Open Access

Eva Daniëlle Schuiten

• Enzymes are well-known for being remarkably selective catalysts. They are often able to catalyse reactions for certain molecules while leaving other similar molecules completely unchanged. Nevertheless, many enzymes are capable of catalysing other reactions and/or transforming other substrates than their physiologically relevant activities. This phenomenon is referred to as enzyme promiscuity and it is thought to play an important role in the emergence of novel functions by providing a starting point for divergent evolution towards different enzymatic activities. It is important for enzymes to be selective to avoid harmful side-products and increase reaction efficiency, but often catalysts are not optimised beyond what is required for their function. Life profits from the cross-reactivity and enzyme promiscuity through accidental discovery of new helpful molecules and pathways, while using regulation to quickly adapt to changing circumstances. Enzymes are grouped together with other similar proteins into structural families and superfamilies. Members of a structural family share significant structural elements and often have similar catalytic mechanisms. However, they often catalyse very different chemical reactions and accept a variety of different substrates. Promiscuous activities are common within superfamilies, where the primary function of one family member is often found as promiscuous activity in other family members. Together with the structural similarities, this prevalent cross-reactivity suggests a common evolutionary origin. One of the largest structural superfamilies is the α/β-hydrolase-fold family. Despite sharing a highly conserved core structure, this superfamily is catalytically diverse and spans several distinct enzyme classes including hydrolases, acyltransferases, oxidoreductases, lyases, and isomerases. Epoxide hydrolases and dehalogenases of the α/β-hydrolase-fold family even share the same Asp/Glu-His-Asp catalytic triad and form similar covalent alkyl-enzyme reaction intermediates, yet they are known for attacking either epoxides or C-X bonds with perfect chemoselectivity. Although promiscuity is often observed within the α/β-hydrolase fold family and despite their mechanistic similarities, no α/β-hydrolases were known that exhibit both epoxide hydrolase and dehalogenase activity simultaneously. The versatility of the catalytic triads used by α/β-hydrolases makes these enzymes attractive targets for the conversion of catalytic activity through protein engineering. Several attempts were made to introduce dehalogenase activity in an epoxide hydrolase, and after several rounds of designing and screening different variants of the epoxide hydrolase PaeCIF from Pseudomonas aeruginosa, minor dehalogenase activity was detected for some of the variants. However, despite promising first results it proved extremely difficult to reliably reproduce the results, primarily due to expression problems and low sensitivity of the halide detection assays that were available at the time. Since the conversion proved to be more difficult than expected (unpublished data), it was decided to investigate other potential protein scaffolds. Considering the prevalence of catalytic promiscuity among members of the α/β-hydrolase-fold superfamily, and the close relationship and catalytic similarities between epoxide hydrolases and dehalogenases, it seemed odd that no enzyme is known to have both epoxide hydrolase and dehalogenase activity. We argued that it is highly probable that a promiscuous epoxide hydrolase-dehalogenase enzyme exists, but it simply has not been found yet due to the absence of sensitive high-throughput halide assays and not screening the right set of enzymes. Although several established assays were available for the determination of dehalogenase activity, these assays suffer major drawbacks. For example, one of the most popular assays, the Iwasaki assay, is not very sensitive and uses extremely toxic chemicals, while pH assays like the phenol red assay are inherently unreliable and insensitive due to the low buffer concentrations employed107,114. Thus, a new assay for the screening of dehalogenase activity through the selective detection of halides was developed115. The halide oxidation assay provides a safer, more reliable, and most importantly, much more sensitive method to detect dehalogenase activity. Using molecular phylogenetics, we studied the evolutionary relationship between epoxide hydrolases and dehalogenases to identify interesting extant epoxide hydrolases. Molecular phylogenetics uses a multiple sequence alignment of the amino acid or nucleotide sequences of extant enzymes to construct a phylogenetic tree. At first, we tried using a large dataset with almost 3,500 putative epoxide hydrolase and dehalogenase sequences, but we quickly realised the resulting phylogenetic tree was impractical. Most of the sequences in this large dataset were not characterised experimentally but annotated automatically based on their sequence similarity to a rather limited number of characterised sequences. Although automated annotations can be used as predictions for catalytic activity, they are often wrong. As we were particularly interested in the interface of both epoxide hydrolase and dehalogenase activities, we needed more certainty and a change in direction was necessary. Instead of trying to filter the α/β-hydrolase fold database, experimentally characterised sequences were collected through literature research. This smaller dataset consisting of characterised sequences resulted in a phylogenetic tree containing 45 epoxide hydrolases, 30 haloalkane dehalogenases and 7 haloacetate dehalogenases from a variety of different organisms. Ancestral sequence reconstruction was attempted for several interesting nodes in this phylogenetic tree. By combining the multiple sequence alignment, the evolutionary relationships from the phylogenetic tree, and evolutionary models, a hypothetical sequence of the theoretical ancestor can be determined. Unfortunately, it was difficult to get good soluble protein expression with the ancestral sequences and despite our best efforts it was not possible to obtain reliable and reproducible screening results. Instead of trying to improve protein expression and purification protocols for the ancestral sequences, we decided to focus on screening extant sequences with the newly developed halide oxidation assay to find a promiscuous epoxide hydrolase-dehalogenase. In addition to reconstructing ancestral sequences, eight extant epoxide hydrolases could be selected for screening towards dehalogenase activity and as promising potential engineering scaffolds from this phylogenetic tree. The eight selected epoxide hydrolases were screened for dehalogenase activity with several haloalkane substrates and the epoxide hydrolase CorEH from Corynebacterium sp. C12 was found to exhibit promiscuous dehalogenase activity. Interestingly, the measured concentrations of bromide for the initial hit with CorEH were only 150-250 nM, well below the lowest detection limit of 20 µM achievable in microtiter plate format with the Iwasaki assay. This means that the dehalogenase activity of CorEH would probably not have been detected were it not for the development of the sensitive halide oxidation assay. CorEH is an epoxide hydrolase that can also catalyse the dehalogenation of haloalkanes, particularly bromoalkanes such as 1-bromobutane and 1-bromohexane. The dehalogenase activity of wild-type CorEH with 1-bromobutane (0.25 nmol·min-1·mg-1) is about 4,000-fold lower than the average activity of several natural dehalogenases with two halide-stabilising residues (1 μmol·min-1·mg-1) and approximately 400-fold lower compared to the dehalogenases with a single halide-stabilising residue. The crystal structure of CorEH was determined to 2.2 Å. Our structure-function studies suggest that the dehalogenase activity of CorEH probably stems from the presence of at least one halide-stabilising residue. Unfortunately, this could not be confirmed experimentally via mutagenesis as the W100A variant lost both the dehalogenase and epoxide hydrolase activity in equal measure, making it difficult to demonstrate that W100 is involved in halide stabilisation. The loss of both activities for variant W100A can possibly be explained by the secondary function of the tryptophan; removal of W100 might lead to the incorrect positioning of the catalytic nucleophile for the nucleophilic attack involved in both epoxide hydrolysis and dehalogenation. Nevertheless, computational modelling of Michaelis-Menten complexes, utilising the crystal structure of CorEH, supports the hypothesis that the tryptophan W100 is involved in halide stabilisation in CorEH. Based on docking studies, the epoxide ring-opening tyrosine is also close enough to form hydrogen bonds to stabilise the substrate. However, it is also possible that like several characterised haloalkane dehalogenases, CorEH only uses a single residue to stabilise the halide. Removal of the tryptophan at the primary halide-stabilising position resulted in the loss of both activities, likely due to the loss of its secondary function to properly position the catalytic nucleophile. Substitution of the uncommon tryptophan in the HGxP-motif with phenylalanine does not completely remove the dehalogenase activity. Nevertheless, it causes a significant drop in both haloalkane dehalogenase and epoxide hydrolase activities, indicating that this residue is important for catalysis or the structural integrity of CorEH. Enzyme promiscuity plays an important role in enzyme evolution and the diversification of enzymes. Several researchers have attempted to interconvert epoxide hydrolase and dehalogenase activity, or to find an enzyme with both activities, without success. It would be hard to maintain the view that promiscuity is a fundamental property crucial to enzyme evolution if we could not observe promiscuity between two enzyme classes with such similar reaction mechanisms. Our findings show that dual epoxide hydrolase and dehalogenase activity can occur in one natural protein scaffold. We believe that we succeeded because we used a phylogenetic analysis of characterised sequences to select the right subset of epoxide hydrolases to investigate and due to the much more sensitive halide assays not available to those before us. The versatility of the catalytic triad in α/β-hydrolases combined with the variety of possible supporting residues found in both epoxide hydrolases and dehalogenases shows that catalytic mechanisms can be flexible. This flexibility allows space for diversification of catalytic residues without loss of function, giving rise to novel (promiscuous) functions and new cross-reactivities.