Rational engineering of a native hyperthermostable lactonase into a broad spectrum phosphotriesterase

The redesign of enzyme active sites to alter their function or specificity is a difficult yet appealing challenge. Here we used a structure-based design approach to engineer the lactonase SsoPox from Sulfolobus solfataricus into a phosphotriesterase. The five best variants were characterized and their structure was solved. The most active variant, αsD6 (V27A-Y97W-L228M-W263M) demonstrates a large increase in catalytic efficiencies over the wild-type enzyme, with increases of 2,210-fold, 163-fold, 58-fold, 16-fold against methyl-parathion, malathion, ethyl-paraoxon, and methyl-paraoxon, respectively. Interestingly, the best mutants are also capable of degrading fensulfothion, which is reported to be an inhibitor for the wild-type enzyme, as well as others that are not substrates of the starting template or previously reported W263 mutants. The broad specificity of these engineered variants makes them promising candidates for the bioremediation of organophosphorus compounds. Analysis of their structures reveals that the increase in activity mainly occurs through the destabilization of the active site loop involved in substrate binding, and it has been observed that the level of disorder correlates with the width of the enzyme specificity spectrum. This finding supports the idea that active site conformational flexibility is essential to the acquisition of broader substrate specificity.

PLL representatives are extremely thermostable [30][31][32][33][34][35][36][37][38] and are particularly promising candidates for biotechnological use 39,40 by virtue of their high stability and resistance towards denaturing agents 41 . PLLs, however, exhibit much lower phosphotriesterase activity than PTEs. Engineering experiments with the aim of increasing PLLs' phosphotriesterase activity were performed. In particular, because PTEs and PLLs' active sites mainly differ in their loop length (7 and 8), sequences and conformations 16,26,42 , loop grafting experiments were performed and proved to be difficult. This approach resulted in aggregation-prone or inactive enzymes 16,34 , but was successful when using a stepwise approach 43 . Recently, more classical engineering techniques were successful in increasing the phosphotriesterase activity of the PLL DrOPH 44 .
In this study, we worked on the PLL SsoPox from the hyperthermophilic archaea Sulfolobus solfataricus 32,45 . SsoPox exhibits high acyl-homoserine lactones and oxo-lactones hydrolysis abilities 11 , and promiscuous, low phosphotriesterase activity 46 . Previous work showed that engineering can successfully improve SsoPox catalytic efficiency against phosphotriesters such as ethyl-paraoxon 11,47 . These previous studies highlighted the role of a key active site loop 8 residue, W263, which modulates the active site loop conformational space 11 . In this study, we aimed to further improve the phosphotriesterase activity of SsoPox, including against phosphothionoesters. SsoPox is a very appealing candidate for the bioremediation of phosphotriesters because of its unique thermal stability, and its ability to resist aging, solvent and protease treatments 48 . On the other hand, SsoPox is a poor phosphothionoesterase, exhibiting a strong preference (100-fold) for methyl-paraoxon over methyl-parathion, a feature referred to as the thiono-effect 46 . We used a rational strategy: taking advantage of the structural similarity between SsoPox and the bacterial PTE from the mesophilic Brevundimonas diminuta (BdPTE) 27 , we designed and produced a structure-based combinatorial library, and screened this library for improved phosphotriesterase activity with the aim of obtaining heat stable, highly active variants with broad specificity. We obtained several improved variants, including against methyl-parathion, with a 2,210-fold improvement in activity. All these variants but variant αsB5 contained a mutation at position 263. Interestingly, the addition of the W263 mutation using site-saturation mutagenesis 12 on an αsB5 background revealed an incompatibility of these mutations, instead of the expected increase in phosphotrieseterase activity. Of these improved variants, we obtained a mutant that lost the thiono-effect against methyl-parathion, exhibiting a >2,000-fold (k cat /K M ~10 4 M −1 .s −1 ) as referred to the wild-type enzyme, and a methyl-paraoxon/methyl-parathion activity ratio of ~1. The X-ray structures of several improved variants reveal that the selected mutations increase the active site loop conformational flexibility and reshape the active site. The best variants were extensively characterized against eight commercially available insecticides, along with previously reported SsoPox monovariants harboring mutation at residue 263 and their specificity spectrum was determined.

Results and Discussion
The lactonase SsoPox was engineered for higher phosphotriesterase activity using structure-based combinatorial libraries. By comparing structures of enzymes with similar topology, it has been possible to redesign, using modelling tools, the active site cavity of SsoPox to mimic as closely as possible that of BdPTE ( Fig. 1; Table S1). This method consisted of two main steps: (i) the identification of mutations corresponding to equivalent positions at the enzyme active sites using structural alignment and (ii) the rational selection of mutations, in a position with no equivalent residue, which mimicked the shape and chemical nature of the target enzyme cavity. A mutation dataset was thus obtained and used to develop a combinatorial library consisting of random combinations of our pre-selected mutations. Structural alignments of SsoPox (red) and BdPTE structures (yellow) guided the mutational design. (C) Active site superposition of BdPTE (green) and SsoPox (cyan) active sites. Residues with structural equivalent (orange sticks) in both structures were used to design mutations of SsoPox's residues into the BdPTE corresponding residues. Positions with no structural equivalent (red sticks) were designed using modelling tools (purple sticks) to mimic the active site cavity size, shape and chemical properties of the BdPTE crystal structure model ( Figure S1). (D) The mutations data base consists of mutations to the BdPTE sequence when there is a structural equivalence (orange), and of mutations designed when there is no structural equivalence (red). The third set of mutations (black) adds more diversity at two selected positions, 258 and 263. (E) Once the library had been validated, primers carrying mutations were used to shuffle them and generate a gene library with random combinations of selected mutations. (F) The library was screened for paraoxonase activity to identify enzymes with improved proficiency.
Scientific REPORTS | 7: 16745 | DOI:10.1038/s41598-017-16841-0 Structure-based design produced mutants with improved phosphotriesterase activity. The structure-based design considerably reshaped the active site cavity of SsoPox, theoretically bringing it closer to that of BdPTE, both in shape and nature ( Figure S1). A total of 14 key positions within the active site were identified, and were mutated to specific residues, or degenerated into several possible amino acids (see Methods). The paraoxonase activity of 184 randomly selected clones was screened and the 14 best were sequenced. Most of these 14 were found to possess the same mutations (e.g. W263F/L/M/A; 73% of cases) and ( Figure S2A; Table S3). Of them, the most active variants, αsA1 (C258L-I261F-W263A), αsA6 (F46L-C258A-W263M-I280T), αsB5 (V27A-I76T-Y97W-Y99F-L130P-L226V), αsC6 (L72I-Y99F-I122L-L228M-F229S-W263L), and αsD6 (V27A-Y97W-L228M-W263M) were subject to kinetic and structural characterization. It should be noted that αsA1 was obtained in a previous engineering effort 47 , and that four of the five selected variants contained a substitution of the key residue W263. We previously highlighted the role of this residue in substrate binding and modulation of the active site loop 8 conformation, resulting in an increase in activity against promiscuous substrates 11 . Variant αsB5 does not harbor any substitution at position W263. We therefore decided to test the W263 substitution on an αsB5 background, using saturation mutagenesis strategy. The five variants (αsA1, αsA6, αsB5, αsC6 & αsD6) were characterized against several phosphotriesters, including ethyl-paraoxon (I), ethyl-parathion (II), methyl-paraoxon (III), methyl-parathion (IV) and malathion (V) (Fig. 2). Overall, all variant exhibits improved catalytic efficiencies against all tested substrates, with the exception of αsA1 with malathion (~2-fold decrease) ( Fig. 3A; Table 2). The best improved variant, αsD6, exhibits a 2,210-fold increase in catalytic efficiency against methyl-parathion ( Table 2). The largest improvements were observed for ethyl/methyl-parathion, two bad substrates for wild-type SsoPox. Interestingly, while the wt enzyme shows a clear preference for small substituents 46 , most selected variants lost this preference, possibly indicating an enlargement of the active site cavity. These improvements were in the range of what was previously obtained but with higher intensive mutation protocols and for only one OP substrate 49 . Finally, it was noted that two variants (αsA6 & αsC6) presented a substrate inhibition for some thiono-phosphotriesters.
The five selected variants were further evaluated for their lactonase activity (Table 2). Interestingly, variants αsA6 and αsC6 exhibit enhanced lactonase activity against γand δ-lactones, while αsA1, αsB5 and αsD6 show reduced lactonase catalytic efficiencies. This emphasizes the fact that the improvement of the phosphotriesterase activity does not necessarily compromise the cognate, lactonase activity of the enzyme. These mutations therefore dramatically increased a new activity (phosphotriesterase) without the complete loss of the native/original function (lactonase) 4 . These mutants are different from those generated on another PLLs, through loop insertion. While the stepwise insertion of residues in loop 7 can significantly increase the ability of PLLs to degrade phosphotriesters, it also drastically decreases their lactonase activity 43 . Insertion in loop 7 has been previously described as a key evolutionary event in the transition from lactonase to phosphotriesterase 50 . The crystal structures of variants αsA1, αsA6, αsB5, αsC6 and αsD6 were solved (Table 1). While most structures were solved at high or medium resolutions (1.4-2.55 Å), mutant asD6 could only be solved at low resolution (2.95 Å). Given this resolution, we limited our interpretations and retained this data for the consistency of the study as it exhibits structural features (loop 8 disorder) that are consistent with all the other mutants. Overall, the mutants' structures are similar to that of wt-SsoPox. However, the active site cavities are larger for the mutants: most selected mutations replace residues with smaller ones: W263 is mutated into M/L/I, V27 into A, I76 into T, L130 into P, Y99 into F. Only two selected substitutions relate to bulkier side chains: Y97W and L228M. This enlargement is difficult to quantify even with the structure of these mutants because of the extremely mobile nature of loop 8 in the variants. Enlarging the active site was a main objective of the active site redesign of SsoPox, to enable the bulkier phosphotriesters to bind within the natural lactonase enzyme. Additionally, the loop 8 conformations differ in mutants, as compared to the wt-enzyme. Because part of loop 8, including position 263, is located at the enzyme dimer interface, altered loop 8 conformations modulate the relative orientation of both monomers. In the case of the mutants characterized in this study, the dimer reorientation yields significant displacement up to 5.2 Å (e.g., αsA1, between equivalent carbon α positions), as compared to wt-SsoPox ( Figure S5). Similar reorientations were previously observed upon substrate binding or mutations of W263 11,42 .
Although the active sites of the improved variant structures superpose well onto the wt-structure, all selected mutants display altered loop 8 conformations. Changes are subtle for some of the variants (e.g. αsA1), but are large for others (e.g. αsC6) (Fig. 4). Notably, mutant αsA6 was crystallized in two different conformations: one in which loop 8 adopts a wt-like conformation, referred to as closed conformation (CC), and another one where loop 8 is unfolded, referred to as open conformation (OC) ( Figure S6). In the two other mutants, αsD6 and αsB5, loop 8 could not be modeled due to the lack of electronic density, which was likely to be due to the high level of motion of this enzyme region. Analysis of the normalized thermal motion B-factor supports this hypothesis. It also confirms that loop 8 is highly mobile in all selected mutants, as compared to wt-SsoPox (Fig. 5). The higher mobility of loop 8 may also partly explain the improved ability of the variants to hydrolyze phosphotriesters with large substituents. Improved mutant αsD6 lost the thiono-effect. SsoPox presents a marked preference for oxono-OP substrates as compared to thiono-ones (>100-fold in catalytic efficiency) 46 . Conversely, PTEs do not exhibit such a drastic preference, paraoxon being a slightly better substrate than parathion for the enzyme 27,51 . Interestingly,  1.00 ± 0.02 × 10 −1 1.13 ± 0.10 × 10 2 -9.08 ± 2.00 × 10 2 163 Continued some selected variants retained a strong preference for oxono-phosphotriesters (~10-fold), whereas others such as αsD6 lost the thiono-effect ( Figure S7). The wild-type enzyme exhibits a low K M value with methyl-parathion (121 µM), but also a very low k cat value (1.10 × 10 −3 s −1 ) ( between the thiono moiety and the metal cations (low K M value), which would subsequently impair catalysis (low k cat values) 52 . Interestingly, the improved variants do not exhibit significant difference in K M values for thiono-phosphotriesters, but show a dramatic increase in k cat values, as compared to wild-type enzymes (e.g. k cat value of 6.89 s −1 and 1.10 × 10 −3 s −1 , as compared to methyl-parathion for the αsD6 variant and wild-type enzymes, respectively). The two selected variants, namely αsB5 and αsD6, which lost the thiono-effect have only two mutations in common: V27A and Y97W. Both residues are located relatively close to the bi-metallic active site, yet do not interact directly with the metals (V27-Fe: 6.7 Å, Y97-Co: 4.3 Å; Fig. 6). Because other PLLs have been shown to exhibit high charge coupling between the β metal and the conserved tyrosine residue 33,53 , Y97 is a likely candidate for the thiono-effect, and will be examined in future studies.
Mutations of W263 and αsB5 are incompatible. αsB5 (V27A-I76T-Y97W-Y99F-L130P-L226V) is the only selected mutant that does not contain a substitution of position W263. Because W263 substitutions were previously shown to be key for increasing SsoPox's phosphotriesterase activity, we used saturation mutagenesis to introduce W263 mutations to the background of αsB5 12 . The best selected mutants against paraoxon ( Figure S2C) were found to be αsB5-W263I/L/M. Surprisingly, none of these mutants demonstrated an increase in phosphotriesterase activity in comparison with αsB5, but rather a decrease in activity ( Table 2). This is intriguing because, taken individually, these mutations have been shown to greatly improve phosphotriesterase activity. For example, with ethyl-paraoxon as a substrate, W263M and αsB5 produce a catalytic efficiency improvement of 14 and 67-fold, respectively, while the combination αsB5-W263M increases activity only 11-fold ( Table 2). This example of negative epistasis could be due to the mode of action of these mutations. Structures reveal that αsB5 mutations have a destabilizing effect on loop 8, as illustrated by the B factor analysis. The same was observed for W263M in a previous study 11 . Combining these two destabilizing sets of mutations may have harmed the active site integrity, as previously described with another enzyme as "conformational active site disorder" 54 , and in particular the necessary alignment and pre-organization of the catalytic residues. The destabilizing effect of these mutations can also be observed on the overall enzyme stability, with αsB5, W263M, αsB5-W263M exhibiting T m values of 70.4 °C, 85.3 °C and 69.1 °C, respectively, as compared to 106 °C for the wt enzyme (Table S4). Here, the effects of the combined 5 mutations harbored by variant αsB5 together with the previously identified positive substitutions at residue W263 were not additive underlining that single beneficial mutations do not necessarily induce positive epistatic effects. Synergistic effects of mutations are often complex to predict and most engineering strategies usually lead to an optimization plateau difficult to overcome 55 .

Improvement of engineered variants over W263 mutation.
In a previous report, we noted that mutation of W263 into all the other 19 residues lead to an increase of the promiscuous paraoxonase activity 11 . Here we note that most of the improved variants (i.e. αsA1, αsA6, αsB5, αsC6 & αsD6) (i) exhibit higher paraoxonase activity than W263L/M, and (ii) foremost are capable of hydrolyzing phosphotriesters that are not substrates for the wild-type enzyme or W263 mutants, such as ethyl-parathion, chlorpyrifos or fensulfothion ( Table 2). The activity spectrum of these engineered variants against the tested phosphotriesters is therefore much wider than the wild-type enzyme and W263 mutants.

Relation between active site loop disorder and broad enzymatic specificity. The organophospho-
rus compound bioremediation potential of several mutants was investigated. In particular, the best variants, αsD6 and αsB5, alongside other improved variants αsB5-W263I/L/M, W263L/M and the wt-enzyme, were characterized for their phosphotriesterase activity with ethyl-paraoxon (I), ethyl-parathion (II), malathion (V), chlorpyrifos (VI), diazinon (VII), fenitrothion (VIII), fensulfothion (IX) and coumaphos (X) (Fig. 2). Interestingly, we found that most mutants were able to hydrolyze fensulfothion, previously reported as an inhibitor of the wild-type enzyme 46 . Fensulfothion was indeed previously shown to bind head-to-tail into the wild-type enzyme active side 46 , a non-productive binding mode. Changes in the active site loop 8 conformational ensemble of engineered mutants might have allowed for the productive binding of fensulfothion. All mutants selected and used in this study are destabilized as compared to the wt-enzyme ( Fig. 7; Table S4). Structures reveal that these mutations, located on loop 8, contribute to increasing the mobility of the loop. Interestingly, the ability of the tested mutants to hydrolyze a wide range of phosphotriesters, including compounds that were inhibitors for the wt-enzyme, correlates with a lower T m (Fig. 7). This correlation suggest that the active site loop 8 degree of disorder may modulate the enzyme's ability to bind and hydrolyze a variety of phosphotriesters, a wider conformational ensemble of loop 8 being correlated with a broader enzyme specificity. This observation is consistent with previous studies highlighting the importance of the destabilization of active site loops to evolve new activities and new substrate specificity 54,56 . Such flexible active site loops involved in enzymatic specificity supports the notion of fold polarity: a portion of the active site (e.g. the loop) is weakly connected to the enzyme scaffold and thereby makes it possible for new functions to evolve with few changes 57 .

Mutant αsD6 is an active, broad spectrum phosphotriester biodecontaminant. Mutant αsD6
was further investigated for bioremediation considerations. Organophosphate-based pesticides are usually spread with concentrations in the millimolar range leading to micromolar-ranged contaminations. We thus investigated the potential of αsD6 for decontaminating pesticide solutions at 250 µM to simulate a contamination of groundwater or runoff waters. We investigated two S [E]/[ ] ratios, 10 −2 and 10 −3 respectively, for decontaminating OP solutions and determined the time required to hydrolyze 95% of pesticide preparations ( Fig. 3B; Figure S4). When considering the S [E]/[ ] ratio of 10 −3 , 95% degradation was achieved for four substrates (ethyl-paraoxon, diazinon, fenitrothion and coumaphos) within 10 minutes. Two substrates were decontaminated within an hour (ethyl-parathion and chlorpyrifos) and fensulfothion was degraded in three hours.
When increasing enzyme concentration 10-fold ( S [E]/[ ] ratio of 10 −2 ), five substrates (ethyl-paraoxon, ethyl-parathion, diazinon, fenitrothion and coumaphos) were decontaminated within two minutes, while fensulfothion needed ~20 minutes and the two other tested (chlorpyrifos and malathion) required 40 minutes. Given both its catalytic proficiency and high thermal stability (T m = 82.5 °C), αsD6 is a promising candidate for organophosphorus compound bioremediation and, in particular, the decontamination of water runoffs, soils, food products and materials. Moreover, SsoPox variants were previously shown to be compatible with immobilization steps including alginate beads and polyurethane-based coatings 48,58 , and could be of prime interest for the development of filtration devices for water treatment purposes.

barrel topology, SsoPox and
BdPTE structures (Pdb ID 2vc5 and 1dpm, respectively) were superimposed (Fig. 1) using PyMol 59 , making it possible to identify structurally equivalent residues at their respective active sites. In an effort to reshape the SsoPox active site, side chains that were well superimposed between the two enzymes were mutated using Coot into the residue present in BdPTE 60 , with the exception of V27 which was mutated in Ala (and not Gly) to avoid increased entropy ( Fig. 1; Table S1). However, due to the major differences in loop 8 and 7 between PLLs and PTEs, numerous residues were not superimposable. These active site residues (Y99, L228, F229 and W263) were therefore mutated in an effort to mimic the BdPTE active site cavity in terms of shape and chemical nature, as illustrated by the chimeric reconstruction of SsoPox carrying all these mutations ( Figure S1). Others mutations were also implemented in the dataset, e.g. C258A and C258L and different substitutions of the key position W263 (F/L/M/A) that have been shown to improve the phosphodiesterase activity 11 .
Synthesis of the combinatorial library. The SsoPox coding gene was amplified from the previously described pET22b-SsoPox plasmid 46 . For each of the 14 mutations, a mutagenesis primer of 30-33 bp (~10-15 bp from each mutation side) was synthesized (Table S2). For close mutations, primers can share several mutations and different primers were synthesized and mixed to keep an equivalent statistical probability for each of the 14 mutations. Typically, 2 pmol of primer mixture and 100 ng of DNaseI (TaKaRa)-generated fragments of the SsoPox gene were assembled as previously described 17 , followed by nested PCR with external cloning primers (SsoPox-lib-pET-5′/3′) ( Table S2). The PCR product was then cloned into a customized version of the pET-32b(-ΔTrx) plasmid and then electroporated into E. cloni cells (Lucigen, USA). After agar-plate growth, plasmid extraction was performed to create the plasmid bank. Variant genes were PCR amplified using T7-prom and   (Table S2) and sequenced. The combinatorial library shows an average of 5.3 ± 3.3 library mutations by sequencing 10 randomly picked colonies, and 0.43 ± 0.53 random mutations per gene.
Screening of the library. The plasmid library was used to transform the Escherichia coli strain BL21(DE 3 )-pGro7/GroEL (TaKaRa) to obtain colonies expressing a library of SsoPox variants. Randomly picked clones (184), representing a coverage of 3.8% of the library, were grown on a 96-well plate in 500 µL of ZYP medium as previously described 11 . Production of proteins and chaperones was induced after five hours of culture at 37 °C by reducing the temperature to 25 °C, adding CoCl 2 (0.2 mM) and adding arabinose (0.2%, w/v). After overnight growth, cell lysate was used to perform activity screening with 100 µM of paraoxon substrate ( Figure S2B) after partial purification of the protein with a heating step of 15 minutes incubation at 70 °C 35,49 . The screening was performed in 50 mM HEPES pH 8, 150 mM NaCl, 0.2 mM CoCl 2 . Kinetics of paraoxon/ methyl-parathion hydrolysis were monitored by following absorbance at 405 nm for 10 minutes using a microplate reader (Synergy HT, BioTek, USA) and the Gen5. Screening was performed as described above ( Figure S2C). Clones were produced in liquid ZYP medium and their activity was screened with 100 µM of paraoxon substrate. The plasmids corresponding to the most interesting variants were extracted and the SsoPox variant encoding genes were sequenced. Kinetic assays. Generalities. Catalytic parameters were evaluated in triplicate at 25 °C, and recorded using a microplate reader (Synergy HT, BioTek, USA) and the Gen5.1 software, in a 6.2 mm path length cell for 200 µL reaction in 96-well plate, as previously explained 30 . Catalytic parameters were obtained by fitting the data to the Michaelis-Menten (MM) equation using Graph-Pad Prism 6 software. When V max could not be reached in the experiments, catalytic efficiency was obtained by fitting the linear part of MM plot to a linear regression using Graph-Pad Prism 6 software. For some SsoPox variants, the MM plot was fitted to the substrate inhibition equation using Graph-Pad Prism 6 software enabling us to determine a K I for several substrates. Consequently, the calculated catalytic efficiencies in these conditions are true only at low substrate concentrations. In some other cases, saturation could not be reached, therefore k K / cat M values were determined using linear regression. For Coumaphos, data were fitted to one-phase decay non-linear regression.
Lactonase activity characterization. Lactonase kinetics were performed using a previously described protocol 30 . The time course hydrolysis of lactones were performed in lac buffer (2.5 mM Bicine pH 8.3, 150 mM NaCl, 0.2 mM CoCl 2 , 0.25 mM Cresol purple and 0.5% DMSO) over a concentration range 0-2 mM for 3-oxo-C10 AHLs (XI) or 0-5 mM for undecanoic-γ/δ -lactones (XII, XIII). Cresol purple (pK a 8.3 at 25 °C) is a pH indicator used to follow lactone ring hydrolysis by acidification of the medium. Molar coefficient extinction at 577 nm was evaluated recording absorbance of the buffer over an acetic acid range of concentration 0-0.35 mM.
Bioremediation of pesticide solutions at 250 µM. For all OP substrates but malathion, experiments were performed in triplicate at 25 °C using a microplate reader (Synergy HT, BioTek, USA). Wavelengths were chosen as described in the kinetic assays section. Degradation of malathion was followed by GC/MS in triplicate. OP concentration was fixed at 250 µM. All the substrates were soluble at this concentration. The best variants of SsoPox for each substrate were used, namely αsD6 for all the OPs. Two enzyme-to-substrate ratios S [E]/[ ] were used, 10 −2 and 10 −3 , corresponding to enzyme concentrations of 90 µg.ml −1 and 9 µg.ml −1 respectively. The reactions were monitored until the plateau was reached. Experimental measures were obtained using Gen5.1 software, then analyzed with GraphPad Prism 6 software. Curves were then fitted using One-Phase Decay non-linear regression with the equation (1): where Y0 = 0% and Plateau = 100%. The rate constant k was determined and the time required to observe a 95% decontamination was calculated accordingly. The curves and results of the fits are shown in Figure S4. Degradation at 95% was confirmed using a final point by GC/MS for all pesticides with a 10 −3 enzyme-to-substrate ratio. 100 µL of activity buffer solution was first extracted with 100 µL chloroform. Organic extracts were analyzed by using a Clarus 500 gas chromatograph equipped with a SQ8S MS detector (Perkin Elmer, Courtaboeuf, France). 1 µL of organic extract was volatilized at 220 °C (split 15 mL.min −1 ) in a deactivated FocusLiner with quartz wool (SGE, Ringwood, Australia) and compounds separated on an Elite-5MS column (30 m, 0.25 mm i.d., 0.25 mm film thickness) for 12 minutes using a temperature gradient (80-280 °C at 30 °C.min −1 , five minutes' hold). Helium flowing at 2 mL.min −1 was used as the carrier gas. The MS inlet line was set at 280 °C and the electron ionization source at 280 °C and 70 eV. Full scan monitoring was performed from 40 to 400 m/z in order to identify chemicals by spectral database search using MS Search 2.0 operated with the Standard Reference Database 1 A (National Institute of Standards and Technology, Gaithersburg, MD, USA). m/z is the mass-to-charge ratio of the base peak fragment detected for each molecule. In the case of weak pesticide signals, extracted ion chromatograms were generated with base peak ions to confirm the presence of chemicals (Paraoxon (I) and Parathion ethyl (II) m/z 109; Malathion (V) and Fenitrothion (VIII) m/z 125; Diazinon (VII) m/z 137; Coumaphos (X) m/z 97; Fensulfothion (IX) m/z 293; and Chlorpyrifos (VI) m/z 197). Selected Ion Recording using base peaks ions was applied in order to specifically monitor pesticides and collect peak areas for kinetics. All samples were analyzed over short periods of time to avoid signal drift. All data were processed using Turbomass 6.1 (Perkin Elmer).
Crystallization. Crystallization assays were performed as previously described 42,46 . Crystallization was performed using the hanging drop vapor diffusion method in 96-well plates (Greiner Microplate, 96 well, PS, F-bottom) on ViewDrop II seals (TPP Labtech). Equal volumes (0.5 µL) of protein and reservoir solutions were mixed, and the resulting drops were equilibrated against a 150 µL reservoir solution containing 20-30% (w/v) PEG 8000 and 50 mM Tris-HCl buffer (pH 8). Crystals appeared after few days at 4 °C.
Data collection and structure refinement. Crystals were first transferred to a cryoprotectant solution consisting of the reservoir solution and 20% (v/v) glycerol. Crystals were then flash-cooled in liquid nitrogen. X-ray diffraction data were collected at 100 K using synchrotron radiation at the ID23-1 beam line (ESRF, Grenoble, France) and using an ADSC Q315r detector. X-ray diffraction data were integrated and scaled with the XDS package (Table 1) 61 . The phases were obtained using the native structure of SsoPox (PDB code 2vc5) as a starting model, performing a molecular replacement with MOLREP or PHASER 62,63 . The models were built with Coot and refined using REFMAC 60,64 . We note that three structures presented in this work exhibit one disordered monomer (and low corresponding electronic density): αsB5 (monomer D, total of 4 monomers), αsD6 (monomer C, out of a total of 4 monomers) and αsC6 (monomer C, total of 4 monomers). Structure illustrations were performed using PyMOL 48 .
Relative B-factor analysis. The occupancies of all residues in all tested structures were set to 1 for this analysis. For residues with alternate conformations, the sums of occupancies were set to 1. Structures were re-refined with REFMAC 60,64 . The relative B-factor values were obtained by normalizing the B-factor values of each residue by the average B-factor of the whole structure as previously described 11,54,57 .