1 Introduction

The fragrance in rice is considered to be important for the determination of rice quality and results in strong human preference which determines its market price. Investigation of fragrant varieties at a molecular level lead to the identification of an aromatic related locus on chromosome 8, fgr, of rice [22]. More recently, the rice gene fgr/OsBADH2, a homolog of betaine aldehyde dehydrogenase (BADH), was proposed to be accountable for aroma metabolism in fragrant rice varieties [20]. It was reported that an 8-bp deletion and 3 single nucleotide polymorphisms (SNPs) in exon 7 of badh2 created a premature stop codon leading to a truncated BADH2. The partial loss of BADH2 function is proposed to account for the accumulation of 2AP, a principal powerful flavour component in fragrant varieties, while functional BADH2 mature protein is found in non-fragrant varieties [7]. Two pathways of 2AP biosynthesis in rice have been proposed: BADH2-dependent 2AP synthesis [2, 5] and BADH2-independent 2AP synthesis [12]. The former model suggests that functional BADH2 inhibits the biosynthesis of 2AP in non-fragrance rice by converting GAB-ald to GABA whereas in fragrance rice truncated BADH2 results in the accumulation of GAB-ald, which then leads to the formation of 2AP. The latter model involves Δ1-pyrroline-5-carboxylate synthetase catalysing the formation of Δ1-pyrroline-5-carboxylate, which then reacts with methylglyoxal to form 2AP in the fragrant variety. However, there is no direct involvement of BADH2 in the second model.

Domain analysis predicted that BADH2 contains three domains: a NAD+ binding domain, an oligomerization domain and a substrate binding domain [5]. Intact BADH2 showed high betaine aldehyde (Bet-ald), γ-aminobutyraldehyde (GAB-ald) and 3-aminopropionaldehyde (AP-ald) dehydrogenase activities, thereby indicating wide substrate specificities similar to BADH from sugar beet, spinach and oat [5, 13, 25]. Even though rice BADH homologous enzymes have been preliminarily studied [2, 5], the enzymes were expressed in low yield and were not fully characterized. In addition, it has been reported that insertion of three nucleotides, resulting in a tyrosine residue being inserted (badh2.8) within exon 13, increases the level of aroma in rice to levels similar to those found in the Myanmar fragrance rice variety [15]. To shed light on the biochemical pathway for 2AP synthesis, the biochemical characterization of both wild-type and mutant rice BADHs are required.

In the present study, recombinant proteins of BADH homologues from Oryza sativa, the insertion mutant BADH2.8 (called BADH2_Y420), and three BADH2 mutants (namely, BADH2_N162A, BADH2_E260A, and BADH2_C294A) were produced in sufficient yield for biochemical analyses. The biochemical and enzymatic properties of the proteins were examined. It was revealed that BADH2_Y420 mutant exhibited less catalytic efficiency towards GAB-ald but greater efficiency towards Bet-ald, compared to the wild-type enzyme. Finally, homology modelling was employed to elucidate the arrangement of substrates in the enzyme binding pocket, leading to a better understanding of the substrate specificity of these BADH enzymes.

2 Materials and Methods

2.1 Plasmid Construction

Full length OsBADH1 and OsBADH2 genes encoding the putative BADH1 and BADH2 enzymes were subcloned via the NdeI/XhoI restriction sites into pET28b(+) from the plasmids, pET17b-BADH and pUC18-Os2AP (gifts from Associated Prof. Apichart Vanavichit, Kasetsart University), which contained OsBADH1 and OsBADH2, respectively. The constructed plasmids, pET28b-OsBADH1 and pET28b-OsBADH2, encoded fusion proteins containing a N-terminal His-tag and thrombin cleavage site which result in the addition of 20 amino acid residues (MGSSHHHHHHSSGLVPRGSH) prior to the BADHs. Therefore, the recombinant BADH1 and BADH2 contain 525 and 523 amino acids, respectively.

2.2 Site-Directed Mutagenesis

The BADH2 mutants were generated with the QuikChange® site-directed mutagenesis kit (Stratagene) using the wild-type pET28b-OsBADH2 plasmid as a template. The site-directed mutagenesis was performed by PCR amplification to generate the Y420-insertion mutant of BADH2 (BADH2_Y420), BADH2_N162A, BADH2_E260A and BADH2_C294A using the following primers (BADH2_Y420_F: 5′-GGCCAACGATACTCATTATTATGGTCTGGCTGGTGCTGTGC-3′, BADH2_Y420_R: 5′-GCACAGCACCAGCCAGACCATAATAATGAGTATCGTTGGCC-3′; BADH2_N162A_F: 5′-GGTTGATCACACCTTGGG CCTATCCTCTCCTGATGGC-3′, BADH2_N162A_R: 5′-GCCATCAGGAGAGGATAGG CCCAAGGTGTGATCAACC-3′; BADH2_E260A_F: 5′-GTTAAGCCTGTTTCACTGG C ACTTGGTGGAAAAAGTCC-3′, BADH2_E260A_R: 5′-GGACTTTTTCCACCAAGT G CCAGTGAAACAGGCTTAAC-3′; BADH2_C294A_F: 5′-GGACCAATGGCCAGATTG CCAGTGCAACATCGCGTC-3′, BADH2_C294A_R: 5′-GACGCGATGTTGCACTGG CAATCTGGCCATTGGTCC-3′). The mutagenic primers include mutations (underlined) at the corresponding triplets (bold): for the tyrosine insertion mutant BADH2_Y420 (TAT = tyrosine) and alanine substitution mutants BADH2_N162A (GCC = alanine), BADH2_E260A (GCA = alanine) and BADH2_C294A (GCC = alanine). The presence of the mutation was verified by DNA sequencing for each construct.

2.3 Expression and Purification of Wild-Type and Mutant BADH Proteins

Escherichia coli BL21 (DE3) cells (Novagen) were transformed with wild-type or mutant pET28b-OsBADH plasmids using the standard heat shock protocol. The transformed E. coli were cultured at 37 °C on Luria–Bertani medium (LB) containing kanamycin (50 μg mL−1). When the OD600 reached 0.6, isopropyl-β-D-thiogalactopyranoside (0.4 mM) was added. After incubation at 16 °C for a further 24 h, the cells were harvested by centrifugation at 5,000 rpm for 30 min at 4 °C and resuspended in lysis buffer [50 mM Tris–Cl, pH 8.0, 0.5 M NaCl, 5 mM imidazole, 2 mM β-ME, 1 mM PMSF and 1% Triton X-100]. The cells were disrupted by sonication, and the lysate was centrifuged at 15,000 rpm for 30 min at 4 °C. The supernatant was then applied to a 1 mL-HiTrap Chelating HP column connected to an ÄKTA™FPLC™ (GE Healthcare) which was previously equilibrated with five column volumes of buffer A [50 mM Tris–Cl, pH 8.0, 0.5 M NaCl and 30 mM imidazole]. Thereafter, His6-tagged proteins were eluted with an increasing gradient of buffer B [50 mM Tris–Cl, pH 8.0, 0.5 M NaCl and 0.5 M imidazole]. Purified proteins were extensively dialyzed in 50 mM HEPES–KOH, pH 8.0 and subjected to SDS–PAGE to confirm homogeneity. Protein concentration was determined by the BCA protein assay (PIERCE) using BSA as a standard protein.

2.4 Western Blot Analysis

The purified proteins were subjected to 12% SDS–PAGE and subsequently electro-blotted onto a BioTrace™ PVDF membrane (PALL Life sciences) with a constant current of 128 mA for 1.30 h (ATTA system). Non-specific binding sites were first blocked by incubating the membrane overnight in 5% skimmed milk TBS-T buffer [20 mM Tris–Cl pH 7.5, 0.5 M NaCl, 0.05% Tween20 and 0.2% Triton X-100] at 4 °C. The membrane was then washed three times for 10 min with TBS-T buffer at room temperature. The membrane was probed with 1:5,000 Penta-His™ HRP antibody (QIAGEN) in 1% skimmed milk TBS-T buffer for 1 h at room temperature and then washed three times with TBS-T buffer. The immunodetection pattern was analyzed by chemiluminescence using ECL Western blotting reagent (GE Healthcare) and developed on Hyperfilm ECL autoradiography film as described in the manufacturer’s protocol.

2.5 Circular Dichroism Spectroscopy

Circular dichroism (CD) spectroscopy was recorded at 25 °C using a Jasco 710 spectropolarimeter. The CD spectra were measured at a protein concentration of 0.3 mg mL−1 in 10 mM sodium phosphate, pH 7.5, using a quartz cuvette with a path length of 0.1 cm for far-UV CD measurements. Each spectrum represents an average of 5 scans collected from 190 to 250 nm at a rate of 20 nm min−1, a response time of 4.0 s and a bandwidth of 1.0 nm. The baseline was corrected by subtracting the spectrum of a buffer blank obtained under identical conditions. The results were converted to per-residue molar absorption units, [θ] (deg cm2 mol−1) and the secondary structure content was analyzed with the CDPro software package [24].

2.6 Fluorescence Binding Study

Fluorescence measurements were carried out according to the method described in a previous paper with some modifications [1]. All measurements were performed on a luminescence spectrometer LS50B (Perkin-Elmer). The emission spectra were recorded from 300 to 450 nm with excitation at 295 nm. Fluorescence titration of enzymes with co-factors (NAD+, NADP+, NADH and NADPH) was conducted in 50 mM HEPES–KOH, pH 8.0. The protein concentration used was 2.5 μM in 50 mM HEPES–KOH, pH 8.0, and aliquots of ligands were added from stock solutions (5 mM). Ligand titrations were carried out by monitoring fluorescence intensity at an emission wavelength of 350 nm. Fluorescence intensity was corrected for dilution of protein due to addition of the ligand. Data were plotted as ΔFmax (maximum attainable change in fluorescence intensity) at 350 nm versus concentration of cofactor. The data were fitted and standard errors were calculated by non-linear regression analysis using the Microcal Origin 6.0 program.

2.7 Enzyme Assays

Enzyme kinetic assays of BADH activity were measured spectrophotometrically by monitoring the oxidation of Bet-ald and GAB-ald. Bet-ald chloride (Sigma) was dissolved in H2O and directly used in the enzymatic assay while GAB-ald dimethyl acetal (Sigma) was used for GAB-ald. Aliquots of the diethylacetals were hydrolyzed with 1 M HCl in a plugged test tube and heated at 80 °C for 1 h. The hydrolyzate of GAB-ald was stored at −80 °C, and neutralized with KOH just before used. BADH activities were measured by monitoring the increase in absorbance at 340 nm of NADH. Briefly, all enzyme activities were determined using a reaction mixture of 200 μL containing 5 μM BADH in 50 mM HEPES–KOH, pH 8.0, 5 mM NAD+ and various concentration of each substrate. The activity was calculated by using an extinction coefficient of 6,220 M−1 cm−1 for NADH. One unit of enzyme activity was defined as the amount of enzyme that catalyzes the formation of 1 μmol of NADH per minute. The kinetic parameters, Km and Vmax were obtained by fitting the initial rates against the concentrations of each substrate to the Michaelis–Menten equation using Microcal Origin 6.0 program. The data were fitted and standard errors were calculated by non-linear regression.

2.8 Size-Exclusion Chromatography

The association state of recombinant BADHs was estimated by size-exclusion chromatography on a HiPrep 16/60 Sephacryl S-300 HR column (GE Healthcare) connected to an ÄKTA™ prime plus (GE Healthcare). The column was equilibrated and eluted with 50 mM HEPES–KOH, pH 8.0 at room temperature with a flow rate of 0.5 mL min−1. Ovalbumin (43,000 Da), Conalbumin (75,000 Da), Aldolase (158,000 Da) and Ferritin (440,000 Da) were used as standards for calibration. Blue dextran was used to determine void volume, which was 40.2 mL. Protein elution was monitored by absorbance at 280 nm. Solute behavior was expressed as

$$ {\text{K}}_{\text{av}} = \frac{{({\text{V}}_{\text{e}} - {\text{V}}_{\text{o}} )}}{{({\text{V}}_{\text{t}} - {\text{V}}_{\text{o}} )}} $$

where Ve, Vo and Vt correspond, respectively, to the elution volume of the solute, the void volume and the total volume of the bed.

2.9 Mass Spectrometer Analysis

MALDI-TOF was performed using an Ultraflex III TOF/TOF (Bruker, America) coupled with delayed extraction. Sample aliquots (0.5 μL) were analyzed using a matrix of sinapinic acids by dissolving sinapinic acid (10 mg) in 50% acetonitrile and 0.1% TFA and spinning at 10,000 rpm for 2 min. The sample was resuspended with 50% acetonitrile in 0.1% TFA. Mass spectra were analyzed using M-scan Ltd.

2.10 Computational Details

Sequence alignment, homology modeling and molecular mechanics calculation were performed using the Discovery Studio 2.5 software package (Accelrys Inc., CA, USA). Homology models for BADH1, BADH2, and BADH2_Y420 were constructed using MODELLER 9v4 [23]. The crystal structures for bacterial betaine aldehyde dehydrogenase (pdb code 3FG0) and plant amino-aldehyde dehydrogenese (code 3IWJ) with resolutions of 1.85 and 2.15 Å, respectively, were used as templates (Supplementary Information Table S1). The stereochemical quality of the BADH1, BADH2, BADH2_Y420 models was evaluated by PROCHECK v.3.5 [16]. To build the complex between BADHs and NAD+ cofactor, Cα-atom superposition between each template and modeled structure was performed and the NAD+ coordinates from the template (code 3FG0) were transferred to the protein model. Finally the modeled complex was subjected to CHARMM energy minimization (steepest descent and conjugate gradient methods until the model reached 0.001 kcal mol−1 Å convergence) to remove unreasonable atomic contacts [3].

2.11 Determination of GABA

After the enzymatic reactions were carried out as described above, the reaction mixture was separated by TLC to identify the desired product, GABA. In brief, the sample was spotted onto a silica gel 60 F254 aluminium sheet (Merck, Germany) which was then immersed in a developing beaker containing a mobile solvent (n-butanol:acetic acid:H2O, 4:1:1, v:v). After the solvent had reached to a marked solvent front, the TLC plate was sprayed with ninhydrin and GABA was detected as a purple spot under these conditions.

3 Results and Discussion

3.1 Expression and Purification of Recombinant Proteins

The results showed that the His-tagged wild-type BADHs and BADH2 mutants (BADH2_Y420, BADH2_C294A, BADH2_E260A and BADH2_N162A) were expressed and purified to homogeneity and had the expected molecular masses of about 57 kDa. The final yield of purified BADH1, BADH2, BADH2_Y420, BADH2_N162A, BADH2_E260A and BADH2_C294A mutants was 19.2, 11.9, 6.42, 8.2, 6.0 and 7.9 mg per 250 mL culture, respectively. The lower yield of BADH2 may result from the fact that a greater portion of BADH2 was expressed as insoluble inclusion bodies compared to BADH1 (lane 4 and 8, Fig. 1a). This result is consistent with Bradbury et al. [2]. Even though these two enzymes share about 76% sequence similarity, this difference in protein stability may contribute to the different roles of these proteins in rice fragrance. The yield of all mutant enzymes was approximately twofold less than that of wild-type BADH2. This suggests that mutation of catalytic residues and insertion of Y420 may affect enzyme stability. Collectively, we have obtained decent amounts of proteins compared to previous reports [2, 5] in order to perform biochemical and enzymatic characterization.

Fig. 1
figure 1

SDS–PAGE and CD analysis. a Coomassie stained SDS–PAGE showing expression and purification of recombinant BADH1 and BADH2. (Lane 1) molecular weight marker (Fermentus Spectra™ Multicolor Broad Range Protein Ladder), (lanes 2 and 6) E. coli lysate pre-IPTG induction for BADH2 and BADH1, respectively; (lanes 3 and 7) E. coli lysate 24 h at 16 °C after induction with 0.4 mM IPTG for BADH2 and BADH1, respectively; (lanes 4 and 8) insoluble pellet after lysis for BADH2 and BADH1; (lanes 5 and 9) purified BADH2 and BADH1. b CD spectra were measured at a protein concentration of 1 mg mL−1 in 25 mM HEPES–KOH, pH 8.0, at 25 °C, using a 1 mm path length cell. The CD scans were collected from 190 to 260 nm in a Jasco 710 spectropolarimeter

Full size image

3.2 Protein Characterization

Western blot analysis was carried out in order to authenticate all BADHs. Results confirmed that the molecular mass of each purified protein corresponded to the theoretical mass deduced from its amino acid sequence (Supplementary Information Fig S1). CD spectra indicated that the recombinant proteins contained α-helix/β-sheet secondary structure, with a maximum in ellipticity at 202 nm and a minimum at 220 nm (Fig. 1b). Analysis of the secondary structure content with the CDPro suite of the CD spectra analysis program indicated 52.60% α-helix, 11.80% β-strand and 36.20% other for BADH1 and 52.40% α-helix, 12% β-strand and 36% other for BADH2. This is consistent with other members of the aldehyde dehydrogenase (ADH) family. Fluorescence titration experiments were also carried out in order to assess the global folding of the proteins and determine the dissociation constant for the cofactors. Excitation of BADH1 at 295 nm results in a single emission peak at about 350 nm (Fig. 2a) characteristic of a class II tryptophan residue in which a tryptophan residue is partially exposed at the surface of the protein [4]. A decrease in the intensity of tryptophan fluorescence for BADHs was observed upon addition of NAD+ (Fig 2a). This phenomenon was also observed for E. coli BADH (YdcW) [1] and indicates that the nucleotide induces conformational changes upon binding to the enzyme. Table 1 summarizes the dissociation constant (Kd) values for each cofactor. BADH can act as a dual dehydrogenase, serving either catabolic or anabolic roles depending on whether it employs NAD+ or NADP+. Pseudomonas aeruginosa BADH can use NAD+ and NADP+ with similar efficiencies [28] while BADHs from plants, such as barley and spinach, prefer NAD+ to NADP+ [8]. Our results clearly showed that both wild-type enzymes from rice have a marked preference for NAD+ over NADP+. With regard to NADH, Kd values for both wild-type enzymes were about 4–6 times higher than those of NAD+. However, for the mutant enzymes, only BADH2_E260A displayed similar Kd values toward their cofactors compared to BADH2 (Table 1B). While BADH2_N162A and BADH2_E260A presented similar preferences for their cofactors, BADH2_Y420 had a preference for NADH over NAD+ with Kd values of 18 and 27 μM, respectively. It is interesting to note that BADH2_C294A also showed a strongly opposite cofactor preference compared to BADH2. The Kd values of BADH2_C294A towards NAD+ and NADH were 8 times higher and 6.5 times less than those of BADH2. Tylichova et al. [26] have reported that amino acid residue C294 is located between the NAD+ and substrate binding sites but amino acid residues N162 and E260 appeared to be in the substrate channel. Therefore, changing residue C294 may affect cofactor binding greater than substitution of residues N162 and E260.

Fig. 2
figure 2

Fluorescence binding study. a Emission spectra for BADH1 upon the addition of NAD+. Fluorescence titration of BADH1 and BADH2 with NAD+ (b), NADP+ (c), NADH (d) and NADPH (e), [E]0 = 2.5 μM in 50 mM HEPES–KOH, pH 8.0, λex = 295 nm, λem = 350 nm. The data were fitted and standard errors were calculated by non-linear regression analysis using the Microcal Origin 6.0 program

Full size image
Table 1 Dissociation constant (Kd) of wild-type enzymes and cofactors
Full size table

3.3 Enzymatic Activity

Kinetic parameters obtained using the purified recombinant enzymes are shown in Table 2. Similar to other BADHs [9, 13, 28], both BADH1 and BADH2 obeyed Micaelis–Menten kinetics for the two substrates Bet-ald and GAB-ald. The Km values of BADH1 and BADH2 for Bet-ald in this study were 1,381 and 694 μM, respectively. These values were slightly different from those reported by Bradbury et al. [2] which were 3,233 and 63 μM and Mitsuya et al. [18] which were 2,600 and 230 μM for BADH1 and BADH2, respectively. The discrepancy between these values is still unclear. Several factors such as the His-tag or buffer conditions may account for these differences. However, our results showed a similar trend, with the Km values of BADH1 for the two substrates being higher than those for BADH2. Additionally, the catalytic efficiency (kcat/Km) of both enzymes towards Bet-ald showed that the catalytic efficiency of BADH1 was almost 6 times higher than that of BADH2. However, when GAB-ald was used as a substrate, the Km values of both enzymes were slightly lower than those of Bet-ald. This implied that GAB-ald is a preference substrate for BADH enzymes. The catalytic efficiency of both enzymes towards GAB-ald showed the same trend as that towards Bet-ald, in that BADH1 exhibited a greater catalytic efficiency than BADH2. The results here are in agreement with those reported by Bradbury et al. [2] and Mitsuya et al. [18] that both BADH homologues exhibited greater affinity and higher catalytic efficiency towards ω-aminoaldehyde rather than Bet-ald. Other BADHs from some plants, human and E. coli can also oxidize ω-aminoaldehydes [9, 13, 17, 21], while those from mangrove and P. aeruginosa cannot catalyze the NADP+- or NAD+-dependent oxidation of other aldehydes rather than Bet-ald [11, 19]. For the BADH2_Y420 mutant, the insertion of the tyrosine affects the binding affinity of the enzyme towards GAB-ald (Table 2). The presence of Y420 in BADH2 alters its substrate and NAD+ cofactor preference and may be involved in rice fragrance production. Kinetic parameters of the other mutants were also determined (Table 2). The results indicated that mutation of catalytic residues of BADH2 abolished enzyme activity except for BADH2_N162A. This result was expected since C294 is the catalytic residue involved in the formation of the thiohemiacetal intermediate and E260 is proposed to act as the base that deprotonates the catalytic cysteine; therefore, replacing these residues with alanine can result in a non-functional BADH2. Unlike C294 and E260, N162 is involved in the hydride transfer reaction. Hence, the mutation at this residue can only reduce the catalytic activity of the enzyme. In addition to ω-aminoaldehyde, GABA was also used as a substrate in the enzyme reverse reaction utilizing NADH as a cofactor (data not shown). Unfortunately, when GABA was used as a substrate, the enzymatic activity of the wild-type and mutant enzymes could not be detected. Collectively, both BADH1 and BADH2 from rice can utilize both Bet-ald and GAB-ald as substrates with NAD+ as a cofactor but the enzymes cannot catalyze the reverse reaction. Further studies of the kinetic mechanism of these two enzymes are under investigation.

Table 2 Kinetic parameters of BADH1, wild-type and mutant BADH2s
Full size table

3.4 Oligomeric State Determination

Two oligomeric states have been found among BADHs. BADHs from animals [14] and bacteria [6, 27] are tetrameric, while those from plants [13] and firmicutes [1] are dimeric. Therefore, it is necessary to determine the native oligomeric state of BADH1 and BADH2. First, we employed MALDI-TOF MS to investigate the mass of the protein. The major peak was observed at a m/z value of 56,628 for [M + H]+1 and 14,663 for [M + 4H]+4 for BADH1 and BADH2, respectively. The calculated molecular masses for monomeric BADH1 and BADH2 from MS results were 56.6 and 58.5 kDa, respectively. These values corresponded to the mass of monomeric BADH1 and BADH2 determined from the amino acid composition for each protein and from SDS–PAGE (Fig 1a). Size-exclusion chromatography was employed to further assess the native state of the enzymes. Standard protein markers were loaded on the column and a graph of Kav versus elution volume was plotted as shown in Supplementary Information Fig S2. Gel filtration of the purified enzymes revealed that the apparent molecular mass of BADH1 and BADH2 was approximately 102 kDa. SDS–PAGE of both enzymes eluted from the gel filtration column gave bands at the same molecular mass as the monomeric protein, confirming that both enzymes were not degraded during the gel filtration process. Taking into account a monomeric molecular mass of around 57 kDa the higher mass observed by gel filtration (102 kDa) suggests that both enzymes are homo-dimeric under physiological conditions. In this respect, the oligomeric state of rice BADHs resembles other plant ALDH enzymes.

3.5 Structural Model for the BADHs–NAD+ Complex

A comparative modeling approach using double templates was carried out to construct the structural model for rice BADHs. The amino acid sequences of both BADH1 and BADH2 were used as templates to search against the Protein Data Bank (PDB) using the BLAST web-server (http://blast.ncbi.nlm.nih.gov/). The crystal structures of amino-aldehyde dehydrogenese from Pisum sativum (code 3IWJ) [26] and betaine aldehyde dehydrogenase from Staphylococcus aureus (code 3FG0) were used as templates for modeling. The eukaryotic enzyme P. sativum BADH is closely related to rice BADH enzymes with high sequence identity (73.3% for BADH1 and 76.5% for BADH2) and sequence similarity (87.3% for BADH1 and 88.1% for BADH2). The crystal structure of S. aureus BADH in complex with NAD+ was used to provide the complete atomic coordinates for the cofactor NAD+. Homology models for both BADH homologues adopted a typical ALDH structure and exhibited low RMS deviations for main-chain atoms between BADH models and the template structures. Analysis of the Ramachandran plot, ProSa Z-scores and Verify3D scores indicated that highly accurate models were obtained (Supplementary Information Table S1–S3, Fig S3). The energy-minimized BADH-NAD+ complexes were used to understand the substrate specificity of both enzymes. The alignment of the BADH1 and BADH2 models with the template and other aldehyde dehydrogenase structures showed that the catalytic residues N164/162, E262/260 and C296/294 in the substrate binding site were conserved whereas residues nearby vary (Fig. 3a, b and Supplementary Information Fig. S4–S5). The grid-based pocket cavity search application (POCASA) web-server [29], which can predict the substrate binding sites by detecting pockets and cavities of proteins with a rolling sphere, was applied to the BADH models to estimate the size of their binding pockets. Although the cofactor binding pockets were comparable for both predicted models, the substrate binding pocket of BADH1 was found to be larger than that of BADH2. In general, the previously reported substrate channel of BADHs was around 12–14 Å in depth and 5–8 Å in width [10, 26]. This cavity volume is much larger than the volume of Bet-ald and thus it could accommodate bulkier substrate. It was previously noted that unlike rice BADHs, P. sativum BADH poorly oxidized Bet-ald due to the absence of negatively charged residues or aromatic residues in close proximity to the quaternary amino group of Bet-ald molecule [26]. Therefore, the difference in substrate pocket size for rice BADHs may lead to differences in substrate specificity between these enzymes.

Fig. 3
figure 3

Structural models of the substrate and cofactor binding sites for BADH1 (a) and BADH2 (b). The possible site for a substrate is shown as a sphere. The structure of the cofactor NAD+ is shown as balls and sticks with a vdW surface. The solvent-accessible surface of the substrate binding site is highlighted for BADH1 (c) and BADH2 (d). Probe spheres rolling on the binding site surface are shown

Full size image

The homology model of BADH2_Y420 reveals that there is an alteration in the conformation of the loop between α21 and β4 (Fig. 4). This altered loop is located near the NAD+ binding site but may not play a direct role in NAD+ binding affinity and substrate binding affinity. The bulky tyrosine residue may decrease the stability of this enzyme or change the hydrogen bonding network around the NAD+ binding site, hence slightly reducing the binding affinity toward NAD+ and substrates. This remains to be further investigated by biophysical and X-ray crystallographic studies.

Fig. 4
figure 4

A schematic representation of the structural overlay of BADH2_Y420 and wild-type BADH2. Only the loops between α21 and β4 are colored in dark. The Y420 insertion located on the loop is shown as dark sticks. The structure of the cofactor NAD+ is in ball and stick representation

Full size image

3.6 GABA Detection

TLC was used to monitor the product of the enzymatic reaction. As proposed previously, the substrate GAB-ald should be converted to GABA by BADH2. GABA at various concentrations was spotted onto the TLC plate to determine the lower limit of GABA detection by TLC. After developing and visualizing by ninhydrin, at least 25 μM GABA could be detected (Fig. 5a). For the enzymatic reaction, GABA could be detected in the reaction mixture containing wild-type BADH2 while in the reaction containing the BADH2_Y420 mutant, only low amounts of GABA were detected (Fig. 5b, c). The TLC results were consistent with the enzyme kinetic studies which showed a lower kcat/Km value for BADH2_Y420 over that of the wild-type. The results indicate that the insertion of Y420, as observed in BADH2.8 from the Myanmar fragrance rice variety, results in a lower BADH activity towards GAB-ald and that BADH2 is possibly involved in the control of the GABA pool which participates in the production of a fragrance molecule, 2-acetyl-1-pyroline, in fragrant rice.

Fig. 5
figure 5

TLC Chromatogram. Various concentrations of the GABA standard were spotted onto the TLC. a A twofold dilution of 100 mM GABA was used (lanes 1–15), b, c show the detection of GABA in the enzymatic reactions of wild-type BADH2 and BADH2_Y420, respectively. Lane 1 control GABA, lane 2 substrate GAB-ald, lane 3 negative control, lane 4–11 twofold dilution of reaction mixture using a substrate concentration of 1,000–3 μM in the enzymatic reaction, and lane 12 a mixture of the reaction without NAD+

Full size image

4 Conclusion

In this study, kinetic analyses of wild-type and four mutant BADH2 enzymes was carried out. The BADH2_Y420 mutant, similar to BADH2.8 from Myanmar fragrance rice, exhibited a lower enzymatic activity GAB-ald but not for Bet-ald. This may lead to the accumulation of GAB-ald/Δ1-pyrroline which can then be converted to 2AP. The catalytic activity of two mutants (BADH2_C294A and BADH2_E260A) was abolished whereas that of BADH2_N162A was vastly decreased. In addition, the kinetic data of wild-type BADH1 and BADH2 were obtained and were similar to those previously reported. We have also demonstrated for the first time the oligomeric state of rice BADHs which is found to form a dimer. Homology modeling of both BADHs revealed differences in the size of the substrate binding pocket which may lead to variations in substrate specificity. The work presented here provides experimental evidences that the insertion of Y420 may lead to 2AP accumulation. Future studies based on the crystallization of this protein in complex with substrates and cofactors are in progress.