X-ray and cryo-EM structures of inhibitor-bound cytochrome bc1 complexes for structure-based drug discovery

Abstract

Cytochrome bc₁, a dimeric multi-subunit electron-transport protein embedded in the inner mitochondrial membrane, is a major drug target for the treatment and prevention of malaria and toxoplasmosis. Structural studies of cytochrome bc₁ from mammalian homologues co-crystallized with lead compounds have underpinned structure-based drug design to develop compounds with higher potency and selectivity. However, owing to the limited amount of cytochrome bc₁ that may be available from parasites, all efforts have been focused on homologous cytochrome bc₁ complexes from mammalian species, which has resulted in the failure of some drug candidates owing to toxicity in the host. Crystallographic studies of the native parasite proteins are not feasible owing to limited availability of the proteins. Here, it is demonstrated that cytochrome bc₁ is highly amenable to single-particle cryo-EM (which uses significantly less protein) by solving the apo and two inhibitor-bound structures to ∼4.1 Å resolution, revealing clear inhibitor density at the binding site. Therefore, cryo-EM is proposed as a viable alternative method for structure-based drug discovery using both host and parasite enzymes.

Keywords: cryo-EM; cryo-electron microscopy; cytochrome bc1; electron-transport chain; malaria; membrane proteins.

Figure 1

Crystal structure of the Q_i site of bovine bc ₁ with the SCR0911 inhibitor bound. (a) Cartoon representation of cytochrome b (blue) and the Rieske protein (grey). The Q_o and Q_i sites of the cytochrome b subunit are highlighted in black boxes. Haems b _L and b _H are shown as blue sticks. Omit F _o − F _c electron density (in green) is contoured at the 3σ level, showing density for the SCR0911 inhibitor. (b) SCR0911 inhibitor bound in the Q_i site. Residues and inhibitor are shown as sticks. The 2F _o − F _c electron-density map is contoured at the 1σ level, with the inhibitor density coloured green and the protein density in purple. Hydrogen bonds between inhibitors and residues are illustrated by black dashed lines. (c) The crystal structures of the Q_i site in bc ₁–SCR0911 (purple) and bc ₁–GSK932121 (grey) have been superimposed, showing the different binding modes of the pyridone (GSK932121, grey) and quinolone (SCR0911, purple) inhibitors.

Figure 2

The bovine bc ₁ structures determined by cryo-EM. Local resolution maps, coloured on the same scale, of (a) apo bc ₁, (b) bc ₁–GSK932121 and (c) bc ₁–SCR0911, which show that the core of the complex is at the highest resolution and that the Rieske protein is the poorest resolved feature in the map. A comparison of the map quality within each map (coloured grey, gold and cyan for the apo, GSK932121-bound and SCR0911-bound structures, respectively) is shown for two representative transmembrane helices (d, e) and a β-strand within the soluble domain (f). In (d)–(f), the density was contoured at 4σ, with the side chains being better resolved in the two inhibitor-bound structures.

Figure 3

Analysis of the Q_i site in the three different cryo-EM maps with the density contoured at 3σ. (a) The Q_i site in the apo bc ₁ cryo-EM map shows minimal density (purple mesh) which does not correspond to any side chains or the haem b _L group, suggesting noise within the map or that the natural substrate ubiquinone is bound in a small number of bc ₁ molecules. (b) The Q_i site in the cryo-EM map of bc ₁–GSK932121. The inhibitor density (green) suggests there are two modes of inhibitor binding, accompanied by rotation around the oxygen–carbon bond. The binding pose shown in green agrees with the crystal structure, with the trifluoromethyl group pointing towards Met194. There is additional density which suggests that the trifluoromethoxyphenyl group could be rotated and point towards Asp228, revealing an additional mode of binding (shown in blue). (c) The Q_i site in the cryo-EM map of bc ₁–SCR0911, with the inhibitor shown in pink and located in strong density. The inhibitor is expected to make a hydrogen-bond contact with His201 and strongly fits the density.

Figure 4

Global comparison between cryo-EM-derived and X-ray-derived SCR0911-bound structures. (a) The cryo-EM and X-ray bc ₁–SCR0911 structures have been overlaid and the crystal structure is coloured according to the calculated C^α r.m.s.d. value (low coloured cyan and high coloured maroon). The main difference (r.m.s.d. of >2 Å) occurs for the Rieske domain (in the black square), which is known to be mobile during the catalytic cycle. (b) An overlay of the Rieske domains in the cryo-EM-derived (cyan) and X-ray-derived (grey) bc ₁–SCR0911 structures. (c) The cryo-EM-derived map for the bc ₁–SCR0911 structure with the Rieske domain fitted shown in the same orientation as in (a) and (b).

Figure 5

Comparisons of the Q_i site. (a) The cryo-EM structures of bc ₁–SCR0911 (cyan) and bc ₁–GSK932121 (gold) overlaid. The SCR0911 inhibitor is coloured green and the GSK932121 inhibitor magenta. The overlay of the two structures shows there is no difference in the secondary-structure positions in the two maps and only minor differences in the positions of the amino-acid residues. The largest difference occurs for His201, which forms a hydrogen bond to SCR0911 but not GSK932121. (b) An overlay of the Q_i site of bc ₁–GSK932121 in the cryo-EM (gold) and X-ray (purple) models. The density in gold corresponds to the EM map, which shows a difference in the position of His201 compared with the X-ray structure. (c) An overlay of the Q_i site of bc ₁–SCR0911 in the cryo-EM (cyan) and X-ray (grey) models. There are no significant differences between the two models in either side-chain or secondary-structure positions.