Insights into the molecular mechanism of dehalogenation catalyzed by D-2-haloacid dehalogenase from crystal structures

D-2-haloacid dehalogenases (D-DEXs) catalyse the hydrolytic dehalogenation of D-2-haloacids, releasing halide ions and producing the corresponding 2-hydroxyacids. A structure-guided elucidation of the catalytic mechanism of this dehalogenation reaction has not been reported yet. Here, we report the catalytic mechanism of a D-DEX, HadD AJ1 from Pseudomonas putida AJ1/23, which was elucidated by X-ray crystallographic analysis and the H218O incorporation experiment. HadD AJ1 is an α-helical hydrolase that forms a homotetramer with its monomer including two structurally axisymmetric repeats. The product-bound complex structure was trapped with L-lactic acid in the active site, which is framed by the structurally related helices between two repeats. Site-directed mutagenesis confirmed the importance of the residues lining the binding pocket in stabilizing the enzyme-substrate complex. Asp205 acts as a key catalytic residue and is responsible for activating a water molecule along with Asn131. Then, the hydroxyl group of the water molecule directly attacks the C2 atom of the substrate to release the halogen ion instead of forming an enzyme-substrate ester intermediate as observed in L-2-haloacid dehalogenases. The newly revealed structural and mechanistic information on D-DEX may inspire structure-based mutagenesis to engineer highly efficient haloacid dehalogenases.

D-DEXs have been used to produce L-2-chloropropionic acid as a chiral intermediate for the production of herbicides [18][19][20] . However, because of the low reactivity and stability, no commercial enzyme has been produced. Although the immobilization that covalently attaches HadD AJ1 which is a D-DEX from Pseudomonas putida strain AJ1/23 to the controlled-pore glass has been tested to improve its stability and increase its tolerance to a high substrate concentration, only a minor improvement was made 19 . Structural and functional information is necessary to rationally direct enzyme engineering to meet scientific and industrial needs. To unravel the molecular basis of D-DEX catalysis, the structure of substrate-free HadD AJ1 as well as the structure of the enzyme-product complex (E-P) were determined. The structure-guided elucidation of the catalytic mechanism of D-DEX is presented here.

Results and Discussion
Overall structure of HadD AJ1. The crystal structure of the wild-type HadD AJ1 (WT) was determined by the molecular replacement method using the structure of a DL-DEX, DehI (PDB: 3BJX) from Pseudomonas putida PP3 as the search probe 1 . HadD AJ1 is refined at 2.64 Å resolution with an R free of 25.1 (Table 1). HadD AJ1 is an α-helical hydrolase similar to DL-DEX. However, it completely differs from L-DEX, which is an α/β type hydrolase 1,12,14 . Each asymmetric unit of HadD AJ1 includes four monomers. Each monomer presents a compact fold featuring twelve α-helices and one 3 10 -helix η 1 (Fig. 2B). Two structurally axisymmetric repeats are observed with 20% sequence identity and a superposition RMSD of 1.24 Å in each monomer. Repeat 1 and 2 consist of N-terminal α-helices 1-6 and C-terminal α-helices 7-12, respectively, linked by a lengthy loop with 33 residues including a 3 10 -helix η 1 (Fig. 2). Each repeat is composed of a three-helix-bundle formed by the first three helices and a three-helix-triangular arrangement formed by the second three helices (Fig. 2B). The two repeats are organized by van der Waal forces, salt bridges, hydrogen bonds and hydrophobic interactions of α 6 /α 12 and α 4 /α 10 , respectively. Helices α 4 and α 10 are spatially oriented parallel to each other. Helices α 6 and α 12 mutually interlace at their bulges, located in the midst of the helix. The similar symmetric architecture of the structurally repeated folds is also observed in the phosphorylation-coupled vitamin C transporter 21 . So far, the structurally internal repeats have been reported in many proteins, which are considered to be caused by genetic processes such as fusion and fission of domains, and gene duplication in protein evolution 22 .
HadD AJ1 folds into a homotetramer in the crystal (Fig. 3), which is consistent with the functional state previously reported in solution 18 . In the tetramer, two types of interfaces including A/C (or B/D) and D/C (or A/B) are observed in the assembly of the monomers with the online interactive tool Protein Interfaces, Surfaces and Assemblies (PDBePISA, http://www.ebi.ac.uk/msd-srv/prot_int/cgi-bin/piserver). The interfaces are stabilized by hydrogen bonds and salt bridge interactions ( Fig. 3B and C). Both interfaces comprise approximately 1000 Å 2 (~12%) of total buried surface area, which is considered to be a weak association for oligomeric proteins 23,24 . Comparison of D-DEX and DL-DEX structures. D-DEXs share 30% similarity with DL-DEXs in primary structure ( Supplementary Fig. S1). As shown in Fig. 4A, the superposition of HadD AJ1 with DehI reveals a very high degree of structural similarity with an RMSD of 1.05 Å for 207 C α atoms 1 . The similar structural repeats are regarded as a "pseudo-dimer" in DehI. The "pseudo-dimers" of both enzymes overlap perfectly, also the linker. These similarities reveal the close evolutionary relationship between HadD AJ1 and DL-DEXs. However, two significant differences are observed. First, the η 1 helix and the subsequent loop of the linker between the two repeats are buried in the interface assisting the assembly of the tetrameric HadD AJ1, while the linker wraps around the outside of the homodimeric DehI molecule, without involving the dimer interface. Second, helices α 3 and α 4 of HadD AJ1 are connected by a loop fragment; whereas, a helical bend links the α 3 and α 4 of DehI (Fig. 4B). The two joints contain a strictly conserved Thr between D-DEXs and DL-DEXs. The replacement of Thr by Ala completely damages DL-DEX activity 1,25 , from which Thr is considered as essential for the dehalogenation. However, the same mutant of HadD AJ1 still retains 87.8% activity ( Supplementary Fig. S2A), which suggests that Thr76 is not crucial for the conformational stabilization and HadD AJ1 catalysis. The distinct effects on dehalogenation resulted from the same mutation of the conserved Thr indicates some differences in mechanisms of D-DEXs and DL-DEXs, which can be explained by the corresponding structural differences of the loop and the helical bend.

D-DEX dehalogenation mechanism.
To unveil the dehalogenation mechanism of D-DEX by exploring the molecular details of substrate recognition, HadD AJ1 was crystallized in the presence of the substrate D-2-chloropropionate (D-2-CPA) at pH5.5~6.4 and 277 K. In addition, co-crystallization and crystal soaking trials were undertaken with other reactive substrates including chloroacetate, bromoacetate, and 2-bromopropionate as well as with nonreactive substrates including 2-fluoropropionate, L-2-chloropropionate, 3-chloropropionate, 2,2-dichloropropionate and 2-bromo-2-methylpropionate. Unfortunately, no E-S complex crystals were obtained. However, it turned out that HadD AJ1 was able to bind the product L-LA in the active site. The E-P complex was captured at 2.18 Å resolution (Fig. 5). Further inspection into the complex structure shows that L-LA is caged in the enclosed binding pocket of 71 Å 3 . The binding pocket is oriented at the interface of the two repeats and framed by the structurally repeat-related α 2 , α 8 , α 6 and α 12 (Fig. 4A). There are 13 amino acids arranged in the binding pocket which are Trp48, Lys50, Val51, Ile52, Asn131, Tyr134, Asn203, Ser204, Asp205, Phe281, Met284, Leu285 and Leu288 (Fig. 5B). The product-bound pocket has an identical spatial location to the active site of DehI ( Supplementary Fig. S2B) 1 . By comparison of the active sites of HadD AJ1 and DehI, the residues corresponding to Trp48, Asn131, Tyr134, Ser204 and Asp205 are highly conserved between HadD AJ1 and DL-DEXs, whereas the remaining residues of the active sites are not ( Supplementary Fig. S1), and their functions are discussed below.
Although another product of the catalyzed reaction -the departing Cl − is missing in the complex, a water molecule adjacent to L-LA is trapped in the active site (Fig. 5B). This water molecule interacts with the carbonyl atom of the side chain of Asn131 and the carboxylic hydroxyl of Asp205 by forming hydrogen bonds. The distance between water and the C2 atom of L-LA is 3.0 Å. Considering the conservation of Asp205 with the catalytic residue Asp189 of DehI 1,25,26 , we propose that Asp205 and Asn131 are key residues that activate water required for the hydrolytic dehalogenation of HadD AJ1. Therefore, the mutation of Asp205 and Asn131 was performed (Fig. 5A). Notably, the replacement of Asp205 by Asn completely abolishes the enzymatic activity, which reveals that it's a critical catalytic residue in HadD AJ1. Furthermore, residue Asp is strictly conserved among D-DEXs and DL-DEXs ( Supplementary Fig. S1). The mutation of Asn131 to Asp in HadD AJ1 loses approximately 96% of catalytic activity, which shows Asn131 is an important residue for the dehalogenation. The location of the water molecule and its interactions with the two crucial residues Asp205 and Asn131 indicate that it is likely to be involved in the hydrolytic dehalogenation. Therefore, we hypothesize that the C2 atom of the substrate is directly attacked by a nucleophilic water molecule (Fig. 1B).
To confirm our hypothesis, a single turnover reaction was conducted in H 2

18
O with the WT enzyme that was ten times higher concentration than D-2-CPA. As seen in Fig. 6B, 5.2 times more labelled 18 O-L-LA was produced compared to 16 O-L-LA, which indicates that the dehalogenation is directly mediated by a water molecule. Furthermore, this finding confirms that the nucleophilic water molecule is activated by the Asp205 residue with the assistance of Asn131. The carbonyl oxygen in the side chain of Asp205 forms hydrogen bonds with the -NH 2 group of Asn131, which drives Asp205 to attract a proton from a proximal water molecule. The activated water molecule is poised to attack the C2 atom of the substrate without involving an E-S ester intermediate, which is identical to DL-DEX but different from L-DEX 15 .
As stated above, the E-P complex structure is captured after releasing Cl − and before L-LA departure from the active site. Hydrogen bonds and hydrophobic interactions predominantly stabilize the product in the active site. Hydrogen bonds are formed between the carboxyl of L-LA and the backbone amide nitrogen of Val51, the amino nitrogen of the side chain in Asn203, the backbone amide nitrogen of Ser204, and the carboxyl of Asp205 (Fig. 5B). Carboxyl is invariant in the substrate and the product before and after the dehalogenation. Therefore, it is hypothesized that Val51, Asn203, Ser204 and Asp205 contribute to stabilizing the substrate as well. Correspondingly, mutagenesis of these residues would be expected to affect enzymatic activity. As shown   Fig. 6A, 63.4%, 98.8%, 99.6%, 95.1% and 58.2% of enzymatic activity is lost for mutants V51F, N203A, N203S, S204A and S204T, respectively, towards D-2-CPA. The replacement of non-conserved Val51 and Asn203 by the equivalent residues found in DL-DEX led to a significantly reduced HadD AJ1 activity, which sheds light on different roles of these positions in the two types of enzymes. The increased K m of the mutants V51F and N203A demonstrate their contribution to the binding interactions with the ground state substrates ( Table 2). Consistent with the altered activity, N203A exhibits an obviously decreased k cat /K m which is 355-fold lower than that of the   (Table 2), which suggests the side chain of Asn203 is crucial for stabilizing the E-S transition state. By contrast, V51F exhibits a 4-fold lower k cat /K m value than the WT enzyme (Table 2), which indicates that the increased steric hindrance in the mutant V51F does not severely perturb the transition state. The different loss in the enzymatic activity resulted from S204A and S204T reveals that the hydroxyl group of Ser204 and its orientation affects the interaction with the substrates. The D205E mutant retains 10.1% of activity (Fig. 6A), suggesting the carboxylic hydroxyl of Asp205 is essential for HadD AJ1 dehalogenation. Simultaneously, the increased K m proves that Asp205 participates not only in water activation but also in the stabilization of the substrate (Table 2). Additionally, the thermodynamic contributions of Val51, Asn203 and Asp205 are demonstrated by the elevated transition state free energies of 3.5 kJ/mol, 14.8 kJ/mol and 6.49 kJ/mol resulting from V51F, N203A and D205E mutants, respectively, in comparison with that of WT (Table 2).
Residues Ile52, Phe281, Met284 and Leu285 form a hydrophobic pocket, which engages in hydrophobic interactions with the -CH 3 group of L-LA. The mutations of Ile52 to Gly and Met284 to Cys resulted in 98.5% and 93% activity loss, respectively, suggesting the hydrophobic side chains of these two residues make a great contribution to the interaction with the substrate. The mutantations F281A and L285I destroy 45.7% and 75.6% of enzymatic activity, respectively, which indicates that hydrophobic residues of appropriate sizes are required in these positions for interacting with the substrate. The Cl − is released towards Phe281 after the C2 atom is attacked from the opposite side of the halogen atom by the nucleophilic water molecule. Therefore, Phe281 was speculated to be involved in stabilizing the departing Cl − , which is likely to be trapped by an anion-π interaction between it and the phenyl group of Phe or a non-classical hydrogen bond, resulting from its "side-on" position at the plane of the benzene ring of Phe 27,28 . Consequently, the dehalogenation is affected by the mutation of Phe to Ala by destroying this interaction. The residues Trp48, Gly50, Tyr134 and Leu288 indirectly interact with L-LA, but they participate in the hydrogen-bond network among the residues in the active site. Residues Trp48 and Tyr134 are strictly  conserved between HadD AJ1 and DL-DEXs. In agreement with the effects resulting from the mutation of the two residues in DL-DEXs 15 , HadD AJ1 mutants W48A and Y134F are almost inactive (Fig. 6A). Although Gly50 and Leu288 are replaced by the corresponding residues of DL-DEXs, the mutations damage 94.9% and 62.8% of HadD AJ1 activity, respectively (Fig. 6A). This result shows that the polarity and the steric hindrance of the residues at these positions have an impact on the dehalogenation catalyzed by HadD AJ1. These different contributions are likely to be caused by natural selection of the residues in the active site among D-DEXs and DL-DEXs.

Conclusions
By combining structural, biochemical and site-directed mutagenesis analyses, we have gained insights into the dehalogenation catalyzed by D-DEX. HadD AJ1 is wholly α-helical and was crystallized as a homotetramer. Each monomer contains two N-terminal and C-terminal repeats, linked by a loop with a 3 10 -helix η1. The substrate binding pocket of HadD AJ1 is locatedat the interface between two repeats. Residues lining the active pocket participate in the E-S interaction, the mutation of which severely impairs the enzymatic activity. Dehalogenation catalyzed by D-DEX is directly mediated by a nucleophilic water molecule activated by Asp205 and Asn131 without forming the E-S ester intermediate, which is the same as DL-DEX but different from L-DEX. These findings enrich the knowledge on haloacid dehalogenases.

Methods and Materials
Cloning X-ray diffraction analysis. X-ray diffraction data were collected on BL 18U1 and BL19U at Shanghai Synchrotron Radiation Facility (China). The diffraction data were processed and scaled using the HKL 2000 software package 29 . Then, the scaled output was converted into MTZ files using scalepack2mtz from the CCP4 suite 30 . The molecular replacement method was used to determine the structure of HadD AJ1 using Phaser-MR from the Phenix 31 suite with DehI (PDB: 3BJX) as a search model 1 . Model rebuilding was performed in Coot 32 . Subsequently, further rounds of restrained refinement were performed, and water molecules were added and adjusted using the Phenix refine program. Structure figures were prepared with PyMol (http://www.pymol. org). Sequence comparison was generated by ESPript 3.0 33 . The structural information of all crystals is shown in Table 1.
Site-directed mutagenesis of HadD AJ1. The residues involved in the substrate binding pocket were mutated. The residues Trp48 and Thr 76 which are strictly conserved between D-DEX and DL-DEX were mutated to Ala. The residues Gly50, Val51, Ile52, Asn203, Phe281, Leu285 and Leu288 which are not conserved between D-DEX and DL-DEX were mutated to the corresponding amino acids found in DehI PP3. The residues Asn131, Tyr134, Ser204, Asp205 and Met284 which are identical between HadD AJ1 and DL-DEX were mutated on the basis of their charge and polarity. The primers containing mutations were designed using DNAstar and are summarized in Supplementary Table S2 Enzymatic activity. The enzymes in Buffer D were used to assay the activity towards D-2-CPA. The standard assay system was used unless stated otherwise 6 . The reaction mixture (1 mL) contained 10 mM D-2-CPA, 100 mM glycine-NaOH (pH 10.0), and enzyme. To ensure that at least 5% of the substrate is degraded, the enzyme concentrations were 0.02 mg/mL for WT, V51F, S204T, F281Y, L285I and L288I, and 0.3 mg/mL for other mutants. The reactions were terminated by the addition of 10 µL of phosphoric acid (85% w/w) following the incubation at 30 °C. After the precipitates were removed by centrifugation (14,000 × g, 10 min), the supernatants were analyzed by HPLC to determine the D-2-CPA contents. One unit of dehalogenase activity was defined as the amount of enzyme that catalyzed the hydrolysis of 1 µmol D-2-CPA per min.
Steady-state kinetic measurements. Kinetic study of WT and several mutants was carried out by assaying the activity of the enzyme under standard condition at varying initial substrate concentrations. The initial rates of the reaction at different substrate concentrations were measured. Kinetic equation fitting was conducted using Origin 7 software. K m and k cat values were derived from the non-linear regression analysis ( Supplementary  Fig. S4).
Data Availability. All data generated or analysed during this study are included in this published article (and its Supplementary Information files).