Introduction

The cocoa tree (Theobroma cacao) is a perennial plant originated from the Amazon River basin in Central America. Cocoa beans are the main raw material used in the multi-billion-dollar chocolate industry and the cocoa pod is used in cosmetics, food thickeners, beverages, and others. On the other hand, fungal pathogens are one of the main causes of production losses in cocoa plantations, estimated at 30% worldwide on average, but affected farms can be completely ruined if left untreated. The basidiomycete fungus Moniliophthora perniciosa is the causal agent of the witches’ broom disease of cocoa (WBD) in Latin America, and a WBD outbreak in Brazil’s main cocoa-producing region—the state of Bahia—in 1989 led to a substantial loss in Brazilian competitiveness. Brazil’s cocoa production fell by 75%, and Brazil has consequently shifted from a net exporter to a net importer of cocoa beans, resulting in an economic, social and environmental crisis1,2,3.

M. perniciosa is a hemibiothtophic pathogen with a complex lifestyle and two physiologically distinct infective stages. The infection begins when M. perniciosa basidiospores reach meristematic cocoa tissues, whereupon infective hyphae penetrate through stomata and wounds. Initially, M. perniciosa develops without killing its host, which is known as the biothrophic stage of WBD2. It is evident that M. perniciosa employs effective molecular tools to suppress or circumvent the plant defence responses and of particular interest is the identification of an alternative oxidase (AOX)-coding gene. We have shown that AOX is highly expressed during the M. perniciosa–cocoa interaction and is critical for disease progression4,5.

There is substantial evidence that the AOX, a stress-induced respiratory chain protein, ensures a degree of metabolic plasticity to the cell, both during regular metabolism and stressful conditions, such as the inhibition of complexes III or IV, which halts cellular respiration and ATP production. Such oxidative imbalance also leads to the generation of harmful reactive oxygen species. AOX, however, acts as a bypass to complexes III and IV, thereby partially maintaining ATP synthesis through enhanced complex I activity6.

Infected cocoa plants produce nitric oxide, a potent complex IV inhibitor, and we have demonstrated that chemical inhibition of AOX prevents WBD progression5,7. Furthermore, M. perniciosa AOX (MpAOX) is a key resistance factor to commercial fungicides, such as strobilurins, complex III inhibitors5, and the same has been observed for other fungal plant pathogens which threaten global food security, such as Botrytis cinerea, Magnaporthe grisea, Mycosphaerella graminicola, Venturia inaequalis and others8,9. This clearly makes AOX a desirable target for the development of novel crop protection agents, however currently available AOX inhibitors are not suitable for commercial use. For instance, salicylhydroxamic acid (SHAM) and n-propyl gallate inhibit M. perniciosa growth only at millimolar concentrations5,7. Colletochlorin B and structurally related compounds, such as ascofuranone and ascochlorin comprise a novel class of AOX inhibitors with potential use as antimicrobial agents10,11. Nonetheless, AOXs from different organisms display varying degrees of sensitivity to those inhibitors owing to differences in the ligand-binding site12, which can severely affect the outcome of a treatment against WBD.

The study of fungal AOX has focused primarily on the interplay between the main and alternative respiration pathways throughout their life cycle, during pathogenesis or in response to stress13. In general, AOX is induced after treatment with main respiration inhibitors or reactive oxygen species (ROS)-generating compounds, as seen in Ustilago maydis14, Sclerotinia sclerotiorum15 and Aspergillus fumigatus16. AOX expression may increase during cell morphology changes, such as the development of biotrophic hyphae in M. perniciosa5, microsclerotia formation in Nomuraea rileyi17 and the transition from yeast-like to filamentous (infective) forms in the opportunistic human pathogens Candida albicans and Paracoccidioides brasiliensis18,19. Interestingly, genetic ablation of AOX activity has led to reduced virulence or survival rates in Cryptococcus neoformans and A. fumigatus20,21. Finally, Aspergillus nidulans and A. fumigatus exhibit a correlation between higher AOX activity and higher production of secondary metabolites, such as sterigmatocystin and cephalosporin C22,23.

The heterologous expression of AOX in eukaryotic or prokaryotic cells has been useful to investigate biochemical features of the enzyme, such as resistance to traditional main respiratory inhibitors, ROS-scavenging activity, determination of the substrate-binding site and post-translational regulation of AOX24,25,26,27,28,29,30,31,32. To our knowledge, works pertaining to the development of novel AOX inhibitors are scarce with the exception of the human parasite Trypanosoma brucei—the best-studied AOX in that regard and the only one for which a crystal structure is available—and a few others, such as Trypanosoma vivax, Antonospora locustae and Blastocystis hominis10,33,34,35,36,37,38. Moreover, because AOX is a membrane-bound protein, it is usually purified and biochemically characterised in the presence of detergents which may not accurately reflect the phospholipid bilayer biophysical properties and are not compatible with the natural AOX substrate—ubiquinol-10 (Q10H2)29,33,34.

The goal of this study was to kinetically characterise MpAOX and determine its sensitivity to CB and analogues in order to develop specific novel antifungals targeted at M. perniciosa. To that end, we have heterologously expressed and purified the recombinant MpAOX (rMpAOX) and demonstrate that it is active in Escherichia coli membranes. This system was used to evaluate putative AOX activators and putative AOX inhibitors in dose–response assays. The analysis of an MpAOX structural model and comparison with the known structure of T. brucei AOX (TAO) allowed the identification of key amino acid residues that influence protein-ligand affinity. Finally, we present and discuss the kinetic characterisation of the purified rMpAOX before and after incorporation into a self-assembled proteoliposome system that better reflects the natural hydrophobic phospholipid membrane environment.

Results

Membrane-bound rMpAOX

The rMpAOX construct used in this work corresponds to the mature MpAOX enzyme without the predicted mitochondrial leading peptide (first 40 N-terminal amino acid residues), which is cleaved following import into the mitochondrial matrix39. The removal of the mitochondrial leading sequence increases recombinant TAO stability35.

Initially, membrane-bound rMpAOX activity was measured using a high-resolution oxygraph. Fungal AOXs are activated by mono- and diphosphate nucleosides, such as AMP, GMP and ADP, whereas plant AOXs are activated by organic acids, such as pyruvate, succinate and others40,41. As can be seen from Fig. 1a, 1 mM GMP increased membrane-bound rMpAOX oxygen uptake 9.8 times, to a maximum of 0.67 µmol O2 min−1 mg total protein−1 and 5 µmol O2 min−1 mg AOX−1 (Supplementary Fig. 1). AMP and ADP stimulated rMpAOX by 8.1 times and 4.5 times, respectively. A dose–response assay with AMP, ADP and GMP reveals that AMP activates rMpAOX as much as GMP at higher concentrations, whereas 6.5 mM ADP was not enough to sustain maximum rMpAOX rates. EC50 values for GMP, AMP and ADP are 166 µM, 918 µM and ca. 6.5 mM (Fig. 1b). Interestingly, GDP did not have any noticeable effect on rMpAOX at 1 mM (Fig. 1a) and at 6.5 mM it increased rMpAOX basal rate by only 1.5 times (Supplementary Fig. 2).

Fig. 1: Testing of rMpAOX activators.
figure 1

a Fold-change in membrane-bound rMpAOX O2 uptake after addition of putative activators (1 mM each): ADP, AMP, glyoxylate, GMP, oxaloacetate, pyruvate and succinate. The dashed line indicates a fold-change of 1 (e.g., no change after addition). GMP caused the greatest increase in enzymatic activity. Horizontal bars indicate the mean of three independent measurements (circles). b Dose–response assay with GMP. Rates were normalised with respect to the maximal rate achieved. Points represent two independent measurements and the continuous lines are the best fit for a 4-parameter sigmoid function. EC50 = 166 µM (GMP), 918 µM (AMP) and ca. 6.5 mM (ADP).

Full size image

Next, we evaluated four AOX inhibitors on membrane-bound, GMP-activated rMpAOX and compared the results to conventional AOX inhibitors such as salicylhydroxamic acid (SHAM) and octyl gallate (OG). As shown in Fig. 2b, SHAM did not exert any effect up to the maximal tested concentration of 25 µM, which is in accordance with previous results obtained with intact E. coli cells7. On the other hand, IC50 values were successfully determined for ascochlorin (AC), ascofuranone (AF), colletochlorin B (CB), colletochlorin D (CD) and OG. Interestingly, AC, AF, CB and CD possess identical head groups, allowing the effect of the distinct carbonic tail on inhibitor potency to be tested. Amongst those four analogues, AF and CB exhibited the lowest IC50 values (8.8 and 14.8 nM, respectively), followed by CD (115 nM) and AC (5246 nM). OG IC50 is 1673 nM.

Fig. 2: Structure–activity relationship of MpAOX inhibitors.
figure 2

a Representation of ascofuranone (AF), colletochorin B (CB), colletochlorin D (CD), ascochlorin (AC), octyl gallate (OG) and salycilhydroxamic acid (SHAM). b Dose–response assay with membrane-bound rMpAOX. Symbols represent the mean and standard deviation of three independent measurements. The continuous lines are the best fit of a 4-parameter sigmoid function used to determine IC50 values which are follows. AF: 8.8 nM, CB: 14.8 nM, CD: 115 nM, OG: 1,673 nM and AC: 5,246 nM. c Overview of MpAOX (red; computer-generated) and TAO (blue; crystal structure PBB ID 3w54) structures. The asterisk marks an insertion of 18 residues in MpAOX (green loop) found in fungal AOXs and related to GMP activation. The arrow indicates the entrance to the substrate-binding pocket (grey shading). d Comparison of MpAOX and TAO-binding pockets. Amino acid residues that differ between those enzymes are shown as red (MpAOX) and blue (TAO) sticks and listed in Table 1. MpAOX residue numbering is used as reference. Yellow sticks depict CB present in the TAO crystal structure, and orange and red spheres represent two Fe ions and one OH− ion in TAO.

Full size image

Overall, the IC50 values obtained with rMpAOX are higher than those reported for the membrane-bound TAO10, which can be explained by differences in total AOX content. Most striking, however, was the broader range of inhibitor potency observed with rMpAOX. For TAO, IC50 values ranged from 0.2 nM (CB) to 1.5 nM (AC) with a ratio of 7.5; but this ratio is much greater for MpAOX and the AC IC50 is 354 times larger than that of CB. This indicates that AC is a much weaker ligand for MpAOX—a valuable piece of information for the design of compounds targeted at M. perniciosa.

MpAOX structural model

In an attempt to gain further insights regarding the structure–activity relationships of MpAOX inhibitors, we compared MpAOX to TAO. Sequence alignment between both proteins revealed a complete conservation of amino acid residues involved in ligand binding42 (Supplementary Fig. 3), which prompted us to create an MpAOX three-dimensional structural model (Fig. 2C) based on TAO. The MpAOX model is in good agreement with TAO crystal structure (PDB ID 3w54) with a Cα root mean square deviation of 1.6 Å. Importantly, the enzymatic core, which comprises six α-helices, which accommodate the hydrophobic substrate-binding site and the catalytic residues, is highly conserved. The main differences between the MpAOX model and TAO are seen in the flexible N-terminal region and in the extra loop in MpAOX (between α-helices α1 and α2), which corresponds to the insertion common to all fungal AOXs.

A comparison of amino acid residue composition, which make up the AOX ligand-binding cavity reveals six differences between MpAOX and TAO (Fig. 2D, Table 1). In general, the MpAOX residues are bulkier than the TAO counterparts rendering the MpAOX cavity smaller (165.89 and 225.72 Å3, respectively). A visual inspection of CB bound to TAO in the crystal structure shows that the end of the geranyl tail is close to residues V125 and M190, which in MpAOX correspond to I167 and F232, respectively. It is plausible that the bulky trimethylated cyclohexanone ring in AC clashes with those MpAOX residues and hence explains the large drop in potency observed with this inhibitor. Interestingly, the furanone ring in AF is not detrimental for MpAOX-ligand interaction, suggesting that the longer geranyl tail of AF (8C atoms in comparison to 6C in AC) is sufficient to prevent a clash with I167 and F232.

Table 1 Differences in amino acid composition in the substrate-binding pocket between MpAOX and TAO.
Full size table

The activation of fungal AOXs by mono- and dinucleotides seems to be a direct result of an insertion of ~20 amino acid residues in those proteins40. In our structural model, this insertion is found between α-helices α1 and α2 (Fig. 2c) and is depicted as a disordered loop possibly due to a lack of homology with the template protein (TAO). However, several secondary structure prediction algorithms indicate that the fungal insertion would be exposed to the solvent and adopt a helical conformation (Supplementary Fig. 4). Such ordered structure could form a binding site for allosteric regulation of fungal AOXs.

rMpAOX purification and reincorporation into proteoliposomes

AOX is a protein that naturally attaches to hydrophobic phospholipid membranes, hence the purification in aqueous medium is carried out with surfactants (detergents). An initial screening with various detergents indicated that DDM and FC-12 were the most efficient for membrane removal whilst maintaining rMpAOX activity and those detergents were selected for further evaluation. C12E8, which improved TAO activity when included in the enzyme reaction buffer33, was also tested. As seen in Fig. 3a, the use of 1% FC-12 during protein solubilisation yielded the highest rMpAOX enzymatic activity regardless of the detergent subsequently used for purification. Based on these results, we selected 1% FC-12 for membrane protein solubilisation and 0.05% DDM (~3× the critical micellar concentration) for downstream steps. Therefore, a highly pure protein of the expected size for rMpAOX was obtained (Fig. 3b).

Fig. 3: Purification and kinetic characterisation of solubilised rMpAOX.
figure 3

a Effect of detergents DDM, FC-12 and C12E8 on pure rMpAOX activity. Horizontal bars depict the Q1H2 oxidation rate of rMpAOX purified with the indicated detergents during membrane protein solubilisation (top label) and affinity chromatography (bottom label). Measurement of three technical replicates. Membrane protein solubilisation with FC-12 allowed for the highest rMpAOX activity regardless of the detergent used downstream. b Coomassie brilliant-blue-stained SDS-PAGE (left) and western blot with anti-AOX antibody (right) of protein samples collected during rMpAOX purification. M: molecular weight marker; 2: total E. coli protein extract; 3: clarified cell lysate; 4: E. coli membrane faction; 5: affinity chromatography flow-through; 6: affinity chromatography eluate. c Kinetic characterisation of rMpAOX with respect to ubiquinol-1 (Q1H2). The assay was performed in 50 µL 20 mM Tris-HCl buffer with 0.0125% C12E8 and 100 ng rMpAOX. Circles and error bars depict the mean and SD of seven independent measurements (diamonds). The continuous line is the best fit for the Michaelis–Menten kinetic model with KM = 68.5 µM and VMax = 504.1 nmol Q1 min−1 mg−1. d Dose–response assay with CB. Points represent two independent assays (circles or diamonds) and the continuous line is the best fit for the Morrison equation43. CB IC50 = 59.3 nM and KIapp = 13.9 nM.

Full size image

Purified rMpAOX was characterised with respect to the hydrophilic substrate ubiquinol-1 (Q1H2). A titration of C12E8 in the reaction buffer showed that 0.0125% resulted in maximal enzymatic activity with an increase of 3.5 times the activity when compared to the non-detergent condition and was used in all subsequent Q1H2 assays (Supplementary Fig. 5). The KM and VMax for rMpAOX were therefore determined as 68.5 ± 8.8 µM and 504.1 ± 28.7 nmol Q1 min−1 mg−1, respectively (Fig. 3c, Supplementary Fig. 6A). Further addition of GMP up to 0.5 mM did not alter the rMpAOX kinetic parameters such as KM and VMax. Moreover, the CB IC50 on the purified rMpAOX was determined as 59.3 nM (Fig. 3d). As the IC50 value is within a 10-fold difference with respect to the enzyme concentration in the assay (49.1 nM, yielding a CB-to-rMpAOX ratio of 1.2), we used the Morrison equation for calculating the KIapp of tight-binders, which gave an KIapp for CB of 13.9 nM43. Note that CB is a mixed-type AOX inhibitor and therefore KIapp is a function of KI and KI′ parameters33,36.

In addition to the detergent-solubilised rMpAOX, we employed a recently developed proteoliposome system to characterise rMpAOX that better reflects the hydrophobic environment of the phospholipid membrane44. This also allowed us to test the natural quinol analogue found in mitochondria, namely ubiquinol-10 (Q10H2), which is too hydrophobic to be used in aqueous solution. Therefore, we prepared self-assembled proteoliposomes (PLs) containing rMpAOX, Q10 and a bacterial NADH:ubiquinone oxidoreductase (NDH-2) as an electron donor to Q10. A range of Q10 concentrations up to 17 mM was tested, which allowed the determination of kinetic parameters of membrane-bound rMpAOX, resulting in KM = 0.96 ± 0.76 mM and VMax = 59.4 ± 6.6 µmol Q10 min−1 mg−1 (Fig. 4, Supplementary Fig. 6B). The CB IC50 in our PLM preparation was 19.5 nM, which is similar to values reported for TAO in a similar defined PL system using complex I as the electron donor45.

Fig. 4: Characterisation of rMpAOX incorporated into proteoliposomes.
figure 4

a Kinetic characterisation of rMpAOX with respect to ubiquinol-10 (Q10H2). The assay was performed in 200 µL 65 mM MOPS pH 7.5 with a starting concentration of 300 µM NADH and ~50 ng rMpAOX. Each point represents one PL preparation for which a different amount of Q10 was added with the SD of technical triplicates. Protein and Q10 quantifications were performed as described in ref. 44. The continuous line is the best fit for the Michaelis–Menten kinetic model. KM = 0.96 ± 0.76 mM and VMax = 59.4 ± 6.6 µmol Q10 min−1 mg−1. b Dose–response assay with CB. Points represent two independent measurements and the continuous line is the best fit for a four-parameter sigmoid function. CB IC50 = 19.5 nM.

Full size image

Discussion

Chocolate is appreciated worldwide and its industry grows by the year with the top 10 chocolate manufacturers reaching net sales of $78 bn in 2018. Cocoa beans are essential for chocolate production but cocoa farms in Latin America are threatened by M. perniciosa. M. perniciosa, in addition to other pests and diseases, can decimate farms if left untreated, and there is a substantial risk of this pathogen reaching other cocoa-producing countries in Africa and Asia with disastrous consequences46. We have demonstrated that a class of compounds initially identified as TAO inhibitors are also effective against MpAOX. Such information is valuable for the future design of antifungal agents specifically targeted at M. perniciosa.

Of particular importance was the observation that AC was less potent against MpAOX than observed for TAO, which led to insights regarding structure–activity relationships of CB analogues on MpAOX. The identification of features that confer inhibitor specificity between different AOXs is especially important in agricultural settings where both plant and pathogen express AOX because—ideally, only the pathogen’s enzyme should be inhibited to reduce the chances of side-effects on the plant. It has been previously demonstrated that CB analogues may show distinct inhibitory profiles between different AOXs, and now we present a mechanistic explanation for such an observation12. Specifically, differences in the entrance of the hydrophobic channel, which leads to the catalytic core may alter inhibitor affinity. MpAOX amino acid residues I167 and F232 clash with the AC trisubstituted hexanone ring and prevent favourable protein–ligand interactions. This in turn can be avoided by increasing the linker length, as seen with AF or avoiding bulky substituents such as in CB. In other words, the length of the aliphatic tail may not only directly modulate the AOX–ligand interaction through the number of contact points made in the hydrophobic tunnel—as shown for TAO10,35,47—but also indirectly through the positioning of large substituents at the tunnel entrance.

Inhibitor potency is an important parameter for drug development projects. Derivatives of salicylic acid and of gallic acid are well-known AOX inhibitors48,49,50, but are only effective against M. perniciosa at millimolar concentrations5,7. We have also recently shown that another AOX inhibitor, N-(3-bromophenyl)-3-fluorobenzamide, has a protective effect in plants against WBD at 200 µM even in the absence of a main respiration inhibitor7. The high concentrations of molecules necessary to inhibit fungal growth are not ideal from an agricultural and commercial perspective. Therefore, we anticipate that AF and CB analogues, which are tight-binding inhibitors, will lead to more effective antifungal agents against M. perniciosa and other fungal pathogens.

We observed a drop in rMpAOX enzymatic activity after removal of the bacterial membrane from 5 µmol O2 min−1 mg AOX-1 (equivalent to 10 µmol Q min−1 mg AOX−1) to a VMax of 504.1 nmol Q1 min−1 mg−1 with the solubilised rMpAOX. This is opposed to results from TAO and SgAOX for which the specific activity increases after purification, which prompted us to investigate the dependence of rMpAOX activity on the phospholipid membrane29,33. Therefore, rMpAOX was incorporated into defined PLs with Q10, the natural substrate present in mitochondria and allowed rMpAOX to reach a considerably higher rate VMax of 59.4 µmol Q10 min−1 mg−1. In fact, the rMpAOX catalytic rate with 180 µM Q10H2 in the phospholipid membrane is calculated to be 9.3 µmol min−1 mg−1 based on KM and VMax values, which is 18.4 times higher than observed for Q1H2.

The difference between rMpAOX catalytic rate before and after PL assembly might have been explained by substrate availability. The low solubility of Q1H2 in aqueous medium may hinder AOX kinetic characterisation, which can be overcome with the use of detergents to prevent Q1H2 micelle formation at concentrations up to 600 µM33,51. Although we have included C12E8 in the enzymatic reaction medium for the pure rMpOAX, we did not immediately discard the possibility of Q1H2 aggregation, which may explain the lower rMpAOX rates after solubilisation due to differences in our experimental setup compared with previous experiments. For instance, Hoefnagel et al.51 describe that micelle formation is time-dependent and our reaction in microtiter plates might take longer to prepare and read than assays previously performed in quartz cuvettes. In order to investigate this hypothesis, we used Q1H2 as substrate in PL assays, which yielded essentially the same result as the detergent-solubilised rMpAOX (Supplementary Fig. 7) demonstrating that the phospholipid membrane is not sufficient to sustain maximal rMpAOX activity with Q1H2. This is in agreement with results obtained with complex I showing that, even in the PL system, the ubiquinone isoprenoid tail length influences the enzyme turn-over rate with Q1 and Q2 yielding lower rates than Q4, Q6, Q8 and Q1045. In fact, structural features of the AOX quinol-binding site and biochemical data obtained with inhibitors (e.g., compare CB and CD) have shown that molecules with short lipophilic tails interact less effectively with AOX6,33,42. Furthermore, CB dose–response assays indicate that the rMpAOX–ligand interaction (and therefore the rMpAOX ligand-binding site) was not disturbed after solubilisation. The CB-to-rMpAOX ratio at the IC50 concentration is 1.2 and similar values have been reported for solubilised TAO (1.3) and for TAO incorporated into PLs (1.4)34,45. Overall, those results show that the lower activity of the solubilised rMpAOX is not due to Q1H2 solubility issues and is instead caused by a weaker protein-substrate interaction. This highlights one advantage of using the PL system for AOX characterisation—namely the ability to use the physiological substrate at controlled concentrations up to saturation where it would otherwise prove problematic in aqueous solution. Another advantage of the PL system is the fact that AOX becomes the limiting step for electron flux thus reaching maximal catalytic rate, as evidenced by the fact that in E. coli membranes rMpAOX is only ~16% as fast as in PLs (Fig. 4, Supplementary Fig. 1C). Most likely, the bacterial NADH dehydrogenase cannot provide sufficient amounts of reduced Q to sustain maximal activity of the overexpressed rMpAOX.

The effect of GMP on rMpAOX was also altered after removal of the bacterial membrane. The ectopically expressed enzyme retained the ability to be stimulated by GMP whilst attached to E. coli membranes, much like what is observed with other fungal AOXs in situ. For instance, the GMP and AMP EC50 values for rMpAOX (166 and 918 µM) are in the same range of concentration necessary for half-maximal activation of Hansenula anomala mitochondria52,53,54,55. However, it is still unclear how variations of those nucleotides might regulate AOX activity in vivo under normal development and during stress response. The intracellular concentration of GMP, AMP and ADP are in the micromolar range in several cell types (bacteria, yeast and mammalian cells) suggesting that GMP has a greater role in AOX activation56,57,58. However, there might be spatial or temporal variations in the concentration of those nucleotides in mitochondria such as an increase in ADP levels caused by the impairment of the main respiratory chain. To our knowledge, there is no information with respect to variations in intramitochondrial GMP and AMP concentration.

No rMpAOX stimulation was observed with the purified protein even in the PL system and this prevented the determination of dissociation constants, ka. The 18-amino acid fungal AOX insertion is close to α-helix 1 which forms—together with α4—the hydrophobic surface responsible for attachment to the phospholipid membrane (Fig. 2c). Hence, it is possible that the lipid environment is necessary for the correct conformation of the fungal loop or that an as-of-yet unknown molecular intermediate or accessory in the cell membrane is required for AOX stimulation. The sensitivity of rMpAOX to GMP was not regained in the PL system, which might indicate that either rMpAOX had already reached maximum catalytic rates or that the molecular intermediate found in mitochondria and in E. coli membranes is absent in our defined PLs.

Methods

Chemicals and reagents

Ubiquinonol-1 (Q1H2) was obtained after the chemical reduction of ubiquinone-1 (Q1; Sigma), as described elsewhere59. In summary, 10 mg Q1 were dissolved in 5 mL ethanol, to which 4 g Na2S2O4 in 30 mL of water were added. The reaction was left for 1 h in the dark and Q1H2 was extracted with n-hexane (3 × 20 mL). The pooled n-hexane fractions were washed with 20 mL water, dried with solid MgSO4 and filtered. The n-hexane was removed by evaporation and the solid residue was resuspended with CDCl3 for nuclear magnetic resonance spectroscopy to confirm the identity and purity of Q1H2. The chloroform was removed by evaporation and Q1H2 was dissolved with acidified DMSO and stored at −80 °C until use. To determine the Q1H2 concentration, the stock solution was diluted 100-fold in water and left to spontaneously oxidise for 24 h, after which the absorbance at 278 nm was determined (ε278 = 15 mM−1 cm−1)33. Detergents were obtained from Anatrace and putative AOX activators from Sigma.

rMpAOX cloning and expression

The MpAOX sequence used in this work corresponds to UniProt ID A8QJP85, which was cloned without the predicted mitochondrial leading sequence, as determined with the Signal P server60. Strep-tagged rMpAOX was synthesised and cloned into the pET15a (Genscript). rMpAOX expression, E. coli membrane fractionation and protein purification were performed essentially as described for TAO and Sauromatum guttatum AOX29,33. FN102 E. coli cells transformed with pET15b:rMpAOX plasmid were inoculated into 10 mL of l-broth supplemented with 100 μg mL−1 ampicillin, 100 μg mL−1 kanamycin and 50 μg mL−1 5-aminolevulinic acid (ALA). This was left to grow overnight at 37 °C. On the following day, 1 mL of overnight culture was transferred to 50 mL of l-broth supplemented with 100 μg mL−1 ampicillin, 100 μg mL−1 kanamycin, 50 μg mL−1 ALA, 0.2 % (w/v) glucose, 50 µg mL−1 MgSO4, 25 µg mL−1 FeSO4, and 25 µg mL−1 FeCl3. The starter culture was left at 37 °C in a shaking incubator (180 rpm) until the OD600 reached 0.6. The cells were harvested by centrifugation, resuspended with ~5 mL of fresh K-broth and inoculated in 4 × 1 L of K-broth supplemented with 100 μg mL−1 carbenicilin, 100 µg mL−1 kanamycin, 50 µg mL−1 MgSO4, 25 µg mL−1 FeSO4, 25 µg mL−1 FeCl3 and 0.2 % (w/v) glucose, with an starting OD600 of 0.01. The culture was left at 30 °C with 180 rpm shaking until the OD reached ~0.6, then 50 μM isopropyl β-d-1-thiogalactopyranoside (IPTG) was added for induction of rMpAOX expression. Cultures were left at 30 °C for 14-16 hours before cell harvesting.

E. coli membrane fraction preparation

E. coli membrane preparation was performed as described with minor modifications29. The cell pellet was thoroughly resuspended in 10 mL of 65 mM MOPS buffer pH 7.5 per gram of cells supplemented with protease inhibitor tablets (Roche; 1 tablet per 50 mL), 2.5 U mL−1 of benzonase (Sigma) and 100 mM MgSO4. The cells were disrupted using two passes at 30 kPa through a pre-cooled Constant cell disruption system (Constant Systems Ltd). The lysate was centrifuged for 15 min at 16,000 rcf, and the pelleted unbroken cells and large cell debris were discarded. The supernatant was centrifuged for 90 min at 200,000 rcf to collect the fragmented membranes. After centrifugation the supernatant was discarded and the membrane pellets were thoroughly resuspended with 65 mM MOPS pH 7.5 with the aid of a Dounce homogenizer.

rMpAOX purification

The E. coli membrane fraction was resuspended with 50 mM Tris-HCl pH 7.5, 20% glycerol, 200 mM MgSO4 and 1% Fos-Choline-12 (FC-12) instead of MOPS buffer. The sample was left under gentle rocking for 1 h at 4 °C, after which it was centrifuged at 200,000 rcf for 30 min. The supernatant was then subjected to affinity chromatography with a 5-mL StrepTrap HP column (GE Healthcare) in an Äkta FPLC (GE Healthcare) at 4 °C. The column was equilibrated with 65 mM MOPS pH 7.5 before passing the sample, after which the resin as washed with 10 volumes of buffer A (65 mM Tris-HCl pH 7.5, 50 mM MgSO4, 60 mM NaCl, 20% glycerol and 0.05% DDM, unless otherwise stated). rMpAOX elution was carried out with buffer A added of 2.5 mM desthiobiontin. Protein purification was followed by SDS-PAGE and coomassie brilliant-blue staining61 and western blot with anti-twin-strep (IBA GmbH, 2-1509-001) and anti-AOX62 which was kindly provided by Prof. T. Elthon, Nebraska, USA. Protein content was estimated through the Bradford method63 and bovine serum albumin as standard.

NDH-2 expression and purification

Expression and purification of a bacterial type II NADH dehydrogenase (NDH-2) from Caldalkalibacillus thermarum was performed as described elsewhere64. The plasmid pTRCndh2 was kindly donated by Prof. Greg Cook (Otago University, New Zealand). In summary, overexpression was carried out in E. coli C41 for 4 h with 1 mM IPTG. Cells were harvested and lysed, the membrane fraction was collected and NDH-2 was solubilised with 2% n-octyl-β-d-glucopyranoside (OG). NDH-2 was purified through metal-affinity chromatography64.

Self-assembled proteoliposome preparation

Proteoliposomes (PL) were prepared as previously described44,45, except that NDH-2 was used instead of complex I and 0.5% DDM was used instead of OG. For IC50 measurements, PLs were prepared with 5 µg of rMpAOX and 500 µg of NDH-2. For kinetic studies, PLs were prepared with 50 µg rMpAOX and no NDH-2—in this case, pure NDH-2 was added directly into the enzymatic reaction mixture at a saturating concentration. Briefly, lipids and Q10 stocks in chloroform were mixed in a glass vial and the chloroform dried under a nitrogen stream. The lipids were hydrated and extruded through a 100 nm NanoSizer MINI Liposome Extruder (T&T Scientific) followed by the addition of protein. After treatment with the Bio-Beads SM-2 Resin (Bio-Rad), proteolipososmes were harvested by ultracentrifugation and resuspended in 150 µL of buffer. Enzymatic assays were performed on the same day.

Enzymatic activity measurements

Oxygen uptake from E. coli membrane-bound rMpAOX was measured in an Oxygraph-2k (Oroboros Inc.) high-resolution respirometer in 65 mM MOPS pH 7.5 with 1.25 mM NADH and 0.1 mg mL−1 potassium cyanide. Respiration rates were recorded before and after the addition of putative AOX activators, such as 5′-ADP, 5′-AMP, 5′-GDP, 5′-GMP, glyoxylate, oxaloacetate, pyruvate or succinate (1 mM each unless stated otherwise). NADH oxidation measurements of membrane-bound rMpAOX were performed spectrophotometrically (ε340 = 6.22 mM−1 cm−1) with 0.1 mg mL−1 KCN, 1 mM GMP and 300 µM NADH in 200 µL reactions. Data were acquired with the Multiskan FC Microplate Photometer (ThermoFisher). For the pure rMpAOX, Q1H2 oxidation was measured spectrophotometrically (ε284 = 14.1 mM−1 cm−1) with 100 ng enzyme in 50 µL of 20 mM tris-HCl pH 7.5 and 0.0125% octaethylene glycol monododecyl ether (C12E8) in 384-well microplates (CLARIOstar Plus Microplate Reader, BMG LABTECH). Absorbance readings were taken periodically and a blank reactions (substrate without enzyme) were included in every assay and were used to determine and subtract the autoxidation rate (background). Proteoliposomes were assayed as described for the E. coli membrane fraction except that no KCN was included and 0.5 µL of the prep (~50 ng rMpAOX) were added into the reaction. When tested, inhibitors were preincubated with the enzyme in reaction buffer for 2–5 min before starting the reaction by the addition of substrate. The path length (i.e., the height of the reaction medium inside the well) was calculated based on the reaction volume and the well cross-section area informed by the supplier. GraphPad Prism 7.0 (GraphPad Software Inc.) was used for function fitting.

MpAOX sequence alignment and structural modelling

The amino acid sequence of AOXs used here are from Trypanosoma brucei (GenBank ID BAB72245.1), Moniliophthora perniciosa (ABN09948.3), Septoria tritici (XP_003851917.1), Neurospora crassa (XP_962086.1), Aspergillus niger (BAA32033.2), Arabidopsis thaliana (NP_188876.1), Oryza sativa (XP_015635413.1) and Zea mays (NP_001105180.1). The sequence alignment was performed with the MUSCLE algorithm built in JalView65 with default parameters. The MpAOX protein sequence was used as input in the SWISS-MODEL protein structure modelling server66 and TAO crystal structures PDB ID 3vv9 and 3w5442 were selected as templates. The quality of the output model was checked within parameters provided by SWISS-MODEL server. The MpAOX model and TAO structure 3w54 were overlaid with PyMOL molecular viewer 2.3.3 (Schrödinger, LLC) and visually inspected, in particular residues that form the active site in TAO and residues with known functions6. The volume of the ligand-binding pocket was calculated with the KVfinder plugin67 with “probe in” and “probe out” radii of 1.1 and 5 Å, respectively. The secondary structure prediction and solvent accessibility prediction were performed with NetSurfP 2.068, Jpred 469, PSIPRED70 and PredictProtein71.

Statistics and reproducibility

Measurements were performed in technical triplicate and each experiment was performed three independent times (e.g. distinct protein purification batches) unless otherwise stated and similar results were obtained each time. No data points were excluded from the analyses. Data fitting and plotting were performed with GraphPad Prism 7.0 (GraphPad Software Inc.).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.