovementioned observations in response to saturated fatty acids, but not unsaturated fatty acids, and invokes a unique role for the TLR4 receptor in fatty acid-, yet not glucose-, responses (795). One other plausible interpretation of NOX4’s effects in adipocytes could be that excess extracellular H₂O₂ from NOX4 mediates a shift in phenotype of adipose tissue-resident macrophages to one with an elevated capacity to generate increased and sustained tumor necrosis factor alpha (TNFα), which participates in a prolonged inflammatory response (796).

4.10.3. Type II diabetes (T2DM)

Elevated and sustained hyperglycemia, glucose intolerance and insulin resistance are primary features of diabetes mellitus and NOX and ROS are recognized triggers for high glucose and high-fat diet (HFD-induced insulin resistance. For instance, monocytes of T2DM patients exhibit increased levels p22^phox, suggesting that hyperglycemia promotes oxidative stress by way of NOX activation (797). As for vascular smooth muscle, HFD-fed mice with vascular smooth muscle-specific p22^phox overexpression exhibit exacerbated obesity, reduced HDL cholesterol, increased levels of leptin and the proinflammatory chemokine MCP1 as well as excessive ROS production (798). It should come as no surprise then that these mice develop a NOX-mediated glucose intolerance and, concordantly, one report shows a reduction of visceral fat accumulation (inguinal fat depot) in HFD fed NOX2^-/- mice compared to HFD WT (799). On closer look, however, macrophages appear to be a primary population of NOX2-positive cells in visceral adipose tissue (799). This raised the likelihood that macrophages are responding to inflammatory chemokines and cytokines released by a growing number of adipocytes in obesity and diabetes in an innate immune-like response at least initially (799) with destructive consequences like adiposopathy, exaggerated cytokine release and lipodystrophy to follow. Thus, it should be evident that the consequences of NOX2 upregulation can range from moderate to excessive depending on the graded participation of resident adipocytes vs. infiltrating or resident macrophages. Predictably, this cumulative cellular response contributes to acute and chronic inflammation even at low subthreshold levels (low grade) whose effects over a lifetime on tissue system-wide increase morbidity and mortality as with organ damage and “inflammaging” (438).

With the knowledge that NOX2 oxidase is present and upregulated in diabetes, came a series of other studies that showed protective effects of NOX2 suppression or deletion. For instance, NOX2^-/- mice were protected against streptozotocin (STZ)-induced beta cell apoptosis, leading to increased beta cell mass and preserved insulin secretion (800). In other experiments, very low-density lipoprotein (VLDL)-instigated NOX2-derived ROS caused beta cell dysfunction and apoptosis and reduced insulin production via JNK/Akt1 signaling (801, 802), which was rescued by NOX2 suppression.

On the other hand, accounts of the exposure of human- and mouse islets to a pro-inflammatory cytokine cocktail containing TNFα, IL1β, and IFNγ have invoked the role of NOX1 in beta cell dysfunction that is promoted by Src kinase activation; and those studies infer that NOX1 inhibition might be protective in that regard (803). Interestingly, in at least one study, global NOX4 deletion sets the stage for HFD-induced insulin resistance in mouse adipose tissue (804). The latter likely stems from the fact that NOX4 plays a central role in adipocyte differentiation (405). From these findings, one might speculate that adipose NOX4 deficiency under normal conditions could partially curtail differentiation and predispose for T2DM (582) and begs the question of whether partial or full loss-of-function of NOX4 in patients with NOX4 polymorphisms might exhibit overt or imperceptible signs of diabetes from an early age.

In ECs, NOX4 appears to promote insulin signaling and vascular angiogenesis, and global NOX4 deletion seemingly contributes to an accelerated loss in insulin sensitivity (805). In sharp contrast, adipocyte-specific NOX4 null mice exhibited a delayed onset of insulin resistance and inflammation during the development of obesity (806). This effect might be explained by a proinflammatory paracrine role of NOX4 in adipocytes, i.e. insulin resistance could be manifested via the recruitment of immune cells (793) and a result of suppressed insulin signaling caused by their secreted oxidants and inflammatory cytokines, such as TNFα, IL6, and IL8 (807). In summation, it seems that NOX4-derived ROS are important for proper adipocyte function under basal homeostatic conditions. However, NOX4 cellular re-localization, chemoattraction of leukocytes containing NOX2, and increased activity in obesity and metabolic disorders might trigger inflammatory responses and insulin resistance in adipocytes (582). Not surprisingly, feature clinical outcomes of diabetes include nephropathy, retinopathy, and neuropathy in which deleterious role of oxidation has been documented in plethora of studies (808).

4.10.4. Diabetic retinopathy

Diabetic retinopathy/neuropathy is a common complication of diabetes causing vascular and sensory neuron damage of the retina; and NOX1-derived ROS participate in retinal EC apoptosis in diabetic rats, an effect that is tempered by post-translational modifiers small ubiquitin like modifier 1, SUMO1, and/or ubiquitin conjugating enzyme E2, UBC9 (809). In high glucose-treated rat retinal ECs, NOX1 contributes to lasting mitochondrial oxidative damage even after reversion to normal glucose, which may be the instigator of what the authors dub “metabolic memory” in retinopathy (810). NOX2 is among the most highly investigated isoforms in diabetic retinopathy, wherein it has been found to contribute to microvascular dysfunction in the early stages of the disease (811). Adding to that, NOX2 upregulation is observed in retinas of diabetic mice, where it accelerates EC senescence resulting in a defective mechanism of vascular repair (812). Accordingly, NOX2 suppression and inhibition prevents this premature senescence in diabetic mice as well as in HG-treated ECs (812). Moreover, NOX2-derived ROS are ascribed at least partial culpability to high-mobility group box-1 (HMGB1)-elevated PARP1 and cleaved caspase 3 in the upregulation of retinal apoptotic markers in diabetic rats (813). Intriguingly, in studies further invoking key inflammatory responses in diabetic retinopathy, hallmark retinal hyperpermeability is significantly lower in STZ-induced diabetic mice lacking the 12/15-lipoxygenase gene compared with diabetic WT mice, and the pro-permeability effect of 12/15- lipoxygenase metabolite 12-HETE is significantly reduced in NOX2^-/- mice (814). From those findings, one might reasonably conclude that NOX2-derived ROS (811) and 12-HETE both linearly and independently play an augmentative role in retinal inflammation and hyperpermeability. On the other hand, suppression of NOX2 completely prevented blood-retinal barrier dysfunction in STZ-induced diabetic rats (638, 815) providing greater support for the linearity of NOX2:12-HETE signaling.

NOX4 oxidase, on the other hand, has been shown to be elevated in diabetic mouse corneal epithelia concomitant with damage (816). It is logical therefore that a close correlation between putative gain-of-function mutations in NOX4 and severe diabetic retinopathy has been revealed in patients with T2DM (817). Indeed, NOX4 expression at the protein and mRNA levels is upregulated in retinas of leptin receptor–deficient db/db mice (816, 818), as well as in human retinal ECs in response to HG (819). In support of its deleterious role in the retina, knockdown of NOX4 reduced retinal vascular leakage in db/db mice (816) and, in accordance with this observation, adenoviral expression of dominant-negative NOX4 abolished hypoxia- and HG-induced ROS and VEGF expression in human retinal capillary endothelial cells (816). Further corroborating these findings, in HG-treated human retinal ECs, inhibition of NOX4 or NOX4 siRNA significantly decreased cell death (819). Lastly and by extension, NOX4 in all likelihood promotes retinal neovascularization via an H₂O₂/VEGFR2/ERK signaling axis activated over the course of mouse retinopathy (820).

4.10.5. Diabetic nephropathy

Salient pathological features of diabetic nephropathy involve mesangial-cell proliferation, thickened glomerular basement membrane, glomerular hypertrophy, accumulation of ECM proteins, as well as chronic inflammation and oxidative stress caused by hyperglycemia (811, 821). Importantly, implications for the role of p22^phox-containing NADPH oxidases in human diabetic nephropathy is supported by meta-analysis showing that T allele carriers of p22^phox C242T gene polymorphism (rs4673) associate with overt diabetic nephropathy (822). Indeed, the authors of the study tacitly imply that this is related to a p22^phox gain-of-function in vivo (822). It stands to reason then that any one of NOXs 1, 2, 3 and 4 requiring p22^phox could be involved. However, in a type II diabetes model of db/db mice at least with respect to NOX2, lower gene expression was observed in the kidney cortex compared with their heterozygous littermates (725) although it is impossible to know from the study whether the decrease in NOX2 was a cause or consequence of type 2 diabetes mellitus. On the other hand, NOX4 oxidase could be the culprit as suggested in the same study by a rather unique NOX4 upregulation (vs. other NOXs) and its demonstrable functional role in fibrotic nephropathy (725). This notion has been corroborated in a very recent report in mice and humans (823).

Further, STZ-induced diabetic nephropathy in rats is accompanied by urinary albumin protein and glomerular histomorphic alterations that coincide with an increase in NOX4 levels and activity (824). Similarly, in a STZ mouse model, NOX4 transcript and cytochemical detection were significantly elevated in glomeruli of diabetic vs. control mice (825). Notably, these results were reproduced in vitro in mouse mesangial cells treated with high glucose for 72 h (825). On closer investigation of clinical relevance, Xu et al. identified increased NOX4 transcripts accompanied by the elevated number of NOX4-positive glomerular podocytes in diabetic patient kidneys (826). Intriguingly, a causal inter-relationship between NOXs and other oxidases in diabetes is corroborated by the demonstration that xanthine oxidase inhibition can suppress expression of NOX1, NOX2, and NOX4 mRNA and protein (827).

There is evidence for NOX5 playing a role in the disease as its expression is increased in human diabetic compared with nondiabetic glomeruli (828) and, stimulation of cultured human podocytes with AngII upregulated NOX5 expression and increased NOX5-dependent ROS generation (828). Interestingly, podocyte-specific knock-in of human NOX5 in mice elicited albuminuria, podocyte foot process effacement, and elevated systolic blood pressure. Moreover, treatment of these animals with STZ to induce diabetes exacerbated the phenotype (828). Similarly, mice with NOX5 expression directed to VSMCs and mesangial cells that were subjected to STZ-induced diabetes showed accelerated glomerulosclerosis, elevated albuminuria, macrophage infiltration and higher expression of profibrotic marker genes (718). More insights into the role of NOX5 in diabetic kidney disease come from studies by Jha et al. wherein the authors also discussed NOX4’s role in diabetic kidney disease, asserting that, in addition to its deleterious effects, NOX4 also has beneficial effects in other settings like vascular protection (719). In that regard, the authors challenge the wisdom of NOX4 being selected as a therapeutic target for diabetic complications (719). In fact, although GKT137831, the NOX1/NOX4 inhibitor proved to be protective in experimental model of diabetic kidney disease, it failed in clinical trials (829). Further to this issue, Jha and coworkers looked at the effect of EC NOX5 overexpression in the absence of NOX4 and demonstrated that NOX5 was independently capable of inducing the tell-tale indicators of diabetic nephropathy (719) and suggest that NOX5 may be a more effective therapeutic target in diabetic kidney disease.

4.10.6. Diabetic neuropathy

Diabetic neuropathy has been tested via ROS/NOX inhibition. Apocynin peculiarly blocked p47^phox and NOX2 expression (assembly did not appear to be tested) in STZ-induced rat diabetic neuropathy (830). That notwithstanding, apocynin decreased oxidative stress-mediated pathogenesis and pain in the sciatic nerve (830). Similarly, p47^phox and NOX2 expression were lower with curcumin treatment which alleviated neuropathic pain in STZ rats (831). In a model of high fat diet feeding in combination with STZ injection and right sciatic nerve blockade by bupivacaine, DPI prevented functional and neurohistological damage (832) underscoring a potential role for NOX2 oxidase in neural damage.

4.10.7. The thyroid

Examination of DUOXs in the thyroid is essential to appreciating their primary known biological function, thyroid hormone synthesis. Despite the presence of a peroxidase homology domain in DUOX enzymes (originally proposed to confer peroxidase activity), a requisite peroxidase reaction is mediated by an independent enzyme, namely, thyroid peroxidase, critical to the generation of thyroid hormone from thyroglobulin. The thyroid is the only tissue to our knowledge in the body wherein both DUOX1 and DUOX2 are expressed under physiological conditions and appear to work together in thyroid hormone synthesis (833, 834).

In support of functional interdependence of the two DUOXs in the thyroid, biallelic inactivating mutations in the DUOX2 gene results in total blockade of thyroid hormone synthesis that is connected with severe and permanent congenital hypothyroidism. Monoallelic mutations which are associated with milder, transient hypothyroidism (835). DUOX2 sequence variants associated with mild to severe hypothyroidism are described at greater depth in Chapter 4.16 and (150, 836–838). On the other hand, thyroidal DUOX1 and DUOX2 gene expression is upregulated in diabetic rats while in the rat PCCL3 cell line, glucose causes an increase in DUOX1 activity via PKC, pointing to a pathological role of this enzyme in the thyrocyte of diabetes patients (839). In the same study, NOX4 gene expression was increased concomitantly with DUOX1 and DUOX2 in diabetic rat thyroids and in a glucose-dependent manner and H₂O₂ generation was PKC dependent (839). In aggregate, while the short and long term of effects of elevations in these NOXs and their collective ROS, in response to glucose, is not well described to our knowledge, it would be logical to assume that they could contribute to thyroid oxidative stress, dysfunction, genomic instability (840) and even cell-cycle progression and cancer (841, 842). Indeed, a series of additional studies showed that NOX4 and p22^phox expression are elevated in thyroid cancers, connecting NOX4-dependent H₂O₂ generation to tumor onset and progression (833, 843). What’s more, NOX4 and Poldip2 transcript levels are higher in adult rat female thyroid glands compared to their male littermates (844) as well as in PCCL3 cells treated with 17β-estradiol. Importantly, this sexual dimorphism is not described in prepubertal thyroid glands, supporting a direct role of sex steroids in the expression of thyroid NOX4 (844). Nevertheless, in normal and cancerous human thyroids, NOX4 is shown not normally to be found on the apical membrane like DUOX but is wide ranging in location from the ER and mitochondria to the perinuclear/nuclear compartments of the cell. Although the significance of its presence in these subcellular locations has not been definitively addressed, its proximity to the nucleus has suggested an outsized role in normal and aberrant cell cycle control (78, 833, 841, 844, 845). Contrary to the above, basal and below basal DUOX2 expression levels are reported for the majority of thyroid cancer cells suggesting that DUOX2 may be involved in maintaining a differentiated airway epithelium (833, 846, 847).

4.11. Cardiovascular system

NOXs1–5, DUOX1–2, as well as p22^phox, p47^phox, p67^phox, p40^phox, Rac1/2, NOXO1, and NOXA1 are expressed in or more of the four major cell types of the cardiovascular system: endothelial cells (ECs) (848–853), vascular smooth muscle cells (VSMCs) (58, 850–852, 854), cardiomyocytes (855–859) and fibroblasts (49, 53). Although DUOX1 has been detected at low levels in VSMCs DUOXs have not to date been characterized to play a role in cardiovascular pathophysiology.

4.11.1. Neointimal growth & atherosclerosis

Vascular NOXs are upregulated during the early stage of atherosclerosis and have, therefore, been characterized collectively as an early trigger of the disease (860). Indeed, atherosclerosis is characterized by the rise in the residence of fat-laden macrophages, cholesterol and calcium concomitant with neointimal proliferation, fatty streaks (lesions) and plaque formation; and leads to a number of serious cardiovascular maladies such as coronary artery disease, stroke and peripheral artery disease (860). Interrogation of NOXs’ roles in the disease has been widely achieved in high fat diet (HFD) models. For example, ApoE^−/− high fat (cholesterol rich) diet-fed mice display elevated aortic mRNA and proteins levels of NOX1, NOX2, and NOX4 (compared with ApoE^−/− mice on a regular chow diet) along with O₂^·- generation, which purportedly all give rise to an increase in fatty vascular lesions (861). From a signaling standpoint, Janus-tyrosine kinase 2 (Jak2) inhibitor tyrphostin (AG490) reduces expression of each of the NOX subunits, overall NOX activity, and atherosclerotic lesions in the aorta of AG490-treated animals (861), suggesting a pivotal role of the kinase in the upstream control of the NOX in the disease.

Interestingly, moreover, NOX1^-/y ApoE^−/− mice exhibit ~20%-30% smaller lesions in a HFD model, in the ascending and descending aorta, while macrophage infiltration is decreased upwards of 50% consistent with reduced resident foam cell formation and a lipid core (862). In line with this, overexpression of NOXA1 enhances neointimal hyperplasia in a murine model of carotid injury and NOXA1 is increased in atherosclerotic lesions of ApoE^−/− mice and patients with the disease (863). Along similar lines, NOX1/NOX4 inhibitor GKT136901 attenuated ROS (864) and NOX1 deletion reduced neointimal development in response to arterial injury (418, 865). Moreover, NOX1 knockout mice infused with AngII are protected from aortic dissection in comparison to AngII-WT mice and reveal role for NOX1 in suppression of MMP inhibitor TIMP-1, which expectedly leads to increased MMP-9 activity and ECM turnover (866).

With respect to NOX2 oxidase, p22^phox and NOX2 expression are overtly higher and coincide with increased O₂^·- generation in atherosclerotic lesions, especially in the shoulder region of the plaque. Consistent with these findings, risk factors for atherosclerosis such as hypertension, oscillatory shear stress, and diabetes promote NOX2 expression in the vasculature (418, 867–870). Conspicuously, NOX2 expression (associated with elevated O₂^·-) is upregulated in aortic endothelia before the appearance of lesions. Whereas no change was observed in plasma lipids, reduced atherosclerosis was found in the descending aortae of HFD NOX2^−/y ApoE^−/− compared with WT ApoE^−/− mice (871). In another HFD atherosclerosis model, NOX2-selective inhibitor NOX2ds-tat inhibited NOX-derived O₂^·- caused a significant decrease in atherosclerotic plaques in aorta as compared to controls (872). Furthermore, increased carotid artery VEGF, HIF1α, MMP9 and adipokine visfatin expression in response to HFD were significantly attenuated by NOX2 inhibition which also decreased MMP9 protein expression and activity (872).

Interestingly, aortic SMCs from p47^phox-deleted, but not NOX2-deleted, mice displayed lower ROS and proliferation in response to growth factors in vitro (873). Furthermore, p47^phox deletion resulted in markedly lower lesion formation in the descending aorta. As such, the authors inferred that canonical NOX2 plays no role in atherosclerosis (873). However, in view of our current knowledge, the data appear to suggest that NOX1 could supplant NOX2 in a hybrid constellation with p47^phox (740, 873). Further to that point, a definitive verdict on the role of NOX2 will ultimately depend on careful assessment of temporal, cell-specific contributions of NOX2 protein.

As for the role of NOX4, comparison of ApoE^-/- and NOX4^-/- ApoE^-/- mice fed HFD revealed demonstrably distinct neointimal responses in the partially ligated carotid artery (874). In particular, the study invokes the role of TLR5-mediated NOX4 and inflammatory proteins in SMC migration to neointimal plaques in response to bacterial flagellar motility protein. After recombinant bacterial flagellar motility protein injection, ApoE^-/- HFD-fed mice presented with significantly augmented VSMC migration into the intimal layer of the carotid artery compared to ApoE^-/- mice on a normal diet and WT mice fed a normal diet. In contrast, SMC accumulation and inflammation observed in ApoE^-/- was obliterated in ApoE^-/- NOX4^-/- mice (874).

In a model of femoral arterial injury and neointimal growth on the other hand, NOX2 expression and ROS production were increased in WT mice, and attenuated in NOX2 nulls (875). Expectedly, NOX2^-/- exhibited reduced neointimal proliferation compared to WT mice in response to injury and reduced lesion formation in NOX2^-/- mice was attributed to both decreased cellular proliferation and leukocyte infiltration (875). In a model examining carotid artery injury, adenoviral-mediated knockdown of NOX4 in Zucker obese rats reduced the extent of SMC SERCA oxidation compared to lean rats and inhibited neointimal growth. Intriguingly, this was consistent with an ability for NOX4-derived H₂O₂ to oxidatively inactivate SERCA at cysteine 674, and, in turn, augment SMC migration (876). As such, the authors describe SERCA oxidation on C647 as pivotal to blocking NO’s anti-migratory function in these cells. On the other hand, contrary to this notion and suggestive of a protective role for NOX4, mice carrying a human NOX4 P437H dominant negative mutation in ECs display an increased vascular stiffness and extent of atherosclerosis. (877).

Additionally, NOX5 levels and its calcium-dependent H₂O₂-generating activity are markedly increased in coronary arteries from patients with coronary artery disease compared to healthy patients suggesting a potentially deleterious role for the isoform (878, 879). Contrary to this notion, NOX5^-/- rabbits, while displaying the same increase in plasma cholesterol as WT animals following an atherogenic diet, exhibited a significantly greater number of aortic plaques compared to WT (90). These findings suggested a protective role for NOX5 against atherosclerosis, albeit in young male rabbits (90). In humanized EC NOX5 knock-in, ApoE^-/- mice on the other hand, in which the human NOX5 gene was overexpressed in ECs, NOX5 expression did not alter atherogenesis or incidence of aneurysm. Rather, diabetes was found to be a determinative factor in these mice that exposed NOX5-related aneurysm but not lesion formation (880). Taken together, the overwhelming majority of reports to date appear to support the contribution of NOXs 1, 2, 4, and 5 to neointimal growth and atherosclerosis with minor exception. This potentially suggests that all of these NOXs either concurrently or disjointedly participate and that redundancy of signaling is at play much like that in other neoplastic diseases.

4.11.2. Cardiac remodeling, fibrosis and heart failure

NOXs have also been interrogated as the underlying causes of other cardiovascular pathologies involving cardiac remodeling, fibrosis, and heart failure. In that vein, NOX2 and NOX4 have long been shown as the most abundantly expressed NOX isoforms in the heart with cardiac NOX2 protein levels upregulated in response to AngII, aortic banding or myocardial infarction injury (881–886). Consistent with those findings, upregulated NOX2 appears to at least temporarily enhance cardiac contractility as well as exacerbate cardiac dysfunction in mice in response to AngII (886, 887). Indeed, clinical relevance is strengthened by the detection of high NOX2 expression in dilated left ventricles in human patients with myopathy and suffering from end-stage heart failure (882).

Importantly, elevated AngII (often a result of heart failure and reduced cardiac output) stimulates NOX activity, NFκB activation, interstitial fibrosis, matrix protein expression (as observed in cardiomyopathy) and MMP2 activity in WT mice, all of which are inhibited in NOX2^−/y (888). Intriguingly, even subpressor AngII (in WT mice) increased heart/body mass ratio, myocyte area, and cardiac collagen content, while these indicators were lower in NOX2 null mice (857). And, the demonstration that cardiomyocyte-specific deletion of Rac1 reduces O₂^·- levels and cardiomyocyte hypertrophy in response to AngII (889) is consistent with NOX2 oxidase involvement. In aggregate, studies indicate that NOX2 oxidase is pivotally involved role in AngII-induced cardiac interstitial fibrosis and remodeling in cardiac failure. Furthermore, despite no apparent differences in infarction size, NOX2^-/- mice with permanent left coronary ligation at least in one study appeared to benefit from decreased post myocardial infarction tissue remodeling compared to WT (883, 890). When it comes to pressure overload, NOX2 deletion improved cardiac interstitial fibrosis and left ventricular function (compared to WT), but had no bearing on myocardial hypertrophy and cardiac mass (855, 891, 892). However, NOX2 does appear to play a role in cardiac interstitial fibrosis and thus be a possible cause of contractile dysfunction as a result of increased afterload.

In contrast, considerable controversy has surrounded the participation of NOX4 in cardioprotection in response to increased afterload as reported by two independent laboratories. Indeed, in 2010, the Shah laboratory observed that global or total body NOX4^-/- mice experience exacerbated cardiac dysfunction in comparison with WT in a model of aortic constriction-induced pressure overload, cardiac hypertrophy, contractile dysfunction, and ventricular dilatation along with impaired myocardial angiogenesis (893). This effect seemed to be due to NOX4-dependent activation of HIF1 and VEGF pathway (893). One might argue that the effect of global nullification of NOX4 in this study was confounded by a potential paracrine role of NOX4 from non-cardiomyocytes. That notwithstanding, in the same study, cardiomyocyte-specific NOX4 overexpressing mice displayed lessened cardiac dysfunction and thus were protected from these effects (893).

Meanwhile, that same year, Sadoshima and colleagues reported that pressure overload-induced cardiac hypertrophy and contractile dysfunction are lessened in cardiomyocyte-specific NOX4^-/- but exacerbated in cardiomyocyte-NOX4 overexpressing mice (894). The fact that the study examined both suppression and overexpression in the cardiomyocyte, per se, argues for a credible phenotypic effect of NOX4 at least from the perspective of a role for the cardiomyocyte. The discrepancies between the two perceptibly conflicting studies could be ascribed to one of multiple variants between the studies including mouse genetic backgrounds (C57BL/6 vs. FVB), the age of the mice at the time of experimentation or the extent to which the mice were in cardiac failure.

4.11.3. Hypertension

Systemic hypertension is one of the best-studied diseases implicating NOXs (895). A plethora of rodent models have been employed to this end including AngII infusion models, deoxycorticosterone (DOCA)-salt, dexamethasone-treated, DAHL-salt sensitive or spontaneously hypersensitive rats.

In mineralo- and glucocorticoid models of hypertension, p22^phox levels and O₂^·− generation in murine and rat aortae are increased and indicate a role for one of NOXs 1 through 4 (84, 896, 897). In mineralocorticoid DOCA salt-treated rats, the use of apocynin to lower blood pressure initially hinted at a role for NOX2 in systolic blood pressure maintenance; but growing evidence over the years supporting broad-spectrum NOX inhibition by apocynin made it inconclusive as to which NOX, per se, was involved (896). Indeed, in mice treated with dexamethasone, p22^phox deficiency partially ameliorated afterload-induced left ventricular dysfunction (84). Incidentally, glucocorticoid treatment-induced systemic hypertension is paralleled by increased pulmonary mRNA and protein levels of p22^phox, NOX2 and NOX4 (84); and in line with those findings and indicative of a role of vascular NOX, smooth muscle-specific overexpression of p22^phox in mice elicited an increase in blood pressure in response to AngII that was reversed by ROS scavenger ebselen (898).

Beginning with NOX1 oxidase, the subtype has been shown to be upregulated in distinct hypertension models, such as renin transgenic- (708) and 2-kidney 2-clip rats (418, 899). Testing causality, AngII infusion elicited an initial spike in pressure (up to 3–5 days) that was unaffected by NOX1 deletion whereas considerably lower sustained elevated pressure in NOX1 nulls compared to WT pointed to a role of NOX1 in blood pressure maintenance over extended periods (900, 901). Importantly, in mice with SMC NOX1 overexpression, AngII caused a potentiated increase in O₂^·−, systolic hypertension and aortic hypertrophy compared to littermate controls (902). Furthermore, endothelium-dependent relaxation of aortas was restored to normal in those taken from NOX1 knockout mice infused with AngII compared to AngII-treated WT, establishing a role of NOX1-derived O₂^·− in AngII-induced endothelial dysfunction (900). Additionally, treatment of 12- to 14-month-old spontaneously hypertensive rats exhibited marked elevations in NOX1 and impaired endothelium-dependent vascular relaxation (vs. WKY) and application of a pan-NOX inhibitor VAS2870 improved the relaxation response beyond that of basal WKY responses (903).

As for NOX2, its levels have been demonstrably augmented in the vascular wall in multiple forms of hypertension and in multiple studies (56, 361, 418). As such, Cohen and colleagues showed in NOX2^-/- mice that although basal blood pressure was reduced, AngII-induced pressure elevations remained unchanged (361). Strikingly, Harrison and colleagues showed that in mice lacking the canonical p47^phox subunit of NOX2, hypertension and aortic O₂^·− production in response AngII infusion were entirely abolished (59). In yet another study, in contrast, Touyz and coworkers report that NOX2 knockout mice developed AngII-dependent hypertension (904). On the face of it, admittedly the overt difference between the latter two studies knocking down two essential subunits of equal importance in the canonical NOX2 (p47^phox and NOX2, respectively) may seem perplexing. However, this difference may be explained by a possible involvement of p47^phox in a NOX1 hybrid system (utilizing p47^phox in lieu of NOXO1) in the study by Harrison rather than the canonical (or hybrid) NOX2 oxidase in the study by Touyz. In contrast, evidence in support of the canonical NOX2 oxidase was provided by another study. That is, systemic delivery of selective NOX2 inhibitor NOX2ds-tat, that selectively disrupts the pivotal interaction of p47^phox with NOX2, was shown to attenuate vascular ROS and to delay, but not attenuate maximal, systolic blood pressure elevation in response to AngII hypertensive mice (57). Perhaps the latter study suggests that p47^phox-NOX2 is involved in the development of blood pressure but not in its sustenance. Intriguingly, in Dahl salt-sensitive rats infused with NOX2ds-tat, O₂^·− production and endothelium-dependent relaxation are normalized, however, blood pressure is unchanged (905). From a slightly different perspective, endothelium-specific overexpression of NOX2 did not augment basal blood pressure, but exacerbated diastolic blood pressure rises in response to high dose AngII (69). It is tempting to speculate then that NOX2, per se, is not essential to basal control of blood pressure under normal conditions but reveals its involvement when physiological levels of NOX2 are on the rise.

Hypertension models are yet one more example for the inconsistency regarding NOX4 knockout mice in the literature. For instance, one study showed that NOX4 knockout mice display an attenuated increase in blood pressure compared to their WT littermates following an infusion of AngII (906). In that study, NOX4 appeared to be deleterious because its suppression limited the pressure response to the agonist (906). However, that finding runs contrary to the work by Schröder et al. in which employment of an inducible deletion of NOX4 revealed no difference in basal or AngII-induced blood pressure between NOX4^-/- and WT (907). On the other hand, in still another study in which NOX4 was overexpressed specifically in endothelial cells, decreased AngII-induced hypertension was observed which the authors attributed to increased H₂O₂ levels, vascular depolarization and vasodilatation (908). This notion is congruent with the authors’ proposed and previously described vasodilatory effect of H₂O₂ (908–910).

The study of NOX5 in traditional animal models of hypertension has been challenging because of its lack of expression in rodents. Still, its likely importance in the development of human hypertension cannot be overestimated (911). Human endothelial cells do express NOX5 that is activated by AngII (912). In fact, genome-wide association studies for hypertension risk genes point in particular to NOXs 4 and 5 (913). In agreement with this, NOX5 has been shown to be robustly upregulated considerably in renal proximal tubules cells of hypertensive patients as compared to normotensive subjects, and the marked increase in NOX5 is much larger than for NOX1, NOX2 or NOX4 in the same patients (722). Also, an increase in NOX5 in circulating endothelial microparticles from hypertensive subjects has been reported (914). For an experimental assessment of NOX5 in hypertension, NOX5 humanized knock-in mice models have been employed. In one scenario, expression of NOX5 specifically in endothelial cells caused a marked increase in systolic- and mean arterial blood pressure in aged knock-in mice compared to WT controls, while no difference was observed in young counterparts (914). Similarly, knock-in expression of human NOX5 in podocytes caused an increase in systolic blood pressure compared to littermate controls (828). Figure 7 schematizes some of the NOX pathways described above.

Figure 7. NOX signaling in vascular tone maintenance and diseases of the cardiovascular system.

Stimuli such as hypoxia, AngII or a high fat diet increase expression and enzymatic activity of NOX1, NOX2 and NOX4. O₂^·- production from NOX1 and NOX2 decreases nitric oxide (NO) availability causing an increase in vascular tone. Furthermore, increased oxidative stress leads to activation of HIF1 and its target genes such as VEGF, a master regulator of proliferation that underpins vascular remodeling seen in many vascular pathologies. NOX1 and NOX2 and NOX5 also acting via the activation of NFκB and decreased NO promote pro-inflammatory effects on vascular tissue leading to vascular remodeling, neointimal growth and atherosclerosis. Distinctly, NOX4 is generally deemed vasculoprotective in part the consequence of H₂O₂’s vasorelaxant properties and, in part, via activation of Nrf2. NOX5, activated through calcium-binding and acting via redox activation of kinases such as c-Src is involved in vascular hypercontractility and vascular dysfunction in hypertension and atherogenesis. A combination of more than one NOX is observed in cardiovascular diseases. MMP2/9: Matrix Metalloproteinase 2/9; NFκB: Nuclear Factor Kappa B; VEGF: Vascular Endothelial Growth Factor; HIF1: Hypoxia inducible factor 1. c-Src: Proto-oncogene, non-receptor tyrosine kinase Src.

4.11.4. Pulmonary Hypertension

Pulmonary hypertension (PH) is a disease characterized by pulmonary vascular hyperplasia and remodeling concomitant with a sustained increase in mean pulmonary artery pressure (>20 mmHg) that leads to right ventricular hypertrophy (RVH) and can result in right ventricular failure (915). Assuredly, PH has been studied extensively in connection with NOXs. Pulmonary VSMCs are shown to express NOX1, NOX2 and NOX4 (p22^phox-associated NOX isoforms) in PH (916, 917). It follows logically then that mice with a loss in p22^phox are protected against the development of hypoxia-induced PH (918). That is, the findings implicate the participation of any one or more of the above-mentioned NOX isoforms that utilize p22^phox. In the context of PH, p22^phox promotes vascular proliferation, migration and capillary formation in an exaggerated propulsion of aberrant vascular remodeling and plexiform lesion formation (918). Concordantly, mice deficient in p22^phox protein were protected from these effects (84).

In a model of persistent PH in lambs (PPHN), elevations in pulmonary artery pressure were induced by ante-natal ligation of the ductus arteriosus at 128 days gestation. Intriguingly, increased pulmonary p22^phox and NOX4 were detected along with increased H₂O₂, thiol oxidation and decreased extracellular SOD in isolated lungs, pulmonary arteries and/or PASMCs from PPHN lambs compared to controls (919). In vitro studies of PPHN PASMCs substantiated the role of NOX4/p22^phox in NFκB and cyclin D1 activation consistent with proliferation and overall vascular remodeling of pulmonary arteries.

In one of the initial studies exploring the role of NOX1 in hypoxia-induced PH, Mittal et al. (920) show no difference in the expression levels of NOX1, NOXO1 and NOXA1 in lung lysates from WT mice exposed to 3-wks hypoxia (10% O₂). On the other hand, increased pulmonary vascular resistance and PAP and RVH arose in NOX1-/γ mice from 9 to 18 wks old consistent with a repressor effect of NOX1 during PH development (921). Primary PASMC from the same NOX1^-/y mice also showed decreased vascular cell apoptosis consistent with increased vascular hyperplasia and remodeling (921). Finally, in piglets, the data are correlative and not causal in nature. Thus, it is not clear what an observed increase in NOX1 expression and reported membrane localization of p67^phox at 10 days of hypoxia signifies. One might deduce that the NOX1 hybrid oxidase is activated but can only speculate as to whether the active complex propagates or arrests PH (922). Perhaps, one can deduce instead that the NOX1 hybrid oxidase is causal at other timepoints of the study that were not examined. Paramount, perhaps, to all of the above, are findings in human tissue and cells. That is, in vessels from PAH compared with non-PAH patients, upregulation of NOX1 at the transcript and protein level was discovered (417). Upon further inquiry in human pulmonary ECs in vitro in the same report, NOX1 inhibition disrupted pro-proliferative signaling exposed to hypoxia consistent with a proposed pro-proliferative role of NOX1 in PH (417). Thus, the data in humans appear to supersede the controversy of whether or not NOX1 (canonical or hybrid) is fundamental to PH and come down in favor of a causal role for NOX1, thereby warranting further studies as to the mechanisms involved (416).

Pulmonary arteries from subjects with idiopathic pulmonary arterial hypertension (iPAH) also express NOX4; however, NOX4 transgenic loss-of-function and gain-of-function animal models offer mixed results. For one, Mittal and colleagues report that NOX4 does not participate in vascular remodeling associated with the development of hypoxia-induced PAH in piglets (920). On the other hand, NOX4 was detected in three distinct rat PAH models in the endothelium and adventitia of pulmonary arteries; and NOX4 inhibitors improved monocrotaline-instigated rat PAH in both prevention and reversal protocols ameliorating right ventricular hypertrophy and non-invasive indices of pulmonary artery stiffness (923). Furthermore, this study reported changes in RV function as evidenced by elevations in RV systolic pressure and RV_max dp/dt that were eradicated by NOX4 inhibition. Moreover, NOX4 inhibition decreased RV thickness, ameliorated cardiac output, and increased velocity time integral, pulmonary ejection time and pulmonary artery acceleration time (which collectively indicate a vessel stiffness reduction) and enhanced cardiac performance (923). Opposing these findings are studies showing that NOX4 does not play a role in hypoxia-induced PH. One study showed that neither global constitutive nor conditional NOX4 knockout altered the response to acute/sustained- or chronic hypoxia in mice (924, 925). In aggregate, some of the differences observed among the studies might be attributed to the distinct genetic backgrounds of the animals employed.

4.12. The Respiratory system

4.12.1. Lung physiology and homeostasis

Aside from their role in the pulmonary vasculature, NOXs are instrumental in the physiology and pathophysiology of the respiratory airways and gas exchange. At the focal point of this vital function are epithelial cells lining alveoli comprised of thin squamous type I (90–95% of the epithelium) and small cuboidal type II pneumocytes (926). Efficient gas exchange in alveoli depends on the clearance of the alveolar fluid via epithelial sodium channel (ENaC) which is expressed on the surface of both cell types. The predominant NOX isoforms expressed in lung epithelium include NOX2, NOX4 and DUOX1/2 (375, 927) and fluid exchange has been shown to be NOX2/Rac1- and ROS-dependent and by extension stimulated ENaC activity is inferred to be the agent of this exchange in rodent epithelial cells (928). Specifically, NOX2 deletion as well as pharmacologic inhibition of NOX2 oxidase with Rac1 inhibitor NSC23766 increased in vivo alveolar fluid retention and impaired fluid clearance upon treatment with LPS implying a protective, physiologic role for NOX2 via modulation of ENaC (928). Concomitantly, a marked increase in LPS-induced ENaC activity (~5-fold) was eradicated with TEMPO consistent with that role (928). In aggregate, these data suggest that NOX2 plays a fundamental role in normal lung fluid balance under normal conditions. On the other hand, when the system is stressed as with chronic alcohol exposure, NOX2 and Rac1 upregulation can result in ENaC hyperactivation and a thickened alveolar mucus linked with acute respiratory distress syndrome (929, 930). Thus, in sum, NOX2 appears to mediate a biphasic pathophysiological role in alveolar fluid clearance and homeostasis.

In addition, NOX2 can regulate the cell cycle in alveolar epithelial cells. That is, PPARγ induction of NOX2-derived O₂^·− was shown to promote PCNA, cyclin D1 and cell cycle progression from G₀/G₁ into S and G₂/M phases (931). In that sense, restoration of an epithelial barrier after injury might be controlled by NOX2 or NOX4. Indeed, owing to NOX4’s central role in proliferation but also differentiation of myriad cell types (687, 932), it is tempting to postulate that NOX4 is pivotal in the renewal and differentiation of epithelial cells after injury. Often, moreover, considered the third cell type in the alveolus, and playing a role in innate immunity, the alveolar macrophage is critical to host defense in the lung (933). When external factors disrupt the normal balance and innate function of macrophages, such as in alcoholics, it predisposes patients to microbial infection (934). With respect to the latter, ethanol causes a reduction in phagocytosis but increases macrophage ROS generation and offsets the integrity of the alveolar barrier determined by epithelial health and composition (933–936). Coincidentally, ethanol has been shown to markedly increase NOX1 and NOX4 in addition to NOX2 in alveolar macrophages (935) which together can generate large amounts of H₂O₂, induce cytokine release, and decrease phagocytic activity. Predictably therefore, this could hasten sodium reabsorption and dehydrate the luminal surface of the lung resulting in reduced clearance of bacteria.

H₂O₂ generated by DUOX1 in type II epithelial cells reportedly regulates acid release in the process of mouse lung development (937, 938). As well, the maturation of type II cells of mice and humans in vitro as evidenced by the production of H₂O₂, transepithelial resistance and acid release is suggested to occur via the induction of DUOX1 and its maturation factor DUOXA1 (937, 938). In the adult organism, DUOX1 & 2 are expressed at the apical pole of ciliated cells of the trachea and in alveolar type II cells where they produce H₂O₂. In the presence of luminal lactoperoxidase in airway cells, H₂O₂’s reaction with thiocyanate (SCN^-) is catalyzed to generate a potent anti-microbial metabolite - hypothiocyanite (OSCN^-) (939). Parenthetically, it appears to be the predominant heme peroxidase in the alveolus though myeloperoxidase and eosinophil peroxidase can also be found in the local milieu in neutrophils and eosinophils (937). In addition to OSCN^-, the airway surface liquid that lies between the epithelium and inhaled air contains mucins which effectively trap pathogens and particles thus protecting the lower airway (940). Release of mucin-5 subtype AC (MUC5AC) from human bronchial epithelial cells purportedly depends on DUOX1 induction [by way of PKC-delta/PKC-theta (PKCδ/PKCθ) activation] and its effects on neutrophil elastase (941). Additionally, the airway surface liquid anti-microbial characteristic depends on its pH level maintenance and release of H⁺ from tracheal epithelial cells is shown to be mediated by an extracellular H₂O₂-dependent mechanism concomitant with the expression of DUOX1/2, p22^phox, p40^phox, p47^phox, and p67^phox (942). On the other hand, NOX2 silencing in human bronchial epithelial cells or inhibition using DPI inhibits NFκB activation and cytokine synthesis following respiratory syncytial- and the Sendai virus infection suggesting manipulation of NOX2 is pivotal to modulating the inflammatory response (943). Notably, it has been suggested that NOX2, via its proinflammatory actions, is indirectly involved in mucin release (375). Meanwhile, NOX1-derived ROS are described to disrupt the zona occludens of the apical tight junctions adding to barrier dysfunction and facilitation of the transmigration of rhinoviruses (944).

4.12.2. Lung injury

Acute lung injury and its most severe form, acute respiratory distress syndrome, normally unfolds in three phases: the acute, proliferative, and fibrotic phases (945). The acute phase follows the initial injury, wherein the most alveolar septal cells undergo cell death either in the form of necrosis or apoptosis, while inflammatory cells invade the site of the injury (945). One of the most common acute lung injuries and injury models involves LPS-induced inflammation (946), and inquiries into the role of NOXs following LPS-exposure have yielded inconsistent results. On the one hand, for instance, the absence of p47^phox, or NOX2 (while inhibiting lipid peroxidation and O₂^·− generation, respectively) had no bearing on LPS-induced lung damage (947) or pneumocytis-induced pulmonary lung injury (948). In contrast, deletion of p47^phox or NOX2 enhances LPS-induced inflammatory gene expression and is associated with increased neutrophil recruitment to lung and heart and, in turn, an unabated immune response (949, 950). Thus, NOX2 oxidase at least in this scenario appears to play a fundamental role in limiting LPS-induced inflammation. As for signaling, LPS binds to LPS-binding protein forming a complex that activates TLRs (TLR4- and 2), submembranal interleukin-1 receptor-associated kinase-4 and phosphorylation of p47^phox of NOX2 oxidase in neutrophils, macrophages and other inflammatory cells, thereby activating NFκB and MAPK and the transcription of inflammation-related genes collectively important to mounting an effective immune response (273) (Figure 8). For more information on the NOX pro-proliferative response, see section 3.1.

Figure 8. NOX activation, signaling and lung disease phenotypes.

Damaging stimuli, such as microbes, cigarette smoking and pollution activate distinct NOXs beyond their physiological signaling and stimulate their participation in the development of various lung diseases including lung injury, fibrosis, COPD and asthma. NOX2 is reportedly fundamental to the activation of ENaC channels and alveolar fluid clearance by epithelial cells and thus is deemed homeostatic in the lung. On the other hand, NOX2 can effect lung injury via NFκB-mediated pathways. Lung fibrosis is under the effect of NOX4 expression. COPD involves the participation of NOX1, NOX2 and NOX4 and part of that pathology is mediated via the suppression of SIRT1 and disinhibition of MMP9. Asthma is brought about by NOX2, NOX4 and DUOX1 through a variety of factors including CK2α-mediated NFκB as well as, in the case of DUOX, EGFR and activation of pro-inflammatory cytokines. While NOX5 has been correlated with a number of these disorders, no studies to date appear to have demonstrated causality in in vitro or in vivo models. COPD: Chronic obstructive pulmonary disease; CK2 α: casein kinase 2α; SIRT1: member of the sirtuin family; MMP9: Matrix Metalloproteinase 9; EGFR: epidermal growth factor.

4.12.3. Lung fibrosis

NOX-derived ROS in general are implicated as major players in fibrotic processes in multiple organ systems (951, 952). Pulmonary fibrosis is a pathology of distinct etiology that results in increased lung tissue scarring, due to exchange of alveolar epithelium with migrating connective tissue cells – myofibroblasts (953). In turn, these cells do further harm to the epithelium and contribute to an exaggerated pro-fibrotic response. To our knowledge, the best characterized NOX isoform in the lung tissue of patients with idiopathic pulmonary fibrosis is NOX4 (386, 954, 955). Indeed, NOX4 expression is augmented in alveolar type II cells following lung injury, while NOX4-derived ROS instigated apoptosis of alveolar epithelial cells, replacement by myofibroblasts and the promotion of lung fibrosis (956). NOX4 that is readily expressed in human lung fibroblasts has been proven necessary and sufficient to induce fibrosis in a variety of lung fibrosis mouse models (954). To that point, TGFβ1 induces NOX4-derived ROS that in turn promote fibroblast migration (957) and mediate pro-fibrotic myofibroblast phenotypes regulating their differentiation, contraction, and ECM deposition (954). Indeed, myofibroblasts seem to be critical cell effectors of pulmonary fibrosis, due to their secretion, deposition, and remodeling of the ECM and formation of scar tissue.

Logically, pulmonary fibrosis is also an age-exacerbated disease since the increase of senescent and apoptosis-resistant myofibroblast phenotypes promote it (473). Importantly, both pharmacological inhibition as well as genetic targeting of NOX4 in aged mice reverse persistent fibrosis through the induction of myofibroblast apoptosis (473). Furthermore, induction of NOX4 that is not offset by an adequate antioxidant response exacerbates the process (473). In rodent models, NOX4 depletion or NOX1/NOX4 inhibitor mitigate the fibrotic response (386, 954). In addition, metformin attenuates TGFβ1–induced NOX4 and myofibroblast differentiation in lung fibroblasts in vitro. It stands to reason, therefore, that metformin attenuates bleomycin-induced lung fibrosis in this manner in vivo (958). Finally, myofibroblasts that arise in response to both mechanical tension and TGFβ1 stimulation, are key cellular contributors to pulmonary fibrosis (959, 960) (Figure 8).

4.12.4. Chronic obstructive pulmonary disease

Airway smooth muscle cell remodeling and alveolar emphysema are characteristics of chronic obstructive pulmonary disease (COPD) commonly triggered by environmental stressors such as cigarette smoking (961). In the lungs of control human subjects, NOX1 is identified in alveolar endothelial cells and macrophages, while it is absent in epithelial cells (961, 962). However, NOX1 appears to emerge in both bronchial and type II epithelial cells in emphysematous lungs (961). NOX2, in contrast, is detected in alveolar and interstitial macrophages in both control and emphysematous lungs (961). In agreement with and adding to those findings, in a model of acute cigarette smoking (ACS)-exposed mouse lungs, NOX1, NOX2, and NOX4 proteins were increased indicating the role of the three NOXs in the early-stage COPD (963). Although NOX1, NOX2, or NOX4 depletion decreases mouse lung ROS in response to ACS exposure (963), only NOX1 or NOX4 depletion seems to prevent inflammatory cell influx and protect against lung inflammation (963). In humans, NOX1, NOX2, NOX4, and NOX5 protein expression are markedly increased in the lung tissue sections of end-stage COPD patients, with NOX1, NOX4, and NOX5 being detected in bronchial epithelial-, alveolar epithelial-, and vascular cells, as well as macrophages. In this milieu, solely NOX2 was evident in both macrophages and neutrophils but the oxidase was also detected at low levels in bronchial epithelium and alveolar epithelium of normal or COPD lungs (963). No tests of causality of NOXs 1 and 5 in COPD were revealed yet a few recent studies support a role for NOX4 in inflammation, mitochondrial damage mitophagy and collagen deposition in the disease (80, 85, 963–965). The role of NOX2 in COPD is still unclear as one study early on suggested that NOX2 deficiency sensitizes mice to the disease (966) whereas another implicates NOX2 as causal in COPD-related inflammation (967). Finally, one report to our knowledge, indicates a possible role for NOX3 in age-related mouse emphysema (161). Thus, multiple additional studies in animal models and humans will be required to elucidate this predictably complex disease landscape (Figure 8).

Continuing on this theme, an increased number of NOX2⁺ macrophages align with increased epithelial NOX2 and NOX1 transcripts in human emphysematous lungs (961). In mice instilled with elastase, NOX1 and NOX2 mRNA was elevated acutely compared to controls (961). However, elastase-induced alveolar airspace enlargement together with elastin degradation were reported to be lower in NOX2- but not NOX1-deficient mice (961) and the described protective effect was characterized as free of inflammatory processes correlating with reduced ROS levels (961). In the same study, NOX2 deficiency led to an increase in SIRT1 and a decrease in MMP9 levels and activity due to an decreased histone acetylation at AP1, NFκB and Pea3-binding sites on the MMP9 promoter (961). Importantly, the functional deletion of NOX2 brought decreased elastase-induced ROS production and protected against emphysema. Corroborating this finding, ROS production and emphysema were reinstated by specifically overexpressing p47^phox in p47^phox-depleted macrophages. In sum, p47^phox is involved in the development of human emphysema and macrophage NOX2 oxidase overall appears to participate in emphysema through its suppressor effect on SIRT1 and related disinhibition of MMP9 in mice (961).

Adding to those findings, WT mouse lungs exhibit significantly increased O₂^·− after 2- and 6 h following a one-hour exposure to acute cigarette smoking (ACS) (968). Similarly, in human small airway ECs (HSAECs) in vitro, O₂^·− production is increased in a time- and dose-dependent manner following cigarette smoke extract (CSE) concomitant with NOX1, NOX2, NOX4 and NOX5 upregulation (968). Notably moreover, nuclear translocation of p65, IL8 release, and induction of COX2 in CSE-treated HSAECs are all sharply elevated indicative of the expected inflammatory response. Furthermore, ACS- or cigarette smoke exposure-stimulated O₂^·− generation and expression of inflammatory mediators were decreased by putative O₂^·− scavenger TEMPO in vivo and vitro consistent with a role for NOX1, 2 and 5 and perhaps even NOX4 in this response (968). Concurrently, ACS induced significant elevated infiltration of neutrophils and macrophages in mouse lung parenchyma and bronchioalveolar fluid (968) implicating an exaggerated vicious cycle at play between ECs and phagocytes. Thus, the data suggest a complex orchestration of the disease triggered by cigarette smoke involving multiple cell types and a panoply of NOXs and inflammatory pathways that lead to elastin degradation and tissue breakdown in COPD and emphysema.

Intriguingly, on the other hand, a feed-forward protective effect may be triggered by the injury. That is, in a murine emphysema model, exposure of WT mice to CS for 6 months or injection with elastase led to a significant expansion of myeloid-derived suppressor cells in the cigarette smoking-affected respiratory and gastrointestinal tract (969). Thus, the lungs appear to have some capacity to protect against the injurious effects of CS in the short term in response to a moderate insult.

With respect to the DUOXs, downregulation of DUOX1 has been revealed in the airways of COPD patients making it a probable participant of COPD pathogenesis. Thus, one might surmise that an insufficiency of DUOX1-mediated innate injury responses participates in the disruption of epithelial homeostasis (970). As such, small airway DUOX1 levels are reduced in lungs of patients with progressed COPD and correlate with loss of function and markers of emphysema (970). Corroborating those findings, DUOX1 downregulation correlating with ECM remodeling is observed in a genetic model of COPD, transgenic SPC-TNF-α mice (970). Further, subepithelial airway fibrosis in the cigarette smoking model or in alveolar emphysema induced by elastase are both worsened in DUOX1-deficient mice (970). In another study in 2008, the variable expression of DUOX1 and DUOX2 in the lungs of smokers depended on the anatomical location and whether they were current or past smokers. In subjects who were current smokers bronchiolar and tracheal epithelial DUOX1 was downregulated as compared to DUOX2, which was on the other hand upregulated compared to non-smokers (483). However, subjects who were former smokers and had mild or moderate COPD displayed decreased expression of both isoforms (483). On the contrary, alveolar expression of either was low and not affected by smoking or COPD status (483). Unfortunately, in that report (483), no causality was tested either in biopsies or animal models. Clearly, the verdict is still out on the individual roles of DUOXs in distinct cell types in COPD and cell-specific depletion and/or overexpression of DUOXs 1 or 2 in each of the three cell types will be necessary to solve the controversy.

4.12.5. Asthma

Asthma is characterized by chronic inflammation of airways and lungs that includes airway obstruction, hyperresponsiveness, and remodeling. In vitro, significant upregulation of NOX4 has been observed when human lung epithelial cells were stimulated with IL13, and this effect was concomitant with mitochondrial damage-induced apoptosis, and NLRP3/IL1β activation (971). In a murine model of asthma, animals were passively sensitized with i.p. ovalbumin, followed by an airway challenge with ovalbumin (972). The second challenge increased casein kinase 2α in airway epithelial cells and activation of NFκB consistent with the asthmatic phenotype; these responses were significantly impaired in NOX2^-/-, but not in NOX4^-/-, mice (972). In contrast, induction of asthmatic phenotypes through enhanced Th2 differentiation and function has been reported in NOX2-null mice independent of dose and route of different allergens (563). NOX2-deficient T cells also demonstrated hyperactivation, reduced airway-induced cell death or apoptosis, and diminished Treg-cell differentiation through increased Akt phosphorylation and enhanced mitochondrial ROS. Importantly, NOX2 deficiency may result in exacerbated experimental asthma that is a cause of an enhanced Th2 effector function in a T-cell-intrinsic manner (563). In aggregate, findings to date suggest that considerably more work needs to be done to tease out the role of NOX2 in the disease.

In vitro or in vivo allergen challenge results in fast airway epithelial IL33 secretion, dependent on DUOX1 induction of EGFR and the protease calpain-2 via a redox-dependent cysteine oxidation (973). DUOX1 deficiency decreased several EGFR-dependent features of house dust mite-induced allergic airway inflammation, neutrophil infiltration, type 2 cytokine production (IL33, IL13), mucous metaplasia, subepithelial fibrosis, and central airway resistance. Intratracheal administration of DUOX1-targeted siRNA or pharmacological NOX inhibitors reversed most of these outcomes (974). Thus, DUOX1 appears to play a seminal role in the asthmatic inflammatory and airway constrictor response.

Moreover, concomitant increases in DUOX1 and IL33 expression detected in primary nasal epithelial cells patients suffering from allergic asthma align with augmented IL33 secretion following allergen exposure compared to non-asthmatic subjects (973). Additionally, induction of type 2 cytokine responses by direct airway IL33 administration was associated with ST2-dependent activation of DUOX1 and activation of Src and EGFR. This response was decreased in Duox1 ^-/- and Src ^+/- mice implying that IL33-mediated epithelial signaling in the epithelium and subsequent responses in the airways involve DUOX1/Src-dependent pathways (975). (Figure 8).

4.13. Central nervous system

NOXs’ role in CNS homeostasis and neuronal development have been extensively described in recent years with NOXs 1 – 5 and DUOX1 and DUOX2 implicated in neurogenesis and NOX2 largely relegated as being chiefly expressed in bone-marrow-derived macrophages and microglia and as regulating axon growth and regeneration (77, 976–979). Despite this bias, it has become increasingly clear that NOX2 is ubiquitous in the central nervous system, is involved multi-factorially in neural cell growth and physiology and is instigator in a wide range of nerve disorders from dyskinesia and dementia to major depression (980–984). Likewise, various NOXs appear to play a fundamental role in neuronal stemness and differentiation with major focus being placed on NOX4 in this function (464, 499, 503, 504, 978, 985–988). For an in-depth survey of NOXs in the CNS the reader is referred to (77, 976) and below.

In the brain’s vascular system even, cerebral arteries, like system-wide corporal arteries, are thought to notably employ NOX in housekeeping and homeostatic functions. These include the preservation of passive tone and patency, myogenic tone (989), vasodilatory properties (908), self-renewal (464), healthy remodeling of the vessel wall, and angiogenesis (449) among multiple other biological processes (94, 990).

4.13.1. NOX localization in the nervous system

NOX1 is reported in the cerebral cortex, thalamus, hippocampus, cerebellum, substantia nigra, ventral tegmental area and striatum (503, 991–993). NOX2, in addition to these, is detected in amygdala, periventricular nucleus, and hypothalamic supraoptic nucleus, the locus coeruleus, and the subventricular zone across rodents (77, 991, 994–998). By cell type, NOX1 is found in astrocytes, microglia, neurons and neurovascular cells, whereas NOX2, in addition to those, is found in oligodendrocytes where it has been postulated to play a key role in differentiation (503). Subcellularly, NOX1 is not only detected in lysosomal and phagosomal membranes but also purportedly in the nucleus and mitochondrion. In contrast, NOX2 is found not just in the plasma membrane, lipid rafts, phagosomes and endosomes but also putatively in mitochondria-associated ER membranes of brain cells (77). In spite of those assertions of subcellular localizations, it bears repeating that the maturity and function of most of the NOXs in the ER and nucleus requires rigorous analysis. To our knowledge, the only NOX that withstands scrutiny as to its function in both of those locations is NOX4 (396, 451, 999). Nevertheless, the common distribution of NOXs 1 & 2 in microglia and phagosomal and lysosomal membranes appears to suggest a role for both oxidase subtypes in the immune response. That said, more discernment as to their definitive roles in neuronal immunity is warranted (77). Additionally. NOX1, NOX2 and NOX4 are inferred to play a role in transcription factor modulation and protein degradation leading to proliferation or migration of astrocytes and microglia (77). NOX3 is not easily detected in brain immune cells and is mainly reported in neurons, astrocytes, oligodendrocytes, cerebellar granule cell precursors and neuronal ECs of the cerebellum and cerebrum. It is also purportedly involved in functions ranging from stem cell proliferation, differentiation, development to synaptogenesis (77, 1000, 1001). Intriguingly, NOX3 is detected in microglia under normal conditions but, unlike other NOXs, disappears in response to LPS (1002). Nevertheless, NOX3 is best known for its roles in the ear with respect to hearing and deafness as well as balance modulation (see elsewhere in this review).

Distinctly, NOX4 is reportedly found in cortex, Ammon's horn and dentate gyrus of the hippocampus, cerebellum, substantia nigra and subfornical organ. This distribution appears across a wide variety of cells, such as pyramidal cells, Purkinje cells, granule cells, dopaminergic neurons, neural stem/precursor cells, astrocytes, microglia and pericytes. Subcellularly in the brain it is found in the plasma membrane, mitochondria, nucleus, or cytosol (77). All told, a large body of work with respect to NOX4 implicates it in fundamental roles spanning neuronal development, neuronal death (1003) and signal propagation (77, 1004, 1005). NOX5’s limited detectability in primary cells contributes to a dearth of knowledge of its expression and function in the CNS though some reports have associated it with early oligodendrocyte differentiation and regeneration (77, 979).

When it comes to a physiological role in cognitive processes, aside from that described for NOX2 in dementia elsewhere in this review, knowledge of NOXs’ participation is far more scant. However, broadly speaking, it is reasonable to propose that ROS derived from NOXs plausibly participate in long-term potentiation and learning (1006, 1007) and NMDA receptor signaling (1008). In agreement with this, learning and memory are impaired in mouse models of CGD (502) as well as in CGD patients of whom 77% had the gp91^phox (X-linked) phenotype (1009).

4.13.2. Peripheral nervous system

In non-differentiated PC12 cells, NGF-induced neurite outgrowth is negatively regulated by NOX1, which is ten times more abundant in these cells than NOX2 (455). These findings dovetail nicely with recent associations drawn between NOX1, α-synucleinopathy and dopaminergic neuron death (77). Similar participation in neuronal cell death has been demonstrated for NOX1 in cerebral artery occlusion model of stroke (77). Indeed, the NOX is connected to a wide variety of brain disorders including Alzheimer’s disease, ALS, Parkinson’s disease, MS, stroke, traumatic brain injury and major depression (980–982). That said, the broad expression and myriad expression of various NOX isoforms in the brain speak to a predictably fundamental and redundant role of NOXs in neural networks related to cognition including physiological excitatory synaptic inputs and behavioral repetition in the central striatum (1010, 1011). In fact, NOX1’s association with β-arrestin recruitment suggest its involvement in a variety of neuronal functions including synaptic plasticity in the hippocampus (1012) as well as locomotor responses (1013). The downside of all this, of course, is the liability this poses to an easily exaggerated feed-forward oxidase activation in inflammatory brain disorders.

4.13.2. Ischemic stroke and the CNS

One of the most common brain vascular pathologies in the modern world – ischemic stroke, is caused by ischemia/reperfusion injury to the brain resulting in an increased ROS load. In ischemic stroke, oxygen deficiency leads to cellular nutrient and oxygen shortage causing neuronal damage. Reperfusion further aggravates the injury by supplying fresh oxygen that in already damaged cells supports ROS production leading to further damage. As such, NOX1 expression is enriched in the cerebral peri-infarct region after middle cerebral artery occlusion in mice and much higher than NOX2 expression (1014). As such, neuronal cell death and astrocyte activation are significantly reduced by NOX1 deletion in the peri-infarct region caused by ischemic stroke while functional recovery after reperfusion is enhanced (1014). This may be complicated by an activated renin-angiotensin system and the effects of AngII which stimulates NOX1 and increases cerebrovascular O₂^·− (1015). Importantly in that regard, AngII decreased cerebral blood flow in WT mice, but it did not in NOX1 knockouts (1015). Moreover, a moderate reduction in lesion volume and cerebral edema as well as improved brain blood barrier leakage with relatively mild ischemic damage is detected in stroked NOX1 knockout mice (1016). That said, NOX1 knockouts were not protected from cerebral infarction after middle cerebral artery occlusion neither in transient (30 min) (1015–1017) nor in persistent (>2 h) (1016) cerebral ischemia. As such, the reports observed no difference in neurological score, total or subcortical cerebral infarct volume or edema volume between wild type and NOX1-KO mice when a very short period of ischemia, 30 min, was followed by 24 h of reperfusion (1015). On the other hand, after 1h ischemia, but not 2-h or longer, NOX1 deletion led to a 55% attenuation in cerebral infarct size and improved neurological outcome at 24 h reperfusion (1016). It would appear then that the protective effect of NOX1 deletion is temporal in nature and closely dependent on the duration of hypoxia to which cerebral cells are exposed.

More convincing are reports on NOX2 and NOX4 role in ischemic stroke which show that protein and mRNA of both subunits are increased in the region surrounding the infarct in the cortex following reperfusion (1018–1021). Interestingly, elevation of NOX2 and NOX4 levels are sustained up to 48 h following reperfusion (1014, 1018–1021) and NOX activity is elevated in cerebral vs. peripheral arteries (1022, 1023). More to the point, even though NOX1, 2 and 4 all are expressed in cerebral ECs and adventitial cells, NOX2 expression appears to be the highest (1022, 1023). Relatedly, in NOX2 X-CGD mice, stroke infarct size is significantly diminished compared to their littermate controls (1024) which is likely attributed to aggregate NOX2 deletion across cerebral arteries, microglia, astrocytes and neurons. Delineation of each cell type’s contribution will require the employment of cell-specific knockout mice. Until then, from NOX isoform distribution we may be able to glean a contributory role. That is, in rodent models of focal- and widespread cerebral ischemia, NOX2 is likely at play in neurons, astrocytes and microglia where it has commonly been detected (1002, 1025). In post-stroke recovery, NOX2 at earlier time points is primarily localized to neurons and ECs whereas at later time points it can also be found in microglial cells. This points to a potential shift in NOX2’s prevailing role from the parenchyma to immune cells that infiltrate later in the recovery period (1025, 1026).

Moreover, NOX4’s role in infarcted brain tissue after I/R injury has been illustrated in an array of studies (1016, 1027, 1028). That is, after stroke, NOX4 is increased in human cerebral cortex in neurons and vascular endothelial cells and may participate in blood brain barrier breakdown (1029, 1030). More specifically, neuronal NOX4 is upregulated during stroke (1027), and concomitant Rac1 activation (potentially implicating NOX1 and NOX2) in the hippocampus are purportedly linked to apoptosis and cognitive impairment after reperfusion (1031, 1032). In accordance with a primary role for NOX4 in stroke and I/R injury, however, oxidative stress, neuronal death and blood brain barrier damage were attenuated in NOX4^-/- mice, but not NOX1 or NOX2 nulls, of either sex in response to either transient or permanent ischemia (1029).

4.13.3. Inflammatory and autoimmune diseases of CNS

In microglia, the “professional” phagocytes of the CNS, of both humans and rodents, NOX2 is not surprisingly identified as the most abundant NOX isoform, accompanied by lower expressions of NOX1 and NOX4 (1033, 1034). In mouse microglial cell line BV2, NOX2 is detected and expectedly induced to the largest degree after LPS treatment in vitro. Also not surprising is the demonstration that NOX2 levels are higher in microglia than in neurons and astroglia (1035), and in the latter, NOX2 raises the potential for synergistic oxidative damage in neuroinflammation (150, 1036). In neurons, activation of NOX2 characteristically involves membrane translocation of cytosolic p47^phox (83, 1037) and can mediate either necrosis or apoptosis in response to brain-derived neurotrophic factor or serum deprivation (1038, 1039). Indeed, NOX2 is ubiquitous and pleiotropic in its phenotypic CNS effects which also include proliferation and migration (77).

Like their systemic white cell counterparts, microglial NOX2-derived ROS are critical to a phagocytosis-related respiratory burst in the brain wherein they resolve the removal of myelin, apoptotic cells and their debris and maintains an optimal milieu for proper neural function. That notwithstanding, phagocytosis and NOX2/ROS can exceed their housekeeping function and severely damage the myelin sheath (1040). Along these lines, reactive microglia are linked to a host of neurodegenerative diseases including ALS, Alzheimer’s disease and Parkinson’s disease as well as white matter disorders, e.g. periventricular leukomalacia (1041, 1042). The latter is a focal necrosis of white matter in the brain underpinning the majority of cerebral palsy cases developing in premature infants wherein NOX2-derived O₂^·− and NO react at a diffusion-limited rate to form peroxynitrite responsible for oligodendrocyte killing (1043). On the other hand, in autoimmune disorders of the brain, like encephalomyelitis, deletion or truncation of p47^phox leads to exaggerated T cell responses. On closer examination, a couple of reports indicate that O₂^·− from p47^phox-dependent NOX2 neutralizes NO which normally suppresses helper T cells (570, 1044).

4.13.2.1. Multiple sclerosis

In an illuminating study of human subjects with multiple sclerosis, NOX1 is weakly expressed in microglia, astrocytes and ECs of a majority of control subjects. In contrast, in multiple sclerosis patients, NOX1 and NOXO1 expression rose conspicuously and colocalized in astrocytes and ECs around active lesions of acute and relapsing MS (1045). Moreover, in macrophages and microglia, multiple components of the NOX1 and NOX2 complexes were present (1045). These results appear to suggest a role of the NOX1 complex in astrocytes, ECs and microglia, and a predominant role for NOX2 in microglia in the development and progression of MS. As to the latter, p22^phox and NOX2 are both expressed in microglia and macrophages of both MS patients and controls (1045). However, post-mortem analyses of MS-diseased white matter displayed a moderate expansion of p22^phox- and NOX2-positive microglia compared to controls (1045) suggesting that an expansive respiratory burst-like activity is responsible for white matter dysfunction. Interestingly, NOX2 deletion results in amelioration of MS symptoms and prevents astrocyte activation in mice (1046). This has been attributed to abrogated NOX2-mediated induction of pro-inflammatory cytokines and stimulation of the anti-inflammatory cytokines IL4 and IL10 (1046). All told, the data appear to implicate NOX2 oxidase activation being the focal of the disease beginning in microglia and relayed to NOX1 oxidase activation in astrocytes and ECs.

4.13.2.2. Alzheimer’s disease

According to the Alzheimer’s foundation, Alzheimer’s disease (AD) is estimated to be prevalent in 60–80% of dementia cases in elderly individuals (1047, 1048) (Alzheimer’s Association, 2024). As mentioned above, physiological ROS signaling coming from NOXs is important for the maintenance of cognitive functions of the brain, and an increase in NOX activity and the consequential increased ROS load is correlated with the decreased cognitive performance (1010–1013, 1023, 1049). Comparison of subjects’ ante-mortem cognitive testing with postmortem histopathological assessment indicates that overall NOX activity is most pronounced in mild cognitive impairment cohorts and remains elevated in all other tested cohorts of different grades of impairment as compared to a no cognitive impairment group (1050). Interestingly, cytosolic NOX subunits p67^phox, p47^phox, and p40^phox were significantly elevated with human disease progression, while the membrane-integrated NOX2 and p22^phox remain unchanged (1050). Indeed, increased NOX activity is correlated with decreased cognitive performance (1023, 1049, 1050). Accordingly, NOX2 inhibition decreased both oxidative stress and AD pathology in aged mice (1051) and in corroboration of these findings, in mice overexpressing a mutant of human amyloid precursor protein (APP), NOX2 deletion decreased neurovascular dysfunction and cognitive decline (1051). Similarly, a correlation between age-dependent Aβ deposition and NOX was found in a “humanized” APP x presenilin, PS1 knock-in mouse model (1023, 1052). Notably, neutrophils and macrophages of CGD patients do not generate ROS in response to Aβ while cells from non-CGD subjects do (1053); and in agreement with that finding, co-cultures of hippocampal neurons and microglia treated with Aβ peptides resulted in putative activation of NOX2, leading to neuronal cell death (1054, 1055). Thus, the findings suggest a strong link between Aβ and leukocyte NOX2 in AD.

Furthermore, large amounts of Aβ produced in AD may activate microglia and astrocytes and, in turn, augment overall NOX2 assembly, ROS production and inflammatory mediators in brain cortical regions (1051). Indeed, a vicious cycle involving ROS-driven inflammatory mediators such as Aβ, TNFα, interleukins and α-synuclein and their activation of NOXs in microglial cells is claimed to cause neuronal death (1056, 1057). In the Tg2576 mouse model of AD, observed increases in ROS, neurovascular dysfunction, and behavioral deficits were abrogated when these mice were crossed with NOX2 knockout mice (1051) solidifying a pivotal role for NOX2 oxidase in the disorder.

Furthermore, NOXs’ tight connection with inflammatory and immune responses became especially interesting in light of a report that Aβ sensitizes the neuronal response to TLR4 resulting in long-term potentiation deficit in the hippocampus and neuron cell death (1058). This has led to speculation that the interaction between TLR4 and Aβ in neurons and other cells in the brain can also stimulate NOX4 activity similarly to that by TLR4 in other tissues (1059, 1060); e.g. endothelial cells stimulated with LPS, the Toll/IL-1 receptor homology domain (TIR) of TLR4 physically interacts with NOX4’s C-terminus and inhibits NFκB (544). Thus, NOX4, by way of its high level of expression in astrocytes and the cerebral vasculature, could plausibly synergize with NOX2 in AD (544, 1059).

Finally, accumulating Aβ in AD leads to the activation of the receptor for advanced glycation end-products (RAGE), which, in turn, stimulate NOXs. RAGE and NOXs alone and in sequence, in that order, activate Ras signaling, causing activation of MAP kinases and the JAK/STAT pathway, resulting in the potentiated protein accumulation and pro-inflammatory molecule production (1023, 1061). Adding insult to injury, RAGE facilitates transport of Aβ via endocytosis and transcytosis across the blood brain barrier, resulting in exacerbated accumulation of Aβ in the parenchyma of the brain (1023, 1061). These findings appear to support the destructive participation of glucose and diabetes in the propagation of the NOX in dementia.

4.13.2.3. Parkinson’s disease

Increases in NOX1 in dopaminergic neurons of the substantia nigra of paraquat-treated rats correlate with the subsequent dopaminergic neuronal death (1062). Importantly, Parkinson's disease (PD) is a nerve disorder involving such dopaminergic neuron loss in the substantia nigra pars compacta and is marked by tremor, muscle rigidity, slowness of voluntary movements and postural instability (1063). Mechanistically speaking, PKCδ is known to upregulate NOX1 expression in dopaminergic cells (1064). Moreover, NOX1-mediated ROS are reportedly tied to α-synucleinopathy in paraquat-induced PD models, since NOX1 depletion disrupted α-synuclein expression and aggregation and diminishes dopaminergic neuronal death in vivo (1065). On the other hand, mitochondrial ROS while essential, are not sufficient for neuronal death which appears to require the sustained accumulation of ROS from NOX1 (1066). Thus, greater understanding of the interdependence of NOX and the mitochondrion in ROS-mediated PD progression will be necessary for the development of long-term therapies combatting the disease.

Similar to the etiology of AD, chronic neuroinflammation in microglial NOX2 is deleterious and ultimately lethal in PD. Brain neuroinflammatory responses through the activation of NLRP3 inflammasome by NOX1 and NOX2 (and downstream PAK1 and MAP kinases) have been deemed pivotal in the disorder (1067). Indeed, a recent study uncovered pernicious evolution from neuronal NOX2 activation to chronic neuroinflammation involving both microglia and neurons, neurovascular uncoupling, and inflammation-related neurodegeneration (1037). These findings appear to support the notion that a long-term, persistent upregulation of NOX2 participates in a low-grade yet chronic inflammatory component of the disease as has been described for multiple brain disorders.

In a rotenone-based model of PD in rats invoking a mitochondrion-centered pathology, multiple aspects of PD in humans are replicated (1068). As to its pathobiology, Greenamyre and colleagues demonstrate that rotenone-treated nigrostriatal neurons in rats and substantia nigra from idiopathic PD patients display active leucine-rich repeat kinase 2 (LRRK2) which leads to increased phosphorylated α-synuclein and its binding to TOM20 (1069). Moreover, these events lead to mitochondrial dysfunction and NOX2 activation. In turn, phosphorylated α-synuclein causes LRRK2 activation and begins the cycle anew (1069). The authors go on to show that NOX2 is a critical intermediary in LRRK2 activation and this cycle of dysfunction by illustrating disruption of it all by NOX2ds-tat (57, 740, 1069). In the same in vivo rat model, oral administration of brain penetrant NOX2 inhibitor CPP11G (1070, 1071) prevents NOX2 assembly and activity in dopamine neurons of post-mortem substantia nigra sections, which in turn, abrogate oxidized and posttranslationally modified α-synuclein, LRRK2 activation and its downstream degenerative effects (83). Together, the findings further point to a key role for NOX2 in a vicious cycle of neuronal decline involving LRRK2 and improper protein folding and function in PD.

4.13.4. Depression

In the burgeoning role of NOX in mental depression, chronic “social defeat”-related stress or corticosterone administration have been shown to induce depressive-like behaviors that are significantly ameliorated in NOX1^-/- (991). In fact, administration of corticosterone in mice induced significant elevation of ROS in the prefrontal cortex (PFC), and upregulated NOX1 uniquely in the ventral tegmental area of the midbrain where dopaminergic cell bodies originate and are involved in pleasure and reward cognition (991). This rather delimited expression suggests a possible pivotal role for NOX1 in learning, memory, and emotional and addictive behaviors. Corroborating the findings, NOX1 miRNA delivery to the ventral tegmental area, but not the PFC, ameliorated corticosterone-induced depressive-like behaviors (e.g. sucrose drinking) in WT littermates. Furthermore, corticosterone application to WT, but not NOX1^-/y, mice significantly reduced brain-derived neurotrophic factor levels (991). Therefore, in aggregate, NOX1 in the ventral tegmental region of the brain appears to participate in normal cognitive function as well as depressive-related behaviors and argues strongly for closer examination of these linkages in humans.

Along similar lines, repeated stressors promote depressive behavior which are likely to be pronounced with age coinciding with the upregulation of NOX. Indeed, basal expression of p47^phox and NOX2 transcripts are higher in the hippocampus of aged vs. young mice, which appears to sensitize them to mild stressors (984). Significantly, the higher p47^phox and NOX2 expression levels in aged mice subjected to a stressor (2 h restraint) for 5 days correlate with decreased mobility and reduced sociability (984). Additionally, repeated, chronic restraints appear to increase the expression of p47^phox and p67^phox in the brain (1072). That is, brains from stressed mice show upregulation of p47^phox and amplified ROS levels in response to mild re-stress evoked in the post-stress period. Elevating the significance of these results, non-selective NOX2 inhibitor apocynin, p47^phox+/− or molecular knockdown of p47^phox with lentiviral p47^phox shRNA in the hippocampus moderates behavior consistent with depression in mice (1072). These results are consistent, therefore, with repeated incidences of stress promoting depressive behavior through the upregulation of NOX2 oxidase and illustrate that even partial abrogation of p47^phox can provide beneficial anti-depressive effects (1072). Furthermore, it is notable that NOX2 expression and ROS levels are significantly higher in the cortex of asphyctic human suicide subjects vs. control groups (983). Because of the severity of major depression and its often intractable treatment resistance, the above-detailed research could lay the foundation for testing of global or localized administration (991) of both NOX1 and NOX2 inhibitors in preclinical studies (57, 439, 740, 1070, 1071) either as a drug or adjunct therapy.

All told, research on ROS signaling and NOX in nervous system physiology and pathology is a field undergoing extraordinary growth and expansion. With this expansion comes the need for better understanding of NOXs’ roles in varied cell types and the disentanglement of NOX-specific signaling to preserve cognitive function while developing curative treatments. In general, NOX inhibition is expected to be a promising, or no less, supplemental therapy for nerve disorders (such as above discussed AD and PD) and ultimately enormous potential benefits will have to weighed against possible risks for individuals with these often intractable and fatal diseases.

4.14. Aging

Aging is clearly a complex phenomenon that involves numerous biological processes arising from genetic, epigenetic, cellular, physiological, as well as environmental and social factors (1073). Multiple theories of aging have emerged over the past century including the programmed cell theory, the somatic mutation theory and the auto-immunity theory, of which, one of the most intriguing has been the free radical theory of aging (1074–1076). Indeed, many of the diverse theories share ‘oxidative stress’ as a common denominator in the processes they describe (1074). While chronic activation of NOXs has been implicated in various age-dependent diseases, they also are expected to maintain and promote the processes of healthy aging (i.e. immune modulation, autophagy) as we have extensively discussed in previous chapters. With respect to disease, recent studies have implicated NOXs in senescence suggesting it to be at the center of a gradual and broader deleterious effect of secreted inflammatory factors on organ systems, i.e. ‘inflammaging’ (438, 1077, 1078). Thus, it seems logical to conclude that aging of the individual as a whole is a complex process in which ROS play important physiological and pathophysiological roles.

Indeed, NOX1, NOX2 and NOX4 have emerged as pivotal in cellular senescence (438, 1079, 1080) and there is little to no doubt that NOXs participate in senescence via a paracrine, and perhaps even endocrine, feed-forward response that involves multiple sources of subcellular ROS and an exaggerated immune response (438, 1078, 1081–1083). Also, it is increasingly clear that gradual yet chronic activation of NOXs together with decreased cellular antioxidant capacity lays the foundation for various age-related diseases such as Alzheimer’s disease, Parkinson’s disease, cancer, diabetes, cardiovascular diseases, and chronic inflammation (1073). In the interest of focus, we turn our attention chiefly to the salient roles of NOX2 and NOX4-derived ROS in CNS and cardiovascular aging (1022, 1084–1087) with particular focus on senescence.

4.14.1. Aging of the CNS

Aging of the CNS is associated with increased oxidative stress arising from augmented hypoxia-induced NOX in both neural and cerebrovascular cells of the CNS (1088). Additionally, aging adversely affects neovascularization following ischemic injury further exacerbating nourishment deficits and an increase in CNS tissue hypoxia that results in the increased oxidative stress (1089). As described in the subchapter on the CNS, vascular I/R injury is one of the main causes of cognitive decline. To that point, NOX2 oxidase activity is pronounced both in cerebrovascular cells and in immune cells that are recruited to the site of I/R vascular injury; and it should not come as a surprise that NOX2 is thus far the best-described NOX isoform in the context of cognitive decline connected with aging. That is, NOX2 expression is sharply upregulated in the superior and middle temporal gyri of patients with mild cognitive decline and in brains of aging mice (23 months old) (1090–1092). Moreover, aged (12–15 months old) mice overexpressing amyloid precursor protein display signs of cognitive decline along with a doubling in NOX2-derived O2•− levels (1051). Similarly, NOX2 activity and cognitive decline are linearly correlated (1052, 1092, 1093); and further in support of this, enhanced expression of p47phox and p67phox was reported in the temporal cortex of mild cognitive decline patients and correlates tightly with cognitive decline (1050).

In concert with this notion, aging WT brains present with high levels of AngII and ROS production (systemically linked to NOX2) and activation of ERK1/2, p53, and γH2AX, together with loss of capillaries and neurons (1094). Reinforcing the role of NOX2, these pathologies are sharply reduced in aging brains of NOX2 knockout mice (1094). Furthermore, brains of mice overexpressing EC-specific human NOX2 (HuNOX2Tg) display high levels of ROS and activation of stress signaling pathways (activation of ERK1/2 and γH2AX) at middle age (11–12 months) similar to those found in WT brains of more advanced age (1094). Not surprisingly, therefore, aging WT mice (20–22 months) as opposed to young WT controls (3–4 months) exhibit metabolic disorders and loss of locomotor activity (1094), and AngII-induced EC NOX2 activation appears at least partially responsible for cerebral capillary rarefaction and reduced brain function in these mice (1094). These data are clinically corroborated with postmortem midbrain tissues of young (25–38 years) and elderly (61–85 years) individuals. Taken together, findings in both rodents and humans suggest a pivotal role for NOX2 in nervous system decline. For more on aging-related diseases see content in the Central Nervous System Section.

4.14.2. Cardiovascular aging

Not surprisingly, aging is a critical factor for the progression of cardiovascular diseases. Given the fact that cardiovascular tissues express multiple NOX isoforms, it is to be expected that NOX upregulation plays a role in this progression. In fact, on close examination of the literature, it becomes evident that physiological decline and dementia can be associated broadly with increased oxidative stress in systemic vascular and cerebrovascular diseases. In this context, NOX2 appears to be the most studied NOX. For example, age-dependent increases in blood pressure, cardiomyocyte hypertrophy, coronary artery remodeling, and cardiac fibrosis are associated with enhanced vascular and myocardial NOX2 (1095). Furthermore, rises in NOX2-derived ROS are paralleled by increased activation renin–angiotensin–aldosterone system in the myocardium, increased expression of connective tissue growth factor and TGFβ1 as well as activation of matrix metalloproteinase MMP2 and membrane type-1-MMP, leading to age-associated cardiac remodeling and fibrosis (1095). In the vasculature and similar to cerebral vascular ischemic injury, hind limb NOX2 expression is increased in ischemia in aged (10 months old) mice as compared to control. This is accompanied by a significant reduction in blood flow recovery after ischemia in old compared to young mice at day 21 after surgery, and by the reduction in capillary and arteriolar densities in ischemic muscles of old animals (1096). In support of this, NOX2 deletion reduces oxidative stress in ischemic tissues and restores blood flow recovery and vascular density in old animals. Moreover, some of the major mechanisms driving neovascularization including the capacity of endothelial progenitor cells (EPCs) to migrate, adhere and mature are significantly impaired in old compared to young mice. Remarkably, NOX2 deficiency rescues these stem cell capabilities (1096) and collectively the findings suggest that NOX2 plays an increasingly obstructive role in homeostatic processes subserving vascular self-renewal and blood pressure maintenance with age.

In a departure from NOX2, in 2021 the Pagano group explored the role of NOX subtypes in mice in vivo and discovered a causal link between NOX1, per se, and aging-related senescence and functional decline involving “inflammaging”. That is, NOX1 was linked to senescence-associated secretory phenotype (SASP) inflammatory cytokines IL6 and monocyte MCP1, cell cycle inhibition, DNA damage, altered nuclear envelope integrity and tissue dysfunction, i.e. impairment of vascular dilatation (438, 1082). Moreover, on the cellular level, selective NOX1 inhibition and siNOX1 concomitantly abolished IL6 expression and senescence markers as well as suppressed elevated cell cycle inhibitor p21^cip in response to age and an age-related senescence trigger in mouse or human endothelial cells (438). Notably, NOX1 was tightly associated with aging in mouse and human tissue and NOX1i restored the capacity for vessel sprouting and vessel dilatation. Moreover, in vivo, selective NOX1 inhibition normalized (a) senescence markers their related inflammatory response; (b) age-impaired vessel dilatation; and (c) regional blood flow via disruption of a NOX1-IL6 feed-forward proinflammatory signaling loop. And, of particular note, clinical samples from diabetic versus nondiabetic subjects corroborated as operant this NOX1-mediated vascular senescence and “inflammaging” in humans of advanced age (438).

Further studies are clearly warranted with respect to both NOXs 1 and 2 and other NOX isoforms with respect to aging that are missing from the literature. Based on findings to date, combinatorial NOX1 & NOX2 inhibitors could hold significant promise for the prevention of cardiovascular and neural decline and it might be expected that a unifying theme across most of the NOXs could be the capacity to solely or jointly produce sustained low to moderate ROS over long periods of time that gradually lead to tissue senescence and degeneration.

4.15. Cancer

Considering manifold examples of NOXs participation in neoplastic phenotypes, it is rational that NOXs would play a fundamental role in both the regulation of tumor development and in the establishment of a tumor-driving microenvironment. Thus, understanding NOX-dependent growth and survival pathways in cancer is expected to contribute significantly to our full appreciation of its etiology and progression. In a practical sense, a fully comprehensive in-depth survey of the literature on the role of NOXs in tumor progression is not feasible. Our main focus here is to survey NOXs in a subset of cancers as well as their role in creating an inflammatory niche essential for tumor growth and in weakening the immune response. Figure 9 shows the differential transcript expression for NOXs across tumors and adjacent normal tissues obtained from the Cancer Genome Atlas Program (TCGA) database and is intended not to be prescriptive of the role of one or more NOXs in cancer but chiefly to stimulate discussion and inquiry on the part of the reader.

Figure 9. NOXs in human cancer.

Differential expression between tumor and adjacent normal tissues for NOX enzymes across tumor types in The Cancer Genome Atlas (TCGA). Distributions of gene expression levels are displayed using box plots. Genes that are upregulated or downregulated in the tumors are compared to normal tissues for each cancer type. mRNA expression of the indicated NOX isoform in normal (blue) and tumor (red) samples, each dot represents a different patient obtained from the TCGA database (https://portal.gdc.cancer.gov). BLCA=Bladder Urothelial Carcinoma, BRCA=Breast invasive carcinoma, CESC=Cervical squamous cell carcinoma and endocervical adenocarcinoma, CHOL=Cholangiocarcinoma, COAD=Colon adenocarcinoma, ESCA=Esophageal carcinoma, GBM=Glioblastoma multiforme, HNSC=Head and Neck squamous cell carcinoma, KICH=Kidney Chromophobe, KIRC=Kidney renal clear cell carcinoma, KIRP=Kidney renal papillary cell carcinoma, LIHC=Liver hepatocellular carcinoma, LUAD=Lung adenocarcinoma, LUSC=Lung squamous cell carcinoma, PAAD=Pancreatic adenocarcinoma, PCPG=Pheochromocytoma and Paraganglioma, PRAD=Prostate adenocarcinoma, READ=Rectum adenocarcinoma, SKCM=Skin Cutaneous Melanoma, STAD=Stomach adenocarcinoma, THCA=Thyroid carcinoma, UCEC=Uterine Corpus Endometrial Carcinoma. Wilcoxon test showing significance between tumor and adjacent tissue at * p<0.05, **p<0.01, ***p<0.001.

4.15.1. NOX1

As originally termed mitogenic, it should come as no surprise that a great deal of focus has been placed on NOX1 in cancer (142). In this vein, NOX1 is notably implicated in cancers of the colon (1097, 1098), and also, but not limited to, skin (677, 1099) and breast cancer (1100). Predictably, the majority of NOX1-related cancers continue to be explored in the gastrointestinal system (1101) and primarily the colon where NOX1 is enriched; however, NOX1 is increasingly implicated in a variety of cancers (1102, 1103). Broadly speaking, anchorage-independent tumor growth, cell invasion and mutagenesis, some of the most common features of carcinogenesis, are associated with increased NOX1-derived ROS (1104, 1105). Moreover, NOX1 has been shown to alter energy metabolism and contribute to proliferation, angiogenesis, and invasiveness, tumor progression and metastasis. Indeed, a major underpinning of these phenotypic shifts appear to be that NOXs including NOX1 can promote complementary ATP production via aerobic glycolysis (1106). As such, it is well established that tumor-proliferating cells are accelerated by the Warburg effect through increased glycolysis, even in the presence of fully operational mitochondria (1107–1109).

Indeed, NOX1 expression is predictive of poor prognosis of hepatocellular carcinoma (HCC) (1103), as mice treated with pro-carcinogenic diethylnitrosamine receiving the NOX1 inhibitor GKT771 show reduced liver inflammation, fibrosis and tumoral stemness (i.e. tumor aggressiveness). These effects coincide with a disruption in what the authors describe as a pro-tumorigenic microenvironment (1103). On the other hand, a 2012 study reported that Wnt is capable of activating the Rac1-GEF Vav2 via Src kinase and stimulating Rac1-NOX1 controlled ROS, which by oxidative inactivation of nucleoredoxin “releases the brake” on β-catenin signaling and permits cell proliferation. In further detail, nucleoredoxin inactivation unleashes dishevelled-1 (Dvl-1) which disrupts tumor suppressor APC and liberates β-catenin allowing it to promote cyclin D1 and c-Myc (732). In essence, NOX1 appears to be a pivotal upstream modulator of a binary and tightly regulated process of cell proliferation in tumorigenesis (732).

In colorectal malignancies, NOX1 is firmly implicated as capable of multi-faceted actions (1110, 1111). That is, there are several mechanisms through which NOX1 might elicit malignant transformation including dysregulation of cell cycle and induction of apoptosis resistance, exacerbation of the components of innate immunity, and induction of the RAS/MAPK pathway and IL4 signaling (1110). Indeed, NOX1 plays “duplicitous” roles in the intestinal host defense whereas it also modulates colonic cell growth (142). During malignant transformation, NOX1-derived ROS dysregulate epithelial cell apoptosis, on the one hand, and induce angiogenesis on the other (437, 680, 1112, 1113). In fact, NOX1-mediated ROS promote proliferation and survival of HT-29 and Caco-2 human colon cancer cells and NOX1 knockdown leads to cell cycle arrest or apoptosis (1114). It should be expected then that, collectively, these NOX1-compounded actions of cell proliferation and new vessel growth potentiate tumor growth.

Furthermore, in patients with ulcerative colitis at increased risk of developing colon cancer, the expression of NOX1 is significantly enhanced during the activated inflammatory response (1115, 1116). Above and beyond NOX1’s role in surface mucosal cells modulating the innate immune response and mediating the killing of pathogenic bacteria (548), there is clear evidence linking NOX1 to exaggerated cytokine-related ROS production, inflammation, and colorectal malignancies in the context of colitis. Indeed, as alluded to above NOX1 knockdown in HT-29 and Caco-2 colonic cell lines results in G1 cell cycle arrest and apoptosis consistent with a role for NOX1 in colon carcinoma (1114). In parallel, another study reported a significant decrease in the expansion of HT-29 cells upon NOX1 silencing, mediated, in part, via suppression of ADAM17-EGFR-PI3K-Akt signaling (1117). Accordingly, the downregulation of multiple oncogenes, chemokines, and angiogenic factors, including Myb, CXCR4 and VEGF, are observed when NOX1 is suppressed, which consequently prevents angiogenesis and brings about a decrease in xenograft growth (414). Collectively, these represent a complex and pleiotropic role for NOX1 in colitis-related colon cancer with multiple signaling redundancies at play.

NOX1 also plays an essential role in ROS-mediated signal transduction through Ras/MAPK pathway (680, 1113). Indeed, NOX1 mRNA and protein are overexpressed in colon cancer and strongly correlate with activating mutations in K-Ras (1098). In support of this, NOX1 knockdown in HT-29 human colon cancer cells blocked the G₁/S checkpoint and was associated with a significant decrease in cyclin D1 expression and a profound inhibition of MAPK signaling (414). Thus, these findings appear to suggest a strong interdependence between NOX1, Ras and NOX1’s crucial role in cell-cycle modulatory signaling in the colon.

4.15.2. NOX2

NOX2-derived ROS, as well, are implicated in cancer progression, ranging from cell proliferation (1118, 1119) to tumor progression and metastasis (355, 1118, 1120, 1121). Undeniably, myeloid NOX2 is implicated in the dysregulation of macrophage differentiation which can lead to autoimmune disorders (e.g. in response to ‘danger’ signals) and cancer (1122). That is, tumor-associated macrophage (TAMs) ROS promote tumorigenesis due to proangiogenic and immune-suppressive functions which resemble those of M2 macrophages (1123); i.e. O₂^·− scavenging inhibited M1 to M2 macrophage differentiation (reduced TAMs) and suppressed tumorigenesis in mice (468). Predictably and by extension, NOX2 and also NOX1 suppression, via reductions in O₂^·−, JNK and ERK, significantly decrease M2 macrophage differentiation and tumor growth (467). Conjointly, H₂O₂ released by macrophage NOX2, and presumably en bloc from other NOXs, is expected to trigger insensitivity of nearby tumor-specific natural killer T cells. This desensitization, in turn, contributes to augmented neoplasia and melanoma metastasis (1121, 1124). NOX2 is also implicated in the dysfunction of tumor-infiltrating effector CD8⁺ T cells in murine lymphoma and inefficacy of antitumor immune checkpoint inhibitors (1125). Moreover, cells from patients with acute myeloid leukemia expressed high levels of NOX2 (1126) and NOX2 inhibition elicited by histamine reduces the expansion of xenografted human NOX2-positive myeloid leukemia cells (1127). Taken together, these studies implicate a role of NOX2 and potentially other NOXs in the development of leukemia through the stimulation of stem cell populations and impairment of the T-cell-driven immune response and are consistent with selective inhibition of NOX2 being a novel target in the treatment of acute myeloid leukemia.

On the other hand, non-myeloid NOX2-derived ROS may facilitate cancer expansion by promotion of cell proliferation and a phenomenon described as “apoptotic escape”. In other words, an observed high-level NOX2-related expression and ROS appear to evoke the capacity to redirect osteosarcoma cells from programmed death to survival (1119). Likewise, NOX2 is shown to be significantly increased in primary prostate cancer, where its expression coincides with increased cell proliferation and prostate tumor development (355). In support of that finding, NOX2 knockout mice, as well as littermates to which NOX2 inhibitor is applied, show decreased angiogenesis and major disruption in prostate cancer growth (355). Furthermore, mice deficient in NOX2 present with reduced metastasis of melanoma cells to the lung purportedly by boosting natural killer cell function (1121). And, consistent with this, parallel experiments systemically treating mice with histamine dihydrochloride, to suppress NOX2, enhanced the infiltration of IFNγ-producing NK cells into the lung, thereby reducing metastasis (1121). Moreover, p22^phox-, NOX2, p47^phox-, p67^phox- and p40^phox-deficient mice displayed reduced lung metastatic colonization [concordant with the appearance of large granulomas infiltrated with galectin-3 (Mac-2)-positive macrophages and eosinophilic deposits]. Naturally, therefore, NOX2 subunit deletion resulted in augmented anti-tumor immune cell populations (1128). Taken together, these intriguing findings are in agreement with: (a) a fundamental role for NOX2 oxidase in the restraint of granulomatous immune cells in the tumor microenvironment; and (b) how tempering oxidase activity can unleash antitumorigenic activity. Thus, localized delivery of NOX2i’s to this microenvironment could aid in both the arrest and regression of metastatic tumors.

4.15.3. NOX4

In an intriguing study by Mori and colleagues, a governing role for NOX4 in the stability and pivotal role of MMP9 in invasiveness and metastasis of Ras-active mammary epithelial cells (MDA-MB-231) is proposed (1129). In greater detail, under the overarching control of Ras, molecular adaptor Hic-5 (hydrogen peroxide-inducible clone 5) purportedly suppresses NOX4 and MMP9 expression and activity. However, when Hic-5 expression is downregulated, as is apparent in metastatic cancer cells, it unleashes NOX4-mediated MMP9 activity and invasiveness (1129). Fascinatingly, this raises the possibility of a nested negative feedback loop in this process involving NOX4-derived H₂O₂. On closer inspection, Hic-5, also detected in melanoma, pancreatic cancer and fibrosarcoma, is associated with paxillin and focal adhesions (where NOX4 has been described to play an instrumental role in cell migration) and is a strong predictor of cancer cell invasiveness (1130–1132). Collectively, these findings appear to add another level of complexity and control to the well-described role of NOX4 in cell adhesion and motility (1132).

In ovarian cancer, NOX4 has been shown to cause oxidation-dependent upregulation of VEGF and angiogenesis - key factors in tumor growth (1133). Incidentally, numerous NOXs reportedly are involved in signaling pathways such as NFκB signaling (NOX1 and NOX4), Src kinase (NOX1, NOX2, and NOX4), and others leading to cellular invasion and migration (NOX1 and NOX4) in the context of cancer (1134). Additionally, among others oncogenic mutant p53 epigenetically upregulates NOX4, ROS, and TGF-β-induced cell migration in several tumor types (1135, 1136).

Additionally, a number of other studies investigated the role of NOX4 in tumor progression. Indeed, NOX4 has been associated with pancreatic ductal adenocarcinoma (PDAC), K-ras activation and p16 inactivation. These two signature alterations activate NOX4 which significantly enhance glycolysis vis-à-vis the Warburg effect and overcome metabolic checkpoints (1137). Indeed, NOX4 is overexpressed in multiple developing cancers, including malignant melanoma and ovarian, bladder, esophageal, head and neck and prostate carcinomas (1138). More significantly, NOX4 appears to mediate oncogenic transformation in renal cell carcinoma, melanoma, glioblastoma, and ovarian and pancreatic cancer and involve a variety of factors including HIF2, NFκB, CREB, Tks5, and VEGF (341, 682, 1133, 1139, 1140). In gastric cancer, in particular, NOX4 overexpression has been linked to epithelial mesenchymal transition and invasion via JAK2/STAT3 (1141), and is associated with poor prognosis in human colorectal cancer (1142) raising the possibility that NOX4 could serve as a biomarker for the disease. In sum, NOX4 appears to be a seminal player in a broad range of upstream and downstream pro-tumorigenic signaling pathways in EMT and tumor progression.

Still another fascinating series of studies suggests an intricate mechanistic role of NOX4 in melanoma. Collectively, the authors provide keen insight into how HGF/CMET activation and NOX4 expression may contribute to malignant resistance to BRAF inhibitors (6). BRAF is a serine threonine kinase which when mutated causes dysregulated MAPK signaling (MEK and ERK) and represents a crucial lynchpin in the development of several types of cancer (1143). It contains a phosphomimetic residue (D448) and a constitutively phosphorylated serine residue (S445) in its “N”-region, which allows for a single kinase domain mutational change in its activation segment (AS), i.e. V600E, to render the kinase constitutively active. Indeed, this mutated BRAF V600E is well recognized in melanoma progression vis-à-vis its resistance to BRAF inhibitors (1144). Importantly, RAF proteins have 3 conserved domains (CR1, CR2, CR3) (Figure 10) and one (CR1) contains a Ras binding domain (RBD) where NRAS (neuroblastoma ras viral oncogene homolog) can bind as well as a cysteine-rich subdomain (CRD) with the potential for redox amplification of BRAF (1145). Indeed, a recent study reports (1) a strong correlation between HGF-c-MET-induced NOX4, BRAF-inhibitor resistance and metastasis; and (2) evidence for an HGF/c-met/NOX4/ROS in EMT in vitro (70). Putting this all together, it is tantalizing to proffer a role for NOX4-derived H₂O₂ in the oxidation of CR1 in BRAF leading to exaggerated NRAS binding, resistance to BRAF inhibitors and melanoma progression (Figure 10).

Figure 10. c-Met/NOX4 axis contributes to metastasis and drug resistance in BRAF mutated melanoma.

Proposed role of the c-MET axis and NOX4 in melanoma resistance and progression. Putative NOX4-mediated oxidation of BRAF at CR1 heightens NRAS leads to epithelial to mesenchymal transition and resistance to BRAF inhibitors in BRAF-mutated melanoma. BRAF proteins have 3 conserved (CR1, CR2, CR3), CR1 contains Ras binding domain where NRAS can bind, and a cysteine-rich subdomain that can be oxidized by NOX4-derived H2O2 leading to exaggerated resistance to BRAF inhibitors and melanoma progression. CR, Conserved region; RBD, Ras binding domain; CRB, cysteine-rich subdomain; AS, activation segment; NRAS: neuroblastoma Ras viral oncogene homolog.

4.15.4. NOX3 and NOX5

While no clear link has been found between carcinoma and NOX3, there are data which could suggest its role in cancer. For example, NOX3 was detected in cancer cell line HepG2 as part of NOX3 discovery and identification (154). It was later shown to be the culprit of insulin-induced H₂O₂ generation and transcription factor Sp1 DNA binding leading to pro-proliferative VEGF-A expression; the authors, therefore, infer a key role for this signaling cascade in tumorigenesis (1146). NOX5’s role may be deduced from its relatively high expression in an array of human cancers, such as its role in cell survival in prostate cancer (398), SHP-1 inhibition in hairy cell leukemia (1147), activation by CREB in esophageal cancer (177) and apoptosis interference in anaplastic large cell lymphoma (1148) among multiple others (690, 1149). Incidentally, NOX5 expression correlates with 100% of aggressive and with 19% of non-aggressive lymphomas (1150). Additionally, NOX5 in prostate cancer (PC-3 and LnCaP) and esophageal cells (SEG1) appears to involve its induction by CREB and the downstream activation of JNK and PKCζ (177, 398), suggesting various tiers of signaling control in tumorigenesis.

Moreover, a pair of in vitro studies showed that the degree of NOX5 expression differentially modulates cancer cell dynamics. That is, the extent of NOX5 expression appears to determine the balance between cellular proliferation and cell death in skin, breast, and lung cancers (1149). Lastly, high NOX5 expression holds promise as a biomarker for colon cancer in humans as it is closely associated with poor prognosis (1151).

4.15.5. DUOXs 1 and 2

Epithelial carcinoma of the lung airways also appears to be influenced by increased H₂O₂ production from DUOX1 and NOX2 (1152) and the oxidation and sulfenylation of Cys797 within the EGF receptor, which enhances its tyrosine kinase activity (1153, 1154). Whereas DUOX1 appears to be the intermediary between epithelial injury and ligand-independent EGFR transactivation, NOX2 appears to mediate ligand activation of EGFR and sustain transactivation of the receptor (1152) that predictably drives EMT transition and cancer progression (1152). On the other hand, the same group demonstrated that endogenous DUOX1 silencing in lung cancer is associated with loss of differentiated epithelial characteristics like barrier function maintained by E-cadherin in exchange for cancer stem cell markers like CD133 (1155). In agreement with this, when DUOX1 was knocked down in normal epithelial cells or in the lung epithelial cancer cell line H292, it resulted in epithelial-to-mesenchymal transition (EMT) transition from the epithelial phenotype to invasive cells - a hallmark of metastatic cancer (1155). Thus, the authors explain that these seemingly contradictory findings (pro- vs. anti-tumor roles of DUOX1) might be brought to bear by an enhanced resistance to EGFR inhibition or by maintenance of epithelial cancer cell stemness upon decreased DUOX1 and ROS (1155), which could be permissive of EMT and cancer progression.

Accordingly, other studies report the downregulation of both DUOX1 and DUOX2 in lung cancer cell lines when compared to normal epithelial cells (1156). However, the effect of DUOX silencing vs. overexpression is very likely to be far more complex on account of the likely pleiotropic effects elicited by DUOX1 and DUOX2 via variable subcellular targets at different stages of cancer.

Still, DUOX1 epigenetic silencing was described in lung, liver and breast cancers which could be interpreted to mean that the DUOX has a tumor suppressive role to play (1156, 1157). To that point, CpG-rich promoter regions figure 11 of the protein (1156). In fact, when the expression of DUOX1 was decreased, higher proliferation rates and diminished migration and adhesion were found in nontumor mammary cells (MCF12A), indicating that DUOX1 downregulation may be involved in the development rather than progression of breast cancer (1158). However, in thyroid cancer, the downregulation of DUOX1 prevented the development of DNA damage after ionizing radiation and demonstrated a correlation between DUOX1, DNA damage, genomic instability and neoplasia (840).

Figure 11. Salient non-myeloid NOX signaling pathways in cancer development and progression.

Non-myeloid NOX2-derived ROS may facilitate cancer expansion by promotion of cell proliferation and a phenomenon described as “apoptotic escape”; ROS can evoke the capacity to redirect osteosarcoma from programmed death to survival. Adaptor Hic-5, which reportedly suppresses NOX4, is downregulated in metastatic cancer cells and unleashes NOX4-mediated invasiveness. NOX4-derived H2O2 causes oxidation of CR1 in BRAF propagating exaggerated NRAS-mediated MEK and ERK activation, and melanoma progression. NOX4 overexpression is linked to EMT and invasion via JAK2/STAT3. DUOX2 is implicated in PKC-induced Akt/MAPK activation and proliferation, migration, and invasion in colon cancer. NOX1 via MAPK induces cyclin D1 and proliferation as well as activates ADAM17-EGFR-PI3K-Akt signaling and increases the expansion of colon cancer cells.

Alternatively, DUOX2 upregulation appears generally to correlate with progressive malignancy, which is especially seen in cancers of the gastrointestinal tract, such as colon cancer (1159) and colorectal cancer (1114, 1160), gastric cancers and Barret’s esophagus (1160) and pancreatic adenocarcinoma (1161, 1162). Additionally, DUOX2 is elevated in breast, lung and prostate cancers (1159) and in colorectal cancer patients, DUOX2 expression correlates strongly with poor prognosis (1163). On the other hand, like DUOX1, DUOX2 may at times be downregulated in cancers, especially those of epithelial origin, including the lungs, liver, and gastrointestinal tract (1155, 1156, 1164).

Lastly, DUOX2 expression has been associated with chronic inflammation in human malignancies. Selectively increased DUOX2 and its maturation factor DUOXA2 were shown in pancreatic ductal adenocarcinoma cells wherein IFNγ, LPS, IL4, and IL17A upregulate the mature DUOX2 oxidase and cause an augmentation of VEGF and HIF1α and DNA damage (1161, 1165). Depletion of DUOX2 in HCC prevents PKC-induced activation of the Akt/MAPK signaling pathways and proliferation, migration, and invasion (1166) consistent with a pro-inflammatory and pro-carcinogenic effect of DUOX2.

Figure 11 depicts some of the salient NOX-dependent redox signaling pathways in cancer progression and metastasis. Overall, the verdict is still out in multiple cancer subtypes as to whether upregulation or downregulation of NOXs is causal to a particular cancer subtype or specific stage of the disease. Clearly, many more studies are necessary to carefully decipher the role of the NOX depending on the isoform, disease and its context.

4.16. NOX polymorphisms & their relation to disease

Dating to original reports of cloning of CYBA and CYBB, single nucleotide polymorphisms (SNPs) among NOX subunits have intrigued NOX biologists and provided key genetic and functional information on the NOX.

It is necessary to note, however, that many of the polymorphisms that have been studied are, indeed, single nucleotide variations (SNVs) as they are rare or extremely rare and, by definition, occur in less than 1% of the population as indicated in HGMD, dbSNP and/or gnomAD databases. Multiple types of mutations, some loss-of- and some gain-of-function and some disease-causing and others disease-preventative, can arise in the genome. Nevertheless, discussion of all mutations, i.e. including non-coding and deletions, insertions and rearrangement, would be prohibitive. For the purposes of this review, we primarily limit our discussion to coding, non-synonymous mutations in the exome which are SNPs that, by definition, have a minor allele frequency (MAF) of greater than one percent in at least one subpopulation of the global population. In a number of cases, we discuss SNVs if a particular subunit or component presents with no SNPs or if variations of intrigue have been extensively discussed in the medical literature. In a handful of cases, a coding, synonymous mutation (change in nucleotide, but no change in amino acid identity) may be discussed. At any rate, the rates of occurrence in a population are gathered primarily from the curated databases listed above.

Furthermore, we have found that mutant databases are not exhaustive in their content and, therefore, have included other variations of potential significance in the literature. It is also important to point out that as genome sequencing continually improves and SNP databases are updated, what once qualified as a SNV or SNP may no longer be considered as such and others may be found. All SNV and SNP entries are listed in ascending order of their position on the gene/protein from N-terminus to C-terminus (Table 3 and Figure 12). Finally, a serious shortcoming of multiple reports is that they do not determine whether NOX activity is altered or unchanged with the mutant, and the literature can be confusing in this regard. Lastly, in multiple instances when searching the references listed here and in Table 3, documentation of the mutant may not be readily apparent as it is sometimes not in the text of the paper but imbedded in supplemental file tables.

Figure 12. Location of NOX single nucleotide variations (SNVs) and single nucleotide polymorphisms (SNPs).

Diagrammatic representation of the specific locations of SNVs (red dots) and SNPs (black dots) in NOX catalytic subunits discussed in the text. EFh, EF-hand domain (calcium-binding domain); FAD, flavin adenine dinucleotide domain; FRD, ferric reductase domain; NADPH, nicotinamide-adenine dinucleotide phosphate domain; Peroxidase, peroxidase domain; TM, transmembrane domain. Note: For visualization purposes, only NOXs that contain pertinent coding non-synonymous mutations discussed in the manuscript are shown.

4.16.1. NOX1

To date, twelve total NOX1 coding mutations are documented among which two coding, non-synonymous mutations are identified in the HGMD database. The first is an SNP at nucleotide 721 (c.721C>T), which encodes replacement of an arginine with a cysteine at amino acid number 241 (p.R241C; rs142303829) and is associated with very early onset inflammatory bowel disease (1167). While it is rare in the global population and most cohorts, it is expressed in approximately 7% of the Ashkenazi Jewish population (dbSNP; https://www.ncbi.nlm.nih.gov/snp/). In a structural model, this change was found to cause a loss of positive charge in the extracellular loop between the 5^th and 6^th transmembrane domains (1167). The variant is also modeled to interfere with N-glycosylation (236–239) or elicit a structural effect on a potential disulfide bond (C243-C257) and influence the electron donor site for NOX1 to O₂ between transmembrane domains 3 and 5 (1167). Accordingly, expression of the mutant in Caco-2 cells results in abrogation of ROS and reduced capacity to fight off cyto-invasive bacteria (1167). In patients, the mutation is associated with acute ileocolitis and disturbed epithelial barrier dysfunction (1167).

The other NOX1 SNP (rs34688635) occurs at nucleotide 1078 (c.1078G>A; p.D360N) and alters encodement of amino acid 360 from aspartic acid to asparagine and thus represents the loss of a negative charge. This mutation is predicted to impede FAD binding and evoke a NOX1 loss-of-function; it presents in patients with very early onset IBD (777) and ulcerative colitis in Ashkenazi Jews (1168). Colonic epithelial cells expressing this mutant in HCT116 cells display significantly lower (~25%) basal ROS vs. WT controls (777). Allele frequency is 2.5% in the global population, 2.7% in Europeans, 1.2% Ashkenazi Jewish and 1.8% in one of two Latin American cohorts (dbSNP). The approximate location of these discussed mutations can be visualized in Figure 12.

4.16.2. NOX2

For the long-studied, paradigmatic NOX2 (CYBB), there are 768 coding mutations reported and two that qualify as SNPs. The first occurs at nucleotide 686 (G>A) resulting in an arginine to histidine conversion at amino acid 229 (c.686G>A; p.R229H; rs139670417) and a loss of a positive charge in the proximity of O₂ reduction at the hemoprotein’s egress site for electrons, which logically is expected to decrease O₂^·− generation. This mutation is considered potentially damaging in Crohn’s disease (1169). Unique among other polymorphisms found in the entire search for this section, rs139670417 is only found in African populations at a MAF range of 0.5 – 2.3 % (dbSNP).

The second SNP arises at nucleotide 1414 (c.1414G>A) and replaces glycine with a serine (p.G472S; rs13306300) in a stretch in the vicinity of NADPH-binding domains in the C-terminal tail of NOX2 (1170, 1171). It exhibits a very low allele frequency globally and, in most subpopulations, with the exception of Asian and East Asian communities (1.6 and 2.0%, respectively) and is classified as a disease-causing mutation (1170). Incidentally, in this study, a female carrier of X-linked CGD with this mutation revealed dysfunctional NOX and symptoms consistent with CGD including sinusitis, Epstein-Barr virus infection and recurrent pneumonia (1170).

Additionally, two SNVs have been documented as causal of CGD which terminate translation at amino acids 73 and 91 (105, 1172) and 3 others which replace amino acids within the terminal NADPH binding domains (41, 1173–1175).

Figure 13 depicts the four NOX2 mutations discussed above in the broader context of 85 NOX2 mutations across extracellular, TM and intracellular domains causing CGD (108).

Figure 13. Localization of single nucleotide variations (SNVs, black font) and single nucleotide polymorphisms (SNPs, gray font) for NOX2.

Virtual 3D-image was obtained from the protein data bank (PDB) with the ID# 8GZ3, and edited using the software UCSF Chimera, version 1.17.2 (May 2024). Briefly, chain B (cytochrome b-245 heavy chain, NOX2) of PDB 8GZ3 was isolated and colorized according to domain classifications: 6 transmembrane domains (TMDs) are depicted in orange, extracellular regions are depicted in gray, and intracellular regions are in magenta. Location of five mutations discussed in the manuscript are highlighted in cyan and magnified to show the amino acid side chain. These include rs137854588 (R73*, intracellular), rs137854591 (R91*, intracellular), rs139670417 (R229H, extracellular), rs137854585 (P415H, intracellular), and rs13306300 (G472S, intracellular) (Table 2). Bold text indicates an SNP. If the mutations result in the substitution of an amino acid, the residue represented in the structure is the changed amino acid. If the mutations result in a termination codon it is denoted by * and no change in amino acid is represented. Note that rs137854593 and rs151344490 (at amino acids 500 and 505) could not be represented because structural data at the C-terminal region was disordered and remains undetermined. Eighty-five mutations spanning extracellular, TMD and intracellular domains causing chronic granulomatous disease are shown in green (from Magnani, et al. (721), Table S5) are highlighted in green. Note that mutations Ala488Asp, His495Pro, Asp500Tyr/Phe/His/Asn/Gly, and Leu505Arg could not be represented because structural data was disordered and remains undetermined.

4.16.3. NOX3

Only six coding mutations were listed for NOX3 with none being SNPs. Admittedly provocative, a couple of extremely rare coding SNV mutations are included in the HGMD database that are potentially associated with autism (1176, 1177) and developmental disorders (1177, 1178). However, these remain controversial considering the very limited literature for NOX3 outside of its role in hearing, hearing loss and balance maintenance with only a couple of references, to our knowledge, referring to NOX3 elsewhere in the CNS (63, 67, 645, 647, 649, 979, 1179, 1180).

4.16.4. NOX4

No coding non-synonymous mutations appear to occur in NOX4’s exons but, broadening our lens a bit, a coding duplication at nucleotide 7 (rs34495256) and a mutation in the promoter region (rs1836882) are tantalizing SNPs to consider. Initially discovered in a pediatric patient with a lethal phenotype of respiratory distress, failure to thrive, pancreatic insufficiency, liver dysfunction, cardiomyopathy and cellular immune deficiency among other serious maladies, the NOX4 coding duplication at nucleotide 7 (c.7_9dupGAG; rs34495256) (1181), which results in two sequential glutamic acid residues at amino acid 3 and 4, ranges from a high frequency of 59% in the general population to 18% in some African populations (dbSNP). Nafisinia and coworkers reported that the mutation leads to decreased NOX4 expression and function; however, this mutation as the primary cause of disease was excluded based on (a) grandparents who were homozygous for the mutation and one of whom had fibromuscular dysplasia; and (b) the high frequency of the variant in the general population (1181). Thus, although it is unlikely to be the primary causative factor, it might be an influencing factor.

The second, a non-coding SNP (c.-8021A>G; rs1836882) (1182), appears 8021 bp upstream of the start codon in the NOX4 promoter region and occurs at relatively high frequency in the global population at 13% and ranges from 11% in Latin America and 13% in Europe to 29% and 27% in Asia and East Asia, respectively (dbSNP). Note: Because NOX4 is coded on the reverse strand [-], what is actually a T>C mutation on the reverse strand as reported in Liu et al. (1182) is listed as a A>G mutation on the forward strand, as is often convention in databases. Intriguingly, a mutation in this upstream promoter region caused a reduction in HNF3γ repressor binding and increased expression of NOX4 (rs1836882) and increased ROS in peripheral blood mononuclear cells exposed to serum from patients with high caloric intake (1182).

4.16.5. NOX5

No SNPs but two SNVs were found for NOX5 in its exon regions. One occurs at nucleotide 164 and causes an exchange of a histidine for an arginine at N-terminal amino acid 55 (p.H55R; no rs # identified) and is potentially associated with congenital heart disease which appears consistent with NOX5’s presence in ECs, SMCs and the heart (1183). However, no information is provided as to its participation in the disease (1183). Perhaps as part of a haplotype, this condition may be realized.

Another SNV at nucleotide 1336 (c.1336G>A; p.V446M; rs144275394) brings about a substitution of valine 446 to methionine (p.V446M) which might be expected to alter FAD binding (Figure 12). The SNV is putatively part of a large haplotype associated with neurodevelopment disorders (1184).

4.16.6. p22^phox

p22^phox is arguably the most extensively studied subunit of the NOX family for its involvement in disease polymorphisms with a total of 109 total mutations. As the co-stabilizing membranal subunit for NOXs 1 – 4, it has been implicated as a modifier of diseases ranging from CGD (126, 1185) to coronary artery disease (1186–1190), myocardial infarction (1191), hypertension (1192–1195), atherosclerosis (1196–1199), end-stage kidney disease (1200), diabetes (1201–1204), diabetic nephropathy (1201), cerebrovascular disease and stroke (1205, 1206), obstructive sleep apnea (1207), and organ rejection (1208, 1209).

The first and most widely reported on is a SNP discovered a few decades ago by Dinauer and coworkers and occurs in the global population at 66% frequency and as high as 92% in some Asian populations. The SNP represents a change at nucleotide 214 and engenders a change in amino acid 72 from a histidine to a tyrosine (c.214C > T; p.H72Y; rs4673) likely disrupting heme binding at the active site (1196) and deemed disease-altering in a wide range of pathologies ranging from multiple cardiovascular disorders and diabetes to CGD (126, 1210). Two studies demonstrated contradictory results of lower and higher respiratory burst in human polymorphonuclear neutrophils with the mutation, respectively (1199, 1211); the reason for this discrepancy is unclear but could be related to the preparations being freshly prepared neutrophils vs. HL-60 cells, respectively. It appears that the abrogation of oxidase activity due to the mutation is manifest in no change in cerebral infarction and coronary artery disease (1195, 1199), an increase in disease CGD severity (126), increased risk of hypertension in patients with end-stage kidney disease (1200), enhanced stroke risk (1205, 1206), and decreased risk of cardiovascular events post myocardial infarction (1192). In a meta-analysis of a broad swath of the literature including Chinese language publications, the authors compared the relative risk of disease in subjects with the TT vs. CC genotype and found an 87% and 78% higher risk of type II diabetes and diabetic nephropathy in Asians (1212). In a study of French patients with type I diabetes, the authors focused their attention on a significant association of the T-allele with arterial hypertension in this subgroup (1201). Interestingly, hypertension risk was illustrated as not significantly elevated in a metanalysis of a wide array of databases (e.g. PubMed, Embase, Web of Science, Google Scholar and others) consistent with studies by multiple other groups (1213–1215). In still another highly powered study, in a comprehensive metanalysis, the authors report a significant association between this C>T polymorphism with the risk for type II diabetes (1202). In fact, a 74% vs. 26% (1202) elevated risk was shown between the TT and T/C + C/C genotypes in Asians vs. non-Asians and the report went on to attribute the differences as influenced by geography, climate, diet, lifestyle and economic status. In contrast, in a Japanese subpopulation, no evidence of increased susceptibility could be found with respect to type I diabetes (1204).

*Note: C214T in p22^phox was originally referred to as C242T (411), a name still widely used in the literature.

A non-coding mutation which has garnered much attention in the literature but not easily found in databases merits discussion; rs1049255 alters the 3’ UTR region of p22^phox (c.A640G) which hints at altered stability or translation of the mRNA resulting in a change in p22 ^phox levels depending on the biological milieu or disease state (1216). In fact, it may modulate transcription factor binding, transcript stability, splicing, or translation efficiency. The SNP exists at a rather high frequency in the total global population (~50%) (OMIM & Ensembl). With such a high allele frequency, it is quite probable that the mutation’s effect is influenced by: (a) lifestyle/environment/diet; or (b) its inclusion in a haplotype that suppresses predisposition to disease. Unlike most others, the authors of this study do measure NOX activity (peripheral blood mononuclear cells) and find that diabetic patients with the GG genotype have approximately 50% higher activity than those with minor alleles AA and AG concomitant with significantly higher carotid artery intima thickness – a surrogate for subclinical atherosclerosis (1197). Incidentally, the AG allele is more prevalent in diabetic patients and is independent of age and sex (1197). Taken together, these studies are consistent with increased NOX activity in blood-borne cells and vascular cells in diabetic patients with this polymorphism predicting a proclivity for atherosclerotic vascular disease long term. In non-Hodgkin lymphoma, a large clinical exploratory study of patients with the GG genotype had a 72% less-favorable event-free survival rate and a 60% less-favorable overall survival rate concomitant with significantly lower p22^phox mRNA, protein and NOX activity in lymphoblastoid cells (1216). The authors offer no logical explanation for why levels and activity are lower and what the repercussions are for the disease. As such, this leaves open the question of whether lower ROS contributes to or is a consequence of the disease. In any event, the A650G polymorphism may be deemed a biomarker for both diseases.

A third and prevalent mutation worldwide is one which is found in the promoter region of p22^phox substituting A with G at nucleotide 932 upstream of the start codon (rs9932581). The MAF ranges from 28% in the African population to 59% in Latin America and has been attributed as a potential cause of a variety of serious diseases (dbSNP). The first notable disease is diabetic kidney disease wherein rs9932581 is one of a haplotype of five CYBA mutations (1201). rs9932581 is associated with a 59% increase in renal events and much higher prevalence of advanced nephropathy in humans at the end of follow-up associated with a lower glomerular filtration rate (1201). Moreover, a ~2-fold higher risk of arterial hypertension was discovered (1201). Consistent with this finding is a reported biologically significant 6 mmHg higher peripheral blood pressure in normotensive patients with this genotype (rs9932581) but not with rs4673 or rs1049255 (1194). San José and coworkers were rather extraordinary in further examining changes in subunit levels and its repercussions for NOX activity (1217). Indeed, they report a significant rise in phagocyte p22^phox protein levels in hypertensive patients and PMA-stimulated NOX activity with the mutation compared to normotensives. From there, they postulate that based on prior reports of an association of higher p22^phox and NOX activity with atherosclerosis, that this polymorphism would be consequential in that regard (1196, 1218) but do not appear to test a correlation with atherosclerosis (1217). In contrast, this phenotype of increased NOX activity with rs9932581 is linked to a decreased frequency of end-stage renal disease (1208).

4.16.7. p47^phox (Ncf1)

Sixty-three total mutations can be found for p47^phox. The first SNP of note is a substitution of G for A at nucleotide 247 resulting in a glycine to arginine modification in the protein (c.247G>A; p.G83R; rs139225348) in p47^phox is potentially associated with Crohn’s disease (1169). The SNP is found in the total global population at ~1%. Intriguingly, this SNP has been associated with reduced neutrophilic ROS in pediatric IBD (1169) and increased patient susceptibility to skin fungal infection (dermatophytosis) (1169, 1219, 1220). Another consequential mutation (rs201802880) which represents a G to A change at nucleotide 269 and an arginine to histidine shift at amino acid 90 (decreased ROS production) occurs in ~ 3% of the world population and is associated with systemic lupus erythematosus in Asians (3.5 odds ratio; OR), European Americans (2.6 OR), Africans and African Americans (2.0 OR) (1221). The SNP is associated with elevated incidence of Sjogren’s syndrome in Chinese (2.5 OR) and European Americans (2.4 OR) (1221). Moreover, as many as 16 and 17% of Japanese and Koreans, respectively, harbor the mutation (1221) (dbSNP) and it appears to be causal to rheumatoid arthritis in Koreans (1.7 OR) (1221). Li and coworkers report an increased incidence in CGD with this mutation and potential predisposition to often fatal Bacillus Calmette-Guérin disease (mycobacterium bovis infection) in the Chinese population within months after birth (1222).

4.16.8. p67^phox (Ncf2)

Three SNPs out of a total of 109 mutations were identified. The first occurring at nucleotide 606 and resulting in a synonymous outcome: alanine to alanine (rs17849501; c.606G>A; p.A202A) having a total global MAF of 5% and ranging from ~1% in Africa and African Americans to 5% in Europeans has been linked with systemic lupus erythematosus (1223). Curiously, although synonymous, the mutation putatively enhances translation efficiency by interfering with suppressor epigenetic interactions (1223). Second, a C>T mutation at nucleotide 836 resulting in substitution of threonine with methionine (c.836C>T; pT279M; rs13306581) is prevalent in various Asian subpopulations (gnomAD) and correlates with CGD (1174). Still another mutation results in a valine to alanine switch at amino acid 297 (rs35937854) and is associated with systemic lupus erythematosus (1223); its frequency is less than 1% in the entire global population but is in the 2 – 6 % range in Latin Americans and persons of African descent.

4.16.9. NOXO1/NOXA1

For NOXO1, seven total (all coding) mutations were located. No SNPs are reported; however, an intriguing SNV was associated loosely with cardiovascular traits, which is listed as occurring at < 1% in all cohorts. Curiously, as defined in HGMD, the mutation causes the replacement of a histidine with a proline at amino acid 198 resulting in a frame shift in coding (rs200352693). Though not confirmed, this shift is expected to create a non-functional protein and is provocatively described as being associated with an unexpected elevation in systolic blood pressure (1224).

For NOXA1, among seven total (all coding) mutations, one very rare SNV mutation (rs564363346) at nucleotide 952 (c.952G>A) resulting in a valine to methionine alteration (p.V318M) is associated with autism spectrum disorder (1176). Albeit rare and preliminary, this finding is provocative with respect to the possible role of NOXA1 in the regions of CNS associated with autism. One other rare SNV (rs953881310; c.1118C>T; p.S373L) potentially corroborates this association (1178).

4.16.10. DUOX1

For DUOX1, 47 total (44 coding) mutations have been identified with numerous rare to extremely rare SNVs correlating with hypothyroidism. For instance, a mutation arising at nucleotide 149 results in an arginine to glutamine substitution (c.149G>A; p.R50Q; rs371937236) associated with hypothyroidism is present in fewer than 7 in 100,000 individuals (1225). Another one deemed a hypothyroidism-causing mutation at nucleotide 2236 is found in 3 in 100,000 persons (1225). On the other hand, an intriguing SNP mutation of nucleotide 2773 C>T (p.R925W; c.2773C>T; rs145668427) is found to be robustly elevated in females with central serous chorioretinopathy. Chronic central serous chorioretinopathy is characterized by subretinal fluid accumulation leading to vision loss (1226). The association is rare globally but prevalent in the South Asian community at a rate of ~2%.

4.16.11. DUOX2

In contrast to DUOX1, 502 total (450 coding) mutations and four SNPs were located for DUOX2 and are associated almost exclusively with IBD and hypothyroidism. Indeed, a mutation in nucleotide 512 T to C results in a substitution of a leucine at amino acid 171 with a proline (c.512T>C; p.L171P; rs199957468) and correlates with a potential increased risk for IBD and disrupted gut microbiota-immune homeostasis; it is present in the global population at ~1.4 % (1000 Genomes database) (79). rs201197899 represents a very rare mutation that is only found in appreciable numbers in Northern Sweden (1.5%) resulting in a threonine to isoleucine substitution at amino acid 423 and an increased hypothyroidism notably in children with the congenital disease (79, 1225). Another substitution associated with hypothyroidism (and IBD) occurs at amino acid 1450 (tyrosine to histidine) (rs753591292) (79, 1227, 1228) (MAF, 1% in East Asians). Still another at amino acid 1469 (rs376623263) with 1 percent frequency in the South Asian population (79, 1225, 1227, 1229–1231).

4.16.12. DUOXA1/DUOXA2

Two noteworthy SNPs in DUOXA1, one at nucleotide 638 and the other at 937, causing a threonine to methionine and serine to glycine substitution at amino acids 213 and 313, respectively, have been potentially associated with hypothyroidism in the Latin American, African-descent and Asian populations (1232). For DUOXA2, one SNP resulting in a lysine to proline substitution at amino acid 204 has been found in the Danish population (3%) and is implicated in hyperthyroidism and increased risk of IBD (79, 1233) while another causes a translational stop at amino acid 246 and is associated with hypothyroidism at approximately 4 – 5 % in the Asian and American populations (1234, 1235). The latter mutation appears to result in a truncated protein lacking transmembrane helix 5 and the C-terminal tail of DUOXA2 and prevents reconstitution of DUOX2 in HeLa cells, i.e. it is retained in the ER (nondetectable on the cell surface) resulting in a complete loss in DUOX2-dependent H₂O₂ activity (1234, 1235).

5. NOX INHIBITORS, DRUGS, & CLINICAL TRIALS

The field has come a long way since antioxidant vitamins proved unsuccessful in clinical trials. The notion that “wholesale” scavenging or blockade of vital ROS by vitamins or inhibitors is a sensible goal in therapy appears behind us. While genetic knockouts and knockdowns provide seminal insight into the role of NOXs, what those phenotypes tell us may have little to reveal about biology or a specific NOX’s function. That is, one might envision NOX subunits in this context as a scaffold on which myriad non-NOX subunits pivot or are modulated. Thus, deleting a particular NOX or NOX subunit in a cell might be seen as pulling a “lynchpin” card from a “house of cards” and be expected to disrupt an array of many other heretofore unidentified NOX-associated signaling pathways. Out of a desire to avoid such collateral damage, to the extent that it was possible, we rationally designed targeted selective competitive inhibitors to limit off-target effects. Moreover, with the added benefit of druggability, came a quest for selective NOX isoform inhibitors. Despite the growing quantity and quality of structural information (108, 110, 116, 131, 134, 198, 266, 1236), there is still a way to go before isoform selectivity is optimized [for detailed information and a comprehensive overview of NOX inhibitors see (75)]. The following discussion attempts to provide an inexhaustive survey of NOX inhibitors from broad spectrum to isoform-selective inhibitors (Table 4).

Table 4.

Isoform specificity of selected peptides and small molecule inhibitors of NADPH oxidase.

Name	Structure	IC50s (μM) (* = Ki in μM)						Suggested Mechanism of Inhibition	Ref.
		NOX1	NOX2	NOX3	NOX4	NOX5	XO
DPI		0.2	0.1		0.1	0.02	yes	FAD site	(1237)
AEBSF									(1245)
VAS2870		↓	10.6 (Neutrophils) 2 (HL-60)		↓				(1335–1337)
VAS3947		↓	1–2 (HL-60 cells)		↓		ns		(1336)
APX-115 (Isuzinaxib)		1.08	0.57		0.63				(720, 1251, 1338)
Celastrol		0.41	0.59		2.7	3.13	ns	p47phox-p22phox interface	(1252)
Setanaxib(aka GKT137831/GKT-831)		0.11	1.75		0.14	0.41	>30 *		(655, 1339)
GKT 136901		0.160 *	1.530 *		0.17 *	0.45	>30 *		(437, 1255)
GKT-771(aka GKT310771)	Not disclosed	0.065(ki)	Inactive	Inactive	4.3(ki)	Inactive			(1103)
ML171		0.25	5.00		5.00		5.50		(1260)
NOS31									(401)
NoxA1ds	EPVDALGKAKV	0.02	ns		ns	ns	ns	Nox1-NoxA1 interface	(439)
Nox2ds-tat	RKKRRQRRRCSTRIRRQL	ns	0.74		ns	ns	ns	Nox2(B-loop)-p47 phox	(57, 740)
GSK2795039		>100	<2.88	>100	>100	>100	XO(29)		(1269)
Ebselen		0.15	0.5		ns	0.7	ns	p47phox and p67phox translocation	(1340)
ML090		0.025	Inactive		0.02	0.01			(1341, 1342)
CPP11G		ns	20		ns	ns	ns	p47phox-p22phox interface	(1070, 1071)
CPP11H		ns	32		ns	ns	ns	p47phox-p22phox interface	(1070, 1071)
CYR5099		Inactive	2.8–5.2						(628)
PHOX-I1		ND	3		Inactive			Rac1/2-p67 phox	(1274)
PHOX-I2		ND	1		Inactive			Rac1/2-p67 phox	(1274)
Naloxone			2					p47phox binding	(1279)
NCATS-SM7270									(1280)
GLX351322			40		5				(1289)
GLX481372		7	16	3.2	0.68	0.57			(1290)
GLX7013114	Not disclosed	Inactive	Inactive	Inactive	0.3	Inactive			(1290)
GLX481304		Inactive	1.25	Inactive	1.25	Inactive			(1290, 1291)
Grindelic Acid (ACD042)			>20		2.06				(1292)
ACD 084			>5.00		3.08	>5.00			(1292)
ML090		0.025	Inactive		0.02	0.01			(1341, 1342)
Gedunin								Hsp90 site	(1296)
Comp10		ND	3.3			95		p47phox-p22phox interface	(1284)
Comp33		ND	12			94		p47phox-p22phox interface	(1284)

(↓= inhibits but no IC₅₀ determined; ns= not significant; blank space=not determined;

^*= K_i)

5.1. Broad spectrum NOX inhibitors

Among the challenges faced with inhibitor development is the high homology among the NOXs. To that point, iodonium compounds, e.g. diphenylene iodonium (DPI), historically were the inhibitor of choice to inhibit phagocyte oxidase in vitro (1237). Because they irreversibly bind free or bound flavin (1238) or by binding a position on the porphyrin ring of heme (1239), iodonium compounds block electron transfer across NOXs (1240). The conserved mechanism of action of DPI across other enzyme systems i.e. nitric oxide synthase (NOS), mitochondrial complex I, cytochrome P450 and xanthine oxidase (XO) (1238, 1241–1244) render the drug illogical in complex systems. Despite that AEBSF was shown to inhibit p67^phox and p47^phox function (1245), it is a broad-spectrum serine protease inhibitor, thus with obvious limitations. Another commonly employed inhibitor is the plant catechol apocynin originally described to block p47^phox translocation to NOX2 (1246). However, it draws criticism for its dependence on peroxidase for its active form and for its ability to: (1) generate and scavenge ROS; and (2) inhibit multiple oxidases (1247). On the other hand, VAS2870 and VAS3947, are regarded as pan-NOXi’s due to their ability to target a particular highly conserved cysteine residue (differently positioned in different NOX subunits) in the DH domain of NOXs that binds the nicotinamide part of NADPH (1248). The polyphenol S17834 once thought to inhibit the NOX family was later shown to activate AMPK (1249, 1250).

APX-115 (Isuzinaxib) is termed a pan-NOX inhibitor targeting the conserved NADPH binding site on the dehydrogenase domain of NOX and exhibits no xanthine- or glucose oxidase inhibition or ROS scavenging ability (1251). From a preclinical standpoint, APX-115 inhibited NOX5 in a mouse model of diabetic nephropathy (720) and a desirable pharmacokinetic profile in mice offered a promise of success in two recent clinical trials for the treatment of diabetic nephropathy (Trial ID: NCT04534439) and COVID-19-induced pneumonia (Trial ID: NCT04880109). Celastrol displays similar properties. While it was originally characterized as a NOX2 blocker, it inhibits NOX4 and 5 as well albeit at higher concentrations (1252). Used in a variety of experimental models (1253, 1254), concerns have been raised about off-target effects including xanthine oxidase inhibition (1248).

5.2. Specific NOX-subunit inhibitors

5.2.1. NOX1i’s

The complexity of enzymatically competent NOX1 and NOX2 NADPH oxidases sets a stage for more targets for a selective inhibition.

Setanaxib (aka GKT137831/GKT-831), a widely designated NOX1 inhibitor with similar potency for NOX4 is one of the few NOX inhibitors to have made it to clinical trials to date (770, 1255). Its forerunner GKT136901 displayed a similar profile of NOX inhibition (1256). Note: GKT136901’s effect on NOX4 is also discussed in section 5.2.3. Both compounds exhibit weaker inhibition of NOX2 and NOX5 with no demonstrable XO inhibition (655, 770). They enjoyed early success in disease models ranging from fibrosis (770) to diabetes-related atherosclerosis (1257) and cancer (437) laying the groundwork for 3 clinical trials for biliary cholangitis (NCT03226067), type 2 diabetes/diabetic nephropathy (NCT02010242) and type 1 diabetes with albuminuria (U1111–1187-2609). Another GKT compound, GKT-771, has proved effective in HIV-related cardiovascular disease (1258). Over the years, concerns have been raised about whether these are indeed true NOX inhibitors. Unfortunately, a large number of these compounds have been shown to interfere with ROS assays (1259). In that vein, efforts to modify these compounds could narrow chemical properties and off-target effects and it seems appropriate, therefore, to recommend that the term selectivity and specificity be reserved only for compounds that inhibit one NOX isoform and no other enzyme system and do not scavenge ROS at therapeutic doses.

ML171, first introduced as a potent small molecule NOX1-specific inhibitor (1260), was shown to attenuate NOX1-dependent disease processes e.g., colon cancer invadopodia, thrombus formation (1260, 1261) and blood pressure (1262). However, ML171 also inhibits other NOXs and XO and therefore does not qualify as a NOX-selective inhibitor.

Another small molecule, NOS31, originally presented as a NOX1i was introduced in 2018. Displaying an IC₅₀ of 2 μM in a heterologous system, it also inhibited NOX4 (401). In the same study, it inhibited colon carcinoma and gastric cancer cell proliferation. Potentially, therefore, the compound holds promise for malignant diseases in which NOX1 is the dominant NOX and benefit vs. risk evaluations may supersede other factors considered when promoting a drug’s application.

NOXA1ds, to our knowledge, is the only selective NOX1 inhibitor available. Compound design was informed by the activation domain of p67^phox and the inhibitor encompassed a stretch of the subunit consisting of both a homologous and heterogenous flanking sequence in and adjacent to the corresponding putative activation domain of NOXA1 with one minor substitution (EPVDALGKAKV). This rational design yielded a selective, highly potent (low nM IC₅₀) NOX1 inhibitor (439). NOXA1ds has been applied in numerous preclinical models wherein it has been shown to be effective (384, 415, 417, 438, 1082, 1263, 1264). Criticism has been drawn to its peptidic formulation and thus skepticism of its oral bioavailability. However, as new delivery formulations emerge including nanoparticles and peptidomimetics, the unique selectivity of NOXA1ds for NOX1 might position it as a candidate for preclinical to clinical testing in some situations.

5.2.2. NOX2i’s

Unsurprisingly, much attention has been centered on NOX2 as the most extensively studied NOX. Similar to NOX1, the number of subunits and complexity of NOX2 lends itself well for selective targeting. However, unlike NOX1, attempts to target NOX2 come with an inherent bias of the potential downside of compromised host defense. Nevertheless, the bar is reportedly high for the degree of NOX2 inhibition (90–99% decreased ROS) in neutrophils to, in turn, block their antimicrobial function (535); thus, the therapeutic window for NOX2 targeting is predictably wide. As such, there is growing anticipation for widely applicable NOX2i’s.

NOX2ds-tat was the first selective inhibitor, to our knowledge, of any NOX (57, 740). It was designed in 1999 on prior knowledge generated by peptide walking on NOX2 (101, 229) with the intent to inhibit NOX2 (competitive inhibition of p47^phox binding to NOX2 B-loop) in AngII-mediated ROS production and blood pressure elevation (56). The highly selective peptidic inhibitor (740) is comprised of an 18-amino acid peptide (sequence RKKRRQRRRCSTRIRRQL) developed from a 9-amino acid peptide of the NOX2 intracellular B-loop sequence (NOX2ds; that allows the binding to p47^phox) chimerized with 9 amino acids of the HIV-TAT sequence (named tat) to facilitate cell permeation. Indeed, it was highly effective at abolishing ROS and delaying blood pressure development (57). In rather quick succession studies proved it successful in abrogating vascular hypertrophy and hyperplasia (364, 1265, 1266) and other disease processes ranging from diabetes to Parkinson’s and Alzheimer’s disease (1267, 1268). Indeed, it is still widely used in empirical studies and, like NOXA1ds, could hold promise in certain scenarios as a potent and selective inhibitor in diseases involving NOX2.

GSK2795039, a 7-azaindol derivative that displayed low nanomolar NOX2 IC₅₀ (1269) displayed minimal XO suppression. In spite of some observed electron donor activity, GSK2795039 showed strong NOX2 selectivity, pharmacokinetic properties and in vivo efficacy. In this study, the authors demonstrated that the compound competitively blocked NADPH binding raising concern as to how selectivity is conferred. In preclinical models, the drug exhibited protective effects in brain injury, prevented epileptic seizures, and staved off thrombosis during thrombocytopenia (1270–1273).

Ebselen congeners have drawn significant intrigue for an ability to prevent assembly of p47^phox and p22^phox. Despite concerns for intrinsic peroxidase activity, NOX2 inhibition is achieved at 2-log order lower concentrations than required for the peroxidase effect.

CPP11G & CPP11H were rationally designed around a common, unifying pharmacophore gleaned from a variety of broad-spectrum NOX2i’s like DPI, apocynin and AEBSF and were selected as optimally NOX2-effective and specific by way of high-throughput screening in a NOX2 oxidase plate assay. Cross-screening against NOX1, NOX2, NOX4 and NOX5 identified two NOX2-selective and non-ROS scavenging molecules so named as CPP11G & H (1070). The compounds were modeled to competitively inhibit p22^phox PRR domain binding to the supergroove of p47^phox (131, 1071). While IC₅₀s were originally thought to be in the micromolar range in heterologous NOX2 systems, mid-nanomolar IC₅₀s were determined in cells and tissue in vitro and in vivo (83, 1071). Beta testing has displayed a wide range blockade of NOX2-dependent ROS and anti-inflammatory pathways, macrophage adhesion, and restoration of hind-limb flow in response to TNFα (1071). Moreover, CPP11G permeates the blood brain barrier and blocks NOX2 assembly in a Parkinson’s rat model with limited toxicity (83).

Other noteworthy inhibitors include CYR5099, PHOX-I1 & -I2, perhexiline and naloxone.

CYR5099 in early development illustrated promise as an efficacious NOX2 inhibitor; however, to our knowledge it has only been cross-tested against NOX1 (with no effect) and molecular docking analysis predicts targeting of the NOX family-conserved FAD-binding region on the DH domain (628).

The only small molecules known to disrupt the binding of Rac1/2 with p67^phox are PHOX-I1 and PHOX-I2. Both were shown to inhibit ROS generation in cell-based assay yet curiously they did not block PMA-induced ROS (1274). Importantly, neither small molecule was capable of inhibiting NOX4. Until the PHOXs are cross-tested against other NOXs it will be impossible to moniker them as NOX2-selective. With that said, a handful of in vitro studies illustrate their potential (1275–1277). Perhexiline is an approved vasodilator in Australia and New Zealand that achieved success as a NOX2i both ex vivo and in vitro (1278), but might also inhibit NOX4 with greater efficacy (1248).

Naloxone is an opioid antagonist given to treat overdose. While it does bind to NOX2’s catalytic subunit and inhibit ROS in a variety of preparations, some findings suggest that its action is indirect and that it is effective at concentrations higher than that required for opioid receptor blockade (1269, 1279).

Among the newest NOX2i, NCATS-SM7270, derived from GSK2795039 has shown improved specificity, potency, and pharmacologic properties and in vivo, demonstrated decrease in mTBI- induced cortical cell death (1280). Also, recent efforts have focused on targeting the interaction of PRR region of p22^phox with the SH3-supergroove region of p47^phox. One strategy employed triproline mimetics (1281) whereas another used indole heteroaryl-acrylonitrile derivatives, C4 and C16 (1282). A third, using fluorescence polarization, a thermal shift assay and surface plasmon resonance as screening methods identified novel small molecule aminoquinolines that bind to p47^phox and block its interaction with p22^phox (1283). Using this third strategy, two new bivalent aminoquinoline NOX2 inhibitors (Comp10 and Comp33) were developed targeting the p47^phox-SH3 supergroove/p22^phox interaction (1284).

5.2.3. NOX4i’s

After NOX2, NOX4 is perhaps the best studied NOX isoforms and is implicated in a host of diseases. Rather uniquely, however, NOX4 is not always considered deleterious and has been shown to promote differentiation of various cell types. Thus, we and others have approached the targeting of NOX4 with special caution due to its suppression contributing to tumorigenesis or stymied development. Notably, peptide strategies to inhibit NOX4 have been met with varied success (1285, 1286). Most NOX4 inhibitors developed to date share the same issue faced with their use as NOX1 inhibitors. Indeed, Setanaxib and other GKT small molecules slated for NOX1 inhibition also inhibit NOX4 with near equal potency and efficacy. Setanaxib, to be specific, helped to define a dual role of NOX4 & 1 in liver fibrosis induced by TNFα and Ang-II (770) and is employed to block NOX4 in multiple fibrosis models.

GKT Compounds: Multiple GKT136901-related compounds pose similar protection in lungs (1255). In particular, the true efficacy of Setanaxib awaits outcomes from phase II clinical trials for IPF (ID: NCT03865927). Additionally, Setanaxib holds potential for NOX4 blockade in experimental models of diabetes and kidney disease as well as cardiac hypertrophy (714, 1287, 1288).

GLX351322, GLX481372, GLX7013114, and GLX481304. GLX351322 inhibited NOX4 (IC₅₀: 5 μM) in NOX4-overexpressing cells and restored insulin sensitivity in human islets cells (1289). One small molecule generated by SAR of the parent compound, GLX483172, blocked NOX4 and NOX5 with similar IC₅₀s in the mid-nanomolar range (but had no effect on NOXs 1 & 2). The other, GLX7013114 was reportedly selective for NOX4 and protected islet cells form glucose-induced cell death (1290). In addition, GLX481304 showed similar selectivity for NOX4 and NOX2 (IC50: ~1 μM) with virtually no effect on NOX1 (IC50: 100 μM) (1291). The latter did not test for NOX5 inhibition.

Grindelic Acid and ACD084. Grindelic Acid (ACD042) and ACD084 are plant-derived small molecules. In HEK cells expressing NOX4, both compounds selectively inhibited NOX4 whereas other related compounds displayed NOX2 or NOX5 blockade (1292). Enthusiasm for these compounds were dampened by evidence of direct inhibition of the NOX4 DH domain and lack of testing against NOX1.

5.2.4. NOX5i’s

ML090, a relative of ML171, inhibits NOX5 with similar potency to NOX1 and 4 but appears to inhibit NOX5 with greater efficacy (1293). NOX5-mediated blood-brain barrier ROS, interference and leakage in a “humanized” mouse in response to hypoxia/reperfusion was blocked by ML090 (1294). Despite other testing in aortic endothelial cells (1295), no extensive pharmacodynamic/pharmacokinetic profiling or further testing in experimental models could be found. Lastly, Gedunin is the most recent NOX5i and apparently acts via HSP90. It was modeled to bind to the C-terminal HSP90-binding domain and to destabilize the interaction (1296), other findings in the study illustrate that it induces antioxidant enzymes and another report suggests it has multiple effects including induction of heme oxygenase 1 and Nrf2 (1297). These manifold effects on other targets and even HSP90-binding region imply a wide spectrum of targets and thus lack of specificity for NOX5.

5.2.5. DUOX1 and DUOX2i’s

Physiological effects of DUOXs including thyroid hormone synthesis and host defense notwithstanding, DUOXs can drive adverse inflammatory responses, allergic reactions and certain types of cancer. DUOX1 participates in persistent EGFR activation, IL33 secretion, and airway allergic reaction (72, 974, 975). DUOX2 is overexpressed in a variety of cancers (1298) and also plays a role in airway inflammation (1299) and in a coordinated effect with iNOS in severe asthma (118, 150, 212, 1300). DUOX2 is also implicated in diabetic nephropathy and inflammatory diseases of the gut and bowels although in the case of the latter, reduced expression appears to be the culprit (735, 1301). Importantly, because of the ostensible bifurcation of the role of DUOXs in processes involving host defense and thyroid function, there has been no conspicuous drive to generate curative DUOX-selective inhibitors and limited attempts to apply existing pan-NOXi’s to the problem (1302, 1303) with one exception. A patent has been filed for the use of peptidic blockers to block a key cysteine’s alkylation in DUOX1, preventing its association with DUOXA (US Patent 20170128517). As with all NOXs, the advent of cryo-EM structures and a more rapid pace of X-ray crystallography foretells confidence in the generation of selective DUOX inhibitors or activators.

Table 4 summarizes known information of NOX inhibitors specificity, structure and possible mechanism of action.

6. CONCLUSIONS AND PROSPECTIVE

In summary, NOXs have been found in virtually all cell types and have evolved over billions of years. They are the only known class of enzymes that seemingly evolved for the purpose of a concerted and directed assault on microbes and not as wayward byproduct of biochemical process. Our hopes are that this review has aided in shifting conventional thinking to a central position for the NOX in biological signaling and in a wide array of physiological roles from innate and adaptive immunity to differentiation, angiogenesis and stem cell renewal. The phenotypic changes elaborated by NOXs are likely to be as varied as the distinct combination of isoforms and cell type or lineage in which they function and the degree to which they are induced.

From a disease standpoint, the participation of the seven mammalian NOX isoforms in pathology is ever-expanding and now prominently includes intractable brain disorders, failing cognition and depression. With this growing appreciation of the role of NOX in disease remains the challenge of holding harmless the role of NOXs in biology, organ development and renewal. Addressing these important issues will undoubtedly aid in the optimal development of therapeutic strategies that exploit selective NOX inhibitors.

Even though the breadth and depth of discovery has expanded exponentially over the past few decades, the study of NOXs is predictably still in its formative stages and expected to grow at a rapid pace in the coming decades. A far broader role of these monomeric and multimeric proteins’ participation in biology via oxidative and non-oxidative impacts are likely to be revealed and knowledge gleaned from orthologs across prokaryotic and eukaryotic organisms will predictably generate new hypotheses to be tested in mammals. Likewise, genomic, proteomic and bio-panning insights will reposit intriguing new insights in bioinformatic databases. Moreover, the details gained from newly resolved structures by cryo-EM and X-ray crystallography in varying degrees of complex formation, activation and stability will undoubtedly bring clarity to essential interactions between NOX subunits and auxiliary proteins. Accordingly, exciting opportunities abound for the targeting of known and yet-to-be discovered interactions that will inform the development of selective and effectual inhibitors for the treatment of cardiopulmonary to brain disorders mindful of the need to delimit inhibition to exaggerated and non-physiological roles of the NOX in distinct contexts.

Indeed, on the heels of a burgeoning wealth of information on NOX structure, we have entered a new era of discovery in the NOX field limited only by the imagination, inspiration and grit of scientists inveterate and new to the field. These exciting prospects are limited only by the imagination and wisdom of a new generation of scientists.

Clinical Highlights.

1. The NADPH Oxidase (NOX) family of multimeric enzymes comprises major oxidant generators in cells. Seven members of the family have been identified in widely diverse and tissue types in mammals.

2. NOX-derived oxidants are known to contribute to numerous physiological processes. However, an overabundance of oxidants, due to elevated NOX levels and/or excessive NOX activation, as well as decreased antioxidant defenses leads to tissue dysfunction and disease.

3. Regulation of NOXs, essential for proper cell signaling and evasion of host toxicity, is tightly controlled by a complex milieu of co-membranal and cytosolic modulating factors.

4. Given NOXs’ involvement in many diseases, ardent pursuit of selective NOX inhibitors is ongoing. With a few drugs currently in clinical trials, there is a promising preclinical landscape for both pan- and isoform-specific NOX inhibitors on the horizon.

5. The aim of this review is to provide a detailed overview of the NOX family members, their biochemistry and upstream and downstream modifiers in physiological and pathophysiological signaling.

GLOSSARY

αSMA

Alpha-smooth muscle actin

Aβ

Amyloid beta

ACS

Acute cigarette smoking

AIR

Auto-inhibitory region

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid

AMPK

AMP-activated protein kinase

AngII

Angiotensin II

CGD

Chronic Granulomatous Disease

COPD

Chronic obstructive pulmonary disease

DH

Dehydrogenase domain

DHE

Dihydroethidium

DOCA

Deoxycorticosterone

DPI

Diphenylene iodonium

EC

Endothelial cell

EGFR

Epidermal growth factor receptor

EMT

Epithelial-to-mesenchymal transition

eNOS

Endothelial nitric oxide synthase

ER

Endoplasmic reticulum

FAD

Flavin adenine dinucleotide cofactor

FBD

FAD binding domain

FDR

Ferric reductase

HCC

Hepatocellular carcinoma cells

HFD

High fat diet

HGMD

Human Gene Mutation Database

HMEC

Human microvascular endothelial cell

HSAEC

Human small airway endothelial cell

HUVEC

Human umbilical vein endothelial cell

iPSCs

Inducible pluripotent stem cells

LPS

Lipopolysaccharide

MAF

Minor allele frequency

MATER

Maternal antigen required by embryos

NAFLD

Non-alcoholic fatty liver disease

NASH

Non-alcoholic steatohepatitis

NBD

NADPH binding domain

NMDA

N-methyl-D-aspartate

NOX

NADPH Oxidase

NO^·

Nitric oxide radical

NOS

Nitric oxide synthase

NRROS

Negative regulator of reactive oxygen species

O₂^·-

Superoxide anion

ONOO^-

Peroxynitrite

OR

Odds ratio

PAP

Pulmonary artery pressure

PDI

Protein disulfide isomerase

PKA

Protein kinase A

PKC

Protein kinase C

PMA

Phorbol 12-myristate 13-acetate

Poldip2

Polymerase delta-interacting protein 2

PPHN

Persistent PH in lambs

Prdx

Peroxiredoxin

PRR

Proline rich region

PTP

Protein tyrosine phosphatases

PVR

Pulmonary vascular remodeling

RAGE

Receptor for advanced glycation end-products

ROS

Reactive oxygen species

RVH

Right ventricular hypertrophy

SNP

Single nucleotide polymorphism

SOD

Superoxide dismutase

SODD

Silencer-of-death domain

SNV

Single nucleotide variation

SSCs

Spermatogonial stem cells

STZ

Streptozotocin

STEAP

Six-transmembrane epithelial antigen of the prostate enzymes

TAM

Tumor-associated macrophages

Tks

Tyrosine kinase substrate

TLR

Toll-like receptors

TM

Transmembrane domain

TKS4/5

Tyrosine kinase substrate 4/5

TPR

Tetratricopeptide repeat

Tregs

Regulatory T cells

TRX

Thioredoxin

UPR

Unfolded protein response

UV

Ultraviolet

VSMC

Vascular smooth muscle cells

WKY

Wistar Kyoto rat

WT

Wildtype

XO

Xanthine oxidase