Structural insights into the production of 3-hydroxypropionic acid by aldehyde dehydrogenase from Azospirillum brasilense

3-Hydroxypropionic acid (3-HP) is an important platform chemical to be converted to acrylic acid and acrylamide. Aldehyde dehydrogenase (ALDH), an enzyme that catalyzes the reaction of 3-hydroxypropionaldehyde (3-HPA) to 3-HP, determines 3-HP production rate during the conversion of glycerol to 3-HP. To elucidate molecular mechanism of 3-HP production, we determined the first crystal structure of a 3-HP producing ALDH, α-ketoglutarate-semialdehyde dehydrogenase from Azospirillum basilensis (AbKGSADH), in its apo-form and in complex with NAD+. Although showing an overall structure similar to other ALDHs, the AbKGSADH enzyme had an optimal substrate binding site for accepting 3-HPA as a substrate. Molecular docking simulation of 3-HPA into the AbKGSADH structure revealed that the residues Asn159, Gln160 and Arg163 stabilize the aldehyde- and the hydroxyl-groups of 3-HPA through hydrogen bonds, and several hydrophobic residues, such as Phe156, Val286, Ile288, and Phe450, provide the optimal size and shape for 3-HPA binding. We also compared AbKGSADH with other reported 3-HP producing ALDHs for the crucial amino acid residues for enzyme catalysis and substrate binding, which provides structural implications on how these enzymes utilize 3-HPA as a substrate.

and this seriously hamper the cell growth and 3-HP production as well [21][22][23][24][25] . Although some studies to screen better ALDHs or to improve the existing ALDHs by protein engineering have been conducted, poor performance of ALDH still remains a significant challenge for successful 3-HP production 26 .
The ALDH family enzymes have been extensively studied and several crystal structures of ALDHs have also been determined. Amino acid sequence and structure analysis of various ALDHs revealed that this family The secondary structure elements are drawn based on the structure of AbKGSADH. The catalytic residues of AbKGSADH are indicated by red colored triangles, and the residues involved in the binding of aldehyde-and hydroxyl-groups of 3-HPA are indicated by green colored triangles. The residues involved in the formation of hydrophobic pocket are indicated by blue colored triangles. The residues involved in the adenine ring and 2′-hydroxyl-group of NAD + are indicated by orange colored residues, and the residues involved in the formation of pyrophosphate, ribose ring, and nicotinamide are indicated by purple colored triangles. Ab, Bs, Cn, Ec, Kp, and Sc are abbreviations of Azospirillum basilensis, Bacillus subtilis, Cupriavidus necator, Escherichia coli, Klebsiella pneumonia, and Saccharomyces cerevisiae, respectively.
Scientific RepoRts | 7:46005 | DOI: 10.1038/srep46005 of enzymes have different substrate specificities, although the overall structures of the enzymes are very similar. Among the eight ALDHs which have been tested for the conversion of 3-HPA to 3-HP, AbKGSADH was first discovered as an enzyme that catalyzed the conversion of α -ketoglutarate-semialdehyde (α -KGSA) to α -ketoglutarate (α -KG) in an alternative pathway of L -arabinose metabolism 27 . In comparative studies, AbKGSADH was identified as a highly efficient enzyme for the conversion of 3-HPA to 3-HP 20 . However, the crystal structures of 3-HP producing ALDH, including AbKGSADH have not yet been reported, and the structural features that determine its 3-HPA binding ability have been veiled.
Here, we report the first crystal structure of the 3-HP producing ALDH from A. basilensis (AbKGSADH) in its apo-form and in complex with NAD + cofactor. On the basis of the docking simulation of 3-HPA binding to the AbKGSADH active site and pertinent biochemical studies, we reveal the structural features of substrate specificity 3-HPA. We also analyzed the amino acids that constitute the substrate binding pockets of known 3-HP producing ALDHs. These studies may provide valuable structural information for the development of engineered ALDHs with high 3-HP producing activity.
As observed in many other ALDH structures, AbKGSADH forms a tetramer. Although there are two AbKGSADH molecules in the asymmetric unit of our present structures, the tetrameric structure can be easily generated by one of the two folds from the P4 3 22 crystallographic symmetry operation (Fig. 2b) is mainly mediated by the OGD and the CTD of AbKGSADH. Three β -strands in the OGD and a β -sheet in the middle of the CTD form big β -sheet, and tetramerization of AbKGSADH is mediated by the OGD and two α -helices (α 2, α 3) (Fig. 2b). Using the PISA software 28 , we calculated that a 5028.9 Å 2 area of solvent accessible interface per monomer is buried, and the percentage of participating residues is 20.9%.
NAD + binding mode of AbKGSADH. Previous research has indicated that the ALDH family of enzymes utilize NAD + or NADP + as a cofactor 29 . First, to identify the cofactor specificity of AbKGSADH, we performed an ALDH activity assay using NAD + and NADP + . AbKGSADH showed less than 10% ALDH activity when NADP + was used as a cofactor as compared to that when NAD + was used (Fig. 3a). This result indicates that AbKGSADH utilizes NAD + as a cofactor instead of NADP + . To elucidate the cofactor binding mode of AbKGSADH, we then determined the crystal structure of the protein in complex with NAD + at a 2.6 Å resolution (Fig. 3b). The NAD + cofactor is bound to an inter-domain space between the NTD and the CTD (Fig. 3c). The NAD + -binding pocket is constituted by seven loops (β 7-α 4, β 8-α 5, β 10-α 7, β 11-β 12, α 8-β 13, α 9-α 10, and α 11-β 16) and four α -helices (α 4, α 6, α 7, and α 10). The adenine ring is stabilized in the hydrophobic pocket that is formed by Phe151, Pro211, Ala212, Phe229, Val235 and Leu239, and a hydrogen bond with Ser215 also contributes to the binding of the ring. Residues Lys178, Glu181, and Pro211 constitute a suitable space for binding of the ribose ring, and stabilize the 2′ -hydroxyl-group of the ring (Fig. 3d). The formation of the ribose ring binding site does not seem to be large enough to accommodate the phosphorylated ribose ring. This observation indicates that AbKGSADH cannot utilize NADP + as a cofactor, which is consistent with the results mentioned above. The pyrophosphate moiety is stabilized by residues Asn331, Arg333, and Arg334 through directly and water-mediated hydrogen bond networks. Residues Arg334 and Glu384 stabilize the ribose moiety of NAD + , and the nicotinamide ring is stabilized by residues Gln160 and Glu253 by hydrogen bonding (Fig. 3e).

Substrate binding mode of AbKGSADH.
AbKGSADH is known to utilize both α-ketoglutaratesemialdehyde (α -KGSA) and succinate-semialdehyde (SSA) as substrates 27 . To elucidate how AbKGSADH accommodates these substrates, we performed molecular docking simulations of AbKGSADH with α -KGSA and SSA. The molecular docking simulations revealed that these two substrates fit well into the somewhat positively charged substrate binding pocket (Fig. 4a). The aldehyde-groups of these substrates, which are the sites of enzyme reaction, are located in the same place around the catalytic residues (Fig. 4a). The aldehyde-group of α -KGSA is stabilized by Gln160 and Arg163 through hydrogen bonds, and two catalytic residues, Glu253 and Cys287, also assist the binding of the molecule (Fig. 4b). The 4′ -keto-group of α -KGSA is stabilized by hydrogen bonds with Arg281, and the carboxyl-group of the molecule is stabilized by Glu106 and Gln160. The substrate binding pocket is also formed by several hydrophobic residues, such as Phe156, Val286, Ile288, Pro444, and Phe450, which seem to contribute to the stabilization of the hydrophobic part of α -KGSA (Fig. 4b). The binding of SSA is similar to that of α -KGSA, however, the stabilization of the carboxyl-group of SSA is quite different. Arg281, a residue that is involved in the stabilization of the 4′ -keto-group of α -KGSA, forms a hydrogen bond with the carboxyl-group  of SSA instead (Fig. 4c). These observations explain how AbKGSADH can accommodate both α -KGSA and SSA as real substrates.

3-HPA binding mode of AbKGSADH.
To elucidate the 3-HPA binding mode of AbKGSADH, we attempted to determine the crystal structure of AbKGSADH in complex with 3-HPA substrate or the 3-HP product, however, both cocrystallization and soaking experiments were unsuccessful. The molecular docking simulation of AbKGSADH with 3-HPA did allow us to speculate how AbKGSADH accommodates an unnatural substrate 3-HPA. The 3-HPA molecule is bound at the same position as the α -KGSA and SSA molecules (Fig. 4a). Moreover, the aldehyde-group of 3-HPA is located in the same position as those of α -KGSA and SSA, and is stabilized by the same residues (Fig. 4a). The 3′ -hydroxyl-group of 3-HPA is stabilized by residues Asn159, Gln160, and Arg163 through hydrogen bonds, and of these three residues, Gln160 and Arg163 are also involved in the binding of the aldehyde-group of 3-HPA (Fig. 4d). As observed in the binding of α -KGSA and SSA, hydrophobic residues, such as Phe156, Val286, Ile288, and Phe450, seem to also contribute to stabilization of the hydrophobic part of 3-HPA (Fig. 4d). One interesting observation is that Arg281, which is a crucial residue for the binding of α -KGSA and SSA, is located distal from the bound 3-HPA and does not participate in its stabilization (Fig. 4d). We speculate that this observation is derived from the fact that 3-HPA has two and one fewer carbon than α -KGSA and SSA, respectively. To confirm the involvement of these residues in the enzyme catalysis and the binding of the 3-HPA substrate, we then performed site-directed mutagenesis experiments. First, we mutated the two catalytic residues, Glu253 and Cys287, to alanine residues, and observed that these mutants exhibited an almost complete loss of enzyme activity (Fig. 4e), indicating that AbKGSADH has the same enzymatic mechanism as other ALDH family enzymes. Second, we mutated residues that form hydrogen bonds with the aldehyde-and hydroxyl-group of 3-HPA to alanine residues. The AbKGSADH R163A mutant showed 90% of the enzyme activity that was present with the wild-type enzyme (Fig. 4e), and the result indicates that 3-HPA can be stabilized sufficiently by hydrogen bonding with other residues even without Arg163. Interestingly, the AbKGSADH N159A and the AbKGSADH Q160A mutants showed approximately 30% higher enzyme activities than the wild-type enzyme (Fig. 4e). These results imply that there is no significant issue in stabilizing 3-HPA even if one of the hydrogen bonding residues is absent, as observed in the AbKGSADH R163A mutant. Rather, substitutions of Asn159 and Gln160 to alanine seem to increase the hydrophobicity of the substrate binding site and consequently increased the stabilization of 3-HPA. Finally, we mutated residues involved in the constitution of the hydrophobic substrate binding site to alanine residues. All of these mutants, AbKGSADH F156A , AbKGSADH V286A , AbKGSADH I288A , and AbKGSADH F450A , exhibited decreased or complete loss of activities compared with the wild-type enzyme (Fig. 4e). We propose that the formation of substrate binding sites with optimal size and shape by these hydrophobic residues is very important for accommodating 3-HPA as a substrate.

Comparison of 3-HP producing ALDHs. So far, eight 3-HP producing ALDHs, including AbKGSADH,
have been reported. To structurally analyse how these enzymes utilize 3-HPA as a substrate, we compared the key residues of AbKGSADH in enzyme catalysis and substrate binding for 3-HPA with those of the other seven 3-HP producing ALDHs (Fig. 1b, Table 2). The amino acid sequence similarities within eight 3-HP producing ALDHs are 31% to 83%, and those between AbKGSADH and the other seven 3-HP producing ALDHs are 32% to 50%. As expected, the two catalytic residues, Glu253 and Cys287, in AbKGSADH are completely conserved in all 3-HP producing ALDHs. However, the three residues involved in the stabilization of the aldehyde-and hydroxyl-group of 3-HPA, Asn159, Gln160, and Arg163 in AbKGSADH, are variable in other 3-HP producing ALDHs (Fig. 1b,  Table 2). Combined with the previously described results that mutations of these residues to alanine did not significantly affect enzyme activity, we propose that, for the stabilization of the aldehyde-and hydroxyl-groups of 3-HPA, the combination of several residues rather than a particular residue is important. Interestingly, the four hydrophobic residues forming the hydrophobic pockets, Phe156, Val286, Ile288, and Phe450 in AbKGSADH, are highly conserved throughout all 3-HP producing ALDHs (Fig. 1b, Table 2). As described above, a single amino acid mutation of any of these four hydrophobic residues to alanine showed decreased or almost complete loss of enzyme activity (Fig. 4e). Taken together, we propose that formation of hydrophobic pockets with optimal size and shape is critical for 3-HP producing ALDHs to accept 3-HPA as a substrate.  In summary, we report the first crystal structure of the 3-HP producing ALDH, AbKGSADH, and provide structural insight into how the 3-HP producing ALDHs utilize the unnatural substrate 3-HPA. For accepting 3-HPA, the location of the appropriate residues for hydrogen bonding to the aldehyde-and the hydroxyl-groups of 3-HPA is important. Moreover, a hydrophobic pocket with optimal size and shape to bind the hydrophobic portion of 3-HPA is also critical. This structural information might be used for developing 3-HP producing ALDHs that possess a higher 3-HP production activity.

Methods
Cloning, expression and purification of AbKGSADH. The AbKGSADH coding gene was amplified through polymerase chain reaction (PCR) using synthetic gene in a pBHA vector by Bioneer. The PCR products were digested by NdeI and XhoI restriction enzymes, and sub-cloned into the pProEX-HTa expression vector (Thermo Fisher Scientific) which contained a 6xHis tag and rTEV protease cleavage site at the N-terminus of the target protein. The pProEX-HTa:AbKGSADH was transformed into a E. coli BL21(DE3)-T1 R strain, which was grown to an OD 600 of 0.6 in LB medium containing 100 mg L −1 ampicillin at 310 K and AbKGSADH protein expression was induced by 0.5 mM 1-thio-β -D-galatopyranoside (IPTG). After 20 h at 293 K, the cell were harvested by centrifugation at 4,000× g for 15 min at 277 K. The cell pellet was resuspended in ice-cold buffer A (40 mM Tris-HCl pH 8.0) and disrupted by ultrasonication. The cell debris was removed by centrifugation at 13,000 g for 30 min, and the lysate was applied onto a Ni-NTA agarose column (Qiagen). After washing with buffer B (40 mM Tris-HCl pH 8.0 and 25 mM Imidazole), the bound proteins were eluted with buffer C (40 mM Tris-HCl pH 8.0 and 300 mM Imidazole). Finally, trace amounts of contaminants were removed by size-exclusive chromatography using Sephacryl S-300 prep-grade column (320 ml, GE Healthcare) equilibrated with buffer A. The eluted protein had a molecular weight of about 200 kDa, indicating a tetrameric structure. The protein was concentrated to 50 mg mL −1 using a spin column (Amicon Ultra Centrifugal Filter, 30 kDa pore size), and kept at 193 K for further experiments. All purification steps were performed at 277 K.
Crystallization and data collection of AbKGSADH. Crystallization of the purified AbKGSADH protein was initially tried with commercially available sparse-matrix screens, including Index, PEG ion I and II (Hampton Research), and Wizard Classic I and II (Rigaku Reagents), using the sitting-drop vapor diffusion method on the MRC Crystallization plates (Molecular Dimensions) at 295 K. Each experiment consisted of mixing 1.0 μ L protein solution (60 mg mL −1 , 40 mM Tris-HCl pH 8.0) with 1.0 μ L reservoir solution and then equilibrating against 50 μ L reservoir solution. AbKGSADH crystals were observed from several crystallization screening conditons. After several steps of crystal improvement, the best quality crystals appeared in 16% polyethylene glycol 3350, 0.1 M sodium cacodylate pH 6.5, and 0.2 M Magnesium chloride hexahydrate. The crystals were transferred to cryoprotectant solution containing 25% polyethylene glycol 3350, 0.1 M sodium cacodylate pH 6.5, 0.2 M Magnesium chloride hexahydrate, and 30% (v/v) glycerol. The crystals were fished out with a loop larger than the crystals and flash-frozen by immersion in liquid nitrogen at 100 K. Data were collected to a maximum resolution on the detector of 2.18 Å at 7A beamline of the Pohang Acclerator Laboratory (PAL, Pohang, Korea), using a Quantum 270 CCD detector (ADSC, USA). All data were indexed, integrated, and scaled together using the HKL2000 software package 30 . The crystals of AbKGSADH belonged to the space group P4 3 22 with unit cell parameters a = b = 129.12 Å, c = 118.12 Å, α = β = γ = 90° Assuming two AbKGSADH molecules in asymmetric unit, the crystal volume per unit of protein mass was 2.46 Å 3 Da −1 , which means the solvent content was approximately 50% 31 .
Structure determination of AbKGSADH. The structure of apo-form of AbKGSADH was determined by molecular replacement with the CCP4 version of MOLREP 32 , using the structure of Succinic-semialdehyde dehydrogenase (SSADH) from Homo sapiens (PDB code 2W8R) as a search model. Further model building was performed manually using the program WinCoot 33 , and refinement was performed with CCP4 refmac5 34 . The structure of AbKGSADH in complex with NAD + was solved by molecular replacement using the crystal structure of the apo-form of AbKGSADH. The data statistics are summarized in Table 1. The refined model of the apo-form of AbKGSADH and that in complex with NAD + were deposited in the Protein Data Bank with PDB codes of 5X5T and 5X5U, respectively. Molecular docking simulations of AbKGSADH. Molecular docking simulations of α -KGSA, SSA and 3-HPA to AbKGSADH structure were performed by AutoDock Vina software 35 . AbKGSADH structure in complex with NAD + cofactor (PDB code of 5X5U) and the α -KGSA, SSA and 3-HPA ligands were prepared using the JLigand software. For the docking simulation, the pdbqt files were generated using AutoDock Tools, and all steps for simulation and grid box creation were performed according to the AutoDock Vina manual. The grid size for α -KGSA was x = 40, y = 50, z = 36, and grid center was designated at x = − 35.377, y = − 49.707, z = − 4.816. And the grid size for SSA was x = 28, y = 50, z = 34, and grid center was designated at x = − 37.77, y = − 49.853, z = − 4.29. Last, the grid size for 3-HPA was x = 26, y = 32, z = 34, and grid center was designated at x = − 37.903, y = − 46.826, z = − 4.534. The final conformations produced in this simulation were checked using PyMOL software.
Activity assay of AbKGSADH. The activity of AbKGSADH was determined by measuring the increase of absorbance at 340 nm (extinction coefficient of 6.22 × 10 3 M −1 cm −1 ). Enzyme reaction was performed with a reaction mixture of 1 mL total volume at 303 K. The reaction mixture contained 100 mM Tris-HCl, pH 8.0, 10 mM 3-HPA, and 1 mM NAD(P), and the background rate of the assay in the absence of enzyme is zero. The reaction