NOX1, 2, 4, 5: counting out oxidative stress

ROS assays

To determine if ROS are formed in a given system, for example cells, a tissue or an organ, a variety of ROS assays can be applied. The most commonly used ones are based on spectrophotometry (cytochrome c reduction, aconitase, nitro blue tetrazolium), chemiluminesence (e.g. lucigenin, luminol, L-012), electron-spin resonance and fluorescence [e.g. dihydroethidium (DHE), and its mitochondrially targeted derivative, MitoSOX, DCF-DA and Amplex Red]. For details of these and further assays, we refer to previous publications on this topic (Munzel et al., 2002; Daiber et al., 2004; Dikalov et al., 2007; Rhee et al., 2010).

In general, measuring ROS is confounded by several factors: (i) ROS levels are often very low. (ii) ROS, as the name suggests, are highly reactive, and short-lived. As such, (iii) their formation is influenced by, for example transition metals, which are often contaminants of buffers used in assays, resulting in analytical flaws. (iv) ROS can be highly localized and are not uniformly distributed within an organ or cell. This may not be adequately assessed by an ROS assay of the entire cell or an organ segment. It may thus remain impossible to accurately localize and quantify ROS in all relevant subcellular compartments. (v) For the same reason, scavengers of ROS, if not properly targeted, may not block the effect, as they do not reach the relevant subcellular compartment(s). (vi) Scavenging of ROS by antioxidants may result in the formation of other ROS with biological effects, for example scavenging of superoxide by SOD and some antioxidant vitamins results in increased levels of H₂O₂, another ROS with many biological effects (Dikalov et al., 2007) (Munzel et al., 2002).

Another consideration is that identification of the enzymatic source of ROS is often accomplished by substrate addition. For instance, with respect to NADPH oxidases, addition of NADPH to cell/tissue lysates or subcellular fractions may indeed be a tool to investigate NADPH oxidase activities. However, it is not known how NADPH can enter the cells after NADPH is added to intact cells or even tissue segments (Dikalov et al., 2007). Thus, pharmacokinetic issues may play a role when adding substrates but also pharmacodynamic factors (e.g. effects of the substrate) may influence ROS assay results.

These considerations explain why none of the available ROS assays is ‘perfect’. All available assays have limitations, such as limited specificity, sensitivity, extracellular versus intracellular detection, transient versus cumulative, validity, costs, reproducibility, etc. It is therefore generally recommended to use at least two different techniques, which have to show similar results (Daiber et al., 2004; Dikalov et al., 2007). But this is only a rule of thumb, of course, and in some studies the results of a third assay not showing the same result may not have been included in the published paper.

Finally, to proof the presence of ROS in the system of interest is not sufficient to demonstrate any involvement of ROS in disease development and progression. Experimental approaches should rather focus on the principle that removal of ROS by scavenging or, even better, deletion or specific inhibition of the potential relevant source can improve the condition, and that re-introduction of the ROS or its source causes the pathophysiological changes investigated (Dikalov et al., 2007). Unfortunately, such data are very rarely included in publications on ROS biology.

One of the most popular probes to measure ROS is DHE. Initially, it was thought that DHE reacted with superoxide to form ethidium, which intercalates with DNA, resulting in fluorescent signals, which reflected levels of superoxide. However, it is now known that the reaction between DHE and superoxide specifically yields 2-hydroxyethidium. In parallel, ethidium can be formed in the assay, but this is not a specific superoxide product. Rather, it reflects the redox status of a cell (Fink et al., 2004; Dikalov et al., 2007; Zielonka et al., 2009). The two products display different fluorescent spectra with an extensive overlap in fluorescence. Therefore, the commonly used fluorescence filters do not allow distinguishing between ethidium and 2-hydroxyethidium, but this is possible by separation of the two products with HPLC (Zhao et al., 2005). Still, in many publications it is wrongly stated that fluorescence recorded with DHE in tissue sections reflects superoxide. Nevertheless, we believe that DHE is an excellent tool for cellular localization and semi-quantitative analysis of ROS in tissue sections. Indeed, we believe that it is of minor relevance to distinguish between and quantify individual ROS (e.g. superoxide vs. H₂O₂), as they are rapidly converted inside a living cell, and it is unknown, which ROS is the relevant type.

Oxidative stress biomarkers

Studies on ROS-dependent molecular mechanisms pointed to specific mediators that play a key role in the development of CVD. Some of these can be seen as circulating blood biomarkers in order to assess cardiovascular risk and/or to monitor the efficacy of cardiovascular drug therapy.

As an example, plasma total antioxidant status (TAC) is measured as an indicator of the general antioxidant status of an individual (Dhamrait et al., 2004). There are, however, major concerns about its usefulness (Young, 2001), as for example in biological fluids the major contributor in most TAC assays is urate (>50% of the TAC activity). However, urate is of limited importance as an antioxidant in vivo (Young, 2001).

Isoprostanes, which have also been proposed as biomarkers, are prostaglandin-like compounds produced primarily from arachidonic acid catalysed by reactive oxygen and nitrogen species. They are classified as the ‘gold standard’ for the measurement of oxidative stress (Uno and Nicholls, 2010). However, most studies have used single spot measurements that can be misleading as the kinetics of isoprostanes in plasma and urine are different (Halliwell and Lee, 2010). Also, they should be standardized, but there is no agreement yet on how to do this (Halliwell and Lee, 2010).

Thiobarbituric acid reactive substances (TBARS) and malondialdehyde (MDA) are the most commonly used biomarkers of lipid peroxidation (Lykkesfeldt, 2007; Niki, 2009). Again, the validity of TBARS/MDA in bodily fluids has been criticized, for example for a lack of specificity, post-sampling MDA formation, antioxidants that can interfere with the assay procedure, and MDA derived from the diet.

Oxidation of lipids such as low density lipoproteins (LDL) is suggested to play a key role in the initiation and progression of atherosclerosis (Uno and Nicholls, 2010). The heterogeneity of oxLDL results in a large diversity of biomarkers, possibly with different clinical implications. Further, lipid peroxidation can probably not be used as a universal criterion of oxidative stress (Dotan et al., 2004).

As all of the above described markers have drawbacks, they are unreliable as an index of oxidative stress. However, emerging biomarkers may prove to be more reliable. For example, asymmetric dimethylarginine (ADMA), an endogenous inhibitor of NOS (Boger et al., 1998; Boger, 2006) is increased in conditions of oxidative stress, which may ultimately lead to endothelial dysfunction. ADMA may therefore be suitable for (pre-)clinical screening of atherosclerosis (endothelial dysfunction and systemic atherosclerosis).

The NO receptor soluble guanylate cyclase (sGC) (see Figure 1) generates cyclic GMP (cGMP) when NO binds to it. A biomarker of the NO-sGC signalling cascade is the phosphorylation of the cGMP-dependent protein kinase (cGK) substrate vasodilator-stimulated phosphoprotein (VASP) (Melichar et al., 2004). P-VASP can therefore be used as a marker for NO bioavailability. VASP is already used as a biomarker to monitor the efficacy of treatment with antiplatelet drugs (Weber et al., 2008). It is also a candidate biomarker to monitor the efficacy of sGC activators and stimulators in restoring the NO-sGC pathway (see also ‘Repairing ROS damage’ below).

By interacting with NO, NO₂^-, NO₂, etc., ROS can be modified to reactive nitrogen species (e.g. peroxynitrite, ONOO^-) that both oxidize or covalently modify proteins and DNA (Rizvi, 2009; Stephens et al., 2009). Nitrosylation of tyrosine (NO₂-Tyr) will ultimately lead to an altered structure and function of the protein, thereby changing the signal transduction pathway involved. Thus, NO₂-Tyr can be used as a general marker of nitrosative stress.

In conclusion, oxidative stress is ill defined in quantitative terms (Dotan et al., 2004) and cannot be accurately assessed with any of the current biomarkers and assays. It is more likely that a combination of methods and biomarkers is needed, and different types of ROS should be evaluated in parallel to generate patterns or indices of oxidative stress.

The ultimate pathomechanistic proof of principle though has to come from pharmacological modulation and cause–effect relationships. We will therefore discuss three approaches, antioxidants, inhibitors of ROS formation and repair of oxidative damage.

Antioxidants: a near death experience for the oxidative stress hypothesis

Laboratory studies showed protective effects of antioxidants and their association with various chronic diseases, including CVD. Consequently, an impressive body of literature has implicated modifiable lifestyle factors, including the diet, in oxidative stress and CVD (Stampfer et al., 2000; Lichtenstein et al., 2006; King et al., 2007; Chiuve et al., 2008; Ford et al., 2009; Imamura et al., 2009; Mozaffarian et al., 2009), and large cohort studies have shown an inverse relationship between plasma levels of antioxidants and the risk of CVD, other diseases and mortality (Riemersma et al., 1991; Gale et al., 1995; Singh et al., 1995; Sahyoun et al., 1996; Nyyssonen et al., 1997; Loria et al., 2000; Khaw et al., 2001). It thus appeared to make sense that CVD could be prevented by supplementation of antioxidants. Due to heavy promotion by supplement manufactures (Miller and Guallar, 2009) and the lay press, today millions of consumers use antioxidant vitamin supplements, often at high doses.

Nevertheless, after many large clinical trials and meta-analyses that have studied this issue, it became obvious that antioxidant supplementation did not result in any reduction in CVD morbidity and mortality. On the contrary, some large clinical trials came to the conclusion that antioxidant supplementation can even be harmful (Vivekananthan et al., 2003; Miller et al., 2005b; Bjelakovic et al., 2007; Dotan et al., 2009; Miller and Guallar, 2009). Considering the many millions of people take antioxidant supplements, and if, as an example, high dose vitamin E supplementation increases mortality even by a small amount, such as the estimated 4% (Miller et al., 2005b), these supplements would be responsible for many deaths (Miller and Guallar, 2009). The same holds true for other antioxidant regimens. It is now, for example, well established that β-carotene supplements increase the risk of cancer, CVD and mortality (Bjelakovic et al., 2007). Likewise, vitamin E supplements can no longer be considered safe for the general population (Dotan et al., 2009; Miller et al., 2005b; Miller and Guallar, 2009).

Consequently, not only antioxidants but the entire oxidative stress theory has been questioned despite the protective effects of antioxidants in pathophysiological animal models. One likely reason for the contrasting results between animals and humans is that much higher doses of antioxidants have been used in most of the animal studies compared to those administered in clinical trials (Griendling and FitzGerald, 2003).

Despite the disappointing results of antioxidant clinical trials, we will focus on evidence that the oxidative stress hypothesis is correct, but that the redox imbalance plays a different role than previously appreciated. In addition to possibly causing reductive stress (see Redox balance), several other mechanisms could explain adverse effects of antioxidant supplements. These include for example: (i) Inhibition of physiological ROS functions. In the vasculature, ROS play important roles as signalling molecules that regulate inflammation, cell proliferation, migration and differentiation, as well as vascular constriction and relaxation (Williams and Griendling, 2007). Untargeted scavenging of all ROS by antioxidants is likely to interfere with the physiological functions of ROS (Ristow et al., 2009; Schafer et al., 2009; Ristow and Zarse, 2010). (ii) Antioxidants are not targeted to the precise localizations where ROS concentrations are elevated, as oxidative stress is not evenly distributed throughout an organism. (iii) The rates of reaction between antioxidants and ROS are often lower than the reaction rates between ROS and their targets (Gotoh and Niki, 1992; Thomson et al., 1995). (iv) After their reaction with ROS, antioxidants can become radicals themselves, initiating new radical chain reactions, and thus cause harm (Vertuani et al., 2004) (Bowry et al., 1992). (v) The association between antioxidant plasma levels and the incidence of chronic diseases may not be due to antioxidative effects. Indeed, beneficial non-antioxidant actions of ‘antioxidants’ have to be considered, such as their influence on gene expression, which cannot be mimicked by supplementing single or a few antioxidants.

Nevertheless, the failure of clinical studies does not disprove the role of oxidative stress in CVD. The negative data have only shown that the antioxidant supplementation approach was apparently the wrong way to counteract CVD, and better approaches are needed.

One alternative approach that has been proposed recently is a more targeted supply of antioxidants to key subcellular locations, for example mitochondria (Dikalova et al., 2010a). Interestingly, the most abundant vascular NOX isoforms, NOX4 (see Preventing oxidative stress), has been located in the mitochondria (Block et al., 2009; Ago et al., 2010; Graham et al., 2010) and may thus itself represent a source of mitochondrial ROS.

As such, targeting antioxidants to mitochondria in angiotensin II (Ang II)-induced and deoxycorticosterone acetate (DOCA) salt hypertensive mice reduced blood pressure (Dikalova et al., 2010a). Despite these intriguing findings, the contribution of mitochondria to oxidative stress remains controversial. On the one hand, it was suggested that mitochondria are the initiating source of ROS under some circumstances, with NADPH oxidases secondarily activated by mitochondrial ROS (Dikalova et al., 2010a). On the other hand, it was proposed that NADPH oxidases stimulate mitochondrial dysfunction resulting in mitochondrial ROS release (Doughan et al., 2008; Ago et al., 2010). While the order remains to be determined (and may be different in different disease settings), both concepts are in line with the hypothesis of ROS-induced ROS release resulting in a vicious cycle. Nevertheless, the strategy of mitochondria-targeted antioxidants is still in its infancy. Whether this concept is translatable into human health benefits remains to be shown, keeping in mind that the translation of general antioxidant supplementation from animals to humans obviously failed. A more promising approach is to tackle oxidative stress at its roots by inhibiting the disease-relevant sources of ROS. One such source is the family of NADPH oxidases, which holds great promise for future treatment of CVD.

NOX5: a special case

NOX5 was the last NOX isoform to be identified (Banfi et al., 2001; Cheng et al., 2001), and much less is known about this NOX compared to the other isoforms. This lack of knowledge is mainly due to the absence of NOX5 in rats and mice making studies on NOX5 in these species impossible. Genetically, NOX5 is the most distinct of the NOX isoforms and exists in at least five splice variants (Cheng et al., 2001). It is present in the human vasculature (BelAiba et al., 2007; Guzik et al., 2008; Jay et al., 2008). NOX5 activity apparently does not depend on any subunits. Another important feature of NOX5 is its N-terminal EF-hand regions, which results in direct activation of NOX5 by an increase in intracellular calcium (Banfi et al., 2001). NOX5 can also be activated without elevating intracellular calcium by protein kinase C-dependent phosphorylation. This modification increases the calcium sensitivity of NOX5 and thus permits a higher level of activity at resting levels of intracellular calcium (Jagnandan et al., 2007). Another mode of calcium sensitization occurs via binding of calcium-activated calmodulin (Tirone and Cox, 2007). Importantly, NOX5 mRNA protein levels were increased in coronary arteries from patients with coronary artery disease (CAD) compared to patients without CAD. In correlation, calcium-dependent NADPH oxidase activity, which is very likely to reflect NOX5 activity, was increased sevenfold in CAD coronary arteries. While NOX5 was mainly localized in the endothelium in early lesions, its protein levels were up-regulated in VSMC of advanced coronary lesions and lost in the endothelium (Guzik et al., 2008). These data suggest a prominent role of NOX5 in human CAD.

We believe that the complexity of the NADPH oxidase family will allow to specifically inhibit the formation of pathophysiologically relevant ROS as opposed to untargeted scavenging. Indeed, an impressive amount of data shows that NOX isoforms are up-regulated in many cardiovascular as well as other chronic diseases and their risk factors, for example atherosclerosis, heart failure, myocardial infarction, fibrosis, diabetes mellitus, aging, smoking, inflammation, as well as sepsis, cancer and neurodegeneration. We here focus on the role of NADPH oxidases in hypertension and stroke. These are two examples, where there is compelling evidence for the involvement of individual NOX isoforms that is based on the use of specific pharmacological NADPH oxidase inhibitors and/or NOX knockout mice: NOX1 in hypertension and NOX4 in stroke. With respect to the possible involvement of NADPH oxidases in other disorders, we refer to recent reviews (Cave et al., 2006; Bedard and Krause, 2007; Brandes et al., 2010; Lassegue and Griendling, 2010).

Current drugs that interfere with NADPH oxidases

Statins, or 3-hydroxy-3-methyl-glutaryl reductase inhibitors act by inhibiting geranylgeranylation of Rac (Wassmann et al., 2002; Chen et al., 2008), which is an important component for the activation of some NOX isoforms (Bedard and Krause, 2007; Opitz et al., 2007). This inhibition prevents the translocation of Rac to the cell membrane and results in inhibition of Rac-dependent NADPH oxidases (Wassmann et al., 2002; Chen et al., 2008). This may explain at least some of the pleiotropic, cholesterol-independent atheroprotective effects of statins. It was indeed reported that, for example, atorvastatin decreased aortic superoxide levels in SHR. Besides, it also decreased NOX1 and p22phox, but not NOX2 mRNA levels in vessels of treated SHR, and catalase mRNA was up-regulated (Wassmann et al., 2002). Cerivastatin and atorvastatin treatments in mice improved endothelium-dependent relaxation to acetylcholine. Interestingly, withdrawal of statins attenuated endothelium-dependent relaxation compared to control animals, and the relaxation was restored by the SOD mimetic, tiron. While vascular ROS were unaffected by statin therapy, which may be due to low NADPH oxidase activity under resting conditions, ROS levels increased during withdrawal. These effects did not occur in NOX2 knockout mice. In human umbilical vein endothelial cells (HUVEC), statin treatment reduced NADPH oxidase activity, and withdrawal resulted in profound translocation of Rac to the membrane and transient increase in NADPH oxidase activity (Vecchione and Brandes, 2002). Thus, NOX2 may not only be inhibited by statins, but also play a role in rebound phenomena after withdrawal of statins, resulting in an overshoot activation or translocation of Rac.

In humans, statins have been shown to improve endothelial function (Treasure et al., 1995; O'Driscoll et al., 1997; Tsunekawa et al., 2001). Further, pravastatin has been reported to have blood pressure lowering effects (Glorioso et al., 1999; Kawano and Yano, 2006). Nevertheless, statins are definitely not specific NADPH oxidase inhibitors.

Angiotensin converting enzyme inhibitors and angiotensin receptor blockers (sartans) also have beneficial effects to which indirect inhibition of NADPH oxidases is likely to contribute, as Ang II is a potent stimulus for NADPH oxidase activity in vascular cells (Griendling and Ushio-Fukai, 2000; Wingler et al., 2001; Williams and Griendling, 2007; Datla and Griendling, 2010). Indeed, the hypertensive effects of Ang II appear in part to be mediated by NADPH oxidase-derived ROS (Dikalova et al., 2005; Matsuno et al., 2005; Gavazzi et al., 2006).

Considering these effects from established drugs and the amount of evidence for the involvement of NADPH oxidases in endothelial dysfunction, associated CVD and other disorders, inhibition of NADPH oxidases as novel therapies is a promising strategy to causally treat and prevent these conditions and/or suppress associated end-organ damage.

Old and new NADPH oxidase inhibitors For many years, specific small molecule inhibitors were not available, and this represented a major bottleneck of research into oxidative stress. However, more recently, several novel, apparently more specific and drug-like NADPH oxidase inhibitors have been published.

The most commonly used NADPH oxidases inhibitors, diphenylene iodonium (DPI) and apocynin are unspecific (O'Donnell et al., 1993; Majander et al., 1994; Vejrazka et al., 2005; Riganti et al., 2006; Williams and Griendling, 2007; Aldieri et al., 2008; Heumuller et al., 2008; Schluter et al., 2008; Selemidis et al., 2008; Jaquet et al., 2009; Tazzeo et al., 2009; Brandes et al., 2010; Castor et al., 2010; Wind et al., 2010b). Thus, their effects cannot be solely attributed to inhibition of NADPH oxidases. Also, 4-2-amino-ethyl-benzolsulfonyl-fluoride (AEBSF) is not a reliable NADPH oxidase inhibitor, as it irreversibly inactivates serine proteases (Diatchuk et al., 1997) and interferes with the most commonly used assays for ROS (Wind et al., 2010b). Studies performed with these compounds must therefore be interpreted with greatest caution. With respect to other rather rarely used and also unspecific inhibitors, we refer to a detailed review on NADPH oxidase inhibitors (Jaquet et al., 2009).

A more reliable tool to inhibit NADPH oxidases are cell-permeable peptide-based inhibitors such as gp91ds-tat (Rey et al., 2001), whereas another peptide inhibitor, the naturally occurring peptide PR-39 (Gudmundsson et al., 1995; Shi et al., 1996), is rather unspecific as it binds to SH3 domains of many proteins (Cai et al., 2003). Despite their value in experimental studies, peptides have the disadvantage of low bioavailabilities. Therefore, they have limited potential as therapeutic agents. More details about these peptides are reviewed in Selemidis et al. (2008). More recently, professional screening programmes for NADPH oxidase inhibitors have resulted in the discovery of several novel small molecule, non-peptidergic NADPH oxidase inhibitors.

Interestingly, many novel inhibitors show some common structural features (Table 1), that is, the inhibitors generally are flat and lipophilic aromatic heterocyclic compounds.

Table 1. Novel small molecule NADPH oxidase inhibitors

Inhibiting phosphodiesterases

Inhibition of phosphodiesterases (PDE), in particular PDE type 5, augments NO-cGMP signalling irrespective of whether it was pathophysiologically reduced beforehand or not. Its first indication was erectile dysfunction, an early marker of CVD (Thompson et al., 2005), where NO signalling may indeed be dysfunctional. In the more recent indication for PDE5 inhibitors, pulmonary hypertension, this does not seem to be the case (Kirsch et al., 2008). Nevertheless, PDE5 inhibitors are effective in pulmonary hypertension (Wilkins et al., 2008; Galie et al., 2009).

Stimulating sGC

For the same indication, another approach has been developed also augmenting NO-cGMP signalling, sGC stimulation. This compound class allosterically enhances the NO-induced cGMP formation by Fe(II)sGC, so that low, submaximal concentrations of NO are potentiated and exert the same cGMP formation as higher concentrations of NO. In addition, these compounds have a small direct stimulating effect on sGC. In principle this compound class is applicable to all disease states where reduced bioavailability of NO, for example by oxidative stress, plays a pathomechanistic role. The NO levels will remain unchanged and low, but their effects will be augmented. The original compound, YC-1 had still some off-target effects (Galle et al., 1999; Li et al., 2008). However, the successor compounds from Bayer, BAY 41-2272 and BAY 63-2521/riociguat (Mittendorf et al., 2009) were devoid of this at therapeutically relevant concentrations (Bischoff and Stasch, 2004). Other indications for this compound class may include systemic hypertension (Zanfolin et al., 2006).

Activating oxidized and apo-sGC

Another principle, sounding confusingly similar, is sGC activation. However, mechanistically this is a profoundly different principle. Under conditions of oxidative stress, the NO receptor Fe(II)sGC can be oxidized to Fe(III)sGC and eventually looses its heme. It then becomes ubiquitinylated and degraded (Meurer et al., 2009). Heme-free apo-sGC is essentially unresponsive to NO, but can bind sGC activators that occupy the empty sGC heme binding site and re-activate apo-sGC to the same V_max as Fe(II)sGC in the presence of NO (Evgenov et al., 2006; Stasch et al., 2006). The first compound of this class, BAY 58-2667/cinaciguat, does this by also preventing sGC ubiquitinylation and thereby preventing sGC degradation. Later compounds such as HMR-1766/ataciguat (Zhou et al., 2008) are devoid of this effect (Hoffmann et al., 2009). Cinaciguat is currently developed for acute heart failure, an indication that may benefit from an unusual vasodilatory profile, that is, preference for microvasculature over large conducting blood vessels and diseased over healthy blood vessels. Thus, sGC activators are the first vasodilator compound class that specifically dilate diseased blood vessels making steel phenomena highly unlikely. Other development targets include peripheral artery disease and chronic pain (Schmidt et al., 2009).

Re-coupling uncoupled eNOS

Finally, a third target of oxidative stress is uncoupled eNOS, a process that involves oxidation of tetrahydrobiopterin (Reif et al., 1999; Kotsonis et al., 2000), an essential NOS cofactor, and accumulation of ADMA (Luo et al., 2010), a competitive inhibitor at the enzyme's substrate binding site for L-arginine (Cardounel et al., 2007). Uncoupling means that oxygen is activated but no longer transferred onto L-arginine. Thus, uncoupled eNOS releases ROS, similar to NADPH oxidases. Substitution or regeneration of tetrahydrobiopterin alone or in combination with supplementation of L-arginine can recouple eNOS and restore NO synthesis (Settergren et al., 2009).