Structural and Enzymatic characterization of the lactonase SisLac from Sulfolobus islandicus

Background A new member of the Phosphotriesterase-Like Lactonases (PLL) family from the hyperthermophilic archeon Sulfolobus islandicus (SisLac) has been characterized. SisLac is a native lactonase that exhibits a high promiscuous phosphotriesterase activity. SisLac thus represents a promising target for engineering studies, exhibiting both detoxification and bacterial quorum quenching abilities, including human pathogens such as Pseudomonas aeruginosa. Methodology/Principal Findings Here, we describe the substrate specificity of SisLac, providing extensive kinetic studies performed with various phosphotriesters, esters, N-acyl-homoserine lactones (AHLs) and other lactones as substrates. Moreover, we solved the X-ray structure of SisLac and structural comparisons with the closely related SsoPox structure highlighted differences in the surface salt bridge network and the dimerization interface. SisLac and SsoPox being close homologues (91% sequence identity), we undertook a mutational study to decipher these structural differences and their putative consequences on the stability and the catalytic properties of these proteins. Conclusions/Significance We show that SisLac is a very proficient lactonase against aroma lactones and AHLs as substrates. Hence, data herein emphasize the potential role of SisLac as quorum quenching agent in Sulfolobus. Moreover, despite the very high sequence homology with SsoPox, we highlight key epistatic substitutions that influence the enzyme stability and activity.


Introduction
SisLac (also known as SisPox [1]) is an enzyme isolated from the archaeon organism Sulfolobus islandicus, which is found in extreme environments like the Yellowstone natural park (U.S.A.) or the Mutnovksy volcano in Kamchatka (Russia) [2]. SisLac belongs to an enzyme family called Phosphotriesterase-Like Lactonase (PLL) that encompasses members from mesophilic organisms (PPH, AhlA, MCP, DrOPH) [3,4,5] as well as thermophilic (SsoPox, SacPox, GsP, GkL) [6,7,8,9] representatives. The PLL family is structurally and biochemically related to bacterial Phosphotriesterases (PTEs) [3]. Indeed, some representatives of PLLs were primarily isolated by virtue of their phosphotriesterase activity towards the insecticide paraoxon, and were named paraoxonases (Pox) as was the case with SisLac's closest homologue, SsoPox, an enzyme isolated from Sulfolobus solfataricus [6]. However, further phylogenetic and biochemical studies has revealed that these enzymes are native lactonases endowed with promiscuous paraoxonase activity and more generally with organophosphate hydrolase activity [3]. Hyperthermophylic PLLs (hPLLs) are appealing enzymes in biotechnology because they possess an intrinsically high stability that often confer high resistance towards harsh conditions and proteases activity [10], which constitute useful properties for storage and large scale purification.
Interestingly, PTEs exhibit diffusion limit-like second order rates with paraoxon as a substrate [11], and are also endowed with promiscuous lactonase activity [3,12]. The particular link between these two families raises the hypothesis that PTEs have diverged from native lactonases like PLLs [3,13,14,15]. Indeed, both PTEs and PLLs belong to the amidohydrolase superfamily [3]. Despite the relatively low sequence identity between these two families (,30%), PTEs and PLLs exhibit the same (b/a) 8 barrel fold or socalled TIM barrel [16]. At the C-terminus of the barrel, two divalent metal cations constitute the active site [13,17,18]. The phosphotriesterase activity of PTEs and PLLs is modulated in the presence of various divalent cations, the highest activity being achieved in a cobalt-containing buffer for Pseudomonas diminuta PTE [19], OpdA [20] and SsoPox [6].The active site metal cations' chemical nature has been investigated using anomalous X-ray scattering and has revealed that both the PTE from Agrobacterium radiobacter (OpdA) and the PLL SsoPox possess an iron/cobalt heterobinuclear center when the expression media is supplemented with cobalt ions [13,20].
The catalytic mechanisms for both the lactonase and phosphotriesterase involve a nucleophilic attack by a water molecule activated by the bi-metallic center. The major difference between the two activities consists of the respective transition state geometries: bi-pyramidal for the phosphotriesters and tetrahedral for the lactones. The fact that these two activities can be catalyzed with significant rates within the same active site suggests an overlap between the stabilization of the corresponding transitions state species, from which the enzymatic promiscuity would stem from [14,21]. The active site of PLLs possesses three sub-sites that are remarkably adapted for the lactone binding: a small sub-site, a large sub-site and a hydrophobic channel [13]. The aliphatic chain of the lactones binds within the hydrophobic channel, the large sub-site accommodates the amide group of the N-acyl chain, and the small sub-site positions the lactone ring. AHLs are molecules that mediate bacterial communication (quorum sensing) for many species. These molecules accumulate in the media to reach a certain threshold for which the transcriptional profile of the bacteria is altered [22]. Some studies report that virulence and biofilm life-style are regulated by quorum sensing [23,24,25,26], and suggest that quenching the quorum sensing (quorum quenching) could be an interesting strategy against multi-resistant pathogen bacteria using AHL based quorum sensing like P. aeruginosa [27,28,29,30,31]. By hydrolyzing AHLs, lactonases like PLLs can quench the AHL-mediated communication between bacteria, as seen for AiiA [28] or for human paraoxonases [30].
Because of their dual catalytic activities, lactonases and phosphotriesterases, PLLs constitute highly attractive candidate for biotechnological utilization as quorum quenching agent [32] or OPs biodecontaminant [33]. However, their precise biological function(s) remain(s) unknown. In many cases, PLLs are found in bacteria that do not produce or sense AHLs, which may suggest a role in interfering with the quorum sensing of other organisms or in the metabolism of the AHL molecules. Furthermore, some members of the PLL family efficiently hydrolyze gamma and/or delta oxo-lactones, but not AHLs [9,34]. However, the structural features that determine the lactonase specificity of these two classes remain unknown.
In this work we present a structural and biochemical analysis of SisLac. We have performed a detailed kinetic characterization and show that this enzyme hydrolyzes AHLs, but also c and d-lactones with high proficiency. Moreover, we provide a characterization of its thermal stability, thermophilicity and structural determinants responsible of its stability. The structural comparison with SsoPox reveals an important dimer interface change. Key amino acids variations between the two close homologues were analyzed by mutagenesis study and revealed their critical involvement in the protein stability and lactonase activity.

Strain, Plasmids and Site Directed Mutagenesis
The plasmids preparations were performed in Escherichia coli strain DH5a (Invitrogen). Protein production was performed in E. coli BL21(DE 3 )-pGro7/GroEL strain (TaKaRa) using plasmids pET22b-StrepTevSisLac and pET22b-SsoPox (provided by Gen-eArt; Germany). Site directed mutagenesis was performed in 50 mL using Pfu polymerase (Invitrogen) on 100 ng of plasmid encoding corresponding genes and primers referenced in Table  S1. The PCR cycle was performed using hybridization temperature of 55uC, elongation time of 12 min during 30 cycles and final elongation of 20 min. The template plasmid was eliminated by Fast Digest DpnI (Fermentas) digestion of 30 min at 37uC followed by inactivation step of 20 min at 80uC. Plasmids were concentrated by classical alcoholic precipitation and then electroporated (Gene-Pulser, Bio-Rad) into E. cloni 5alpha cells (Lucingen), a particularly competent strain of E. coli. Site directed mutagenesis was finally verified by sequencing.

Production-purification of SisLac, SsoPox and SisLac's Variants
SisLac and its variants were heterologously produced and purified from the Escherichia coli strain BL21(DE3)-pGro7/GroEL (TaKaRa) as previously described [1] with the only difference being that 0.2% (w/v) arabinose was added at the start of the overexpression in order to induce chaperones expression. A very similar production protocol was used for wt SsoPox. Briefly, protein production was performed in 2 liters of ZYP medium [35] (100 mg/ml ampicillin, 34 mg/ml chloramphenicol) inoculated by over-night pre-culture at a 1/20 ratio. Cultures grew at 37uC to reach OD 600nm = 1.5. The induction of protein production was made by starting the consumption of the lactose in ZYP medium. Subsequently, 0.2 mM CoCl 2 was added and the temperature was reduced to 25uC for additional 20 hours. Cells were harvested by centrifugation (3 000 g, 4uC, 10 min), re-suspended in lysis buffer (50 mM HEPES pH 8, 150 mM NaCl, CoCl 2 0.2 mM, Lysozyme 0.25 mg/ml, PMSF 0.1 mM DNAseI 10 mg/ml) and stored at 280uC. The frozen cells were thawed and disrupted by three steps of 30 seconds of sonication (Branson Sonifier 450; 80% intensity and microtype limit of 8). Cell debris were removed by centrifugation (12 000 g, 4uC, 30 min). The purified protein being hyperthermostable, host proteins were precipitated by incubation of 30 min at 70uC and harvested by centrifugation (12 000 g, 4uC, 30 min). Other thermostable proteins from the host E. coli were eliminated by ammonium sulfate precipitation (326 g/L), and the overexpressed protein was concentrated by ammonium sulfate precipitation (476 g/L) and suspended in activity buffer (HEPES 50 mM pH 8, NaCl 150 mM, CoCl 2 0,2 mM). The remaining ammonium sulfate was removed by dialysis against the activity buffer and the protein sample was subsequently concentrated prior to the size exclusion chromatography step (S75-16-60, GE Healthcare). The yield of protein production varied between 20 and 100 mg of protein per liter of culture after purification. The purity and the protein quality were verified by SDS-PAGE and mass spectrometry.

Oligomerization State Determination
Dynamic light scattering (DLS): experiments were performed at room temperature using zetasizer nano series apparatus (Malvern, UK) and the Zetasizer software. 30 mL of purified wt SsoPox and SisLac (2.5 mg/mL) was used in the activity buffer to measure the hydrodynamic radius of particles in the protein solutions at 633 nm.
Multi-angle Light Scattering Studies: experiments were performed at room temperature using a size exclusion chromatography (KW803 column (Shodex)) carried out on an Alliance 2695 HPLC system (Waters) at a flow rate of 0.5 ml/min. The buffer used was similar to activity buffer despite pH was adjusted to 7.3 using NaOH. The signal was monitored by a three-angle light-scattering detector (MiniDAWNTMTREOS; Wyatt Technology), a quasi-elastic light-scattering instrument (DynaproTM, Wyatt Technology), and a differential refractometer (Optilab rEX, Wyatt Technology). Molecular weight, gyration, and hydrodynamic radii determination were performed by the ASTRA V software (Wyatt Technology) using a dn/dC (specific refractive index increment of the solution) value of 0.185 ml/g.

Enzymatic Characterization
The time course of ethyl-paraoxon hydrolysis by SisLac at 70uC was monitored following the p-nitrophenolate production at 405 nm (e 405nm = 17 000 M 21 cm 21 ) in 1-cm path length cell with a Cary WinUV spectrophotometer (Varian, Australia) and using the Cary WinUV software. Standard assays (500 mL) were performed in paraoxonase buffer CHES 50 mM pH 9, NaCl 150 mM, CoCl 2 0.2 mM, EtOH 6% (v/v), with pH adjusted with NaOH at 70uC.
At 25uC, the phosphotriesterase, esterase and lactonase activities were analyzed monitoring absorbance variations in 200 mL reaction volumes using 96-well plates (6.2-mm path length cell) and a microplate reader (Synergy HT) using the Gen5.1 software at 25uC. For each substrate, assays were performed using organic solvent concentrations below 1%. The monitoring wavelength, the solvent used, the molar extinction coefficient and the concentration range for each substrate (Fig. 1, S1 & S2) are summarized in Table S2. Phosphotriesterase and esterase activities were performed in activity buffer. When required, DTNB at 2 mM was added to the buffer to follow hydrolysis of substrate releasing thiolate group (malathion (Fig. S1V)). Catalytic parameters for some phosphotriesters were also recorded using SDS at concentrations 0.01 and 0.1% (w/v). Lactone hydrolysis assays were performed in lactonase buffer (Bicine 2.5 mM pH 8.3, NaCl 150 mM, CoCl 2 0.2 mM, Cresol purple 0.25 mM and 0.5% DMSO) using cresol purple (pK a 8.3 at 25uC) as pH indicator to follow the acidification related to the lactone ring hydrolysis. Molar coefficient extinction was measured by recording absor-bance of the buffer over a range of acetic acid concentrations (0-0.35 mM). The absorbance values versus acetic acid concentration were fitted to a linear regression (GraphPad Prism 5 software) with a slope corresponding to molar extinction coefficient (see Table  S2). For all experiments, each point was made in triplicate and the Gen5.1 software was used to evaluate the initial velocity at each substrate concentration. Mean values were fitted to the Michaelis-Menten equation using Graph-Pad Prism 5 software to obtain the catalytic parameters. In the case of C4 AHL hydrolysis for which the substrate concentration that enable to determine the enzyme V max could not be reached, the catalytic efficiency has been determined by fitting the linear part of the Michaelis-Menten plot to a linear regression.

Thermostability Analysis
Temperature dependence analysis. The temperature dependence of the SisLac paraoxonase activity was studied over the temperature range 25-85uC with 10uC increment. The ethylparaoxon (2 mM) hydrolysis was monitored in 500 mL at 405 nm (e 405nm = 17 000 M 21 cm 21 ) in 1-cm path length cell with a Cary WinUV spectrophotometer (Varian, Australia) using the Cary WinUV software. Triplicate experiments were performed in paraoxonase buffer with pH adjusted with NaOH to 9 at each temperature.
Activity-based thermal stability. The residual paraoxonase activity of SisLac after incubation at different temperatures was performed. Incubation time and temperatures tested in this experiment were 390 min at 85uC, 90uC and 95uC in activity buffer. The ethyl-paraoxonase (1 mM) activity was followed at 25uC every 15 min during the first hour of incubation and every 30 min till the end of the experiment (after cooling). Paraoxon hydrolysis was monitored at 405 nm (e 405nm = 17 000 M 21 cm 21 ) with a microplate reader (Synergy HT) and Gen5.1 software in a 6.2 mm path length cell for 200 mL reaction in 96-well plate. Initial velocity at 25uC was used as reference to normalize velocities obtained after incubation at high temperature. Values are represented as fraction of the reference (value 1.0). The halflife of the protein at each temperature was determined by fitting the data to an exponential decay equation using GraphPadPrism 5 software.
Melting temperature determination. Circular Dichroism (CD) spectra were recorded using Jasco J-810 spectropolarimeter equipped with Pelletier type temperature control system (Jasco PTC-4235) in 1 mm thick quartz cell and using the Spectra Manager software. Measurements were carried out in 10 mM sodium phosphate buffer at pH 8 with a protein concentration of 0.1 mg/mL. For wt SisLac, denaturation was first recorded between 190 to 260 nm with a scattering speed of 20 nm/min every 10uC at temperatures ranging between 20 to 90uC. To determine the melting temperature of proteins (wt SisLac and variants), the denaturation was recorded at 222 nm by increasing the temperature from 20 to 90uC (at 5uC/min) in 10 mM sodium phosphate buffer at pH 8 containing increasing concentrations (0.5-3.5 M) of guanidinium chloride. The theoretical T m without guanidinium chloride was extrapolated at the y-intercept by a linear fit using the GraphPadPrism 5 software.

pH Dependence Profile Determination
The pH-dependence ethyl-paraoxonase activity (2 mM) profile of SisLac was monitored at 348 nm (the pH-independent isobestic point of p-nitrophenol and p-nitrophenoxide ion; e 348nm = 5 300 M 21 cm 21 ) [6]. To explore the pH range 5-11, different buffers were prepared containing 50 mM monobasic phosphate over the pH range 5-7, 50 mM HEPES for pH 8, 50 mM CHES over the pH range 9-10 and 50 mM dibasic phosphate at pH 11. The buffers also contained 150 mM NaCl and 0.2 mM of CoCl 2 , and were adjusted with NaOH or HCl. The kinetic measurements were performed at 25uC with a microplate reader (Tecan) and Magellan software in a 6.2 mm path length cell for 200 mL reaction in 96-well plate. Each experiment was made in triplicate and initial velocities were evaluated using Excel software (Microsoft).

Crystallization and Structure Determination
The crystallization procedure of SisLac has been previously described [1]. Diffraction data were collected at the ESRF (Grenoble, France) BM-30A (FIP) beamline using a wavelength of 0.98 Å on an ADSC Quantum Q315 Detector. X-Ray diffraction data were integrated and scaled with the XDS program [36]. The presence of a twin was clearly established using phenix.xtriage [37] from the PHENIX refinement-program suite [38]. The Molecular replacement using the SsoPox structure as model (PDB code 2vc5) was performed with Phaser [39]. The twin operator (-h, -k, l) and a twofold axis (z) arising from the twinning were determined using phenix.xtriage. The solution was then used for refinement performed using REFMAC5 [40] using the twin option and Coot [41] for model improvement. The model and structure factor were deposited under the Protein Data Bank (PDB) code 4G2D. Despite the twinning, the electronic density maps were of good quality ( Fig. S3; R and R free values (0.2649 and 0.2925, respectively; Table 1)).
Structural analysis and comparison, cartoon and ribbon representation were made using PyMOL (www.pymol.org). Surface contacts and interaction analysis was performed using the PROTORP server [42]. Root mean square deviations (r.m.s.d) were computed using Swiss-pdb-viewer software [43].

Sequence Alignment
The alignment was performed using T-coffee server [44,45], manually improved with seaview software [46] and finally drawn with BioEdit 7.1.3.

Biochemical and Biophysical Characterization of SisLac
Oligomeric state analysis. Size exclusion chromatography, dynamic light scattering and multi angle light scattering experiments were carried out to determine the oligomeric states of SisLac and SsoPox. Using a combination of static plus dynamic light scattering, UV spectrophotometry, and refractometry, SisLac and SsoPox appear to be dimeric at room temperature (25uC) (Fig. 3A 3C-D). These results confirm that both proteins are dimeric at 25uC. The existence of homodimers is consistent with the crystal structures of SisLac, SsoPox [47] and other PLLs (DrOPH (PDB ID: 3FDK), GsP (PDB ID: 3F4D) and GkL (PDB ID: 3OJG)).
pH and temperature dependence. We determined the pH and temperature dependency of SisLac's catalytic activities. However, because the lactonase assay utilizes a pH indicator, these characteristics could only be determined for the paraoxonase activity. The optimal pH for SisLac's paraoxonase activity was established by measuring the velocity at pH ranging from 5 to 11 (Fig. 4A). The pH-rate dependence plot displays a bell-shape curve with a wide plateau between pH 7 and 10 (with maximal activity at pH 9). The same dependency what was described for SacPox at 70uC [7], and only a slightly different pattern was observed for SsoPox (optimum at pH 8 [6]). The PTE from P. diminuta also exhibits an activity maximum at a pH range of 8-10 [48]. This observation is consistent with the hypothesis of common mechanism shared by these enzyme families.
The temperature dependency was investigated by measuring the paraoxonase activity at temperatures ranging from 25 to 85uC (Fig. 4B). The highest temperature tested (85uC) presented the highest velocity. However, within the tested temperature (imposed by technical limitations), we did not find a maximum. Similarly, no maxima were found in the cases of SsoPox [6], DrOPH [5] and GsP [8]. In contrast, SacPox presents a maximum activity around 70uC and the activity decrease above this temperature [7].

Thermostability
The thermostability of SisLac was evaluated based on its catalytic activity. The residual paraoxonase activity of the enzyme after different incubation times at 80, 90 and 95uC was measured (Fig. 4C). The enzyme exhibited respective half-lives of 84620 min, 8.561.5 min and 3.660.4 min at 85, 90 and 95uC. In comparison, SsoPox exhibits a half-life of 4 hours at 95uC, and 90 min at 100uC [6], and SacPox exhibits a half-life of 5 min at 90uC [7] (from S. acidocaldarius, living temperatures from 55 to 85uC [49]).
The thermal stability of SisLac was also determined by circular dichroism at temperatures ranging between 20uC and 90uC (Fig.  S5). However, as previously observed for SsoPox [47], the extreme thermostability of this enzyme does not allow to precisely determine a melting temperature (T m ) by this method. Different concentrations of guanidium chloride were required to further destabilize the protein and the T m values were extrapolated to zero guanidinium chloride: T m at 10262uC for SisLac (Fig. S5) (while SsoPox's T m = 106uC [47]). These values are in the range of the other characterized thermostable PLLs, including DrOPH and GsP whose Tm values are 88.1uC [5] and 106.6uC [8], respectively. Notably, the extremophile Deinoccocus radiodurans is a  mesophilic bacteria [5], and DrOPH exhibits an extreme T m , typical for enzymes from hyperthermophiles or thermophiles (Sulfolobus solfataricus living temperatures from 50 to 87uC [49], Sulfolobus islandicus from 59 to 91uC [2], Geobacillus stearothermophilus from 37-70uC [50,51]).
The modulation of SisLac phosphotriester hydrolysis by Sodium Dodecyl Sulfate (SDS), which has been previously shown to act as an activator in the case of SsoPox [53], was also tested ( Table 2). Interestingly, the addition of 0.01% SDS yields to a 2.5 folds increase in the paraoxonase catalytic efficiency (6.36(62.18)610 2 M 21 s 21 ), whereas the addition of 0.1% SDS enhanced the efficiency by 25 folds (k cat /K M = 7.14(64.16)610 3 M 21 s 21 ).

Structural Analysis of SisLac
SisLac is homodimeric in the crystal structure with overall dimensions of the monomers of being approximately 39648656 Å . As for its close homolog SsoPox [13] and related PLLs like DrOPH [5], GsP [8] and GkL [9], SisLac is roughly globular and exhibits a (b/a) 8 barrel topology. The active site consists of a binuclear center located at the C-Terminal of the barrel. Four histidines (His22, His24, His170, His199), one aspartic acid (Asp256) and a carboxylated lysine (residue 137) are coordinating the two metals. The two metal cations (possibly iron and cobalt, as seen for SsoPox [13] and OpdA [20]) are bridged by a water molecule that is presumed to be the catalytic nucleophile. The active site includes a long hydrophobic channel that was revealed by structural studies on SsoPox as the binding region of aliphatic chains for the AHL substrates [13]. Indeed, SisLac's structure is overall very similar to the structure of SsoPox (root-mean-square deviation (r.m.s.d.) for a-carbon atoms (over 314 atoms) of 0.35 Å ).
Salt bridge network analysis. SisLac sequence exhibits approximately the same amino acid content as SsoPox, containing 14.3 versus 16% of uncharged polar residues and 28.7 versus 28% of charged residues. This is no surprising since both enzymes possess high sequence identity (91%) (see sequence alignment, Fig. 2). As described for SsoPox, the charged residues are mainly located at the protein surface, forming complex electrostatic networks [47] that includes 28 salt bridges implicating 46 residues. This charge network mainly differs by the substitution K14E in SisLac that suppress a salt bridge network between E12-K14-D15 of SsoPox,  and consequently creates a local concentration of 3 negative charges within 4 consecutive residues (Fig. 5A). Dimer interface analysis. The dimer interface of SisLac comprises 46 residues (44 in SsoPox). The contacting area is almost identical and comprise a typical value for homodimers [42] (1770 Å 2 for SisLac structure, 1750 Å 2 for SsoPox structure, lower for other thermostable PLLs: 1728 Å 2 for GkL structure, 1632 Å 2 for GsP structure and 1473 Å 2 for DrOPH structure). The nature of this interface is mainly hydrophobic in both enzymes, but SisLac's interface tends to be more hydrophobic (56% for SisLac, 52.6% for SsoPox), more charged (SisLac 23.91%, SsoPox 20.45%) and less polar (SisLac 32.61%, SsoPox 36.36%). Four interface residues differ from SisLac to SsoPox but these substitutions do not fully explain the observed differences. Interestingly, the superposition of SisLac and SsoPox monomers shows a relative movement of the second SisLac monomer of about 5 Å relative to SsoPox's, revealing that the interface area is slightly shifted (Fig. 5B). This reorganization of the dimer interface appears to be due to the substitution Q34Y, although it could also originate from the different crystal packing of both proteins. Whereas the two Q34 interact with each other's in SsoPox structure, the bulkiness of both Y34 in SisLac imposes a reorientation of the dimer (Fig. 5C). This reorientation increases the monomers interpenetration, and makes one closer to the active site of the other. This trend was previously described while comparing the P. diminuta PTE and SsoPox structures [47], although these enzymes are far more divergent in sequence (about 30% identity). Although the biological importance and catalytic influence of dimerization for these enzymes remain unclear, second shell active site residues are notably involved in dimer formation.

Key Substitutions between SisLac and SsoPox
The structural analysis revealed that the positions 14 and 34 seem to be the major impacting variations between SisLac and SsoPox structures, that may relate to dimerization changes (position 34) or stability (positions 14,34). Mutational intermediates between SisLac and SsoPox have been constructed to evaluate the consequences of these substitutions (E14K and Y34Q) for the enzyme stability and activity. All the variants exhibit lower Tm as compared to SisLac (10262uC) and SsoPox (106uC) [6] while double variants (Y34Q-E14K) present the highest Tm among the variants ( Table 7). The analysis of Tm reveals that the substitution Y34Q and E14K are destabilizing on SisLac background but the combination of both variations tends to restore partially the stability ( Table 7). Additionally, the mutants have been characterized for catalytic activity against ethyl/methylparaoxon and for the best AHL, d-lactone and c-lactone substrates of wt SisLac ( Table 8). The efficiency of methyl-paraoxon hydrolysis is similar for wt and the mutants, whereas the mutants exhibit higher catalytic efficiency against ethyl-paraoxon than the wt enzyme. However, the mutants exhibit a dramatically reduced AHLase activity ( Table 8). A similar trend is observed with d/clactones. These results clearly highlight the critical importance of these positions for SisLac stability and activity, and validate our structural analysis.

Catalytic Properties of SisLac
The pH dependence of SisLac was investigated and yields a bellshaped curve with a pH optimum at pH 9, a consistent behavior with previously characterized PLLs [6,7] and PTEs [48]. This pH dependence profile is also in agreement with the commonly accepted hydrolysis mechanism where a water molecule activated   by the bi-metallic active site serves as nucleophile [13]. Additionally, the metal dependence was assayed and the metal nature was found to modulate the catalytic activities, as previously described in PTEs [19] and PLLs [34]. Amongst the tested metals, SisLac shows preference for cobalt cations for both lactonase and paraoxonase activities (supplementary information & Fig.  S6), as previously reported for the paraoxonase activity of SsoPox [6] and the lactonase activity of DrOPH [34], MCP [4] and GkL [9].

Phosphotriesterase Activity
The catalytic efficiency of paraoxon hydrolysis by SisLac at 25uC (k cat /k M = 2.60(60.58)610 2 M 21 s 21 ) shows that SisLac is endowed with one of the highest paraoxonase activity amongst PLLs [3,4,5,6,8,9]. Moreover, this activity can be considerably increased by addition of 0.1% SDS (25 folds), suggesting that the enzyme has an interesting potential for catalytic improvement. The potential of this enzyme for organophosphorus compounds bio-decontamination is further illustrated by its ability to hydrolyze the nerve agent analog CMP-coumarin with significant efficiency (k cat /K M = 4.26(61.86)610 3 M 21 s 21 ).
Interestingly, we observed that methyl-paraoxon is a better substrate than ethyl-paraoxon, the same trend being monitored between methyl and ethyl-parathion. The fact that smaller substituents on the phosphorus center are preferred by the enzyme is consistent with the promiscuous nature of paraoxonase activity in SisLac. Moreover, the higher compaction of the SisLac monomer makes active sites closer one to each other in the dimer as compared to SsoPox structure and thus could explain the substrate preference in disfavor of bulkier substrates.
In addition, SisLac exhibits very low catalytic efficiency towards P = S containing organophosphates (e.g. ethyl and methyl-parathion, malathion). The K M values are in the range of the native substrates (hundreds of mM), but the k cat values are extremely low (about 3 orders of magnitude lower than P = O containing OPs), which is a typical behavior for promiscuous activities [56]. PTEs, albeit preferring paraoxon to parathion as a substrate, do not exhibit such a pronounced thiono-effect [20,55]. PTEs constitutes a protein family that are believed to have diverged from PLLs like SisLac [3] in the last few decades to specifically hydrolyzes manmade insecticides [57]. They thus may have evolved to suppress this thiono-effect in order to hydrolyze the most used pesticides (e.g. parathion).

Esterase and Lactonase Activity
PLLs have been previously characterized as poor esterases [3], and so is SisLac. Amongst the five tested esters, only pNP-acetate is a substrate for SisLac. The natural substrates of PLLs, lactones, being a specific class of esters, it is thus surprising that PLLs exhibit low esterase activity. It might results from a rather good binding to the active site (as suggested by observed K M for pNP-acetate), but with a large fraction of non-productive binding (very low k cat ).
PLLs are natural lactonases that might be involved in quorum quenching [13]. Their precise substrates and biological function(s) are however still unknown. Lactones encompass two major families of compounds, the lipophilic aroma (oxo-lactones) and the Acyl Homoserine Lactones (AHLs) involved in quorum sensing. The quorum sensing is common in bacteria, but its existence in the archaeal world remains unclear, despite the finding of AHL-based quorum sensing stimulating molecules in Natronoccocus occultus [58], the presence of biofilms in Sulfolobus sp. [59] and the recent characterization of complete carboxylated-AHLs quorum sensing system in methanogenic archaeon [60].
Our kinetic experiments show that SisLac prefers long aliphatic chain lactones, exhibiting optimal activity when the acyl chain contains 7 carbon atoms, as seen for SsoPox [52]. SisLac show also preference for 3-oxo-AHLs and hydrolyzes poorly short chain AHLs. It is interesting to notice that other PLLs, like AhlA and PPH, hydrolyze short and long chain lactones with similar catalytic efficiency [3]. It is thus possible that within the PLL family, different sub-groups of enzymes exhibit different specificities, and thus possibly different physiological functions. In addition, we show that SisLac is a proficient enzyme against oxo-lactones (best substrate: undecanoic-d-lactone, k cat / k M = 1.77(60.04)610 6 M 21 s 21 ), and hydrolyzes more efficiently long chain lactones, with a preference for 5 to 6 acyl chain carbon atoms. This preference is similar to that observed for AHLs as substrates, and possibly indicates that oxo-lactones and AHLs acyl chains bind into a similar pocket, most likely the hydrophobic channel connected to the active site that was depicted for SsoPox structure [13,47]. However, interestingly, whereas the short chain C4-AHL is a poor substrate for SisLac, the heptanolide-d-lactone (3 carbon atoms in the acyl chain) shows a 10 4 higher hydrolysis efficiency. In addition, lactones with very short or without acyl chains (dihydrocoumarin, c-butyrolactone, d-valerolactone, ecaprolactone) constitute better substrates than C4-AHL. Altogether, these features might reveal that cand d-lactones utilize an alternate binding mode for the lactone ring than AHLs, and/or that C4-AHL does not bind in a catalytically relevant fashion to SisLac. Figure 5. Structural analysis. A-Close view on position 14 in SsoPox (left side) and SisLac (right side) structures. Negatively and positively charged residues are represented as red and blue spheres, respectively. Black dashed lines correspond to the putative salt-bridges in SsoPox. B-Structural comparisons of the dimers of SisLac (violet), SsoPox (red) and P. diminuta PTE (yellow). One monomer was used to superpose the three structures (shown on the right side), thus indicating the conformational shift in the position of the second monomer (left side). The differences are highlighted by blue arrows (P. diminuta PTE versus SsoPox) or by black arrows (SsoPox versus SisLac). When compared to P. diminuta PTE, the rotational shift observed in SisLac is more pronounced than that of SsoPox. C-Close view on position 34 in SsoPox (grey) and SisLac (violet). Black dashed lines correspond to the hydrogen bond between Q34 of each monomer of SsoPox with a distance of 3.2 Å . The black arrow shows the rotational shift that seems to be induced by this substitution. doi:10.1371/journal.pone.0047028.g005 Structural Determinants for Thermal Stability The major structural determinants explaining the high thermal stability of hPLLs have been documented with the example of SsoPox [47], the comparison with GsP and DrOPH enzymes [8], and are part of the classical properties described for hyperthermostable proteins [61]. The structures of these enzymes, including SisLac, exhibit a high number of salt bridges organized in complex networks of charges at the protein surface that may rigidify the global protein architecture. Moreover, the homodimer interface is larger and more hydrophobic (see Results) and the overall structure is more compact than mesophilic counterparts [47]. Here we observed that the interface area between SisLac's monomers is slightly shifted, as compared with SsoPox structure. This reorganization of the dimer interface is consistent with observations made in solution. Indeed, as observed in crystals, both enzymes are dimeric at 25uC. The importance of hydrophobic contacts within the dimer interface of SisLac and SsoPox explain the dimerization of the proteins. Since the hydrophobic effect increases with temperature, it is highly probable that these enzymes could be dimers at physiological temperatures (50-90uC) [2,49].
Interestingly, both Sulfolobus species from which SisLac and SsoPox enzymes originates lives in similarly extreme environments (S. solfataricus from 50 to 87uC [49], S. islandicus from 59 to 91uC [2]) and exhibit similar thermostability (T m = 10262uC for SisLac and T m = 106uC for SsoPox [6]). Taking advantage from their high sequence identity between the two proteins, we studied the substitution K14E in SisLac (as compared to SsoPox) that breaks a salt bridge network at the C-terminus of the protein, a region concentrating the highest divergences among hPLLs as revealed by sequence alignment (i.e. SsoPox, SacPox and SisLac). E14 engenders in SisLac a cluster of 3 negatively charged residues at the surface of the structure allowing to evaluate the contribution of these electrostatic interactions to the enzyme stability and activity. Moreover, another key substitution occurred between the two enzymes in the homodimerization interface. Q34Y is indeed a key substitution in the interface since it consists in a ''pivot'' residue, i.e. a residue that contacts its equivalent in the second protein molecule while forming the dimer, and seems to be responsible for the observed dimerization shift between SisLac and SsoPox structures. Q34Y is, moreover, the substitution between SisLac and SsoPox that is the closest in space to the active site (second shell). We therefore studied the effects of this variation on SisLac's activity and stability.
Surprisingly, whereas K14 in SsoPox is involved in a large network of charged interactions and may contribute to the overall protein rigidity [47], the variation E14K in SisLac is destabilizing (Tm is decreased by 6uC). The variation of the pivot interface residue Y34Q is also destabilizing on SisLac background (decrease of Tm by 8uC). Interestingly, the double variant E14K-Y34Q that carries two destabilizing mutations exhibits a higher Tm than the single variants, revealing the highly epistatic nature of these positions. Moreover, whereas the promiscuous phosphotriesterase activity is not altered by these substitutions, the lactonase activity, especially the AHLase activity is considerably reduced (by ,100 folds) as compared to wt. The influence of E14K on SisLac's catalytic activity is not obvious from a structural analysis. However, position 34 comprises a second shell residue, and the overall dimer interface is in the vicinity of the active site. Mutation of position 34 highly influences the protein dimerization and thus the degree of monomers interpenetration. Monomer active sites being close one to each other, their interpenetration could influence substrate specificities and catalytic efficiencies by fine steric or dynamic constraints which can't be evaluated by structural analysis. These identified key substitutions, however, does not fully explain the observed different catalytic properties of SisLac and SsoPox. The active site residues and configuration of these two enzymes being similar, these discrepancies might be partly mediated by yet unidentified substitutions distant from the active site. Reconstructed mutational intermediates between SisLac and SsoPox have lower fitness both in term of stability and AHLase activity. Despite the very high sequence identity between both proteins (91%), it may indicate that the evolutionary route that links them already comprise a fraction of highly epistatic mutations. In other words, the mutations that accumulate at a very early stage of divergence might not only be neutral but a fraction of them are highly cooperative.