Molecular modelling studies using co-crystais

In available crystal structures o f cytochrome bc\ complexes from mammalian sources [35-40] no electron density ascribed to ubiquinone has so far been identified, presumably because o f weak binding o f ubiquinone [31]. Therefore, conclusions about the exact location o f quinone/quinol binding sites within the structure remain speculative. However, structures obtained with Qo site-specific inhibitors have revealed the location o f the Qo site and inhibitors occupying different domains (Fig. 4.6). The crystal structures also revealed that binding o f different occupants leads to changes in the configuration and a substantial expansion o f the site [34, 86]. This has led to the claim that there may be sufficient space to accommodate two ubiquinone molecules [73, 74], a view consistent with the report that two MOA-stilbene inhibitors bind to the fully reduced enzyme Qo site [78]. However, the expansion suggests that the inhibitors do not displace an existing occupant, as was the case for the Qj site upon binding o f antimycin A [34].

To test whether two ubiquinone molecules can bind at the same time to the Qo site, they were modelled into the site simultaneously, approximately into the stigmatellin and MOA-stilbene binding domains (Fig. 4.7). The models were built from published structures 3BCC [35] using Cerius2 from MSI. Energy minimised structures o f ubiquinones were used as a starting point for manual ‘tweaking’ to fit them into the solvent accessible surfaces in the site. The best model in terms o f the available volume has the headgroups overlapping as proposed by Brandt et a l (1996) [69, 85] instead o f the edge-to-edge conformation favoured by D ing et a l (1992) [70, 71]. However, only one tail could be accommodated within the access tunnel and the adjacent volume. The structure with one tail has little unfilled space and substantial expansion would have to occur for the site to be able to accommodate another tail in

the access tunnel and also in the adjacent volume. Moreover, it is hard to envisage how the quinone bound at Qow domain could get in and out o f the structure (Qo site) in reasonable physiological timescale. To allow for proper functioning o f the bc\

complex, as proposed by the double Q variant model for the bifurcation reaction at the

Q o site [69, 70]. However, the molecular modelling study presented in this work is

consistent with mechanistic models such as the ‘mobile’ ubisemiquinone model [34, 45, 84, 86] and the ‘thermodynamic’ explanation [44] (Section 1.9.6.2), involving a single ubiquinone occupant o f the Qo site. The large domain o f the Qo site, which can almost be occupied by two ubiquinone headgroups, may be part o f the in-built control to ensure efficient bifurcation o f the two electrons from ubiquinol between the high- and low-potential chains.

Fig. 4.7. Double occupancy of the Qo reaction site: Qos and Qow A) Side view B)

Top view. Molecular modelling o f ubiquinones into the Qo site based on the

stigmatellin and MOA-stilbene bound structures using co-ordinates from Zhang et al.

datafile 3BCC [35]. The ubiquinone molecules are shown as space filling models (Qow, green and Qos, grey). Selected residues are shown as sticks and the rest of the protein as wireframe. ‘Rieske’ ISP cluster atoms are shown as small (iron) green and yellow (sulphur) spheres. The cloud and its different colours are the same as previously described in Fig 4.6B. They represent the solvent accessible surfaces. The visualisation and manual docking were carried out as described in Section 4.2.2.

Other recent modelling studies have suggested that changes o f the order o f only 1.5Â in the atomic positions o f a few neighbouring amino acid residues is required for binding o f two ubiquinone molecules {Bartoschek et a l personal communication, 2001). Movements on this scale are only slightly larger than those, which have been observed experimentally after removal o f ubiquinone from the reaction centre from Rb. viridis [13]. In this model, two ubiquinone molecules were modelled into the stigmatellin site, in contrast to the model presented here, where the ubiquinone molecules were modelled into the stigmatellin and MOA-stilbene binding domains respectively. The molecular modelling presented here and elsewhere is based on the structures in native crystal, which may not necessarily represent the flexibility o f physiological structure. Therefore, It will clearly be o f importance in the future to determine the occupancy under physiological conditions.

In summary, the molecular modelling study presented here show that only a single ubiquinone headgroup can be adequately accommodated within the Qo site and there is a degree o f overlap when a second ubiquinone molecule is modelled into the site. Furthermore, only one ‘tail’ could be accommodated within the access tunnel and adjacent volume. This is fully consistent with a model o f overlapping binding sites o f two classes o f Qo site inhibitors from the inhibitor competition and mutagenesis studies, and o f MOA-stilbene and decylubiquinol from the kinetic studies (Chapter 3).

4.3.3.1

Rational design of inhibitors

Development o f fungicide resistance in the field can be a major problem. This could be overcome by applying two compounds, if they bind to the same active site in a dissimilar way, e.g. myxothiazol and stigmatellin. The target is then less likely to develop resistance to both compounds. However, in the case o f bc\ complex Qo site,

as evident from the studies o f mutants resistant to inhibitors, it is certainly possible for resistance to both compounds to be developed by a single mutation. Apart from resistance in the field, there is a need for competing companies to circumvent existing patents (a process termed ‘patent busting’) and to develop more potent inhibitors. The availability o f a great deal o f information from molecular, biochemistry and biophysical studies coupled with crystal structures o f the bc\ complex in the native and inhibited states makes it possible for rational design o f new inhibitors. The kinetics, inhibitor and mutagenesis studies have indicated that the Qo site has two binding but overlapping domains, a feature some have interpreted as favouring doubly occupancy [73, 75]. Taking this into consideration, it is feasible that a single compound or two compounds could be designed to fit adequately into the two domains. This compound(s) would have all the favourable contacts exhibited by the two inhibitors.

The co-crystals with stigmatellin and MOA-stilbene bound indicate that E271 (chicken numbering) is a ligand (hydrogen bond) to both o f these inhibitors [31, 34]. The backbone NH is modelled as forming an H-bond to M OA-stilbene and mutations at this residue lead to stigmatellin resistance but not resistance to myxothiazol [34]. The side chain presumably plays no role in binding o f class II inhibitors. The side chain o f E271 is also found in different configurations in the two co-crystals (Fig. 4.6) [34]. It projects into the Qo pocket in the stigmatellin structures, close enough to H- bond with the OH group o f the second ring o f the inhibitor. However, in the myxothiazol and MOA-stilbene structures, the side chain is rotated away from the pocket (Fig. 4.6). This residue form part o f the well conserved -PEW Y-sequence and any substitution o f this residue leads to significant loss o f activity [31-34].

H16I