The multifunctional enzyme, type-1 (MFE1) is involved in several lipid metabolizing pathways. It catalyses: (a) enoyl-CoA isomerase and (b) enoyl-CoA hydratase (EC 126.96.36.199) reactions in its N-terminal crotonase part, as well as (3) a 3S-hydroxy-acyl-CoA dehydrogenase (HAD; EC 188.8.131.52) reaction in its C-terminal 3S-hydroxy-acyl-CoA dehydrogenase part. Crystallographic binding studies with rat peroxisomal MFE1, using unbranched and branched 2E-enoyl-CoA substrate molecules, show that the substrate has been hydrated by the enzyme in the crystal and that the product, 3S-hydroxy-acyl-CoA, remains bound in the crotonase active site. The fatty acid tail points into an exit tunnel shaped by loop-2. The thioester oxygen is bound in the classical oxyanion hole of the crotonase fold, stabilizing the enolate reaction intermediate. The structural data of these enzyme product complexes suggest that the catalytic base, Glu123, initiates the isomerase reaction by abstracting the C2-proton from the substrate molecule. Subsequently, in the hydratase reaction, Glu123 completes the catalytic cycle by reprotonating the C2 atom. A catalytic water, bound between the OE1-atoms of the two catalytic glutamates, Glu103 and Glu123, plays an important role in the enoyl-CoA isomerase and the enoyl-CoA hydratase reaction mechanism of MFE1. The structural variability of loop-2 between MFE1 and its monofunctional homologues correlates with differences in the respective substrate preferences and catalytic rates.
The structures have been deposited in the Protein Data Bank under accession numbers: 3ZW8 (MFE1 apo), 3ZW9 (MFE1 2S-methyl-3S-hydroxy-butanoyl-CoA complex), 3ZWA (MFE1 3S-hydroxy-hexanoyl-CoA complex), 3ZWB (MFE1-E123A 2E-hexenoyl-CoA complex) and 3ZWC (MFE1 3S-hydroxy-decanoyl-CoA complex).
Structured digital abstract
isomerase or enoyl-CoA isomerase, monofunctional Δ3,Δ2-enoyl-CoA isomerase
3S-hydroxy-acyl-CoA dehydrogenase (also referred to as l-3-hydroxy-acyl-CoA dehydrogenase)
- hydratase or enoyl-CoA hydratase
monofunctional Δ2-enoyl-CoA hydratase-1, which generates 3S-hydroxy-acyl-CoA
multifunctional enzyme, type-1 (also referred to as l-bifunctional enzyme)
multifunctional enzyme, type-2 (also referred to as d-bifunctional enzyme)
trifunctional enzyme of the β-oxidation pathway
Multifunctional enzyme, type-1 (MFE1) is an abundant mammalian, peroxisomal enzyme . It is a monomeric protein, consisting of ~ 700 residues. The N-terminal part adopts the crotonase fold [2-4] and the C-terminal part has the 3S-hydroxy-acyl-CoA dehydrogenase (HAD; EC 184.108.40.206)-fold [2, 5-7]. The active sites of the N- and the C-terminal parts catalyse, respectively, the second (2E-enoyl-CoA hydration) and third (3-hydroxy-acyl-CoA dehydrogenation) reactions  of the β-oxidation pathway of fatty acid metabolism (Fig. 1). In peroxisomes, the same two reactions are catalysed also by the multifunctional enzyme, type-2 (MFE2) . The crystal structures of rat peroxisomal MFE1, complexed with CoA , and fruit fly peroxisomal MFE2, unliganded , have been described. MFE2 does not share any sequential or structural similarities with MFE1. However, MFE1 is homologous to the α-chain of the trifunctional enzyme (TFE), a α2β2-tetramer, which is membrane bound in mitochondria [10, 11] and soluble in bacteria [10-12].
The 3-hydroxy-acyl-CoA intermediate generated from 2E-enoyl-CoA has different chirality in the MFE1 and MFE2 catalysed hydratase reactions, such that the 3S-hydroxy-acyl-CoA intermediate is formed by MFE1 (Fig. 1) and 3R-hydroxy-acyl-CoA by MFE2 . When tested in vitro, the substrate specificity properties of MFE1 and MFE2 are overlapping. Dysfunctional MFE2 is one of the most common reasons of peroxisomal disorder in humans . The precise physiological function of MFE1 is not fully understood, although studies indicate that MFE1 is involved in several lipid metabolizing pathways, including the degradation of dicarboxylic acids [14-16]. MFE1 can also use 2-methyl-2E-enoyl-CoA as a substrate , which is hydrated into 2S-methyl-3S-hydroxy-acyl-CoA (Fig. 1). It has been proposed that the latter activity is important in the bile acid synthesis pathway . Other experimental work shows that the MFE1-HAD part is functioning also in the synthesis of bile acids from 24S-oxysterols . The HAD part of MFE1 is loosely connected to the crotonase part by a linker helix, such that the catalytic sites are separated from each other via a tunnel that has an excess of positively-charged side chains .
The crotonase part of MFE1 is sequentially related to the hydratases and isomerases (Fig. 2), which also adopt the crotonase fold (Fig. 3). These monofunctional enzymes are either trimeric [20, 21] or hexameric (dimers of trimers) [22, 23] in contrast to the monomeric MFE1. The catalytic rate of MFE1 hydratase (~ 50 s−1)  is lower than the corresponding activity of the monofunctional hydratases (~ 1000 s−1) . The latter hydratase displays also residual isomerase activity (0.2 s−1) .
The crotonase domain of MFE1 also has isomerase activity [1, 24, 26] (Fig. 1). Enoyl-CoA isomerase activity is crucial for the auxiliary pathway of the degradation of fatty acids with Z double bonds or with E double bonds at odd positions . There are two types of enoyl-CoA isomerase (ECI; EC 220.127.116.11): ECI1 (short chain ECI) and ECI2 (long chain ECI) . These isomerases use fatty acyl moieties with either Z or E double bonds . Sequence (Fig. 2) and structural comparisons  show that ECI1 and the crotonase domain of MFE1 have a common catalytic residue, whereas, in ECI2, the corresponding catalytic glutamate protrudes from loop-4 (Fig. 2). Zhang et al.  speculate that the MFE1 isomerase reaction is in particular relevant for the 2,5- to 3,5-enoyl-CoA interconversion, which is the reverse of the Δ3,Δ2-enoyl-CoA isomerase reaction, depicted in Fig. 1. In any case, the rat MFE1, when expressed heterologously in an ECI2-deleted yeast strain, can restore the growth of this yeast variant on oleic acid, demonstrating that the catalysis of the Δ3,Δ2-enoyl-CoA isomerase reaction is functional in vivo .
The crotonase fold (Fig. 3) is observed in enzymes with a wide range of different catalytic functions, as a result of different substrate and reaction specificities [3, 4]. The crotonase fold has a central parallel β-sheet of four strands (A1–4) followed by loops (loops 1–4) and helices (H1–4) (Figs 2 and 3). The active site is at the N-terminus of the buried helix H3. The active site is also shaped by residues of loop-2 and loop-4. These loops, loop-2, loop-3 and loop-4, are continuing into helix H2, helix H3 and helix H4, respectively. The variability of the length and composition of the loop-2 region (Fig. 2) correlates with different substrate specificities . The two catalytic glutamates in MFE1 are conserved in the hydratases [30, 31], as well as in the TFE α-chain (Fig. 2). The common feature of an active site built on the crotonase fold is the presence of an oxyanion hole, which stabilizes the thioester-enolate reaction intermediate anion [32, 33]. This important property makes the crotonase fold an attractive framework for the development of new biocatalytic activities, as reported previously .
The present study focuses on the biocatalytic properties of the N-terminal crotonase part of full-length MFE1. The reported structures concern the apo as well as the complexes of the crotonase domain with bound product molecules, 3S-hydroxy-hexanoyl-CoA, 3S-hydroxy-decanoyl-CoA and 2S-methyl-3S-hydroxy-butanoyl-CoA (Fig. 1). The structures of these complexes are also compared with the structures of the liganded monofunctional enoyl-CoA hydratase (EC 18.104.22.168) and enoyl-CoA isomerase. New insights into the reaction mechanism of the MFE1 crotonase active site with respect to the hydratase and the isomerase reactions are discussed.
Altogether, five structures of MFE1 complexes were determined. The precise crystal handling protocols are detailed in Table 1. The structures were refined at resolutions ranging from 2.3 to 3.1 Å (Table 2). The mode of binding of the ligands and the conformation of the catalytic glutamates are well defined by the respective electron density maps (Fig. 4). For each of the structures, the crystals were grown in the presence of CoA , except for the E123A mutant variant, which was obtained by cocrystallization with 2E-hexenoyl-CoA. Four structures were determined using crystals of wild-type MFE1, including the apo structure. There are two molecules (A and B) in the asymmetric unit (Fig. 3 and Table 2). In molecule A of the apo structure, a glycerol is bound in the active site , although not in molecule B as a result of a different conformation of its nearby loop-2. In molecule B of the apo structure, the crotonase active site has a well defined water molecule bound between the side chains of the two catalytic glutamates, Glu103 (at the N-terminal of H3) and Glu123 (loop-4) (Fig. 5), also referred to as the catalytic water. This active site also contains a bound sulfate ion (Fig. 6). The active site of molecule B has been used for most of the detailed descriptions of the structures, unless specified otherwise.
|apo||2S-methyl-3S-hydroxy-butanoyl-CoA||3S-hydroxy- hexanoyl-CoA||3S-hydroxy- decanoyl-CoA||2E-hexenoyl-CoA|
|Ligand used in crystallization buffera||2 mm CoA||2 mm CoA||2 mm CoA||2 mm CoA||0.2 mm 2E-hexenoyl-CoA|
|Ligands used for soaking with the apo crystals||–||0.2 mm tiglyl-CoA + 0.2 mm NADH||0.2 mm 2E-hexenoyl-CoA||0.2 mm 2E-decenoyl-CoA + 0.2 mm NADH||–|
|Cryo protectant bufferc||15% glycerol in crystallization buffer||15% glycerol + 0.2 mm tiglyl-CoA + 0.2 mm NADH in crystallization buffer||15% glycerol + 0.2 mm 2E-hexenoyl-CoA in crystallization buffer||15% glycerol + 0.2 mm 2E-decenoyl-CoA + 0.2 mm NADH in crystallization buffer||15% glycerol + 0.2 mm 2E-hexenoyl-CoA in crystallization buffer|
|apo (WT)||Complexes with CoA derivative|
|2S-methyl-3S-hydroxy-butanoyl-CoA (WT)||3S-hydroxy-hexanoyl-CoA (WT)||3S-hydroxy-decanoyl-CoA (WT)||2E-hexenoyl-CoA (E123A)|
|Data processing statistics|
|Unit cell parameters (Å)||65.2 125.7 226.4||65.6 126.7 225.4||65.4 125.8 225.8||65.6 126.5 225.0||65.4 125.5 224.2|
|Number of molecules per asymmetric unit||2||2||2||2||2|
|Resolution range (Å)a||44.4–2.5 (2.6–2.5)||49.5–2.9 (3.1–2.9)||64.6–2.5 (2.6–2.5)||44.6–2.3 (2.4–2.3)||65.4–3.1 (3.3–3.1)|
|Completeness (%)a||86.8 (73.6)||99.7 (98.3)||95.9 (88.6)||99.9 (99.9)||97.8 (86.2)|
|<I/σ(I)>a||8.9 (5.4)||14.8 (3.6)||11.9 (5.4)||18.8 (5.7)||4.9 (2.7)|
|Rpim (%)a,b||5.8 (10.2)||4.1 (18.7)||3.9 (9.8)||3.0 (12.5)||13.0 (23.1)|
|Number of unique reflectionsa||56678 (6882)||42965 (6074)||64345 (8561)||89967 (12105)||33670 (4260)|
|Redundancya||5.5 (4.9)||7.4 (7.0)||3.3 (2.7)||6.4 (6.4)||4.2 (3.4)|
|Wilson B-factor (Å2)||36.0||72.6||44.0||32.5||59.7|
|Software used||imosflm, scala||xds, scala||imosflm, scala||imosflm, scala||imosflm, scala|
|Number of reflections||53526||40409||60772||79539||31832|
|Number of protein atoms||11098||11074||11112||11086||11088|
|Number of waters||246||114||363||414||136|
|Number of crotonase active site ligands||2||2||2||2||2|
|Number of other molecules|
|rmsd, bonds (Å)||0.016||0.014||0.017||0.019||0.012|
|rmsd, angles (˚)||1.6||1.7||1.7||1.85||1.4|
|rmsd B (Å2)|
|Protein main chain||0.7||0.5||0.6||0.8||0.5|
|Protein side chain||2.2||1.3||1.8||2.4||1.4|
|Average B (Å2)|
|A-crotonase active site ligand||15.8c||56.6||43.4||61.1||55.2|
|B-crotonase active site ligand||10.1c||92.8||37.3||74.2||34.0|
|Ramachandran plot (molprobity) (%)|
|Protein Data Bank code||3ZW8||3ZW9||3ZWA||3ZWC||3ZWB|
For the crystallographic binding studies different 2E-enoyl-CoA molecules have been used (Tables 1 and 2), including tiglyl-CoA (Fig. 1). Tiglyl-CoA is a modified 2E-butenoyl-CoA, being methylated at the C2 carbon; if extended at its ω-end by a steroid moiety, it resembles the structure of intermediates of the bile acid synthesis pathway. The experiments of the wild-type MFE1 crystals with the three 2E-enoyl-CoA substrates show that these molecules become hydrated and remain bound in the crotonase active site. Tight binding of such 3-hydroxy-acyl-CoA product molecules to the monofunctional enoyl-CoA hydratase has been observed in kinetic studies ; however, structural studies of these enzyme-product complexes have not been carried out.
In the case of crystals of the E123A variant, the crystallographic data confirmed the mutation and indicate that the bound ligand is the unhydrated substrate, 2E-hexenoyl-CoA (Fig. 4), instead of the product. The water structure is different because of the mutation of Glu123 into an alanine. Its fatty acid tail is bound in the same way as that seen for the mode of binding of the product molecules. The isomerase and hydratase activity assays showed that this variant is inactive for both reactions. Apparently, Glu123 is important for catalysis but not for binding.
No substrate or product molecules (CoA derivatives) are bound in the HAD active site in any of the complex structures. In the experiments for the apo structure and the structures with 3S-hydroxy-hexanoyl-CoA and 3S-hydroxy-decanoyl-CoA, no cofactor (NAD+ or NADH) was added in the crystal handling steps and, indeed, in these structures, no cofactor is bound in the HAD active site. In the studies with tiglyl-CoA and 2E-decenoyl-CoA, the reduced cofactor (NADH) has been included in the soaking experiments. In these experiments, which potentially would allow for the formation of a dead end ternary complex in the HAD active site, only NADH is bound in the HAD active site and its mode of binding is similar to that observed for ADP in the structure of the MFE1–CoA complex  and also similar to that observed in the structures of the complexes of NADH bound to the monofunctional HAD [5, 36]. For the latter enzyme, it is proposed that the 3-hydroxy product only binds in the active site when it is complexed with NAD+ . Therefore, these crystallographic binding experiments suggest that such an ordered mechanism could also exist in the HAD active site of MFE1.
The fatty acid binding tunnel of the crotonase domain
The binding pocket of the fatty acid tail is shaped by loop-2 (Fig. 3). Loop-2 of the crotonase superfamily of enzymes has significant structural [22, 37] and sequential variability (Fig. 2). This structural variability is also apparent when comparing the structures of molecules A and B of the asymmetric unit of the crystallized MFE1. In the A and B molecules, the crotonase domain and the HAD domain have adopted a slightly different orientation with respect to each other . This structural variation must be a result of differences in the crystal packing of molecules A and B. Crystal packing effects also cause the different loop-2 conformations of molecules A and B. In MFE1, loop-2 consists of residues 63–74 (Fig. 2). These loop residues have higher than average B-factors but are nevertheless well defined by their respective electron density maps. This structural variability is depicted in Fig. 3, in which the active sites of the A and B molecules of the 3S-hydroxy-decanoyl-CoA complexes are compared. It is observed that the fatty acid tail finds a different exit tunnel in molecules A and B. In molecule B, the exit tunnel is between loop-2 and loop-4. In molecule A, the exit tunnel is between loop-2 and loop-1. In both molecules, the fatty acid tail fits snugly in the respective exit tunnel. Therefore, it is predicted that further conformational changes will be required for binding of the steroid moiety of the bile acid substrate (e.g. similar to the order-to-disorder transition of loop-2 as observed for the mode of binding of octanoyl-CoA to the monofunctional hydratase) . Interestingly, in the crotonase domain of the α-chain of the bacterial TFE , the exit tunnel as seen in molecule A is also observed. In this structure, loop-2 is considerably longer (Fig. 2) and the tunnel is occupied by a n-octylpentaoxyethylene (C8E5) molecule (Fig. 3), possibly mimicking the mode of binding of the fatty acid tail. The conformational flexibility of loop-2 of MFE1 correlates with its broad substrate specificity.
The MFE1 crotonase active site catalyses two reactions: the isomerase and the hydratase reactions. Several detailed reaction mechanism studies of the monofunctional hydratase have been reported [31, 40], as reviewed previously . For the isomerase, a crystal structure of the complex with the octanoyl-CoA is available . In octanoyl-CoA, the fatty acid tail is saturated and therefore it mimics the substrate of the isomerase reaction. For the hydratase, the two crystal structures are the complexes with hexadienoyl-CoA  and with 4-dimethylamino cinnamoyl-CoA . The latter two compounds mimic the substrates of the hydratase reaction; however, the hydration reaction is apparently thermodynamically not favoured because of the conjugated double bonds of the fatty acid tail. The structures of these complexes show that the catalytic water is hydrogen-bonded to the two catalytic glutamates, Glu144 and Glu164 (Fig. 6), providing the geometric data of a competent hydratase-substrate Michaelis complex. Crystal structures of the monofunctional hydratase product complexes have not yet been reported.
The mode of binding of the 3-hydroxy-acyl-CoA molecules in the crotonase active site
The structure of the MFE1 3-hydroxy-decanoyl-CoA complex is compared with the MFE1 apo structure in Fig. 6. This comparison includes the structure of a complexed monofunctional hydratase active site, showing that the fatty acid tails in these complexes point in the same direction. The distance between the catalytic water of the apo structure and the 3-hydroxy moiety of the superimposed product molecule is only 0.6 Å. The mode of binding of the hydrated products of 2E-hexenoyl-CoA, 2E-decenoyl-CoA and tiglyl-CoA are compared in Fig. 6. Indeed, the chirality of the 3-hydroxy moieties conforms to the 3S-geometry. In each of these product molecules, the 3-hydroxy group is in the same position.
The hydration of a 2-methyl-2E-enoyl-CoA substrate molecule by MFE1 generates two chiral centres: at C2 and C3 [17, 42, 43]. Both chiral centers have S-geometry (Fig. 1), as is also seen from the crystallographic experiments with tiglyl-CoA. For the 2-methyl-3-hydroxy-acyl-CoA molecule (Fig. 6), the product of the reaction with tiglyl-CoA, the 2S-methyl group fits in a hydrophobic binding pocket generated by the side chains of Ala61, Phe66, Leu73 (loop-2), Ile128 (loop-4) and Phe255 (helix-9C). In the 2R-geometry, a clash would occur with CG(Glu123); therefore, this binding pocket for the 2S-methyl group explains the unique specificity of MFE1 for 2S-methyl branched acyl-CoA molecules.
The comparison of the active sites of MFE1 and enoyl-CoA isomerase
MFE1 has considerable enoyl-CoA isomerase activity, ~ 1 s−1 [24, 28]; however, larger activities, 10–100 s−1 , are observed for the monofunctional homologues. Therefore, the MFE1 active site has also been compared with the active site of the liganded mitochondrial ECI1 (Fig. 6). Glu123 (MFE1) corresponds to Glu136 of ECI1 (Fig. 2), whereas the residue corresponding to Glu103 (MFE1) is Leu114. The catalytic water is absent in the isomerase active site. Also different in the isomerase is the conformation of loop-4. This loop-4 (residues 131–134 in MFE1) anchors the Glu103 side chain of MFE1 by main chain NH hydrogen-bonding interactions with OE2(Glu103) (Figs 7 and 8). The conformation of this loop is conserved between MFE1 and the enoyl-CoA hydratase. The fatty acid tail is bound in a different conformation in the enoyl-CoA isomerase (Fig. 6), as it points to the differently folded loop-4 . These distinct structural differences correlate with the notion that the respective catalytic rates are not the same.
The crotonase oxyanion hole
The crotonase oxyanion hole provides this fold with unique biocatalytic properties. In MFE1, the oxyanion hole is formed by the NH groups of Ala61 (loop-2) and Gly100 (loop-3) (Figs 2 and 8). The structures of the MFE1 complexes show that the thioester oxygen binds similarly, as seen in the monofunctional hydratases (Fig. 6A) and isomerases (Fig. 6). On binding of the thioester oxygen in this oxyanion hole, the pKa of the C2-proton of the thioester moiety is lowered because the oxyanion hole stabilizes the thioester enolate intermediate [33, 34, 40, 44, 45]. The apo MFE1 active site also contains a sulfate ion bound in the oxyanion hole (Fig. 6). One of the oxygen atoms of the sulfate ion is hydrogen-bonded to both main chain NH groups of the oxyanion hole. Furthermore, hydrogen-bonding interactions of a protonated sulfate oxygen with O(Ala59) and O(Ala61) indicate that the bound sulfate ion is . The mode of binding of the ion mimics the stabilizing electrostatic interactions in the oxyanion hole of the complex with a thioester enolate intermediate. As discussed in detail below, this oxyanion hole is critically important for the catalysis of both the isomerase and hydratase reactions (Fig. 7).
The reaction mechanism of the MFE1 crotonase active site
The crystal structures of the MFE1 3S-hydroxy-acyl-CoA complexes provide the geometry of the 3-hydroxy moieties bound in its crotonase active site. The comparison of these complexed structures with the apo structure (Figs 5 and 6) suggests that the catalytic water molecule of the enzyme–substrate complex becomes the 3-hydroxy moiety of the hydratase–product complex, whereas it remains hydrogen-bonded to the OE1 atoms of Glu103 and Glu123 after the hydratase reaction. There appear to be no structural rearrangements between the active sites of the apo and liganded structures.
The hydrogen-bonding interactions of the carboxylate oxygen atoms of the catalytic glutamates and the catalytic water are schematically shown in Fig. 7, where the the complete catalytic cycle is depicted, as suggested by the crystallographic data. The first conversion of the catalytic cycle concerns the isomerization of the double bond of 3E-enoyl-CoA. It is assumed that, in the formation of the Michaelis complex of this reaction, the active site water is bound as seen in the apo structure and the fatty acid tail of the 3E-enoyl-CoA substrate is bound as seen in the structures of the product molecules. This assumption is supported by the notion that the proposed position of the catalytic water coincides with the position of the active site water in the complex of the monofunctional hydratase with hexadienoyl-CoA (Fig. 6)  and 4-dimethylamino cinnamoyl-CoA .
For the isomerase reaction (Fig. 7), it is proposed that, in the competent complex, Glu103 is likely protonated and Glu123 is deprotonated, as deduced from the hydrogen-bonding networks. OE1(Glu123) abstracts a proton from C2 of the substrate (3E-enoyl-CoA). This reaction is facilitated by the oxyanion hole for the thioester oxygen, stabilizing the thioester enolate intermediate. The interactions of the thioester oxygen in this oxyanion hole are visualized in Fig. 8. The geometry of the active site (Figs 5 and 6) suggests that, subsequently, this intermediate is reprotonated at C4 by a proton from the catalytic water, which by itself is reprotonated by Glu103. The 2E-enoyl-CoA intermediate then becomes the substrate of the hydratase reaction. This proposal implies differences in the isomerase reaction mechanisms of the monofunctional enoyl-CoA isomerase and MFE1. For the catalytic cycle of this enoyl-CoA isomerase, it appears likely that the active site glutamate (Glu136 of ECI1; Fig. 6) abstracts a proton from C2 and subsequently shuttles it to the C4 carbon to complete the reaction cycle [20, 32]. Apparently, in the enoyl-CoA isomerase, there is no water molecule involved in the proton transfer reaction mechanism.
The reaction mechanisms of the MFE1 hydratase and the monofunctional hydratase  appear to be very similar. The key active site geometry, shaped by the two catalytic glutamates, the catalytic water and the oxyanion hole, is preserved. The proposed protonation state (Glu123 protonated, Glu103 deprotonated) for the MFE1 hydratase reaction is also consistent with reaction mechanistic studies of the monofunctional enoyl-CoA hydratase [30, 31, 41, 43] and the enzyme kinetic studies of the α-chain of the TFE-enzyme , suggesting that at least one of the catalytic glutamates should be deprotonated in the active site competent for hydratase catalysis. As in the monofunctional enoyl-CoA hydratase, the catalytic water is tightly anchored between the OE1-carboxylate atoms of Glu103 and Glu123, which act jointly to achieve catalysis. In apo MFE1, this water is 2.9 Å from OE1(Glu103) and 2.5 Å from OE1(Glu123). This water position is near the C2, C3 and C4 atoms of the hydrated product, at distances of 2.1, 1.1 and 1.9 Å, respectively, when comparing the MFE1 apo structure and the MFE1-3S-hydroxy-decanoyl-CoA complex structure (Figs 5 and 6). In this comparison, it is found that there is no other protic residue nearby the substrate C2 atom, except for OE1(Glu123) (at 2.8 Å). From this superposition, it can also be noted that the OE1(Glu103) atom, at 4.6 Å, is the closest protein protic side chain atom near the C4 atom. In the proposed MFE1 hydratase reaction mechanism, the deprotonated Glu103 activates the catalytic water for reaction with the C3 carbon. This reaction is facilitated by the oxyanion hole for the thioester oxygen atom (Fig. 7) because it favourably interacts with the thioester enolate intermediate. The enolate is reprotonated by Glu123 at the C2 atom. The 3-hydroxy moiety of the product is hydrogen-bonded to OE1(Glu103) (at 2.7 Å) and to OE1(Glu123) (at 2.9 Å) (Fig. 8). Now Glu123 is deprotonated and Glu103 is protonated and, once the product is released and a new water and a new substrate bind, the active site becomes ready for the next round of the catalytic cycle (Fig. 7).
It is well known that the hydratase reaction of the monofunctional hydratase proceeds according to the syn-addition of a water molecule to the substrate 2E double bond [31, 40, 46]. In this mechanism, the hydroxyl group and the hydrogen atom are added both to the same side of the double bond . The three-dimensional arrangement of the catalytic water moiety and the OE1(Glu123) atom with respect to the thioester enolate moiety, as depicted in Fig. 8, is as expected for the syn-addition mechanism. The perpendicular arrangement of the catalytic water and the OE1(Glu103) with respect to the thioester enolate moiety (Fig. 8) is also in agreement with the stereoelectronic rules [33, 47] for this chemistry.
The crotonase active site of MFE1 catalyses both the enoyl-CoA isomerization and enoyl-CoA hydration reactions. The comparison of the structures of the MFE1 apo and the MFE1-product complexes highlights the rigidity of the two catalytic glutamates, Glu103 and Glu123, and the oxyanion hole geometry of the catalytic site. The crotonase active site of MFE1 is rather similar to the active sites of the monofunctional enoyl-CoA hydratases. However, the corresponding active sites of the monofunctional enoyl-CoA isomerases are more different because they have no catalytic water and only one catalytic glutamate (Fig. 2). Therefore, it appears that the enoyl-CoA isomerase reaction mechanism of MFE1 is distinctly different from the reaction mechanism of the monofunctional enoyl-CoA isomerases. It is proposed that Glu123 is the catalytic base in the MFE1 isomerase reaction. The catalytic water anchored between the OE1-atoms of the two catalytic glutamates plays a key role in both the enoyl-CoA isomerization and enoyl-CoA hydration reactions.
Materials and methods
Tiglyl-CoA (Fig. 1) and CoA were purchased from Sigma (St Louis, MO, USA). The quality of the tiglyl-CoA sample was checked by MS. 2E-hexenoyl-CoA, 2E-decenoyl-CoA and 3E-decenoyl-CoA were synthesized as described previously .
The expression and purification protocols of recombinant rat peroxisomal MFE1 were the same as reported previously . The hydratase activity of the enzyme was measured using 2E-decenoyl-CoA as substrate [8, 24]. For the isomerase assay, 3E-decenoyl-CoA was used as substrate.
The E123A mutation was made using the overlap extension PCR method  and confirmed by sequencing. The E123A variant was expressed and purified using the same procedure as that for the wild-type.
The wild-type MFE1 crystals were grown in the presence of CoA as described previously . The E123A variant crystals were obtained by the same method as that employed for the wild-type but, during co-crystallization, 2E-hexenoyl-CoA was used instead of CoA (Table 1). To generate apo crystals, the wild-type MFE1 crystals were soaked overnight in a mother liquor lacking CoA to remove the bound CoA from the active site of the crotonase part. Subsequently, these apo crystals were transferred into a mother liquor containing 0.2 mm 2E-hexenoyl-CoA, 0.2 mm 2E-decenoyl-CoA or 0.2 mm tiglyl-CoA, respectively, and kept for overnight soaking. The precise crystal growth conditions and the respective soaking protocols are summarized in Table 1. For each of the datasets, 15% glycerol was used as a cryoprotectant during crystal freezing (Table 1).
The data were collected at the European Synchrotron Radiation Facility (ESRF, Grenoble, France). Data processing was performed using xds , imosflm  and scala  (Table 2). Other data handling software of ccp4 were also used . Initial phases for each of the datasets were obtained by a rigid body refinement using the previously solved CoA-liganded MFE1 structure (2X58; 2.8 Å resolution) , stripped of all its waters and bound ligand molecules, as a starting model. The structures were refined with refmac5  using NCS restraints (excluding loop-2). The crystallographic data show that, in the three substrate binding studies with wild-type MFE1 crystals, the three product molecules (3S-hydroxy-hexanoyl-CoA, 3S-hydroxy-decanoyl-CoA and 2S-methyl-3S-hydroxy-butanoyl-CoA, respectively) are bound at the crotonase active site of each of the two molecules of the asymmetric unit. Model building and structure analysis were performed using coot . The structures were also analyzed with Molprobity . The refinement statistics are summarized in Table 2.
For comparisons, all superpositions were performed using the SSM  or LSQ options of coot . The structures of the complexes that were used for structural comparisons are the CoA-liganded MFE1 reference structure (2X58; 2.8 Å resolution), liganded human mitochondrial ECI1 enoyl-CoA isomerase (1SG4; 1.3 Å resolution), liganded rat mitochondrial enoyl-CoA hydratase (1EY3, 2.3 Å; 1 MJ3, 2.1 Å; 1DUB, 2.5 Å; 2DUB, 2.4 Å resolution) and the crotonase domain of the bacterial TFE α2β2-complex (1WDK; 2.5 Å resolution). For the structure-based sequence alignments, the structures of yeast ECI2 (1PJH; 2.1 Å resolution) and human ECI2 (2F6Q; 1.95 Å resolution) were used. Images were constructed using pymol .
This work was supported by the Academy of Finland (grant number 122921). We are very grateful to the beamline scientists of ESRF beamlines ID14-1 and ID29 for their expert help with the data collection. We thank Ulrich Bergmann for carefully performing the MS experiments; Rajaram Venkatesan and Kristian Koski for help with the initial protein crystallographic data collection and computing efforts; and Tiila Kiema and Petri Pihko for valuable discussions.