6.2. The structures of the unliganded and the octanoyl-CoA-complexed yeast Δ3-Δ2-enoyl-CoA isomerase and the differences between them (III, IV)
The structure of the yeast Δ3-Δ2-enoyl-CoA isomerase determined in this study is the fifth published structure within the hydratase/isomerase superfamily. The other previously known structures are those of the rat 2-enoyl-CoA hydratase-1 (Engel et al. 1996, Engel et al. 1998), the 4-chlorobenzoyl-CoA dehalogenase from Pseudomonas sp. strain CBS-3 (Benning et al. 1996), the rat Δ3,5-Δ2,4-dienoyl-CoA isomerase (Modis et al. 1998) and the methylmalonyl-CoA decarboxylase from E. coli (Benning et al. 2000). Very recently, the structure of the human AUH protein, the RNA-binding homologue of hydratase-1, has also been published (Kurimoto et al. 2001). Liganded structures are known for all of them, except the dienoyl-CoA isomerase and the AUH protein. The overall structure of the yeast Δ3-Δ2-enoyl-CoA isomerase was found to resemble closely the other structures within the hydratase/isomerase superfamily, having a similar spiral core domain for catalysis and a helical C-terminal domain for trimerization.
In both the unliganded and the liganded yeast isomerase structures, the C-terminal PTS1-like sequence, HisArgLeuCOOH, is disordered. This is considered to be a common feature of peroxisomal proteins, the only exceptions being the rat dienoyl-CoA isomerase, whose PTS1 is buried in a pocket between the two dienoyl-CoA isomerase trimers (Modis et al. 1998), and the SCP-2-like domain of the human MFE-2, whose PTS1 is solvent-exposed (Haapalainen et al. 2001). In the unliganded yeast isomerase (the low pH form), part of the region preceding helix H2, i.e. the residues 74-88, is disordered (Fig. 13A). In the liganded structure (the neutral pH form), however, this region becomes ordered and appears to adopt a helical conformation. This new region, helix H2A, is still quite flexible, since the B-factors are high and no side chain density can be seen for all residues. In addition, H2 takes a different conformation compared to the low pH form (Fig. 13B). Conformational changes are also seen in the helices H9 and H10 in a comparison of the two pH forms.
In both the low and neutral pH forms of the yeast enoyl-CoA isomerase, three monomers are packed into a tight trimer. The mode of assembly into hexamers is, however, rather different. In the neutral pH form, the two trimers interact closely with 159 atom-atom contacts within a contact distance of 3.5 Å, whereas in the low pH form, there are only 30 intertrimer contacts. The distance between the trimers is longer in the low pH form, and the trimers are also rotated 25° with respect to each other. The structurally most variable regions, i.e. the helices H2 and H9, form the contact surface in the intertrimer space and the changes in their conformations thus correlate with the differences in the mode of assembly of the trimer disks into hexamers in the two pH forms. In the neutral pH form, residues from H2 and H9 make up most of the 159 intertrimer contacts and both regions are involved in forming 10 salt bridges to the other trimer (see original article IV, table 2). The importance of these salt bridges rationalizes the possible pH dependency of the mode of assembly of the trimers into hexamers. Ultracentrifugation sedimentation velocity studies also suggest that the different mode of hexamer assembly could be pH-dependent, since at pH 5.6 the yeast isomerase has a molecular mass corresponding to a trimer and at pH 7.2 it is clearly a hexamer. This agrees with the crystallographic data showing that, at low pH, the isomerase trimers are only loosely assembled into a hexamer. Additional ultracentrifugation experiments to confirm these results could, however, not be performed because, at low pH, the yeast enoyl-CoA isomerase does not stay soluble and aggregates rapidly.
There are no major conformational differences at the catalytic site between the low and neutral pH forms, and the mode of binding of octanoyl-CoA into the active site of the neutral pH form is much like expected on the basis of superposition studies. Octanoyl-CoA could be built in all of the three subunits, but the occupancy of the ligand seems to be relatively low as judged from the high B-factors (unpublished results). The mean B-factor for the ligand atoms is 86 Å2, whereas for the protein atoms it is 37 Å2. The ADP and octanoyl moieties of the ligand were well defined in the electron density map, but the pantothenic acid part was mostly disordered (Fig. 12). The considerably higher B-factors for the octanoyl-CoA atoms in comparison to the protein atoms are also seen in the structure of 2-enoyl-CoA hydratase-1 complexed with octanoyl-CoA (Engel et al. 1998). Near the active site, the side chain of Tyr38 leads to a large empty apolar cavity. It is possible that this cavity could be able to bind the long fatty acid chain in catalysis. The octanoyl chain, unfortunately, is too short to reach this hydrophobic pocket, leaving its role still unravelled. Binding studies with longer fatty acid substrates are needed to determine whether this cavity can be used for binding. Moreover, because of the low resolution of the liganded structure (3.3 Å), no precise and reliable analysis of the protein-ligand interactions and the active site architecture can be made.
The side chain of the catalytic residue, Glu158, is positioned so that it contacts both C2 and C4 of the octanoyl-CoA molecule (Fig. 12). The conformation of the actual substrate molecule, cis/trans-3-enoyl-CoA, is suggested to be similar, making deprotonation of C4 and protonation of C2 possible. Both cis- and trans-3-enoyl-CoA are thought to bind quite rigidly in the same conformation up to carbon C4, in order for the reaction to proceed and the right product, trans-2-enoyl-CoA, to form (Fig. 14). The trans-isoforms bind in a similar, straight fashion as saturated fatty acyl-CoAs, such as octanoyl-CoA. Instead, the cis isoform needs more space in the active site pocket because the cis double bond produces a bend in the acyl chain (Fig. 14). Crystallographic binding studies with different substrates are needed to determine how the substrate-binding pocket adopts to the space requirements of kinked acyl chains. The reaction mechanism of the yeast Δ3-Δ2-enoyl-CoA isomerase can be speculated to occur as follows: Glu158 in its deprotonated form abstracts a proton from C2 of 3-enoyl-CoA. This leads to a transition state with double bonds at C1 and C3 and a negative charge on the thioester oxygen. This negative charge is stabilized by hydrogen bonds in the oxyanion hole formed by the main chain NH groups of Ala70 and Leu126 (Figs 11, 12). Subsequently, the protonated Glu158 donates the proton to C4 of the substrate and the product, trans-2-enoyl-CoA, is released (original article III, Fig. 1). The stabilization of the intermediate state of the reaction by hydrogen bonding in the oxyanion hole is a conserved feature in the hydratase/isomerase family (Holden et al. 2001), and also according to the sequence alignment, the position of the residues forming the oxyanion hole is conserved (see original article IV, Fig. 1).
The role of Glu158 as the only catalytic residue is consistent with the experimental data: [1] Glu158Ala mutant is totally inactive, [2] in the crystal structures Glu158 is optimally positioned for catalysis and no other protic residues are in the vicinity, and [3] the sequence alignment of the mitochondrial and peroxisomal enoyl-CoA isomerases (Müller-Newen & Stoffel 1991, Kilponen et al. 1994, Geisbrecht et al. 1999b) (see original article III, Fig. 2) shows that, in all peroxisomal isomerases, a glutamate at position 158 is conserved. In the mitochondrial isomerases, the conserved catalytic glutamate (Glu165 in the rat mitochondrial isomerase, Müller-Newen & Stoffel 1993) is at the position of Phe150 of the yeast isomerase. Still, it has to be emphasized that a better resolution structure with an active site ligand is needed for an accurate analysis of the protein-ligand interactions as well as the reaction mechanism.