Resistance mutations reveal the atovaquone-binding domain of cytochrome b in malaria parasites

Authors


Akhil B. Vaidya. E-mail vaidyaa@mcphu.edu; Tel. (+1) 215 991 8557; Fax (+1) 215 843 4152.

Abstract

Atovaquone represents a class of antimicrobial agents with a broad-spectrum activity against various parasitic infections, including malaria, toxoplasmosis and Pneumocystis pneumonia. In malaria parasites, atovaquone inhibits mitochondrial electron transport at the level of the cytochrome bc1 complex and collapses mitochondrial membrane potential. In addition, this drug is unique in being selectively toxic to parasite mitochondria without affecting the host mitochondrial functions. A better understanding of the structural basis for the selective toxicity of atovaquone could help in designing drugs against infections caused by mitochondria-containing parasites. To that end, we derived nine independent atovaquone-resistant malaria parasite lines by suboptimal treatment of mice infected with Plasmodium yoelii; these mutants exhibited resistance to atovaquone-mediated collapse of mitochondrial membrane potential as well as inhibition of electron transport. The mutants were also resistant to the synergistic effects of atovaquone/ proguanil combination. Sequencing of the mitochondrially encoded cytochrome b gene placed these mutants into four categories, three with single amino acid changes and one with two adjacent amino acid changes. Of the 12 nucleotide changes seen in the nine independently derived mutants 11 replaced A:T basepairs with G:C basepairs, possibly because of reactive oxygen species resulting from atovaquone treatment. Visualization of the resistance-conferring amino acid positions on the recently solved crystal structure of the vertebrate cytochrome bc1 complex revealed a discrete cavity in which subtle variations in hydrophobicity and volume of the amino acid side-chains may determine atovaquone-binding affinity, and thereby selective toxicity. These structural insights may prove useful in designing agents that selectively affect cytochrome bc1 functions in a wide range of eukaryotic pathogens.

Introduction

There is a clear need to have available antimalarial drugs with targets different from those affected by agents for which resistance is already rampant in the field. Identification of such targets and development of compounds that inhibit their functions have to be a priority in malaria research. Because of the critical role played by mitochondria in cellular physiology, functions assigned to these organelles could serve as targets for antiparasitic drugs, provided that these compounds act in a selective manner. Several unusual properties, such as the lack of complete citric acid cycle (Scheibel, 1988), a highly condensed mitochondrial genome with fragmented rRNA genes (Vaidya et al., 1989; Feagin, 1994) and unique structural features of cytochrome b (Vaidya et al., 1993a), have been associated with the mitochondria of malaria parasites. These features could serve as targets for antimalarial drug action (Vaidya, 1998).

Atovaquone (2-[trans-4-(4′-chlorophenyl)cyclohexyl]-3-hydroxy-1,4-naphthoquinone) is a new drug that appears to target parasite mitochondrial physiology (for a review, see Vaidya, 1998). It was developed in the 1980s by the Wellcome laboratories but traces its origins to studies initiated 50 years ago on hydroxynaphthoquinones as potential antimalarials (Fieser et al., 1948). While this drug has been reasonably successful in the treatment of toxoplasmosis (Kovacs, 1992) and Pneumocystis carinii pneumonia (Hughes et al., 1993), when used as a single agent against malaria, it was met with a high rate of treatment failures (Chiodini et al., 1995; Looareesuwan et al., 1996). This then led to the inclusion of proguanil as a synergistic agent with atovaquone (Canfield et al., 1995); the resulting combination is a promising antimalarial belonging to a new class of agents (Looareesuwan et al., 1996). Atovaquone has been shown to inhibit mitochondrial electron transport selectively at the level of ubiquinone-cytochrome c oxidoreductase (cytochrome bc1 complex; Fry and Pudney, 1992), as well as to collapse the electropotential across the mitochondrial inner membrane (ΔΨm) of malaria parasites (Srivastava et al., 1997). Critical structural features of atovaquone for its antiparasite activity appear to be the hydroxyl group at the 3-carbon position, and the cyclohexyl-chlorophenyl side-chain at the 2-carbon position (Hudson et al., 1991). In addition, the trans configuration of this side-chain is required for the inhibition of parasite bc1 complex, while cis configuration is inactive (Fry and Pudney, 1992). Inhibition by atovaquone of the parasite cytochrome bc1, while sparing that of the host mitochondria, remains unexplained at the molecular level. Selective targeting of mitochondrial functions of a eukaryotic parasite is clearly an attractive approach in the search for a means to combat such pathogens as they become increasingly resistant to the available therapies. Through a study of several independently derived atovaquone-resistant malaria parasites, we report here the structural features of the parasite cytochrome b that may determine selective toxicity of atovaquone.

Results and discussion

Atovaquone-resistant mutants

To assess the molecular basis for atovaquone action in a model system, we derived atovaquone-resistant parasites by infecting mice with cloned P. yoelii 17XL and administering atovaquone orally at a suboptimal dose of 5 mg kg−1. This in vivo derivation of the resistant parasites was chosen to proximate the clinical situation, as bioavailability of orally administered atovaquone may significantly affect pharmacokinetics and, consequently, the development of resistance. As shown in Fig. 1A, atovaquone treatment suppressed parasite growth, whereas untreated mice developed high parasitaemia. However, a significant number of treated mice (ranging from 20% to 50%) had recrudescent parasitaemia (Fig. 1A), and the parasites isolated from such recrudescing mice exhibited resistance to a much higher dose of atovaquone (Fig. 1B). This suggested that high-dose atovaquone resistance can arise readily through suboptimal treatment, reflecting the observations made in clinical trials (Chiodini et al., 1995; Looareesuwan et al., 1996). Through the use of this regimen, nine independently derived atovaquone-resistant P. yoelii lines were obtained and named AR1 to AR9. The following reasons indicated that resistant parasites were not being enriched from a pre-existing population but were emerging during the treatment: (i) a cloned P. yoelii population was used for infections; (ii) treatment of mice infected with the parental parasites with a 20 mg kg−1 dose of atovaquone uniformly resulted in complete cure, whereas the resistant mutants are completely refractory to this dose (Fig. 1B); and (iii) polymerase chain reaction (PCR) analysis with mutant specific oligonucleotides (see below) failed to detect such mutations pre-existing in the parental P. yoelii.

Figure 1.

. Development of atovaquone resistance in P. yoelii. In this representative experiment, BALB/c mice were injected with 105 cloned P. yoelii 17XL, a lethal rodent malaria parasite. A. Progression of parasitaemia determined by Giemsa staining of blood smears in three control mice and three mice treated with 5 mg kg−1 atovaquone on days 5 and 6 after infection (arrows). All control mice developed lethal malaria, and the treated mice appeared to clear their parasites. Two of the treated mice showed recrudescence on days 10 and 11 after infection and were treated with atovaquone again (arrows), but failed to respond. Parasites were recovered from the blood of one of these mice and inoculated into a new set of mice (B). Three of the mice were given 20 mg kg−1 atovaquone daily beginning at day 5 after infection (arrows). None of the treated mice responded to the treatment, and this line was deemed to be atovaquone resistant. Further experiments showed that the resistance phenotype was stable and did not require maintenance of drug pressure. A total of nine atovaquone-resistant lines was derived independently from similar experiments.

Resistance to ΔΨm collapse and electron transport inhibition

We have shown previously that atovaquone collapses the ΔΨm in live intact malaria parasites in addition to inhibiting parasite respiration (Srivastava et al., 1997). The ΔΨm is measured by flow cytometric assay of mitochondrial accumulation of a fluorescent lipophilic cationic probe, 3,3′-dihexyloxacarbocyanin (DiOC6), as described previously (Srivastava et al., 1997). Figure 2A shows profiles of the dose-dependent effects of atovaquone on the ΔΨm of the parental and four representative atovaquone-resistant parasite lines. The effective atovaquone concentrations (EC50) at which 50% of the ΔΨm was collapsed were calculated from these curves. As summarized in Table 1, the EC50 for atovaquone-mediated collapse of ΔΨm in the parental parasites was 15 nM, while for the resistant parasite lines, the EC50 values were ≈ 1000-fold higher, ranging from 10 000 to 25 000 nM. Atovaquone-resistant parasites were also resistant to the drug-mediated inhibition of mitochondrial electron transport, as measured by the rate of respiration (Fig. 2B); indeed, even at the highest concentrations, atovaquone could only inhibit ≈ 50% of the respiration by the resistant parasites. Thus, development of atovaquone resistance in malaria parasites is accompanied by concomitant resistance to atovaquone-mediated ΔΨm collapse and electron transport inhibition.

Figure 2.

. Mitochondrial physiology in atovaquone-resistant P. yoelii. A. Effects of various concentrations of atovaquone on ΔΨm in live P. yoelii infected erythrocytes were assayed as described previously (Srivastava et al., 1997). Curves demonstrating dose-dependent response in the parental (□) parasites as well as the resistant lines AR1 (♦), AR3 (▪), AR6 (●) and AR8 (○) are shown. B. Effects of atovaquone on electron transport were assayed by measuring respiration by live parasites. Atovaquone-mediated inhibition of respiration in the parental P. yoelii (○) and the resistant line AR1 (▴) are shown. C. Effects of myxothiazole on ΔΨm in parental as well as atovaquone-resistant P. yoelii were assayed. Curves demonstrating dose-dependent response in the parental (⋄) as well as the resistant lines AR1 (▴), AR3 (○), AR6 (▪) and AR8 (●) are shown. D. Proguanil does not change atovaquone-mediated response in the resistant parasites. Collapse of ΔΨm by atovaquone by itself in the parental parasites (▵) and in the resistant line AR1 (▪) is compared with the response in the AR1 line in which the indicated concentrations of atovaquone were combined with 1.4 × 10−7 M (▴), 6.8 × 10−7 M (♦), 1.4 × 10−6 M (○) and 1.4 × 10−5 M (□) concentrations of proguanil.

Table 1.  . Effect of inhibitors on mitochondrial membrane potential in atovaquone-resistant parasites. a. Inhibitor concentration at which 50% of the membrane potential was collapsed. Mitochondrial membrane potentials in live intact parasites were measured by a flow cytometric assay using 2 nM DiOC6, a lipophilic cationic fluorescent probe, as described previously (Srivastava et al., 1997). Parasites were isolated from wild-type or mutant P. yoelii-infected mice, enriched for trophozoite and schizont stages on Percoll gradients, and the effect of various concentrations of the inhibitors on mitochondrial membrane potential was determined. Membrane potential loss induced by the protonophore CCCP was used as the value for 100% collapse.Thumbnail image of

We also examined the ΔΨm in the resistant parasites in the presence of myxothiazol, a standard cytochrome bc1 complex inhibitor (Fig. 2C). The EC50 for myxothiazol in the parental parasites was about 180 nM and, in the resistant parasites, the EC50 values increased ≈ 5- to 50-fold (Table 1). Hence, the level of cross-resistance for this inhibitor appeared to be at a lower level than for atovaquone.

As mentioned earlier, atovaquone and proguanil appear to have synergistic antimalarial activity (Canfield et al., 1995; Looareesuwan et al., 1996). In a recent study, we have provided evidence suggesting that proguanil acts synergistically to enhance the ability of atovaquone to collapse ΔΨm without affecting electron transport inhibition (Srivastava and Vaidya, 1999). Hence, it was of interest to see whether proguanil could continue to have synergism even in atovaquone-resistant parasites. However, as shown in Fig. 2D, for atovaquone-resistant parasites, the inclusion of proguanil did not affect the ΔΨm collapse profile for atovaquone. This suggests that the inclusion of proguanil will have no synergism with atovaquone once atovaquone resistance has emerged.

Cytochrome b sequences

The above observations strongly suggested the possibility of structural changes within the site at which atovaquone binds in malaria parasites. In a number of different organisms, essentially all naturally arising mutants resistant to inhibitors of the cytochrome bc1 complex carry mutations in the cytochrome b gene (Brasseur et al., 1996). Hence, we analysed the cytochrome b gene of the unusual 6 kb malarial mitochondrial genome from each of the nine independent atovaquone-resistant P. yoelii lines. Initially, the mitochondrial DNA (mtDNA) was scanned by single-strand conformation polymorphism (SSCP; Vaidya et al., 1993b) analysis of overlapping regions (data not shown). The only region showing SSCP differences was within the cytochrome b gene, which was then PCR amplified and sequenced. As summarized in Fig. 3A, the independent atovaquone-resistant lines AR1, AR2 and AR9 had identical 2 bp changes, resulting in L271V (TTA to GTA) and K272R (AAA to AGA) mutations. Lines AR3, AR4 and AR5 had an identical 1 bp change resulting in I258M (ATA to ATG) mutation; lines AR7 and AR8 had a 1 bp change giving Y268C (TAT to TGT) mutation; and AR6 had a single change resulting in F267I (TTT to ATT) mutation. For one of the atovaquone-resistant lines (AR1), the entire mtDNA was cloned and completely sequenced (data not shown). This revealed total sequence identity with the parental mtDNA except for the 2 bp mutations in the cytochrome b gene noted above.

Figure 3.

. Sequence changes in the cytochrome b gene of atovaquone-resistant malaria parasites. A. A portion of the P. yoelii mtDNA encoding amino acid positions 253–273 of cytochrome b with the altered nucleotide and amino acid positions highlighted in yellow. Sequences of this region in atovaquone-resistant lines AR1 to AR9 are aligned below with the indicated nucleotide and amino acid changes. B. Alignment of the conserved PEWY region of cytochrome b from six organisms. The sequence from Plasmodium cytochrome b (positions 253–277) is aligned with corresponding regions of the related apicomplexan parasites Toxoplasma (protein identification number 2564669) and Theileria (437865) as well as of the yeast (2851445), human (117863) and chicken (117847). For the chicken sequence, the amino acids shown correspond to positions 264–288 of cytochrome b. Arrows indicate amino acid positions altered in atovaquone-resistant malaria parasites. Pink and blue highlighting indicates positions that are completely conserved in all species; blue highlighting is for the three universally conserved positions at which mutations occur in atovaquone-resistant parasites; and the yellow highlighting indicates the positions that are divergent in apicomplexan cytochrome b and are altered in the drug-resistant malaria parasite.

The mtDNA in malaria parasites is remarkably well conserved, with its apparent evolutionary drift being much slower than that for the nuclear genome (McIntosh et al., 1998). Furthermore, there is little sequence variation within a parasite species, as judged from essentially complete sequence identity of the 6 kb mtDNA sequenced from geographically isolated clones; indeed, only a single nucleotide difference was observed over the entire 5966 bp mtDNA sequenced from two geographically distant P. falciparum isolates (Vaidya et al., 1993b; McIntosh et al., 1998). Therefore, the rapid sequence changes observed here in atovaquone-resistant parasites indicate a great selective advantage provided by these mutations in the presence of this drug. The mitochondrial DNA in malaria parasites appears to replicate through a rolling circle mechanism followed by extensive recombination and strand invasions (Preiser et al., 1996). We have suggested that this mode of DNA replication may result in a high level of copy correction of the tandemly arrayed mitochondrial genome (McIntosh et al., 1998); while this would limit sequence divergence under normal conditions, rapid dissemination of an advantageous mutation would ensue once it does arise. We were unable to detect the parental cytochrome b sequence through a PCR-based assay in atovaquone-resistant P. yoelii (data not shown). There are ≈ 100 copies of mtDNA molecules per parasite in P. yoelii (Vaidya and Arasu, 1987); it appears that, in the resistant parasites, none of these molecules displays the parental cytochrome b sequence. This indicates that, in very short order, the entire mtDNA repertoire has been converted to the resistant sequence, suggesting a rapid gene conversion rate for the parasite mtDNA.

Of the 12 nucleotide changes seen in these resistant parasites, 11 replaced an A:T basepair with a G:C basepair. This bias could be caused by the reactive oxygen species-mediated damage to the mtDNA in mitochondria with atovaquone-inhibited electron transport. Electron transport inhibition at the cytochrome bc1 complex would result in the accumulation of ubiquinol that can readily donate electrons to molecular oxygen, resulting in superoxide-generated free radicals. Among the products of oxidative damage are 8-oxo-guanine nucleotides, which can basepair with A when incorporated into DNA; hence, replacement of A:T basepairs with G:C basepairs is a common consequence of free radical mutagenesis (Friedberg et al., 1995). It is tempting to speculate that atovaquone may indirectly induce mutations in mtDNA by causing increased oxidative damage within the mitochondria through electron transport inhibition.

Visualization of the atovaquone resistance-associated residues

Inhibitors of the cytochrome bc1 complex have been proposed to act mainly as ubiquinone/ubiquinol antagonists by interfering with ubiquinol oxidation or ubiquinone reduction steps of the proton-motive Q cycle (Trumpower and Gennis, 1994). Crystallographic evidence has now localized myxothiazol, stigmatellin and antimycin in their respective sites that correspond very well with the locations of the mutations providing resistance in various systems (Xia et al., 1997; Iwata et al., 1998; Kim et al., 1998; Zhang et al., 1998). Atovaquone resistance mutations also localized to the general vicinity of the ubiquinol oxidation region of cytochrome b. They mapped within a highly conserved 15-amino-acid region of this subunit (Fig. 3B), which contains the universal PEWY sequence found in all cytochrome b. Of the five amino acid changes conferring atovaquone resistance, three (I258, Y268 and L271; blue highlight in Fig. 3B) involved residues that are absolutely conserved in all cytochrome b, while the other two (F267 and K272; yellow highlight in Fig. 3B) involved residues that were different between the parasite and vertebrate proteins. The crystal structure of the cytochrome bc1 complex has recently been solved for the bovine (Xia et al., 1997; Iwata et al., 1998; Kim et al., 1998) and chicken (Zhang et al., 1998) mitochondrial proteins. Using the co-ordinates for the chicken cytochrome bc1 complex, we visualized the topology of the atovaquone resistance-conferring amino acid residues within the three-dimensional structure of the complex. Figure 4 depicts a model for the three redox-active subunits (cytochrome b, iron–sulphur protein and cytochrome c1) of the chicken mitochondrial cytochrome bc1 complex and a close-up view of the relevant region homologous to the loop in which resistance-associated amino acids are located. Clearly, as seen in Fig. 4D, the side-chains of these amino acids line a discrete cavity in close proximity to the low potential haem (bL) of cytochrome b as well as the iron–sulphur cluster, the two redox centres critical for ubiquinol oxidation. This region has been postulated to be part of the ubiquinol oxidation site, in which electron transfer to the iron–sulphur protein and the consequent charge separation result in proton translocation. This site appears to be proximal to, but clearly distinct from, the myxothiazol binding region (the cd1 helix highlighted in orange in Fig. 4C). Experimental observation of a lower level cross-resistance of atovaquone-resistant parasites to myxothiazol (Fig. 2C; Table 1[link]) supports this proposition.

Figure 4.

. Structural features of the atovaquone resistance-conferring region. A. Chemical structures of ubiquinone (CoQ) and atovaquone. B. Visualization of the three redox-active subunits of the chicken mitochondrial cytochrome bc1 complex (Zhang et al., 1998). Cytochrome b (yellow), cytochrome c1 (magenta) and the iron–sulphur subunit (green) are shown along with the bL and bH haems (grey). The cyan colour identifies the loop bracketed by positions 264–288 of the chicken cytochrome b. C. A close-up view of the region outlined by a square in (B). The loop homologous to the region shown in Fig. 3B is coloured in cyan with the amino acids positions altered in the resistant parasites highlighted in red (the numbering is for the chicken cytochrome b). The orange-coloured helix corresponds to the region where most of the mutations conferring resistance to myxothiazole are found (Brasseur et al., 1996). D. A space-filling model viewed from the inside of the cavity with the chicken cytochrome b positions 264–288 in cyan, except for the side-chains of Ile-269 (white), Ala-278 (orange), Tyr-279 (dark green), Leu-282(yellow) and Arg-283(magenta) to visualize the topological relationship between atovaquone resistance-associated residues, the iron–sulphur subunit (amino acids 137–149 in green) and haem bL (grey). It is clear that atovaquone resistance-conferring amino acid positions line a discrete cavity proximal to the region critical for ubiquinol oxidation.

In the cavity defined by the atovaquone resistance-associated mutations (Fig. 4D), two amino acid residues that bear different side-chains in the parasite and vertebrate cytochrome b seem to acquire a special significance as to the selectivity of atovaquone binding. These are F267 and K272 in malaria parasites, corresponding to A278 and R283 in the chicken or human respectively (yellow highlights in Fig. 3B). These positions are also conserved in Toxoplasma and Theileria cytochrome b (in Toxoplasma, there is a tyrosine instead of a phenylalanine at position 267). Hence, we would suggest that an aromatic side-chain at position 267 in combination with lysine at 272 is required for sensitivity to atovaquone. If so, then in the vertebrate cytochrome b, alanine at position 278 in combination with the larger arginine at position 283 would appear to be responsible for resistance to atovaquone. This then would explain the selective toxicity of atovaquone towards parasites without affecting the vertebrate mitochondrial functions. Consistent with this proposal is the observation that other atovaquone resistance-associated amino acid changes also subtly alter the hydrophobicity or volume of this cavity at absolutely conserved positions (blue highlighting in Fig. 3B). This may greatly diminish the accessibility and/or affinity of atovaquone for this cavity without significantly affecting that of ubiquinol and effective electron transfer to the iron–sulphur protein. For efficient electron transport, ubiquinone has to move in and out of the catalytic sites of the bc1 complex. The flexible polyisoprenyl side-chain at the 2-carbon position (Fig. 4A) could aid this movement, whereas the relatively inflexible cyclohexyl-chlorophenyl side-chain at the 2-carbon position in atovaquone (Fig. 4A) may hinder it, possibly by interacting directly with the resistance-associated residues of cytochrome b.

Implications for chemotherapy

The structural features of the atovaquone binding site revealed here may assist in further development of selectively toxic drugs against structurally distinct cytochrome bc1 complexes of mitochondria-containing eukaryotes. Such organisms form a vast group, ranging from protists and fungi to nematodes, and exact a major toll on health and economy. Rapid emergence of resistance, however, is likely to be a major problem if such compounds are used as single agents. To counter this, it is conceivable that combination therapy with multiple hydroxynaphthoquinones with subtle differences in their side-chains may provide a means of restricting the rapid emergence of resistance. The acquisition of resistance to such drug combinations would require multiple mutations; such mutations would be rare and may be incompatible with the functional integrity of the parasite cytochrome bc1 complex.

Experimental procedures

Parasites, infection and drug treatment

P. yoelii strain 17XL, cloned by micromanipulation, was used to infect BALB/c mice by injecting 105 parasitized erythrocytes as described previously (Srivastava et al., 1997). Parasitaemia was monitored by Giemsa staining of blood smears, and atovaquone treatment was initiated when parasitaemia reached 5–8%. The drug was administered orally at a 5 mg kg−1 dose. Resistant parasites that emerged were passaged through mice and were resistant to a 20 mg kg−1 dose of atovaquone. For mitochondrial physiology studies, parasites from each line were harvested from five infected mice at parasitaemia levels of ≈ 60%, and enriched by Percoll gradient as described previously (Srivastava et al., 1997).

Mitochondrial membrane potential and respiration assays

Assays of ΔΨm intact parasitized erythrocytes using flow cytometric measurements of DiOC6 were carried out using procedures reported previously (Srivastava et al., 1997). Inhibitors at concentrations ranging from 10 pM to 0.1 mM were tested for their effect on ΔΨm, and the EC50 values were determined from the resulting curves. Oxygen consumption by intact parasitized erythrocytes was measured using Clark's oxygen electrode as described previously (Srivastava et al., 1997). The rate of oxygen consumption was measured in a reaction volume of 1.5 ml containing 108 parasitized red blood cells ml−1 at 37°C and calculated as nAO/108 infected erythrocytes min−1. The rate of respiration ranged from 15 to 20 nAO/108 infected erythrocytes min−1 for these experiments. The difference in the rate of O2 consumption by the infected erythrocytes in the presence compared with the absence of, the compound was calculated as the measure of respiration inhibition.

DNA analysis

DNA was extracted from various parasite lines using standard procedures and subjected to SSCP analysis with PCR primers from the 6 kb mtDNA as described previously (Vaidya et al., 1993b). The cytochrome b gene was isolated by PCR amplification and sequenced by an automated sequencer. For one parasite line (AR1), the entire mtDNA was cloned from an EcoRI library and sequenced in its entirety using an automated sequencer.

Molecular modelling

The atomic co-ordinates (Zhang et al., 1998) of the chicken heart mitochondrial cytochrome b, cytochrome c1 and iron–sulphur protein subunits kindly provided by Dr E. Berry (Lawrence Berkeley Laboratory, Berkeley, CA, USA) were used to visualize the atovaquone resistance-associated amino acid positions of cytochrome b and their location in the cytochrome bc1 complex.

Footnotes

  1. *Present address: Chiron Corporation, Emeryville, CA, USA

Acknowledgements

We thank Dr Ed Berry (Lawrence Berkeley Laboratory, Berkeley, CA, USA) for providing co-ordinates for the chicken heart mitochondrial bc1 complex structure before its release from the Protein Data Bank, and Glaxo Wellcome Pharmaceuticals for providing atovaquone. This work was supported by grants from the National Institutes of Health to A.B.V. (AI28398) and F.D. (GM38237).

Ancillary