Crystal structure of Pelagibacterium halotolerans PE8: New insight into its substrate-binding pattern

Lysophospholipase_carboxylesterase (LPCE) has highly conserved homologs in many diverse species ranging from bacteria to humans, as well as substantial biological significance and potential therapeutic implications. However, its biological function and catalytic mechanism remain minimally investigated because of the lack of structural information. Here, we report the crystal structure of a bacterial esterase PE8 belonging to the LPCE family. The crystal structure of PE8 was solved with a high resolution of 1.66 Å. Compared with other homologs in the family, significant differences were observed in the amino acid sequence, three-dimensional structure, and substrate-binding pattern. Residue Arg79 undergoes configuration switching when binding to the substrate and forms a unique wall, leading to a relatively closed cavity in the substrate-binding pocket compared with the relatively more open and longer clefts in other homologs. Moreover, the mutant Met122Ala showed much stronger substrate affinity and higher catalytic efficiency because less steric repulsion acted on the substrates. Taken together, these results showed that, in PE8, Arg79 and Met122 play important roles in substrate binding and the binding pocket shaping, respectively. Our study provides new insight into the catalytic mechanism of LPCE, which may facilitate the development of structure-based therapeutics and other biocatalytic applications.

Esterases have important physiological roles and biotechnological applications because they can catalyze the hydrolysis of short-chain ester-containing molecules and produce carboxylates and alcohols [1][2][3] . Esterases belong to the lysophospholipase_carboxylesterase family (the LPCE family) 4 and were previously classified in bacterial family VI by Arpigny and Jaeger 1 . This family includes the smallest carboxylesterase (23)(24)(25)(26) found to date, and bacterial carboxylesterases show high sequence similaritiy with their eukaryotic counterparts (~40%) 1 . LPCE family proteins play significant roles in human diseases. For example, human putative Gα-regulatory protein acyl thioesterase (APT1) has been well characterized as a modulator in the Ras signaling pathway and has been confirmed as a target for cancer therapeutics 5,6 . Human lysophospholipase-like 1 (LYPLAL1) might be a triacylglycerol lipase involved in obesity 7,8 . In addition, a bacterial carboxylesterase (FTT258) from Francisella tularensis, a causative agent of tularemia 9, 10 , has been investigated as a novel drug target 11 . Overall, LPCE family enzymes are increasingly pharmaceutically interesting as potential therapeutic targets. Nevertheless, our understanding of the LPCE family is very limited. Currently, the crystal structures of only six LPCE family proteins have been reported, including Rhodobacter sphaeroides RspE 12 , Pseudomonas aeruginosa PA3859 13 , P. fluorescens esterase II 14 , F. tularensis FTT258 11 , human APT1 15 and human LYPLAL1 7 .

Results and Discussion
Overall structure. The diffracting dataset of the PE8 crystal was integrated into monoclinic space group P2 1 with two molecules per asymmetric unit and a resolution of 1.66 Å. However, multi-angle light scattering (MALS) analysis showed that the molecular weight (MW) of PE8 was 26.9 kDa (±2.4%) (Fig. 1A), consistent with the theoretical MW of 6× His fusion PE8 (25.4 kDa), and revealed that PE8 existed as a monomer in solution. Additionally, 329 water molecules, one polyethylene glycol (PEG) monomethyl ether (MME) 550 molecule and one glycerol molecule were modeled. The final refined model had an R work of 16.84% and an R free of 20.52%. The crystallographic statistics for data collection and structure refinement are summarized in Table 1.
Compared with the known structures of other LPCE family members, the main differences were found in the short β-strands, helix α4, the short α helices and the 3 10 helices (Supplementary Figure S1). Helix α4 connects β7 with β8 in PE8, R. sphaeroides RspE 12 and human LYPLAL1 7 . However, in P. fluorescens esterase II 14 , P. aeruginosa PA3859 13 and human APT1 15 , strands β7 and β8 are connected by long loops containing short α helices or 3 10 helices. In addition, on the loop between strand β3 and helix α2 (i.e., loop β3), there are four short antiparallel β-strands (β4, β5, βA and βB) in esterase II and APT1. In contrast, βA and βB of loop β3 are replaced with a short helix in RspE or a winding loop in PE8, PA3859 and LYPLAL1 (Supplementary Figure S1).
The β-strands β6, β8 and β9 provide the framework onto which the catalytic residues (Ser118, Asp169 and His201 in PE8) are placed (Figs 2 and 3A). The remaining β strands and α helices are not directly involved in the formation of the catalytic site, and thus, the differences in the secondary structure mentioned above may not influence the catalysis of the active site directly.
Active site. Sequence analysis and the three-dimensional (3D) structure revealed that the catalytic triad residues of PE8 consist of Ser118, Asp169 and His201, which are located on the C-terminal sides of β strands β6, β8 and β9 in the central β sheets (Figs 2 and 3) and are conserved in esterases 1 . The catalytic site is located on the loops outside of the α/β/α-sandwich structure, and no lid covers the catalytic site. The catalytic residue Ser118 is located in the conserved GFSQG motif (Fig. 2). The hydrogen bond distance within the catalytic triad is 2.7 Å from Ser118-Oγ to His201-Nε2 and 2.7 Å from His201-Nδ1 to Asp169-Oδ1 (Fig. 3B). To confirm the roles of the three amino acid residues, site-directed mutagenesis was performed to replace these residues with alanine. The activities of mutant S118A against p-nitrophenyl (p-NP) acetate, p-NP butyrate, p-NP hexanoate and p-NP octanoate were 19.5 ± 0.2%, 16.3 ± 0.5%, 1.3 ± 0.2% and 0%, respectively, compared with the wild-type enzyme (Fig. 1C). The replacement of catalytic residues Ser118 or Asp169 with alanine led to a complete loss of activity Mutants L73A, R79A, R83A, M122A, V171A and H201A changed slightly, whereas mutants V172A, S118A and D169A showed heterogeneity upon gel filtration. (C) The enzymatic activities of wild-type PE8 and its mutants were determined using the following substrates: p-NP acetate, p-NP butyrate, p-NP hexanoate and p-NP octanoate.
Compared with the known structures of LPCE family homologs, the oxyanion hole of PE8 is likely formed by nitrogen atoms of Tyr29 and Gln119, as observed for RspE (Tyr and Gln) but not for esterase II, PA3859, FTT258 and human APT1 (Leu and Gln) or human LYPLAL1 (Ser and Gln) (Figs 2 and 3B). In PE8, the oxyanion hole is occupied by a water molecule in each chain (data not shown). These water molecules might be candidates for the nucleophilic attack on the acylated enzyme 13 , followed by the release of the enzyme in its resting form.
Superimposing these homologs also revealed their similar features and overall folds. The core α/β-hydrolase fold structure, including the catalytic triad and oxyanion hole, is highly conserved, especially in β strands β6, β8 and β9 and helices α3, α5 and α6 in the loops (Fig. 4A,C). High structural variability can be observed in helices α1, α4, and the β3 loop (Fig. 4A), suggesting that these structures are not essential for the catalytic activity.
Among these regions, the β3 loop shows the most significant variation in its amino acid sequence and secondary and 3D structures (Figs 2,4 and S1). The β3 loop shows a higher B-factor within the PE8 structure (Fig. 4B), implying flexibility in its structure and function. Highly variable loops have also been described in LPCE family homologs previously 11,15 . This winding loop is believed to be responsible for the substrate specificity and conformational changes of these homologs 13,25 and is observed to cause the open and closed conformations and affect the catalytic activity and membrane binding of F. tularensis FTT258 11 .  18 . A docking study of p-NP acetate was performed to explore the interaction between PE8 and the substrate. The substrate successfully docked into the active site of PE8 (Fig. 5A), where a PEG MME 550 molecule was detected in the crystal structure (Fig. 5B). In the enzyme-substrate complex docking model, the alcohol part of the substrate occupied the hydrophobic substrate-binding pocket, which was formed by hydrophobic residues Leu73, Met122, Val171 and Val172 (Fig. 5A), similar to the previously suggested mechanism in the LPCE family 11 . The distance between the residues and the nearest carbon atom of the alcohol part of the substrate was 3.3-4.1 Å. More importantly, we found that the two nitro-O atoms of p-NP acetate might form one and two hydrogen bonds with the side chain N atoms of Arg79 and Arg83, respectively (Fig. 5A). To the best of our knowledge, this hydrogen bond between the alcohol part of the substrate and the binding site of the esterase (i.e., not the catalytic site) has not been reported before [11][12][13][14] . In addition, two different configurations of the side chain of Arg79 were observed in chain A and chain B of PE8 ( Fig. 5B and C). The configuration of Arg79 in chain B was proposed as the substrate-binding configuration because it was occupied by the PEG molecule, whereas that in chain A might correspond to the releasing state of the substrate-binding site. Thus, the substrate might be bound and subsequently released by the configuration switching of Arg79. Arg79 is located on the β3 loop of the structure of PE8. It forms a closed wall between the long winding β3 loop and the loop between strand β8 and helix α5, in which the Asp169, Val171 and Val172 are located (Figs 2 and 6A). This wall makes the alcohol binding pocket of PE8 a relatively closed cavity and forms stronger interactions with the alcohol group of the ester substrate. However, because of the high sequence   variability and structural flexibility of loop β3 (Figs 2 and 4), no similar structure was found in other homologs of the LPCE family. For example, R. sphaeroides RspE has the highest sequence and structural similarity with PE8; however, the substrate-binding pocket of R. sphaeroides RspE is an open, longer cleft (Fig. 6B), similar to other members of the LPCE family 11,13 . Hence, Arg79 may play an important role in substrate binding and the shape of the binding pocket, thereby conferring substrate specificity to PE8. To investigate the relationship between the supposed substrate-binding-related residues and the hydrolysis activity of PE8, mutants L73A, R79A, R83A, M122A, V171A and V172A were constructed by site-directed mutagenesis, and their catalytic activities against p-NP esters and kinetic parameters for the hydrolysis of p-NP acetate were determined (Fig. 1C and Table 3). The Km, kcat and kcat/Km of wild-type PE8 were 0.66 ± 0.049 mM, 30 ± 0.68 s −1 and 45 mM −1 s −1 , respectively. The catalytic activities and Km and kcat values of L73A, R83A and V172A were similar to those of wild-type PE8, indicating that these mutations had little effect on the catalytic activity and substrate affinity of PE8 using p-NP esters. A slight increase in the Km value and decrease in the kcat value of mutant V171A were observed (0.89 ± 0.070 mM and 19 ± 0.42 s −1 , respectively), suggesting that the ability to bind the substrate was weakened and that the turnover rate of the enzyme-substrate complex to the product and enzyme was decreased. Thus, the catalytic efficiency against p-NP acetate of mutant V171A was reduced by nearly half, as indicated by a kcat/Km value of 21 mM −1 s −1 . The kcat and kcat/Km values of mutant R79A were approximately 2-fold higher than those of wild-type PE8. Considering the structure of Arg79 shown above, replacing arginine with alanine probably removed the barrier and expanded the substrate-binding pocket, thereby accelerating substrate access and exit. Interestingly, mutant M122A showed a large decrease in its Km value (0.075 ± 0.0069 mM) toward p-NP acetate, resulting in a 10-fold improvement of kcat/Km (553 mM −1 s −1 ). In addition, the catalytic activities of mutant M122A against p-NP butyrate, p-NP hexanoate and p-NP octanoate also increased relative to those of wild-type PE8 (Fig. 1C). The substitution of methionine by alanine might cause less steric repulsion of the substrates and remove the barrier to substrate access. Hence, further engineering in positions Arg79, Met122 or Val171 may provide high activity, affinity or selectivity mutants to specific substrates to facilitate the development of structure-based therapeutics and other biocatalytic applications.

Conclusion
In this study, a new esterase structure in the LPCE family is presented. The structural information of P. halotolerans PE8 expands our knowledge of the catalytic mechanisms of the LPCE family and provides new insight into the substrate-binding pattern in this family. The results establish a novel approach for developing specific inhibitors of its homologs, which could be used for mechanistic research and targeted therapy. In addition, the results of this paper may help broaden the applications of the LPCE family members as biocatalysts in industry.

Materials and Methods
Sequence analysis. The PE8 coding gene was identified and cloned from the genome of P. halotolerans B2 T16, 18 . Amino acid sequence analysis was conducted by BLASTp against the PDB from the National Center for Biotechnology Information (NCBI). Multiple sequence alignment of homologs belonging to the LPCE family was performed by ClustalX v. 2 26 . Secondary structure assignment was determined by DSSP v. 2.0 27 and PROMOTIF 28 . The alignment result with the secondary structure was visualized using ESPript 3.0 29 .
Mutation, protein expression and purification. Point mutants were generated by site-directed mutagenesis using wild-type plasmid as a template for the polymerase chain reaction (PCR). A 15-cycle reaction was performed with the following steps: 98 °C for 10 sec, 55 °C for 30 sec, and 72 °C for 3 min per cycle with PrimeSTAR HS DNA polymerase (Takara, Dalian, Liaoning, China). After digestion with enzyme DpnI (New England Biolabs, Beverly, MA, USA), the PCR products were transformed into Escherichia coli DH5α cells. The positive constructs were determined by DNA sequencing. The wild-type and mutant forms of PE8 were cloned into the expression vector pET28b (Novagen, Madison, WI, USA) and were expressed in E. coli Rosetta (DE3) cells induced by 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 16 hours at 20 °C, as described previously 18,30 . Cells were harvested and disrupted by a sonicator or high-pressure homogenizer. The lysates were sequentially purified by   Biochemical characterization of PE8 and its mutants. The standard reaction was carried out with the appropriate amount of purified PE8 or its mutants in 1 ml mixtures containing 100 mM Tris-HCl (pH 7.5) buffer and 1 mM p-NP acetate (Sigma-Aldrich, Milwaukee, WI, USA, dissolved in acetonitrile) 18 . The activities were determined at 30 °C and 405 nm using a DU800 UV/Visible spectrophotometer (Beckman, Houston, TX, USA). All experiments were performed in triplicate and corrected for substrate autohydrolysis. Substrate specificity assays were performed with p-NP acetate, p-NP butyrate (Sigma-Aldrich), p-NP hexanoate (TCI, Tokyo, Japan) and p-NP octanoate (Sigma-Aldrich). The kinetic parameters were obtained using p-NP acetate as a substrate at different concentrations (0.05 to 4 mM). The kinetic parameters were calculated by analyzing the slopes of the Michaelis-Menten equation using GraphPad Software (GraphPad Inc., USA).
Crystallization, data collection, and structure determination. Crystals of PE8 protein were obtained using the "hanging drop" method by mixing 1 µl of 20 mg/ml protein with 1 µl of reservoir solution at 20 °C. The reservoir buffer contained 0.05 M CaCl 2 , 0.1 M Bis-Tris, and 25% (v/v) PEG MME 550 (pH 6.5). The X-ray diffraction datasets were integrated, scaled and merged using the HKL3000 program 31 . Phases were obtained by molecular replacement using Phaser 32 with the PDB coordinates 4FHZ (R. sphaeroides RspE) 14 as the initial model. The refinement was conducted by Refmac5 33 in the CCP4 software suite 34 and Phenix 35 . The model was built manually by Coot 36 . A structural similarity search was performed with the DALI server 24 . Docking studies were performed with AutoDockTools program 37 . The ligands for docking were edited by Avogadro software 38 , and the topologies of the ligands were generated using the PRODRG server 39 . The successful docking conformation should be satisfied the following criteria: the distance between the OG atom of serine and the carbonyl carbon atom of the substrate was about 2 Å; the catalytic hydrogen bonds were formed, including that between Hδ of the catalytic histidine and the ester oxygen of the substrate, as well as those between the carbonyl oxygen of the substrate and the nitrogen atoms of oxyanion hole residues (Tyr29 and Gln119) 40,41 . All the structures were drawn using PyMOL software (http://pymol.sourceforge.net).
The structure of PE8 was deposited in PDB with accession number 5DWD.