Catalytic and structural insights into a stereospecific and thermostable Class II aldolase HpaI from Acinetobacter baumannii

Aldolases catalyze the reversible reactions of aldol condensation and cleavage and have strong potential for the synthesis of chiral compounds, widely used in pharmaceuticals. Here, we investigated a new Class II metal aldolase from the p-hydroxyphenylacetate degradation pathway in Acinetobacter baumannii, 4-hydroxy-2-keto-heptane-1,7-dioate aldolase (AbHpaI), which has various properties suitable for biocatalysis, including stereoselectivity/stereospecificity, broad aldehyde utilization, thermostability, and solvent tolerance. Notably, the use of Zn2+ by AbHpaI as a native cofactor is distinct from other enzymes in this class. AbHpaI can also use other metal ion (M2+) cofactors, except Ca2+, for catalysis. We found that Zn2+ yielded the highest enzyme complex thermostability (Tm of 87 °C) and solvent tolerance. All AbHpaI•M2+ complexes demonstrated preferential cleavage of (4R)-2-keto-3-deoxy-D-galactonate ((4R)-KDGal) over (4S)-2-keto-3-deoxy-D-gluconate ((4S)-KDGlu), with AbHpaI•Zn2+ displaying the highest R/S stereoselectivity ratio (sixfold higher than other M2+ cofactors). For the aldol condensation reaction, AbHpaI•M2+ only specifically forms (4R)-KDGal and not (4S)-KDGlu and preferentially catalyzes condensation rather than cleavage by ∼40-fold. Based on 11 X-ray structures of AbHpaI complexed with M2+ and ligands at 1.85 to 2.0 Å resolution, the data clearly indicate that the M2+ cofactors form an octahedral geometry with Glu151 and Asp177, pyruvate, and water molecules. Moreover, Arg72 in the Zn2+-bound form governs the stereoselectivity/stereospecificity of AbHpaI. X-ray structures also show that Ca2+ binds at the trimer interface via interaction with Asp51. Hence, we conclude that AbHpaI•Zn2+ is distinctive from its homologues in substrate stereospecificity, preference for aldol formation over cleavage, and protein robustness, and is attractive for biocatalytic applications.

Aldolases catalyze the reversible reactions of aldol condensation and cleavage and have strong potential for the synthesis of chiral compounds, widely used in pharmaceuticals. Here, we investigated a new Class II metal aldolase from the p-hydroxyphenylacetate degradation pathway in Acinetobacter baumannii, 4-hydroxy-2-keto-heptane-1,7-dioate aldolase (AbHpaI), which has various properties suitable for biocatalysis, including stereoselectivity/stereospecificity, broad aldehyde utilization, thermostability, and solvent tolerance. Notably, the use of Zn 2+ by AbHpaI as a native cofactor is distinct from other enzymes in this class. AbHpaI can also use other metal ion (M 2+ ) cofactors, except Ca 2+ , for catalysis. We found that Zn 2+ yielded the highest enzyme complex thermostability (T m of 87 C) and solvent tolerance. All AbHpaIM 2+ complexes demonstrated preferential cleavage of (4R)-2-keto-3-deoxy-D-galactonate ((4R)-KDGal) over (4S)-2-keto-3-deoxy-D-gluconate ((4S)-KDGlu), with AbHpaIZn 2+ displaying the highest R/S stereoselectivity ratio (sixfold higher than other M 2+ cofactors). For the aldol condensation reaction, AbHpaIM 2+ only specifically forms (4R)-KDGal and not (4S)-KDGlu and preferentially catalyzes condensation rather than cleavage by 40-fold. Based on 11 X-ray structures of AbHpaI complexed with M 2+ and ligands at 1.85 to 2.0 Å resolution, the data clearly indicate that the M 2+ cofactors form an octahedral geometry with Glu151 and Asp177, pyruvate, and water molecules. Moreover, Arg72 in the Zn 2+ -bound form governs the stereoselectivity/stereospecificity of AbHpaI. X-ray structures also show that Ca 2+ binds at the trimer interface via interaction with Asp51. Hence, we conclude that AbHpaIZn 2+ is distinctive from its homologues in substrate stereospecificity, preference for aldol formation over cleavage, and protein robustness, and is attractive for biocatalytic applications.
Aldolases catalyze reversible reactions of carbon-carbon bond formation (aldol condensation) and breakage (aldol cleavage). Based on their different catalytic mechanisms, aldolases can be classified into three groups, including pyridoxal 5 0 -phosphate (PLP)-dependent, Class I lysine-dependent, and Class II metal-dependent aldolases. PLP aldolases employ PLP as a cofactor to react with an amino-containing nucleophilic substrate to form a quasi-stable carbanion and an iminium intermediate. Class I lysine aldolases (also called Schiff baseforming aldolases) utilize an active lysine residue to form a Schiff base with an aldehyde/keto substrate to also result in an iminium intermediate. This imine intermediate is susceptible to C-C bond cleavage or formation. For Class II metal aldolases, the enzyme uses a divalent metal ion (M 2+ ) as a cofactor for substrate binding and stabilization of an enolate intermediate (1)(2)(3)(4)(5)(6), which allows the reaction to proceed through C-C bond formation or cleavage.
The most well-studied pyruvate-specific Class II metal aldolase is HpaI (EC 4.1.2.52) found in the p-hydroxyphenylacetate (HPA) degradation pathway in Escherichia coli (EcHpaI). The enzyme catalyzes the reversible aldol cleavage of 4-hydroxy-2-keto-heptane-1,7-dioate (HKHD) to form pyruvate and succinic semialdehyde (SSA). Crystal structures, steady-state kinetics, and substrate specificity of EcHpaI indicate that the enzyme exists as a hexamer (a dimer of trimers) in which each subunit can bind to an octahedral divalent metal ion such as Mg 2+ , Mn 2+ , or Co 2+ coordinated with substrates pyruvate (20,23,26,27,30,39,40). Results from quantum mechanics/molecular mechanics (QM/MM) calculations and site-directed mutagenesis studies indicate that Arg70 and His45 together with the M 2+ -bound apex water molecule are important for substrate specificity, C-C bond cleavage, and enolate stabilization (27,30,39,40). EcHpaI can also catalyze the aldol condensation of keto donors (pyruvate or 2ketobutyrate) and various types of aldehyde acceptors of different carbon chain lengths (C 2 -C 5 ) to generate the corresponding 4-hydroxy-2-ketoacids with preference toward a longer chain C 5 -aldehyde (pentaldehyde) rather than other short-chain aldehydes (23). Although EcHpaI can use a broad range of aldehydes, its reaction lacks stereospecificity (23). Therefore, a new aldolase with similar catalytic capability as EcHpaI but capable of catalyzing stereospecific reactions with thermostability would be a more preferred biocatalyst.
Our group has identified a new HpaI from the HPA degradation pathway in Acinetobacter baumannii (AbHpaI), which shares 59% amino acid sequence identity with EcHpaI ( Fig. S1) (41). Structures and catalytic properties of AbHpaI have never been investigated. In this work, we investigated the catalytic and biophysical properties of AbHpaI and found that the enzyme has biochemical and biophysical properties significantly different from EcHpaI and other enzymes in this class such as the use of Zn 2+ as a cofactor. We also showed that AbHpaI can catalyze stereospecific aldol condensation to synthesize pure (4R)-2-keto-3-deoxy-D-galactonate ((4R)-KDGal) without producing the contaminating 4S-isomer, demonstrating that AbHpaI can control the stereospecificity of aldol product formation. Steady-state kinetics indicate that the turnover number of aldol condensation to synthesize (4R)-KDGal was about 35-to 40-fold faster than that the cleavage, suggesting that the aldol condensation is a more favored direction of AbHpaI catalysis. Moreover, AbHpaI is also tolerant to various solvents and highly thermostable especially in the Zn 2+ -bound form in which T m is 87 C. We solved 11 X-ray structures of AbHpaI in complex with various M 2+ and substrates to elucidate the structural factors underlying the catalysis of AbHpaI. Structual analysis clearly explains why the 4R-isomer is more preferred over the 4S-isomer for cleavage and how different M 2+ cofactors affect the binding features of both substrates. Arg72 is the key residue governing the stereochemistry of AbHpaI. Together, these properties, which are quite different from EcHpaI, make AbHpaI attractive as a robust biocatalyst for aldol condensation to produce the stereospecific/stereoselective 4-hydroxy-2-ketoacid synthons for further preparation of APIs.

Identification of the native metal ion cofactor for AbHpaI
We first explored the selectivity of metal ion binding in AbHpaI and identified its native cofactor. Using inductively coupled plasma-optical emission spectroscopy (ICP-OES), which can detect a wide variety of metal ions simultaneously, alkaline earth ions (Ca 2+ and Mg 2+ ) and transition metal ions (Zn 2+ , Mn 2+ , and Ni 2+ ) were detected in the purified AbHpaI (Fig. 1A). Quantitative measurements indicated that Ca 2+ ion was the most prevalent, followed by Zn 2+ and Mg 2+ , while Mn 2+ and Ni 2+ were found in very low amounts (Fig. 1A). In contrast to the properties of EcHpaI, which could bind to three metal ions (Mn 2+ , Mg 2+ , Co 2+ ) with Co 2+ giving the highest activity (26), Co 2+ was not found in the purified AbHpaI.
As the presence of metal ions in the purified AbHpaI may not directly relate to the enzyme catalytic activity because their existence may depend on their availability in cells, we thus further investigated the effects of different metal ions on the catalysis of AbHpaI. First, the binding properties of these metal ions to apo-AbHpaI and in the presence of pyruvate substrate were determined. Apo-AbHpaI was reconstituted with each metal ion, namely Zn 2+ , Mn 2+ , Co 2+ , Ca 2+ , and Mg 2+ and ability of these M 2+ to bind to the apoenzyme was determined. Although Co 2+ was not found in the purified AbHpaI (Fig. 1A), we included Co 2+ in this study because it is a native cofactor of EcHpaI (26,27,39). The ICP-OES results (Fig. 1B) showed that the mole ratios of each of the reconstituted metal ions Zn 2+ , Mn 2+ , Co 2+ , Mg 2+ , and Ca 2+ to apo-AbHpaI varied from 1.1, 0.6, 0.8, 0.6, and 0.1, respectively, suggesting that AbHpaI has different affinities and preferences toward these metal ions. Unexpectedly, Ca 2+ has the lowest binding ability to AbHpaI after reconstitution, albeit Ca 2+ was the most detected ion in the purified enzyme. We further investigated the binding constant (K d ) of M 2+ and pyruvate in AbHpaIM 2+ and AbHpaIM 2+ pyruvate complexes using isothermal titration calorimetry (ITC). AbHpaI has three and sevenfold greater affinity for Zn 2+ binding over Co 2+ and Mn 2+ , respectively, while the K d of Mg 2+ and Ca 2+ to AbHpaI could not be determined (Table 1). Together with the finding that Zn 2+ has the highest mole ratios in metal ion reconstitution experiments and in the native purified enzyme, these results suggest that Zn 2+ is the native cofactor for AbHpaI.
As pyruvate alone cannot bind to apo-AbHpaI, we thus further explored the role of the five metal ions in facilitating the binding of pyruvate to AbHpaI by measuring the binding constant of pyruvate using ITC. We found that only the transition metal ions Zn 2+ , Co 2+ , and Mn 2+ could support the binding of pyruvate in which AbHpaICo 2+ has a three and fourfold higher affinity to pyruvate than AbHpaIZn 2+ and AbHpaIMn 2+ (Table 1). Notably, the K d values of pyruvate binding to AbHpaIMg 2+ and AbHpaICa 2+ could not be measured, indicating that pyruvate has poor affinity to these enzyme complexes. However, apparent kinetic results showed that AbHpaIMg 2+ could catalyze the aldol condensation reaction of pyruvate and D-glyceraldehyde with 1.3-fold slower than AbHpaIZn 2+ , while AbHpaICa 2+ could not (Table 2). These suggest that Zn 2+ , Co 2+ , Mn 2+ , and Mg 2+ but not Ca 2+ have properties relevant to being metal ion cofactors. In addition, our work here indicates that an enzyme in the pyruvate-specific Class II metal aldolases can use Zn 2+ as a catalytic cofactor.

Stereoselectivity of the AbHpaI aldol cleavage
To investigate the influence of metal ions on the stereoselectivity of the substrate stereoisomer for aldol cleavage, AbHpaIM 2+ complexes of Zn 2+ , Co 2+ , Mn 2+ , Mg 2+ , and Ca 2+ , prepared by equilibrating the apo-AbHpaI with excess M 2+ , were employed for catalyzing the aldol cleavage of substrates, namely (4R)-KDGal and (4S)-KDGlu (see details of chemical structures in Fig. S2). (4R)-KDGal and (4S)-KDGlu were chosen as model substrates because these compounds are only different in stereo-isoforms (R and S) at the cleavage site of the C 4 -hydroxyl (C 4 -OH) group. The reaction rate of substrate cleavage was measured by coupling with the reaction of AbHpaIM 2+ with lactate dehydrogenase (LDH) to detect NADH oxidation upon pyruvate formation.
Regarding the overall yield of (4S)-KDGlu cleavage (Fig. 2B), 15 to 20% (4S)-KDGlu could be cleaved by AbHpaICo 2+ and AbHpaIMn 2+ , while only 1% cleavage could be catalyzed by AbHpaIZn 2+ . These results illustrate an interesting property of AbHpaIZn 2+ as a biocatalyst, because this form of enzyme exhibits strong stereoselectivity toward the 4R-isomer. Altogether, the data obtained from both (4R)-KDGal and (4S)-KDGlu cleavage reactions confirmed that AbHpaIM 2+ prefers to cleave the 4R-isomer over 4S and the enzyme stereoselectivity is metal-dependent, with AbHpaIZn 2+ exhibiting the highest R/S stereoselectivity ratio, 5to 9-fold higher  Figure 2. Stereochemistry of aldol cleavage and condensation of AbHpaI. Activities of AbHpaI without (black) and with metal ion cofactors Zn 2+ (blue), Co 2+ (green), Mn 2+ (red), Mg 2+ (orange), and Ca 2+ (pink) in aldol cleavage and condensation reactions were analyzed to determine the reaction stereochemistry. In the aldol cleavage reactions, either (A) (4R)-KDGal or (B) (4S)-KDGlu was used as a substrate in the LDH-coupled assay for AbHpaI activity. The relative absorbance decrease at 340 nm refers to the NADH oxidation upon conversion of pyruvate (generated from the AbHpaI cleavage reaction) to form L-lactate by LDH. Therefore, the NADH oxidation can be used to represent the cleavage reaction of the 4R-and 4S-isomer substrates, depending on the substrate added. Rates of substrate cleavage by individual AbHpaIM 2+ compelexes were determined and summarized in Table 2. C, products from the AbHpaIM 2+ catalyzed aldol condensation of pyruvate and D-glyceraldehyde were analyzed by a triple-quadrupole LC/MS to detect aldol products of (4R)-KDGal and (4S)-KDGlu. All AbHpaIM 2+ complexes, except AbHpaICa 2+ , can catalyze aldol condensation to form (4R)-KDGal. Rates of (4R)-KDGal formation by individual AbHpaIM 2+ were determined and summarized in Table 2. Error bars represent standard deviations (S.D.) from three replications of the data. than the Co 2+ -, Mn 2+ -, and Mg 2+ -bound AbHpaI enzymes ( Table 2).

Stereospecificity of AbHpaI aldol condensation
To explore whether AbHpaIM 2+ enzymes have stereoselectivity toward the synthesis of the 4R-isomer in the aldol condensation reaction similar to the aldol cleavage reaction, we examined the product stereoisomer resulting from the condensation of pyruvate and D-glyceraldehyde catalyzed by AbHpaIM 2+ enzymes. In order to differentiate between the two stereoisomeric compounds, (4R)-KDGal and (4S)-KDGlu, which have the same molecular mass of 178.0 g/mol, we used a high-sensitivity LC/MS system (Hi-Plex H cation-exchange column and triple quadrupole MS in a negative mode), which can distinguish between the two compounds (Fig. S3). Peak area of the product at m/z 177.0 ([M-H] − ) was used for calculating the reaction yield.
The results obtained from aldol condensation of pyruvate (4 mM) and D-glyceraldehyde (30 mM) showed that only (4R)-KDGal was produced by all AbHpaIM 2+ complexes of Zn 2+ , Co 2+ , Mn 2+ , and Mg 2+ , while the reaction of AbHpaICa 2+ could not produce any product even after 60 min ( Fig. 2C and Table 2). This was therefore confirmed that Ca 2+ cannot promote pyruvate binding, consistent with the ITC binding result of pyruvate and AbHpaICa 2+ . Similar to the cleavage rates, the Co 2+ containing enzyme catalyzes formation of (4R)-KDGal ≥ 2-fold faster than the Mn 2+ , Zn 2+ , and Mg 2+ -bound enzymes ( Table 2). It should be noted that 0.1% (4S)-KDGlu formation could also be detected after 30 h (Fig. S4). The results clearly showed that AbHpaIM 2+ prefers catalyzing stereoselective synthesis of the 4R-isomer, especially for the Zn 2+ -containing enzyme. Taken together, these findings show that AbHpaI can practically catalyze stereospecific product formation.

Steady-state kinetics of AbHpaIZn 2+
To understand the kinetics properties of the enzyme, steady-state kinetic parameters for aldol cleavage and condensation of AbHpaIZn 2+ were determined as listed in Table 3 and Fig. S5. Results indicate that the Michaelis-Menten constant (K m ) of (4R)-KDGal cleavage was half of that for (4S)-KDGlu, suggesting that AbHpaIZn 2+ requires lower concentrations of (4R)-KDGal to reach the maximum velocity of the reaction. The k cat value suggests that AbH-paIZn 2+ catalyzes the cleavage of (4R)-KDGal 18-fold faster than that of (4S)-KDGlu. In addition, the catalytic constant (k cat /K m ) of (4R)-KDGal cleavage was about 36-fold greater than that of (4S)-KDGlu, indicating that AbHpaIZn 2+ catalyzes the aldol cleavage of (4R)-KDGal more efficiently than that of (4S)-KDGlu. These results agreed well with the activities measured in Table 2 in that AbHpaIZn 2+ is highly stereoselective toward 4R-isomer.
For the aldol condensation kinetics, only the kinetics of (4R)-KDGal synthesis was investigated because (4S)-KDGlu could not be detected ( Table 2). The results in Table 3 showed that the turnover number of (4R)-KDGal synthesis was 100 min −1 and the K m value of pyruvate was approximately sevenfold lower than that of D-glyceraldehyde, indicative for higher affinity of pyruvate to the enzyme. Together, the kinetics results from both aldol cleavage and condensation reactions firmly support that AbHpaI is stereoselective for the 4R-isomer. Further comparing the kinetics of (4R)-KDGal cleavage versus synthesis, it was interesting to note that the condensation reaction was much faster than the cleavage, as its turnover number was 40-fold greater. This property is interesting for AbHpaI application as a biocatalyst because stereospecific aldol condensation is useful for preparation of APIs.
Each protomer contains eight β/α motifs of a TIM barrel fold with an additional α-helix (residues 6-11; N-helix) at the  a Kinetics of the cleavage reaction was investigated in buffer H containing 0.1 mM Zn 2+ , 0.05 to 2 mM (4R)-KDGal and 0.1 to 3.2 mM (4S)-KDGlu, and 5 or 40 μM AbHpaIZn 2+ (for (4R)-KDGal and (4S)-KDGlu, respectively). The amount of substrate was measured by RapidFire high-throughput mass spectrometry coupled with triple-quadrupole mass spectrometer. b The condensation kinetics were analyzed by monitoring product formation using LC coupled with triple-quadrupole mass spectrometry. c The reactions were carried out in 0 to 8 mM pyruvate with a fixed concentration of 30 mM D-glyceraldehyde. As pyruvate concentrations greater than 4 mM showed inhibition, the kinetics data at >4 mM pyruvate were not included in the analysis. d Due to pyruvate inhibition at >4 mM, the reactions were carried out in 0 to 36 mM D-glyceraldehyde at 4 mM pyruvate.
Values in parentheses are for the highest resolution shells. R f = Σ hkl ||F obs | − |F calc ||/Σ hkl |F obs |, where F obs and F calc are the observed and calculated structure-factor amplitudes, respectively. R free was calculated in the same manner as R f but using only a 10% unrefined subset of the reflection data. a Data were processed with Proteum3 except datasets of Co 2+ Pyr and Mn 2+ Pyr using HKL2000.
N-terminus (Fig. 3B). Three protomers are associated in a tight trimeric structure to generate three catalytic pockets, and the trimer is dimerized to form a stable and rigid hexamer with 35% buried area. The hexameric quaternary structure of AbHpaI is stabilized through extensive interactions from αhelices of the (β/α) 1 (residues 27-38), (β/α) 2 (residues 53-64), and (β/α) 8 (residues 236-253). The α-helix (residues 236-253) of the (β/α) 8 of each protomer docks on the surface of the neighboring trimer for holding the trimer of dimers. In the structures of AbHpaI, a C-terminal end at residues 254 to 266 was not built because of no electron density. Each active site, located on a side face of the trimer, was built from two protomer subunits with a main catalytic pocket located in one subunit accompanied by a loop linker of residues 110 to 136 between (β/α) 4 and (β/α) 5 motifs of the shared protomer as a pocket periphery (Fig. 3B).

Metal ion octahedral coordination in AbHpaI is important for AbHpaI reactivity and stability
Understanding how AbHpaI accommodates the M 2+ cofactor could yield biochemical insights into substrate recognition, reactivity, and stereospecificity control by different types of metal ions. Therefore, the coordination geometry of the M 2+ cofactor in each complex obtained was analyzed. The crystal structures of AbHpaI complexed with metal ions Zn 2+ , Co 2+ , or Mn 2+ and pyruvate (PDB codes 7ET9, 7ETA and 7ETB) revealed all types of M 2+ chelate to a carboxyl group of Asp177, a water molecule (W S1 ), and a carboxyl and 2-oxo groups of pyruvate in a square planar arrangement, and with a carboxyl group of Glu151 and a water molecule (W A ) in an axial position, arranged in an octahedral coordination geometry (Fig. 3C). Besides Glu151 and Asp177, Glu46 and His47 on the (β/α) 2 loop also provide watermediated hydrogen bondings via W A and W S1 . Superposition of all AbHpaI complex structures revealed that the bound M 2+ cofactor is at the same position with six atoms in octahedral geometry.
Notably, Glu151 and Asp177 in the Zn 2+ , Co 2+ , or Mn 2+ complexes were more rigid than those of apo-AbHpaI, as reflected by temperature factors (B-factors) of the crystal structure. In the AbHpaIM 2+ complex structure, the B-factors of the (β/α) 4 and (β/α) 6 loop regions including the helix α 6 where Asp177 is located are smaller than that of the apo structure. This suggests that M 2+ can reduce the mobility in this region and strengthen subunit compactness, thereby stabilizing the overall structural architecture. In addition, the formation of the M 2+ octahedral coordination with pyruvate in AbHpaI is important for enzyme reactivity.
Ca 2+ ion neutralizing negatively charged Asp51 at the AbHpaI trimer surface facilitates dimerization of subunits Ca 2+ ion was found abundantly in purified AbHpaI (Fig. 1A). However, it does not act as a cofactor to enhance enzyme catalysis (Table 2). Therefore, the function of Ca 2+ ion in AbHpaI was further investigated by analyzing the enzyme structure in complex with Ca 2+ . Figure 3. X-ray structures of AbHpaI. A, the quarternary hexameric structure of AbHpaI is the result of dimerization of two trimers of AbHpaI. One trimer is shown in ribbons docked on the second trimer shown as a surface model. B, the AbHpaI dimer shows the subunit TIM barrel fold and the active pocket periphery-residues 110 to 136 in yellow contributed from the supporting subunit in red-of the active subunit in green. C, the pocket of Zn 2+ and pyruvate (PYR) binding in Subunit A (green) is located at the interface between Subunit A and B (red and yellow). Two water molecules are drawn as red spheres. The inset shows the distances of Zn 2+ with octahedral coordination in AbHpaIZn 2+ PYR. Ca 2+ ion is bound at the AbHpaI trimer neutralizing Asp51 negative charges. The distance between the catalytic Zn 2+ and Ca 2+ is 23.3 Å.
The crystal structure of apo-AbHpaI (PDB code 7ET8) crystallized in the presence of CaCl 2 only showed Ca 2+ at the defined trimer center (not in the active site) in a distorted octahedral geometry with three Asp51 side chains (2.3-2.4 Å) and three water molecules as observed in all structures of AbHpaI studied here (Figs. 3C and S9). These data, together with the nonfunctional role of Ca 2+ discussed above, confirmed that Ca 2+ does not serve as a cofactor, but rather acts as a stabilizing factor on the dimerization surface of the AbHpaI trimer by neutralizing the negative charges of Asp51 on the trimer surface. In EcHpaI, the equivalent position to Asp51 was found to be Asn (Asn48 Ec ), thus abolishing the ability of this enzyme to bind to a divalent metal ion. Therefore, Ca 2+ functions to prevent repulsive forces and to reduce movement of the (β/α) 2 loop in AbHpaI where Asp51 sits. Consequently, by stabilizing the (β/α) 2 loop where the active residues Glu46 and His47 reside (see proposed mechanisms), Ca 2+ could indirectly aid in the catalysis of AbHpaI.

Insights into stereoselectivity of AbHpaI in the aldol cleavage reaction
To gain insights into why AbHpaI significantly prefers the 4R-isomer over the 4S-isomer in the aldol cleavage reaction (Table 2), crystal structures of AbHpaIZn 2+ in complex with (4R)-KDGal (PDB code 7ETC) and (4S)-KDGlu (PDB code 7ETD) were determined at 1.95 and 1.90 Å resolutions, respectively. Superimposed structures illustrated that for the pyruvate core, the 1-carboxyl and 2-oxo groups of both substrates are directly coordinated to the Zn 2+ site in an octahedral geometry similar to that found during pyruvate binding (Figs. 3C and 4A). However, the major differences are at the C 4 -OH, which interacts with Arg72 and at the binding site of the D-glyceraldehyde moiety. The structures revealed that the 4-OH of (4R)-KDGal forms a hydrogen bond with a guanidinium side chain of Arg72 at a 3.1 Å distance, whereas that of (4S)-KDGlu interacts with a longer distance (3.5-3.6 Å) (Fig. 4A). This therefore affected the arrangement and interactions of the D-glyceraldehyde moieties in the two compounds. The 5-OH and 6-OH functional groups of the Dglyceraldehyde moiety of (4R)-KDGal form hydrogen bonds with the main chains of Val 0 120 (2.6 Å) and Ala 0 122 (3.5 Å) from the pocket site created by the neighboring subunit, while those of (4S)-KDGlu do not form such interactions. Clearly, the observed interaction differences implied that AbH-paIZn 2+ could preferably bind (4R)-KDGal over (4S)-KDGlu, further supported by QM/MM MD simulations. The binding energies of (4R)-KDGal and (4S)-KDGlu were calculated as −158 ± 9 and −131 ± 8 kcal/mol, respectively (Table S2). Furthermore, the configuration of (4R)-KDGal bound in the AbHpaIZn 2+ complex can provide more suitable orientation and decreased motion of the substrate for the aldol cleavage in contrast to the bound (4S)-KDGlu. The structural analysis clearly supports the cleavage activity of the Zn 2+bound enzyme toward the 4R over the 4S substrates (Tables 2  and 3).
In addition, to understanding why the Mg 2+ cofactor gave such a slow cleavage rate, the crystal structure of the AbHpaIMg 2+ (4R)-KDGal complex (PDB code 7ETE) solved at 1.95 Å resolution was compared with the structure of the AbHpaIZn 2+ (4R)-KDGal complex. Superimposed structures revealed a significant difference at the pyruvate core linked to the metal ion cofactors (Fig. 4B). A water W S2 replaced the carboxyl group of (4R)-KDGal to join the Mg 2+ octahedral coordination. This feature gives rise to a longer distance between the 4-OH and Arg72 side chain in the Mg 2+ complex (3.4−3.6 Å), causing a weaker binding interaction compared with the Zn 2+ complex, thereby decelerating the C 3 -C 4 bond breakage ( Fig. 4B and Table 2). Based on structural and kinetics analyses, it could be summarized that the key binding features of the ligand important for stereoselective control in the aldol cleavage reaction of AbHpaI were (i) C 4 -OH anchoring by Arg72, (ii) interaction of the substrate pyruvate moiety in an octahedral M 2+ complex, and (iii) interactions of the aldehyde moiety with the neighboring subunit.  1 Å), whereas the C 4 -OH of (4S)-KDGlu is in longer distance (3.5−3.6 Å), making it less susceptible for C 3 -C 4 bond cleavage. There are also differences in the interactions from the supporting subunit in which the C 5 -OH and C 6 -OH of (4R)-KDGal are directly stabilized by hydrogen bonding with Val 0 120 and Ala 0 122, while the hydroxyl groups of (4S)-KDGlu are more than 6 Å away from these residues. In (B), binding interactions of Mg 2+ (4R)-KDGal are quite similar to the Zn 2+ (4R)-KDGal except that the COOH in the Mg 2+ (4R)-KDGal complex does not directly chelate Mg 2+ , but it is replaced by water (W S2 ), which is mediated by hydrogen bonding via the Mg 2+ octahedral coordination.

Biocatalytic aspects of AbHpaI
AbHpaI catalyzes the aldol condensation reactions with broad aldehyde specificity To investigate whether AbHpaIZn 2+ can use various aldehydes as substrates in aldol condensation with pyruvate, we screened different categories of aliphatic and aromatic aldehydes. The corresponding aldol products were analyzed using liquid chromatography with high-resolution mass spectrometry to measure the exact m/z in a negative mode. The results showed that the selected aldehydes could be used by AbH-paIZn 2+ in aldol condensation with pyruvate to yield various products (m/z values shown in Tables 5 and S3, and Fig. S11). The derivatives of aliphatic aldehydes with C 3 -C 6 chain length could be successfully converted into the corresponding 4hydroxy-2-keto aliphatic acids (Tables 5 and S3, and Fig. S11). Next, we examined with aromatic aldehydes and found that AbHpaIZn 2+ can catalyze aldol condensation of pyruvate with various aromatic aldehydes such as benzaldehyde, HBA, and anisaldehyde, to yield the corresponding 4hydroxy-2-keto aromatic acids (Tables 5 and S3, and Fig. S11). From our data, AbHpaIZn 2+ can use a wide range of aldehyde substrates, suggesting that the enzyme active site is flexible enough to accommodate a variety of aldehydes for aldol condensation.

Space for accommodating various aldehydes in the active site of AbHpaI
To understand how the AbHpaI can accommodate various aldehydes for aldol condensation, we investigated the binding interactions of three aldehydes from the cocrystal structures of AbHpaIMn 2+ PYR and AbHpaICo 2+ PYR with SSA (PDB codes 7ETF and 7ETG, respectively), and AbHpaIZn 2+ PYR with PPA and HBA (PDB codes 7ETH and 7ETI, respectively).
Overlaid structures of AbHpaIMn 2+ PYRSSA and AbH-paICo 2+ PYRSSA revealed that at the aldehyde binding site, the carbonyl group of SSA forms hydrogen bonds with Arg72 (2.7−2.9 Å) and the apex W A water (2.6-2.9 Å) in the M 2+ octahedral coordination, while the carboxyl tail of SSA is hydrogen bonded to the Ala 0 123 or Ala 0 122 NH backbone of the nearby subunit and to a water network via W 3 or W 4 in the pocket tunnel filled with waters ( Figs. 5A and 6A, see later). In addition, a distance between the C 1 atom of SSA and the C 3 methyl group of PYR was in the range of 3.1 to 4.1 Å (Fig. 6A, see later). We found that both metal ion complexes provide a similar binding of PYR but a slightly different configuration of SSA (Fig. 5A). This indicated that space for accommodating aldehyde substrate is larger than a van der Waals sphere of SSA, hence with a flexible hydrocarbon backbone, two configurations of SSA can be docked (Fig. 5A).
Although the substrate binding pocket of AbHpaI is wide open and exposed to outside solvent on the protein surface, the opening narrows down to the bottom of the active site where the reaction takes place. The site for aldehyde docking appeared to be hydrophobic, as most of the residues lining the site are nonpolar except for Arg72 (Fig. 5). Next, we explored the crystal structure of the AbHpaIZn 2+ PYR complex liganded with PPA, which is a more hydrophobic ligand than SSA. PPA, which has only a polar carbonyl moiety, can dock on a hydrophobic cleft, composed of Trp21, Leu214, Val236 of the active subunit, and Leu 0 124 from the shared subunit (Fig. 5, B and C) and arranges the carbonyl moiety to hydrogen  Catalytic and structural studies of a thermostable AbHpaI bond with Arg72 in a reactive trajectory for condensation with PYR, while the pocket tunnel is still full of water. However, the number of long-chain hydrocarbons in the aldehydes may be limited due to substrate solubility if the reaction is carried out solely in an aqueous environment because the long-chain hydrocarbon would protrude out of the hydrophobic cleft toward the water milieu in the direction similar to the SSA trajectory (Fig. 5, A and C). Thus, the nonpolar aldehydes may not be able to move along the water tunnel to reach the catalytic site located deep within the protein. For aromatic aldehydes, HBA was chosen as a representative to explore the binding mode. The cocrystal structure of the AbHpaIZn 2+ PYRHBA complex shows that the Trp21 indole ring, Leu214, and Val236 stabilize the HBA benzene ring through van der Waals interactions, while the carbonyl moiety forms a hydrogen bond to Arg72, which may be crucial for aldol condensation with PYR (Fig. 5, B and C). This structural analysis confirmed that AbHpaI can accommodate aromatic aldehydes well, as long as they can pass through a polar environment to get inside the active pocket.

Proposed mechanism for aldol reactions catalyzed by AbHpaI
The results from structural analysis reveal several structural water molecules at the active site potentially involved in AbHpaI catalysis (Fig. 6). A possible model for the AbHpaI aldol cleavage mechanism is proposed in Figure 7A. The cleavage of a C 3 -C 4 bond in the (4R)-KDGal substrate to yield pyruvate and D-glyceraldehyde products is facilitated by Glu46, His47, and Arg72 together with bound water molecules, W A , W 1 , and W S1 , to deprotonate the 4-OH leading to the bond cleavage to form an enolate intermediate, which then abstracts a proton from W A to yield a pyruvate.
For aldol condensation, via pyruvate carboxylate mediation, a water W 2 likely acts as a catalytic base to abstract a proton from the C 3 -methyl of pyruvate in a similar reaction to that of Glu46, His47, and Arg72 with three waters W A , W 1 , and W S1 to generate the enolate intermediate, which then forms a covalent linkage with the C 1 atom of SSA to produce HKHD (Fig. 7B). This model is supported by a pK a value of pyruvate C 3 -methyl of 6.5 (previously estimated by a pD-profile of pyruvate C 3 proton exchange reaction of EcHpaI (39)).

Thermal and solvent stability
Thermostability and organic solvent tolerance are important requirements for biocatalytic applications (42). Therefore, we determined the effect of M 2+ on the thermostability of AbHpaI using thermofluor stability measurements. The results showed that apo-AbHpaI is quite thermostable naturally with a protein melting temperature (T m ) value as high as 81.3 C. The binding of transition M 2+ , but not alkaline earth M 2+ , can further increase the thermostability of apo-AbHpaI (Table 1). The T m values of AbHpaI were enhanced by 5.7, 3.4, and 1.7 C upon binding to Zn 2+ , Co 2+ , and Mn 2+ , respectively. In contrast, the change of T m values for AbHpaIMg 2+ and AbHpaICa 2+ was negligible. Consequently, the result suggested that Zn 2+ enhances the thermostability of AbHpaI, the highest among all types of AbHpaIM 2+ enzymes. Apart from the T m value, which represents the protein stability, we also measured the remaining activity of AbHpaIZn 2+ incubated at increasing temperatures (25-85 C) for various incubation times (0-24 h) to examine the thermostability of AbHpaI. The result in Figure 8A showed that AbHpaIZn 2+ can tolerate a wide range of temperatures from 25 to 75 C for 24 h (and possibly longer) and can tolerate 80 C for up to 2 h without activity loss.
We then explored the effects of Zn 2+ on solvent tolerance of AbHpaI. As most substrates of aldolase reactions such as aliphatic and aromatic aldehydes are not soluble well in aqueous phases, addition of organic solvents is required to enhance substrate solubility. Therefore, we determined the T m of AbHpaIZn 2+ complex in the presence of organic solvents to represent enzyme stability. The organic solvents generally used in industries were chosen in this study, namely polarprotic (methanol, MeOH; ethanol, EtOH; and isopropanol, IPA) and polar-aprotic (acetronitrile, ACN; and dimethylsulfoxide, DMSO). The T m results showed that upon addition of 20% (v/v) MeOH, ACN, and DMSO, the protein stability of AbHpaI was perturbed by only 10% (Fig. 8B), while EtOH and IPA disrupted protein stability by about 30 to 40%. This implied that 20% (v/v) of MeOH, can, or DMSO can be used as a cosolvent to enhance substrate solubility with small perturbation in protein stability of AbHpaI. Taken together, these findings suggest that AbHpaI is thermostable and solventtolerant enzyme, which can be a promising robust biocatalyst for aldol reaction.

Discussion
Our report here has shown that AbHpaI is distinct among pyruvate-specific Class II metal aldolases for its ability to catalyze stereospecific aldol condensation and to use Zn 2+ as a cofactor. Zn 2+ binding can enhance stereoselectivity in aldol reactions and enzyme thermostability. Comprehensive structural investigation of AbHpaI complexes can explain how the enzyme is more stereoselective toward (4R)-KDGal over (4S)-KDGlu and how a variety of aldehydes can be accommodated.
Zn 2+ is the most abundant transition metal ion found in the purified AbHpaI, binds with the highest affinity, and significantly increases the substrate stereoselectivity and stability of AbHpaI (Figs. 1, 2, 4, 8, S4, S5 and Tables 1-3, S2). In addition, the enzyme can use other divalent ions (Co 2+ , Mn 2+ , and Mg 2+ ) as cofactors, but these metal ions do not mediate AbHpaI stereoselective aldol cleavage functions as effectively as Zn 2+ . This property is different from other HpaI enzymes such as Figure 6. Key residues and structural waters at the active site. A, AbHpaIMn 2+ PYRSSA (PDB code 7ETF) and (B) AbHpaIZn 2+ (4R)-KDGal (PDB code 7ETC) structures are presented on the left and a simple schematic diagram of each structure is on the right. Water molecules are labeled as W A , W S1 , W 1 , and W 2 . Dash lines display H-bonds and distances between atoms. H-bond distances are designated in black digits, while the distances between W A to 4-OH of (4R)-KDGal and W 2 to C 3 of PYR and C 1 of SSA to C 3 of PYR are in blue.
Catalytic and structural studies of a thermostable AbHpaI those from E. coli (EcHpaI) and Sphingomonas wittichii RW1 (SwHpaI), which cannot use Zn 2+ , instead use Co 2+ , Mn 2+ , and Mg 2+ as cofactors (27,30,33,39,40). Zn 2+ binds to AbHpaI in an octahedral coordination with six chelating atoms consisting of the pyruvate core, Glu151, Asp177, and two water molecules (W A and W S1 ), so does the geometry of Co 2+ , Mn 2+ , and Mg 2+ (Figs. 3-6), which differs from Zn 2+ tetrahedral coordination commonly found in aldolase and nonaldolase enzymes (43)(44)(45)(46). However, the Zn 2+ octahedral coordination is similar to those found in E. coli RhuA, a DHAP-specific Class II metal aldolase (47). We noted an interesting cooperative enhancement of ligand (pyruvate) binding to AbHpaI in the presence of Zn 2+ , Co 2+ , and Mn 2+ (Figs. 4 and 5 and Table 1) while weaker interaction or none was observed with Mg 2+ and Ca 2+ . This therefore suggests that the active site is more rigid upon cofactor binding, thereby enhancing the binding of pyruvate and the cleavage activities of (4R)-KDGal (Table 2). Zn 2+ cofactor also promotes the highest R/S stereoselectivity ratio in the AbHpaI aldol cleavage. The preference of (4R)-KDGal over (4S)-KDGlu cleavage by AbHpaIZn 2+ is approximately sixfold greater than the reactions of other M 2+ cofactors ( Table 2 and Fig. 2, A and B). QM/MM MD calculations gave higher favorable binding energy of (4R)-KDGal leading to a more stable complex of AbHpaIZn 2+ (4R)-KDGal better poised for cleavage, explaining the stereoselectivity of AbHpaIZn 2+ (Table S2).
Structural analysis of AbHpaIZn 2+ (4R)-KDGal and AbHpaIZn 2+ (4S)-KDGlu complexes suggested that Arg72,Val 0 120, Ala 0 122, and Ala 0 123 on the pocket border from the nearby subunit (which also defines the pocket size) are key factors for stereoselectivity via facilitating stronger interactions with the 4-OH and D-glyceraldehyde moiety of (4R)-KDGal over (4S)-KDGlu (Fig. 4A). The preferred orientation of the C 4 -OH cleavage site of (4R)-KDGal binding to The key residues Glu46, His47, and Arg72 together with bound water molecules W A , W S1 , W 1 , and W 2 facilitate aldol cleavage and condensation reactions. The aldol cleavage mechanism begins with a cascade of proton abstraction mediated by His47 and W A , general acid/ base. The abstraction of a C 4 -OH proton mediated by W A results in a C 3 -C 4 bond cleavage of (4R)-KDGal, which can be stabilized by Arg72, to generate D-glyceraldehyde and enolate. The enolate intermediate then abstracts a proton from W A , which can be facilitated by His47 and nearby H-bond networks of W 1 , Glu46, and W S1 . For the aldol condensation mechanism, W 2 water mediated by the pyruvate carboxylate anion is proposed to act as a catalytic base to abstract a proton from the C 3 -methyl group of pyruvate to yield an enolate intermediate. The carbanion C 3 of the enolate attacks the C 1 of SSA to yield HKHD, followed by a protonation step from W A to yield a C 4 -OH.
Zn 2+ thus gives rise to a shorter distance between W A water (a catalytic base) and 4-OH of (4R)-KDGal (3.6 Å), compared with that of (4S)-KDGlu (3.9 Å) (Fig. 6A). Comparison of the catalytic pockets between AbHpaI and EcHpaI (As EcHpaI was solved as one protomer per asymmetric unit, thus the dimer was generated by symmetry operation) showed that the EcH-paI pocket was 0.3 Å wider than that of AbHapI. Moreover, HKHD bound in EcHpaI (PDB code 4B5V) was found in two isomeric forms, which both interact with Arg70 and the nearby subunit residues similar to the case of AbHpaI. A wider binding pocket in EcHpaI may be the cause of the lacking stereoselectivity in this enzyme (23,39). Altogether, molecular interactions between AbHpaIM 2+ and ligand, particularly hydrogen bonding with the aldehyde moiety of substrate governed by Arg72 and coordination of the pyruvate core in the M 2+ cofactor geometry described above, promote stereoselectivity and stereospecificity in the AbHpaIM 2+ -catalyzed aldol cleavage reaction.
AbHpaI can catalyze the aldol condensation with stereospecificity and use broad aldehyde spectrum (Tables 2 and 5 and Fig. S11). Our kinetics demonstrated that AbHpaIZn 2+ can synthesize only (4R)-KDGal from pyruvate and D-glyceraldehyde. This finding suggests that AbHpaI has a stereospecific control over aldol condensation reactions. The crystal structures showed that AbHpaI can bind both aliphatic and aromatic aldehydes (PPA, SSA, and HBA) with a proper chemical space to satisfy van der Waals interactions between the substrate and hydrophobic residues (Trp21, Leu214, Val236, and Leu 0 124) (Fig. 5). Recently, aromatic substituted aldehydes have been reported in two aldolases, YfaU (a pyruvate-specific Class II metal aldolase) and 2-keto-3-deoxy-6-phosphogluconate (KDPG) aldolase (a pyruvate-specific Class I lysine aldolase); however, their crystal structures were not available (5,25). Nonetheless, the molecular dockings were performed and showed that N-benzyloxycarbonyl (N-Cbz)substituted aldehydes are surrounded by hydrophobic residues Trp23, Phe174, and Leu216, similar to the aldehyde binding residues identified in AbHpaI. These properties offer an opportunity for AbHpaI to serve as a biocatalyst to catalyze formation of 4-hydroxy-2-ketoacid and 2-keto-3-deoxy-Dsugar acid precursors valuable for the synthesis of APIs (48,49).
Beyond promoting the aldol reactivity and the binding of pyruvate, Zn 2+ also enhances the thermostability of AbHpaI. The enzymatic activity of AbHpaIZn 2+ could be retained under very high temperature, i.e., 80 C with a half-life (t 1/2 ) at 3 h (Fig. 8A). Moreover, AbHpaIZn 2+ is also stable in the presence of organic solvents such as MeOH, can, and DMSO up to 20% (v/v) (Fig. 8B). The data indicate that not only does Zn 2+ serve as a catalytic cofactor, but it is also involved in the quaternary structure stabilization of AbHpaI. Apart from the catalytic Zn 2+ , our structural data demonstrates that Ca 2+ is found at the trimer center on the dimerization interface of the AbHpaI hexamer, neutralizing the negative charges of three Asp51 carboxylate side chains (Fig. S9), which is unique for AbHpaI and is not found in EcHpaI, SwHpaI, other pyruvatespecific Class II metal aldolases (23,33,39).
In conclusion, our results here provide insightful mechanistic and structural understanding in stereoselectivity/stereospecificity control in aldol cleavage and condensation of AbHpaI. As the enzyme has broad aldehyde substrate specificity, high thermostability, and solvent tolerance, the insightful knowledge obtained from this study will serve as a basis for future rational protein engineering of AbHpaI and also other HpaIs in the pyruvate-specific Class II metal aldolases to achieve the capability to synthesize tailor-made, optically pure 4-hydroxy-2ketoacid synthons required for preparation of APIs.

Chemicals and reagents
All chemicals were commercially available and of analytical, high purity, and HPLC grades. Buffers used in this work were (i) Expression, purification, activity assay, and oligomeric state of AbHpaI Recombinant AbHpaI was overexpressed in E. coli BL21(DE3) as previously described (41). Unless otherwise indicated, purification of AbHpaI was carried out at 4 C. The cell paste (24 g obtained from 7.8 l culture) was resuspended in buffer A, and cells were then disrupted by ultrasonication. The broken-cell suspension was centrifuged at 36,000g for 40 min, and the clarified supernatant was collected as crude extract. Polyethyleneimine (PEI), at a final concentration of 0.5% (w/v), was added to the crude extract to remove nucleic acid contents. After centrifugation, the clarified supernatant was fractionated with 20 to 40% (w/v) ammonium sulfate ((NH 4 ) 2 SO 4 ) saturation. The protein pellet was resuspended in buffer B and dialyzed in the same buffer for 16 to 18 h. After dialysis, the dialysate was clarified by centrifugation before loading onto a DEAE-Sepharose column (172 ml, 2.5 cm × 35 cm) pre-equilibrated with buffer B. The column was washed with buffers B and C, respectively, and then eluted with a linear gradient of buffers C and D. Fractions containing AbHpaI were pooled and concentrated by ultrafiltration. The enzyme solution was further purified on a Phenyl-Sepharose column (45 ml, 1.5 cm × 25 cm) pre-equilibrated with buffer E. After loading the enzyme solution, the column was washed with buffer E and then eluted with a linear gradient of buffers E and A. Fractions containing AbHpaI were pooled and concentrated as described above. The concentrated enzyme solution was exchanged into buffer F using a Sephadex G-25 column. The concentration of the purified AbHpaI was determined using the molar absorption coefficient of 30,035 M −1 cm −1 at absorbance 280 nm (A 280 ), which was calculated from the deduced amino acid sequence using the online tool on the ProtParam program of ExPaSy Proteomics Server (http://web.expasy.org/protparam/). The aliquots of enzyme solution were then stored at −80 C until used. The amount of protein was determined by Bradford assay using BSA as a protein standard. The AbHpaI purity and subunit molecular weight (MW) were analyzed by 12% (w/v) SDS-PAGE.
SEC was used to determine the oligomeric state of AbHpaI as described previously (50). Briefly, a Superdex 200 Increase 10/300 GL gel-filtration column equipped with an ÄKTA FPLC system (GE Healthcare) was equilibrated with buffer G at a 0.5 ml/min flow rate at 25 C and A 280 was monitored for protein elution. Protein standards with known MWs (12.4-440 kDa) were used to construct a calibration curve. The elution volume (V e ) of each protein was measured, while that of blue dextran was used as the void volume (V o ). The protein mass was determined from a calibration curve of the relative ratio of V e /V o versus the logarithm of the protein standard MWs. The oligomeric state of AbHpaI was then estimated based on the calculated subunit MW.
An LDH-coupled assay was used to determine the aldol cleavage activity of AbHpaI at 25 C using oxaloacetate (OAA) as a substrate. Briefly, the reaction contained NADH (0.2 mM), LDH (30 μg/ml), Zn 2+ (0.5 mM), OAA (1 mM) or (4R)-KDGal (0.2-0.3 mM), and AbHpaI (0.1 μM) in buffer F. The control reaction without enzyme was used for a background subtraction. The decrease of NADH absorbance at 340 nm can be used to infer pyruvate release from AbHpaI aldol cleavage. One unit of AbHpaI was defined as the amount of enzyme that consumes 1 μmol of NADH per minute.

Measurement of metal ions in AbHpaI
The M 2+ species in the purifed AbHpaI were measured by Agilent 700 Series ICP-OES (Agilent Technologies). In total, 270 μM of the purified enzyme in buffer F was subjected to ICP-OES. The emission intensity of Zn 2+ , Mn 2+ , Co 2+ , Ni 2+ , Cu 2+ , Fe 2+ , Ca 2+ , and Mg 2+ was detected and quantified by subtraction from the background emission intensity of a buffer blank. The concentration of each M 2+ was determined by a calibration curve of varying concentrations (0.1-10 mg/l) of each standard M 2+ prepared in 2% (v/v) HNO 3 versus the emission intensity.
To determine the mole ratios of AbHpaI and M 2+ , the apo-AbHpaI was reconstituted with each of the M 2+ ions. Briefly, the purified enzyme was treated with chelating agents, EDTA, EGTA, and Chelex 100 (Merck KGaA), to strip off the M 2+ ions. The fivefold excess concentrations of EDTA and EGTA (5 mM) and 0.5 g of Chelex 100 were added into a 10-ml AbHpaI solution (1 mM). The solution mixture was thoroughly mixed and incubated at 4 C for 16 to 18 h to complete the metal ion chelation. The excess chelating agents and the metal ion chelation complexes were then removed by a Sephadex G-25 gel-filtration column equilibrated with buffer F to obtain apo-AbHpaI. To assure that the M 2+ ions were completely removed, the apo-AbHpaI was first analyzed by ICP-OES. For the reconstitution process, each M 2+ ion in a chloride salt form, namely ZnCl 2 , MnCl 2 , CoCl 2 , CaCl 2 , and MgCl 2 was dissolved in the Chelex 100 treated Milli-Q Type I ultrapure water, and a fivefold excess of each M 2+ chloride (1.5 mM) was added into a 2-ml apo-AbHpaI solution (0.3 mM). All samples were mixed thoroughly and incubated at 4 C for 16 to 18 h to reconstitute the apo-AbHpaI. The excess M 2+ in each sample was removed by a PD-10 desalting column equilibrated with buffer F to obtain a holoenzyme. The mole ratio of AbHpaI and M 2+ was then determined.

Measurement of the K d values for the AbHpaIligand complex
The

Thermal and solvent tolerance assay
To examine the thermal tolerance of AbHpaI, thermal stability and activity measurements were carried out by thermofluor and LDH-coupled assays, respectively. Thermofluor stability assays were performed as previously described (50). The T m values were determined for apo-AbHpaI and AbHpaIM 2+ complexes of Zn 2+ , Co 2+ , Mn 2+ , Mg 2+ , and Ca 2+ . For thermal stability measurements of the AbHpaIZn 2+ complex, the purified AbHpaI was first incubated at various temperatures from 25 to 85 C for 0 to 24 h. The activity was then measured by LDH-coupled assay.
To investigate the solvent tolerance, the T m values of the purified AbHpaI in the presence of 0 to 50% (v/v) of polarprotic (MeOH, EtOH and IPA) and polar-aprotic (ACN and DMSO) solvents were determined as above. The percentage of relative protein stability was calculated from the T m of AbHpaI in the absence of solvents as 100%.

Aldol cleavage reactions
The aldol cleavage assays were carried out by LDH-coupled assay in buffer F containing 0.2 mM NADH, 0.

Aldol condensation reactions
The aldol condensation steady-state kinetics of pyruvate and D-glyceraldehyde catalyzed by AbHpaIZn 2+ were monitored Catalytic and structural studies of a thermostable AbHpaI The aldol condensation of pyruvate and D-glyceraldehyde catalyzed by AbHpaIM 2+ complexes of Zn 2+ , Co 2+ , Mn 2+ , Mg 2+ , and Ca 2+ was monitored as described above. The assay reactions were carried out at 25 C for 1 h in buffer H containing 4 mM pyruvate, 30 mM D-glyceraldehyde, 0.1 mM metal chloride, and 0.05 μM of each AbHpaIM 2+ complex. The rate of (4R)-KDGal and (4S)-KDGlu formation was determined for each AbHpaIM 2+ complex. Time-course synthesis of (4R)-KDGal was performed with 0.5 μM AbH-paIZn 2+ and the product was monitored for 90 h.
For analysis of a broad spectrum of aldehydes, different aldehydes PPA, SSA, butyraldehyde, pentanal, glutaraldehyde, hexanal, benzaldehyde, HBA, and anisaldehyde were used as substrates. The reactions were carried out at 25 C for 1 h in buffer H containing 10 mM pyruvate, 5 mM aldehyde, 2 mM ZnCl 2 , and 10 μM purified AbHpaI. The exact m/z ([M-H] − ) of the product was monitored by high-resolution Compact QTOF (Bruker Daltonics) in negative mode, equipped with a Zorbax-eclipse C 18 column (5 μm, 4.6 × 250 mm) ultrahighperformance liquid chromatography (Thermo Scientific) operating at 30 C with a flow rate of 0.5 ml/min of 0.5% (v/v) formic acid.
Crystallization and X-ray data collection and structure determination For crystallization, apo-AbHpaI (0.7 mM) was incubated for 10 min at 25 C in buffer F containing 52 mM pyruvate and 11 mM divalent metal chloride (ZnCl 2 or CoCl 2 or MnCl 2 ). Crystals were grown at 18 C in microbatch drops containing 1 μl of apo-AbHpaI complex with 1 μl of buffer I containing 20 mM CaCl 2 and 30% (v/v) 2-methyl-2,4-pentanediol (MPD) as a crystallizing agent. For SSA soaking, crystals of apo-AbHpaIM 2+ PYR complexes were soaked in a crystallizing agent containing 18.5 mM pyruvate, 7% (v/v) glycerol, 3.6 mM divalent metal chloride (ZnCl 2 , CoCl 2 , or MnCl 2 ) and 36 mM SSA in a microbatch well at 27 C for 5 to 10 min. For PPA complex formation, crystal soaking was performed in a sitting drop well containing 25 μl of a crystallizing agent plus similar concentrations of pyruvate and ZnCl 2 , and 2.5 μl of 13.9 M PPA with 7.5 μl of 13.9 M PPA in the reservoir for 3 days at 18 C. All soaking solutions contained 7% (v/v) glycerol for cryoprotection. Crystals of AbHpaI cocomplexed with (4R)-KDGal and (4S)-KDGlu compounds and Zn 2+ /Mg 2+ cofactors were obtained from the wells containing 1 μl crystallizing solution ( Phases were calculated with Phaser (53), using EcHpaI (PDB code 2V5J) (30) as a search template for molecular replacement in CCP4 suite (54). Other AbHpaI structures were solved with Phaser MR using the apo-AbHpaI (PDB code 7ET8) as a template. Model building and refinement were performed using Coot (55) and Refmac5 (56). The ligand dictionary was prepared using ProDrg (57). Structures were validated in Procheck (58) and the wwwPDB validation server. Data collection and refinement statistics of the AbHpaI complexes were listed in Table 4. Superposition of structures was done by SSM Superposition (59). EPS was calculated by APBS-PDB2PQR tools, v2.1 (60,61). Figures were prepared with the PyMol Molecular Graphics System, v1.8 Schrödinger, LLC. Surface area was calculated using Pisa v1.48 (62). 2mF obs -DF model maps were calculated using Refmac5 (56), and mF obs -DF model OMIT maps with the compounds omitted were calculated using polder maps (63) in Phenix suite (64). Sequences were aligned with ClustalW v2.1 (65). The alignment was drawn with ESPript (66).

Computational calculations
To investigate the binding interaction of AbHpaIZn 2+ with (4R)-KDGal and (4S)-KDGlu, QM/MM MD simulations were performed. The structures of AbHpaIZn 2+ (4R)-KDGal (PDB code 7ETC) and AbHpaIZn 2+ (4S)-KDGal (PDB code 7ETD) complexes were employed and prepared as follows. The system was truncated to a 25 Å sphere with the center on the C 4 atom of (4R)-KDGal. The positions of the hydrogen atoms were located in the enzyme using the CHARMM procedure HBUILD (67). Hydrogen atoms of amino acid residues were added based on the results obtained from the PropKa (68). The atom types in the topology files were assigned according to the setup CHARMM27 parameters (69). For investigation of the enzyme-substrate interactions, the system was divided into two parts, QM and MM. The QM part consisted of substrate (either (4R)-KDGal or (4S)-KDGal), which was minimized using 1000 steps of Adopted Basis Newton-Raphson (ABNR) minimization with the AM1/CHARMM27 method. Next, the AM1/CHARMM27 MD using the leapfrog Langevin dynamics with a time step of 0.001 ps was performed at 300 K. The rest of protein, Zn 2+ , and water molecules were treated as the MM part. The system was equilibrated with QM/MM MD for 120 ps. The structures of this equilibration were collected at every 20 ps. Distances between substrate, metal ion, and the surrounding residues were determined. The binding energies of both substrates to the AbHpaIZn 2+ were calculated and compared.

Data availability
Data of X-ray structures are available at Protein Data Bank under PDB codes indicated.
Supporting information-This article contains supporting information.