The alternative oxidases (AOXs) are ubiquinol-oxidoreductases that are members of the diiron carboxylate superfamily. They are not only ubiquitously distributed within the plant kingdom but also found in increasing numbers within the fungal, protist, animal and prokaryotic kingdoms. Although functions of AOXs are highly diverse in general, they tend to play key roles in thermogenesis, stress tolerance (through the management of radical oxygen species) and the maintenance of mitochondrial and cellular energy homeostasis. The best structurally characterised AOX is from Trypanosoma brucei. In this review, we compare the structure of AOXs, created using homology modelling, from many important species in an attempt to explain differences in activity and sensitivity to AOX inhibitors. We discuss the implications of these findings not only for future structure-based drug design but also for the design of novel AOXs for gene therapy.
The alternative oxidases (AOXs) have long been considered enigmatic with respect to their nature, function and indeed importance as a respiratory chain protein . This arises from the early observations that respiration in certain thermogenic plant tissues was resistant to conventional inhibitors of the cytochrome pathway such as KCN and azide, leading to the suggestion that such activity arose as a result of incomplete inhibition of cytochrome oxidase (COX), i.e. the excess oxidase hypothesis . It is now well established that such resistance is in fact due to the presence of a terminal oxidase which is located on the substrate side of the cytochrome bc1 complex. AOXs are monotopic proteins which in mitochondria are attached to the inner surface of the inner membrane and act as ubiquinol oxidases catalysing the 4-electron reduction of oxygen to water in a non-protonmotive fashion [1,3,4].
AOXs belong to a subfamily of the diiron carboxylate proteins, which include methane monoxygenese, ribonucleotide reductase R2 subunit, desaturase and rubrerythrin . For decades, the AOX was considered only to be important as a respiratory chain complex in thermogenic plants such as the Amorphophallus titanum (the titan arum) and their smaller cousins in the Arum family (Arum maculatum). However, research by McDonald and colleagues over the years have clearly established that AOXs are ubiquitously distributed not only throughout the plant kingdom but also within the fungal, protist, animal and prokaryotic kingdoms [4,5]. Of particular significance was the finding that many protist organisms which contain an AOX, such as Trypanosoma brucei, Cryptosporidium parvum, Blastocystis hominis and the microsporidia, are highly pathogenic to humans . In these organisms, AOX plays a critical role in their respective life cycles and since the protein is absent from mammals, it has become a novel therapeutic target for the treatment of diseases caused by such organisms [6,7]. Although AOX is not found in mammals, the further development of bioinformatics tools has revealed that the protein is also distributed throughout the animal kingdom. Notable invertebrate species expressing AOX include marine worms, Arenicola marina and Sipunculin nudis, mussel, Genkendia demissa, the pacific oyster Crassostrea gigas and the sea squirt Ciona intestinalis . Although the full extent of the distribution of AOX within prokaryotes is still relatively uncertain, it is becoming evident that AOX is quite widespread amongst proteobacteria, including Novosphingobium aromaticivorans, members of the Vibrio genus and Thiobacillus denitrificans .
In addition to its wide distribution in prokaryotes, AOX is known to be present in numerous fungi of agronomic importance such as Blumeria graminis, Magnaporthe oryzae, Septoria tritici, Ustilago maydis and Botrytis cinerea [10–12]. Fungal AOXs are known to be involved in stress responses similar to those observed in plants, namely maintaining radical oxygen species (ROS) levels and the management of cellular energy [13,14]. Typical economically important fungicides tend to target the fungal respiratory chain, thereby preventing ATP generation and ultimately resulting in cell death. Such fungicides include thiabendazole , which targets fumarate-reductase, and strobilurins such as azoxystrobin, which target the Qo (ubiquinone-binding site outer surface) site within the cytochrome bc1 complex . Recently, there has been a rise in the emergence of fungicide-resistant strains of phytopathogenic fungi, such as Magnaporthe grisea and B. graminis, which is posing a serious threat to the agricultural industry and worldwide food security . In several species, including S. tritici, fungicide treatment increases the expression of AOX [18,19]. Magnaporthe graminicola , M. grisea  and Sclerotinia sclerotiorum  all display an increased expression of the AOX in the presence of Qo inhibitors, suggesting that this may be a mechanism to allow fungal growth to continue thereby mimicking so-called fungicide resistance .
What is evident from the above discussion is that the emergence and success of both phytopathogenic and parasitic pathogens appear to be a reliance on the presence of an active AOX enabling such organisms to survive under challenging environmental conditions. If such pathogens are to be controlled in the future, then the development of more specific AOX inhibitors is urgently required.
The goal of the present review is to investigate if there are common structural features within the AOX protein that would aid in future structure-based rational design of potent and safe antiparasitic and phytopathogenic inhibitors. We have used the crystal structure of the trypanosomal AOX (TAO)  to enable, through homology modelling, a detailed examination of the structure of AOXs from many important species. In particular, comparisons of sequence alignments of multiple AOXs and AOX homology models have revealed that although there is a high degree of conservation of residues involved in the ubiquinol/inhibitor-binding site and in the mechanism of oxygen reduction, significant differences in the nature of the hydrophobic cavity leading to the active site are apparent. We suggest that such differences may not only explain experimental differences in inhibitor sensitivities, but may also prove important in the design of species-specific AOX inhibitors.
Overall structure of the AOXs
We have previously elucidated the crystal structure of TAO which confirmed many aspects of existing proposed models [1,22]. As predicted from sequence homology , AOX is part of the diiron carboxylate superfamily, consisting of a diiron core surrounded by a four-helix bundle, with the core ligated by four fully conserved glutamate (E123, E162, E213 and E266) and two histidine residues (H165 and H269). TAO exists as a homodimer where each monomer consists of six long and four short α helices, with helices α2, α3, α5 and α6 forming the four-helix bundle (Figure 1A). There is a large hydrophobic region found on one side of the monomer at helixes α1 and α4, proposed to be the membrane-binding interface (Figure 1B). This region is scattered with conserved hydrophilic residues thought to interact with the negatively charged phospho-head group of the lipid bilayer. The dimer interface of TAO (see Figure 1B) contains multiple conserved residues, suggesting that all AOXs may exist as dimer, contrary to previous findings . Alongside the native crystal structure of TAO, the protein has also been co-crystallised with ascofuranone and colletochlorin B, both potent inhibitors of AOX activity . These inhibitors reside within a hydrophobic cavity leading from the membrane side of the protein to the diiron core (Figure 1C), which we have previously postulated to constitute the ubiquinol-binding pocket, given the structural similarities of the inhibitor to the substrate .
Crystal structure of TAO depicting helices, surface and hydrophobic cavity.
The guanidino groups of R96, R118 and the hydroxyl group of T219 are predicted to be involved in interacting with the hydroxyl group of quinol orientating the para-hydroxyl group between the iron core and Y220 for reaction . The other highly conserved residues within this region play a more structural role; i.e. E215 and D100 stabilise the position of R96 for reaction with ubiquinol, whereas L122 and L212 form a bottleneck within the hydrophobic cavity (Figure 1C), ensuring that the ubiquinol head group is in the correct orientation to interact with the diiron core [25,26].
Interestingly, Pennisi et al.  recently identified some protist species which contained genes for an AOX, the sequence of which indicated that the key ubiquinol-interacting ligands, D100 and R118, appear not to be conserved. Pennisi et al. found that, within the AOX sequences of both Chondrus crispus and Naegleria gruberi (protists), D100 and R118 were both substituted as histidines. Such a finding would cast grave doubts upon the role of D100 and R118 in ubiquinol binding . Since neither of these protist AOXs have, however, been characterised, there is little evidence to confirm whether or not such genes produce functional AOX proteins.
One of the final regions of high conservation within all AOX sequences published to date is a proposed proton-coupled electron transport (PCET) network within the secondary ligation sphere. The PCET pathway comprises N161, D265, W246 and W65; these residues are, in the majority of cases, fully conserved across the AOX family and have been shown to fully align with similar networks within the diiron group of proteins . Furthermore, site-directed mutagenesis suggests that they are critical for the oxygen reduction pathway catalysed by AOX .
Does high conservation mean comparable activities?
Although the AOX family contains a highly conserved active site, striking differences are seen in ubiquinol oxidase activity between different AOXs. Of those AOXs which have been purified to date, TAO has the highest Vmax [600 µmol QH2 (ubiquinol) oxidised min−1 mg−1 ), followed by AOX from thermogenic plants such as A. maculatum and Sauromatum guttatum (30 µmol QH2 oxidised min−1 mg−1  and 37 µmol QH2 oxidised min−1 mg−1 ), whereas both N. aromaticivorans and C. intestinalis possess low Vmax values (∼3 µmol QH2 oxidised min−1 mg−1 ). Given the high levels of conservation of the primary and secondary ligation spheres throughout the AOX family, the cause for such high levels of variability in quinol oxidation is unlikely to be mechanistic in nature . Whilst there is evidence in the literature for various small-molecule activators of plant AOXs (such as pyruvate or α-keto acids) [32–35], the purified AOX proteins described above tend not to display sensitivity to pyruvate and analysis of structural models across the AOX family suggests that it may well be the nature of the hydrophobic cavity itself, which plays a key role in regulating overall activity.
We have recently shown that the leucine gate, comprising L122 and L212 (see Figure 1C), appears to direct ubiquinol (and analogous inhibitors) into the active site by allowing the head group to ‘dog-leg’ towards the diiron core . Mutation of either of these residues to alanine resulted in not only a significant reduction in ubiquinol oxidase activity, but also sensitivity to quinol mimetic inhibitors . Conversely, it was found that bulkier inhibitors increased in potency following such mutations when compared with the wild-type protein, which is more than likely due to an increase in the ability of the inhibitor to traverse the tunnel. Such results have led us to explore the importance of the residues lining the cavity leading to the active site.
Characteristics of the hydrophobic cavity
Examination of the structure of TAO reveals that the residues lining the tunnel leading to the active site are primarily hydrophobic. To determine if there is a common trend in residues lining the ubiquinol channel amongst the AOXs, the structures of Saromatum guttatum AOX (SgAOX), Ciona intestinalis AOX (CiAOX) and Novosphingobium aromaticivorans AOX (NaAOX) were created via homology modelling and compared with the original trypanosomal structure.
Exploration of the tunnels and cavity of each AOX illustrated in Figure 2 reveals that hydrophobic residues, which are highly conserved, constitute the backbone of all of the cavities (Figure 2A). Although polar residues only constitute ∼17% of the residues lining the hydrophobic cavity of TAO and SgAOX, they are nevertheless the only AOX proteins investigated with polar residues close to the active site (Figure 2B). We explored the possibility that residues with ring-based side chains could interact with the ubiquinol head group through potential π–π stacking or orientate the head group by steric hindrance. These residues constitute the second largest group within the channel, although their location is dependent on the species of interest. As shown in Figure 2C, TAO has the highest density of ring-based residues around the diiron core in contrast with the other AOXs. However, on closer inspection, none of the residues appear to be in a beneficial orientation for quinol interaction (see refs [22,24–26] for the specific side chains that interact with inhibitors and quinone). In combination with the leucine gate, we suggest that polar residues may act as a guide to direct the head group of quinol towards the active site. Hence, a mixture of polar and hydrophobic residues could direct quinol into the correct orientation for binding within the active site [24–26].
Homology models of the hydrophobic cavity of a variety of AOXs viewed from the membrane-binding interface.
As previously indicated, TAO has the highest Vmax of all published AOXs, and it is interesting to speculate that the structural architecture of the TAO cavity may be the optimal design for directing quinol into the active site. Although CiAOX and NaAOX have similar areas of conservation in terms of functionality, both have restricted entry into the active site. CiAOX has the narrowest cavity (∼5 Å), due to the presence of L168 and I202 (CiAOX numbering and typically substituted as a valine in other AOXs), which we suggest will restrict quinol entry to the active site. In NaAOX, a phenylalanine substitutes for L212, which may also restrict quinol entry through steric hindrance. Although SgAOX does not have either of these limitations in terms of restricted entry, it still possesses a significantly lower Vmax. We suggest that this could be due to the lower levels of polar residues surrounding the cavity entrance, thereby reducing the attraction of quinol into the active site.
Such findings highlight the importance of the environment lining the cavity. The recently published structures of complex I (Thermus thermophilus , Bos taurus  and Ovis aries ) have all identified long (39 Å) and relatively narrow (12 Å) hydrophobic cavities leading from the phospholipid bilayer to the ubiquinone-binding site, in contrast with that observed in AOX (which is only ∼12 Å in length and ∼6 Å at its narrowest point; Figure 3). Another difference is that the complex I channel is highly hydrophobic but, unlike AOX, is lined with a mixture of charged residues (Figure 3C) . Differences in areas of hydrophobicity and polarity could influence the movement of not only ubiquinol but also of AOX inhibitors and thus such considerations are important to take into account during the design of novel AOX therapeutics.
Comparison of hydrophobic cavities in AOX and complex I.
Variation in inhibitor sensitivity
There is a general presumption that if an organism possesses an AOX, then its activity should be equally sensitive to all AOX inhibitors. This is obviously not the case and hence the question arises as to why AOXs sourced from different organisms have varying sensitivity to the same inhibitors (Table 1). For instance, it is apparent from Table 1 that although it is well documented that ascofuranone is the most potent inhibitor of TAO [7,39,40] with an IC50 of ∼5 nM, CiAOX has relatively low sensitivity to this inhibitor (an IC50 of ∼500 nM). Does the variation in inhibitor sensitivity occur because there are problems with the inhibitor accessing the hydrophobic cavity? From the above discussion, we suggest that does indeed appear to be the case, for whilst TAO and CiAOX both appear to have comparable residues lining the hydrophobic channel, CiAOX has the narrowest of channels, prior to the leucine gate, which could severely limit both substrate supply and inhibitor penetration into the active site. Interestingly, as indicated above, in NaAOX, a phenylalanine replaces L212 within the leucine gate in TAO. Placing a bulky residue such as phenylalanine into the leucine gate area could severely limit substrate supply into the active site. Although NaAOX does appear to have a significantly lower quinol oxidase activity, incorporation of such a bulky residue appears not to be detrimental to inhibitor sensitivity (see Table 1). Our current hypothesis is that the phenylalanine increases the binding affinity of the quinol mimetic inhibitors via π–π stacking, but obviously further experimentation is required to confirm such an idea.
Abbreviations: AF, ascofuranone; AC, ascochlorin; CB, colletochlorin B; CD, colletochlorin D; HfAOX, Hymenoscyphus fraxineus AOX; OG, octyl gallate; SHAM, salicylic hydroxamic acid.
Future challenges and opportunities
The challenge of emerging pathogens
The emergence of fungal infectious diseases is increasingly recognised as not only posing a worldwide threat to food security, but also more recently as posing a major threat to human and ecosystem health . As previously indicated, the impacts of fungal diseases are clearly seen in cereal crops, such as wheat and rice in addition to maize, potatoes and soybean, and more recently, emerging fungal infections are devastating worldwide cacao  and banana  crops. Fisher et al.  calculated that if severe epidemics were to occur in all five cereal crops simultaneously, this would result in a catastrophic effect on food security, leaving only sufficient food to feed 39% of the world's population! From the above discussion, it is apparent that certainly in some of the phytopathogenic fungi attacking both cereal and other crops, AOX is used as a strategy to enhance pathogenicity.
In human diseases, in addition to the role of AOX in T. brucei (the causative agent of trypanosomiasis), apicomplexans such as C. parvum (cryptosporidiosis) and parasites such as B. hominis, AOX also plays a key metabolic role in opportunistic human fungal pathogens such as Candida albicans (candidiasis) and Cryptococcus neoformans (cryptococciosis). Immunocompromised individuals are not only particularly susceptible to infection by such opportunistic pathogens, but also to other emerging human pathogens possessing AOX such as the microsporidia .
Given the important cellular role of AOX in the metabolism of such organisms, the need to control AOX activity is obvious; hence, the identification of new potent inhibitors possessing specificity and a broad spectrum of species sensitivity is urgently required.
Opportunities for gene therapy
In addition to the role of AOX in enabling human and fungal pathogens to seek opportunistic ways of causing diseases, there are also opportunities for AOX to have a positive impact on human disease. One such emerging opportunity is within the area of AOX gene therapy in attempts to overcome mitochondrial dysfunctions. The ability of AOX to maintain mitochondrial homeostasis with respect to energy demand and ROS production is currently being exploited to treat specific mitochondrial disorders. Mitochondrial diseases are caused by dysfunctions or deficiencies in any component of the oxidative phosphorylation system and can lead to an over-reduction in the quinone pool, thereby substantially increasing ROS [44–46]. To date, gene therapy has been successful in expressing AOX in human fibroblast cell lines , HeLa cell lines , Drosophila  and mouse models . CiAOX is currently being used for gene therapy since the use of an animal-based AOX was considered to more likely increase the chance of successful expression in mammalian mitochondria . Introducing CiAOX into mammalian mitochondria has been suggested to alleviate the symptoms of the disease through oxidising the quinol pool . Restoration of electron flow through the quinone pool prevents the accumulation of reduced intermediates and hence oxidative damage [46,50]. This type of gene therapy does not directly restore ATP production, but can reactivate respiratory complexes indirectly, thereby re-establishing a protonmotive force and oxidative phosphorylation . To date, CiAOX has been shown to efficiently bypass chemical inhibition of the respiratory chain, rescue COX-deficient hypertrophic cardiomyopathic cell lines and COX-deficient Drosophila models, partially rescue locomotor degeneration in Drosophila models exhibiting Parkinson's-like symptoms [44,47,51] and more recently mitigate β-amyloid production and toxicity in Drosophila and human cultured cells .
Although CiAOX has been shown to be efficient in human cell lines and Drosophila models, the question arises as to whether it is the best AOX to use in such environments. C. intestinalis lives in cold water (<20°C) and oxygen-rich environments , which is not necessarily compatible to that found within mammalian cells. T. brucei, however, exists within the human blood stream and TAO functions effectively at 37°C . Furthermore, the key role of TAO in the T. brucei bloodstream form is to maintain high glycolytic activity in the absence of a classical respiratory pathway , and this is achieved through TAO possessing a Vmax 200-fold higher than that of CiAOX. We suggest that future investigations into the use of AOX for the treatment of mitochondrial diseases may well wish to consider using a TAO-like oxidase rather than CiAOX for a more efficient oxidase better suited to the mammalian environment.
Engineering novel AOX proteins
As indicated above, the question arises as to whether or not CiAOX is the most appropriate of AOX to be effective as a strategy for gene therapy or whether the gene should be engineered to account for the different environment the protein encounters in the mammalian system. For instance, the differing temperature and oxygen tension found in mammalian systems may affect the overall functionality of CiAOX when expressed in eukaryotic cells. Although data published to date do indeed confirm that AOX expression alleviates many mitochondrial deficiencies, it is interesting to speculate what effect engineering CiAOX would have on the alleviation of such deficiencies within the respiratory chain. As indicated above, if the hydrophobic cavity to CiAOX was to be modified to mirror that observed in TAO, thereby enabling QH2 to more favourably enter the active site, then the maximal rate of substrate oxidation should be enhanced considerably. This would be helpful where there is an impairment in the respiratory chain at the level of complex III or IV, TAO has certainly proved adept at enhancing respiratory flux through complex I in proteoliposomes . Furthermore, the possibility of incorporating functional elements within the CiAOX gene, such that its activity can be regulated by α-keto acids, is also conceivable. For example, it has been shown that highly conserved cysteines  in conjunction with an ENV motif [33–35] are necessary for pyruvate stimulation within plant-type AOXs. The inclusion of such elements into a CiAOX gene would potentially result in an AOX protein that could be turned ‘on’ and ‘off’ by cellular pyruvate levels which could be advantageous under conditions where ROS homeostasis is impaired .
Research into the structure and function of the AOX is currently undergoing a resurgence of interest. The work outlined in this review clearly shows that AOX is not only playing an increasingly important role in pathogens attacking economically important crops but also becoming an emerging threat to human health. We show here that although the primary and secondary ligation spheres within the active site of AOX are highly conserved and display little diversity amongst a wide variety of AOXs, the hydrophobic cavity leading to the active site and the location of the substrate and inhibitor-binding sites do show diversity. We suggest that variations in the lining of the cavity may not only explain observed differences in maximal activity, but could also account for differences in inhibitor sensitivity. We envisage that more extensive research into these differences may well result in the future structure-based development of more potent, specific and safer set of anti-pathogenic drugs to treat the plethora of emerging diseases in which AOX plays an important metabolic role. However, not all is gloom and doom with respect to AOX; for instance, the expression of novel AOXs in mammalian cells may well prove to be an effective strategy to combat diseases associated with mitochondrial dysfunctions.
Ciona intestinalis AOX
Novosphingobium aromaticivorans AOX
proton-coupled electron transport
ubiquinone-binding site outer surface
ribonucleotide reductase R2 subunit
reactive oxygen species
Saromatum guttatum AOX
Trypanosoma brucei AOX.
Research in A.L.M.'s laboratory has been supported by the Biotechnology and Biological Research Council [BB/L022915/1 and BB/N010051/1]. We also acknowledge funding from Agform Ltd (Southampton) and ongoing collaboration.
We acknowledge all of the many students, postdoctoral fellows and visitors who, over the years in A.L.M.'s laboratory, have contributed enormously to our current understanding of the nature and function of the AOXs. Our thanks also go to our many collaborators and friends in laboratories all over the world, particularly colleagues in London, Helsinki and Tokyo, who not only have widened our horizons on what can be achieved with AOX but also have made research into AOX a great deal of fun.
B.M., L.Y. and A.L.M. hold patents and financial interests in the development of phytopathogenic fungicides.