Novel glyoxalases from Arabidopsis thaliana



We examined six Arabidopsis thaliana genes from the DJ-1/PfpI superfamily for similarity to the recently characterized bacterial and animal glyoxalases. Based on their sequence similarities, the six genes were classified into two sub-groups consisting of homologs of the human DJ-1 gene and the PH1704 gene of Pyrococcus horikoshii. Unlike the homologs from other species, all the A. thaliana genes have two tandem domains, which may have been created by gene duplication. The six AtDJ-1 proteins (a–f) were expressed in Escherichia coli for enzymatic assays with glyoxals. The DJ-1d protein, which belongs to the PH1704 sub-group, exhibits the highest activity against methylglyoxal and glyoxal, and Km values of 0.10 and 0.27 mm were measured for these two substrates, respectively, while the corresponding kcat values were 1700 and 2200 min−1, respectively. The DJ-1a and DJ-1b glyoxalases exhibited higher specificity towards glyoxal. The other three proteins have either no or extremely low activity for glyoxals. For the DJ-1d enzyme, the residues, Cys120/313 and Glu19/212 at the active site and His121/314 and Glu94/287 at the oligomeric interface were mutated to alanines. As in other enzymes characterized to date, mutation of either the Cys or the Glu residues of the active site completely abolished enzyme activity, whereas mutation of the interface residues produced a variable decrease in activity. DJ-1d differs from its animal and bacterial homologs with respect to the configuration of its catalytic residues and the oligomeric property of the enzyme. When the wild-type DJ-1d enzyme was expressed in E. coli, the bacteria became resistant to glyoxals.


Nucleotide sequence data are available in the GenBank database under accession numbers NM_112361 (DJ-1a), NM_104206 (DJ-1b), NM_119563 (DJ-1c), NM_111140 (DJ-1d), NM_090959.1 (DJ-1e) and NM_115317.5 (DJ-1f).


glyoxalase I/II/III






Mutations in DJ-1 (or PARK7) cause early-onset Parkinson's disease in humans [1]. DJ-1 has been reported to have several functions, including as an antioxidant [2], a redox-dependent chaperone [3] and a protease [4], and in neuronal protection performed via autophagy and mitochondrial cycling [3]. Recently, the bacterial and animal homologs of DJ-1 have been characterized as enzymes that convert glyoxals to carboxylic acids [5, 6]. Glyoxals are reactive 2-oxoaldehydes that are found in cells under various stress conditions [7]. Despite their chemical similarity to other carbohydrates, the reactivity of glyoxals is much higher than that of other sugars due to their inability to form hemiacetals [8]. Under normal physiological conditions, methylglyoxal (MG) is primarily produced by glycolysis as well as by amino acid catabolism, and the oxidative breakdown of glucose also generates glyoxal (GO) [9]. A high concentration of MG is harmful because it reacts with macromolecules such as DNA, RNA and protein. In addition to MG, the glyoxalase II substrate S-d-lactoylglutathione is toxic at high concentrations through its inhibitory effect on DNA synthesis [7].

The glyoxalase pathway consists of two steps involving glyoxalase I and II. Glyoxalase I (GlyI, lactoylglutathione lyase; EC catalyzes the formation of S-d-lactoylglutathione from MG and glutathione. In the second step, S-d-lactoylglutathione is hydrolyzed by glyoxalase II (GlyII, hydroxyacylglutathione hydrolase; EC, thereby releasing glutathione and d-lactate [7]. GlyI has been cloned and characterized in Brassica juncea, and it was suggested that it plays a role in stress tolerance [10]. Eleven homologs of GlyI have been listed in the TAIR database ( [11]. Although GlyI enzymes in Arabidopsis thaliana are highly expressed in response to stresses such as salinity, drought and cold [12], their cellular roles are poorly understood. Five different isoforms of GlyII were initially reported in A. thaliana, of which GlyII-4 and GlyII-5 appear to be expressed in mitochondria, while GlyII-2 is localized to the cytoplasm [13-15]. However, the remote homologs GlyII-1 and GlyII-3 were later identified as metallo-β-lactamase and a member of the human ethylmalonic encephalopathy protein family [16, 17].

Recent studies of three A. thaliana DJ-1 homologs, DJ-1a, DJ-1b and DJ-1c, indicated that DJ-1a is localized to both the nucleus and cytosol, while DJ-1b and DJ-1c are localized to plastids [18]. Further experiments revealed that DJ-1a is induced by several stresses, including oxidation, and that the protein interacts with copper superoxide dismutase 1 and glutathione peroxidase 2. Loss of the DJ-1a gene accelerates cell death in aging plants. A homozygous knockout of the DJ-1c gene was also shown to result in non-viable albino seedlings [19]. We have performed a series of experiments characterizing the enzymatic function of DJ-1 homologs in various species, including bacteria [5], Caenorhabditis elegans, and mammals [6]. The proteins exhibit enzymatic specificities for 2-oxoaldehydes, including MG and GO, without the involvement of any co-factor. This suggests a novel enzymatic mechanism involving cysteine and glutamic acid residues at the active site. In the present study, we analyzed the DJ-1 homologs in A. thaliana in terms of their amino acid sequence similarities and by molecular modeling, characterizing them as novel glyoxalases. The proteins were expressed from cDNA clones and purified to study their glyoxalase activity. We also demonstrated by site-directed mutagenesis that the invariant Cys and Glu residues play a crucial role. Although the important catalytic residues are conserved, the 3D organization of the catalytic structure of the DJ-1 enzymes in A. thaliana is different from that of other species.


Structural modeling-based prediction of glyoxalase enzymes in A. thaliana

We assessed the function of six DJ-1/ThiJ family members in A. thaliana [20] as glyoxalases. Because all the members contain two tandem DJ-1 domains, which were presumably created by gene duplication [19], these domain sequences were analyzed for amino acid sequence similarities (Fig. S1 and Table S2). The modular sequences are connected by a linker of 7–18 amino acids, and have additional N- or C-terminal peptides of < 55 amino acids, which are assumed to be flexible rather than forming a defined secondary structure. The 12 modular sequences were submitted to the Swiss Model Repository to obtain 3D structures based on homology modeling. Although, by and large, the coordinates obtained predicted whole tertiary structures superimposed onto the backbone structures, some structures were found to have a partial deletion on the basis of their lower predictive scores.

Because the structural modeling was based on the two reported 3D structures of the DJ-1 [21] and PH1704 [22] proteins, we attempted to align the modules as two different sub-groups (Fig. 1A,B). Using PyMOL, the putative catalytic sites were visualized with their topological diagrams superimposed on the backbone structures of DJ-1 and PH1704 (Fig. 1C). Except for DJ-1c, the proteins contain an invariant Cys residue in the first (N) and/or second (C) module. Thus, we tested whether the proteins exhibit glyoxalase activities. There is an additional Glu residue (Fig. 1C) in contact with the catalytic histidine located in the other pairing monomer of the dimeric PH1704 protein, which has its catalytic residues located in the dimeric interface. This additional Glu is present in all six modules of the PH1704 homologs.

Figure 1.

Amino acid sequence alignments and predicted catalytic structures of the AtDJ-1 proteins. Sequence alignment-based comparisons of AtDJ-1 proteins with homology to the human DJ-1 (A) and PH1704 (B) proteins are shown. The targeting sequences (cleavage site; blue arrows) in DJ-1b and DJ-1c were predicted using the Predotar [44] and Mitoprot programs [45]. The linker regions of the AtDJ-1 proteins are indicated by green boxes; the homologs of human DJ-1 have 17–18 amino acid linkers, while those of the homologs of PH1704 consist of 7–11 amino acids. The catalytic residues (red arrows) for the AtDJ-1 proteins were predicted based on comparisons between the human DJ-1/Hsp31 structure and the predicted AtDJ-1 models. (C) DJ-1 homologs (Protein Data Bank ID 1pdv [21]) and PH1704 homologs (Protein Data Bank ID 1g2i [22]) have conserved putative catalytic residues. The putative catalytic residues in DJ-1a, DJ-1b and DJ-1c are similar to those of DJ-1. Both DJ-1a and DJ-1b have tyrosines instead of histidine in the catalytic sites in the N domain, as is also found in the DJ-1β protein of Drosophila melanogaster. Notably, the C domain of DJ-1a has the same catalytic residues as human DJ-1. The cysteine residues typically found in the predicted catalytic sites of the AtDJ-1 enzymes are not conserved in DJ-1c. The catalytic residues found in PH1704 are conserved in the N domains of the PH1704 homologs. However, only the DJ-1d enzyme exhibits conservation of the catalytic residues of the C domain. Additional Glu residues (pink) are expected to contact the catalytic histidines in the paired domains of PH1704 protein dimers and their A. thaliana homologs because the catalytic residues are located at the dimeric interface. The structures were illustrated using the PyMOL Molecular Graphics System, Version Schrӧdinger, LLC.

Expression and enzymatic characterization of the putative glyoxalases

For the six AtDJ-1 genes, we attempted to purify their encoded proteins by cloning the corresponding cDNAs into the pET15b and pET21a vectors, which contain the T7 promoter and a His tag at the N- or C- terminus of the recombinant protein. The clones expressed considerable amounts of soluble protein (data not shown). The purified homologs of DJ-1 (DJ-1a and DJ-1b) and PH1704 (DJ-1d, DJ-1e and DJ-1f) were analyzed by gel filtration chromatography and were found to have dimeric and trimeric forms, respectively (Table 1). We were unable to recover DJ-1c protein due to its unstable nature in Escherichia coli. We performed analytical ultracentrifugation of the DJ-1d protein, and obtained a molecular mass of approximately 112 kDa (Fig. S2). The trimeric form was also observed under an electron microscope (Fig. 2). As a means of verifying the integrity of the purified proteins, the DJ-1 homologs were analyzed using far-UV CD spectroscopy, and similarities in secondary structure were observed (Table 2). DJ-1a and DJ-1b exhibit significant specificity toward GO rather than MG (Table 3). The enzymatic activities of the glyoxal-specific glyoxalases were further analyzed by determining their kinetic constants as shown in Table S3. DJ-1d showed much higher specific activities with MG and GO than found for the other AtDJ-1 proteins (Table 3). We determined the enzyme kinetics of the DJ-1d protein with MG and GO substrates; a Km of 0.10 mm and a kcat of 1700 min−1 were determined for MG, and a Km of 0.27 mm and a kcat of 2200 min−1 were determined for GO (Fig. S3 and Table 4) The glyoxalase activities of DJ-1e and DJ-1f with MG were extremely low, while the activities of DJ-1c with glyoxals were undetectable (Table 3).

Table 1. Size estimation of AtDJ-1 proteins by gel filtration chromatography
AtDJ-1Molecular mass (kDa)aEstimation (kDa)bPrediction
  1. a

    Calculated molecular weight.

  2. b

    Estimated molecular weight.

Table 2. Secondary structure content estimated by circular dichroism
Secondary structure (%)AtDJ-1aAtDJ-1bAtDJ-1dAtDJ-1eAtDJ-1f
β-sheet, anti-parallel17.116.917.317.517.4
Random coil33.131.434.833.033.2
Total (%)100.0100.0100.0100.0100.0
Table 3. Enzymatic activities of the AtDJ-1 homologs. AtDJ-1 proteins were assayed at 25 °C with 5 mm glyoxals. The specific activities of the negative controls measured using 2,4-dinitrophenylhydrazine were 3.1 ± 0.6 and 1.5 ± 0.8 nmol·min−1·mg protein−1 for MG and GO, respectively. The activities were determined in triplicate. Values are means ± SEM
AtDJ-1Specific activity (nmol·min−1·mg protein−1)
  1. a

    These activities were essentially the same as those of the negative controls. The remaining activity values were obtained by subtracting the activity value of the corresponding negative control from the value measured for each experimental sample.

a15 ± 4250 ± 12
b13 ± 3310 ± 5
d8600 ± 20011 000 ± 600
e6.2 ± 1.7a
f5.5 ± 1.1a
Table 4. Comparison of kcat values between DJ-1d and its catalytic mutants. The values were obtained at 45 °C in triplicate, and the mean values are presented
DJ-1dkcat (min−1)
  1. a

    The Km values for wild-type DJ-1d with MG and GO were 0.10 and 0.27 mm, respectively.

  2. b

    These activities were essentially the same as those of the corresponding negative controls. The kcat values of the negative controls with the MG and GO substrates were 0.1 and 0.1 min−1, respectively. The activity values of the appropriate negative controls were subtracted from the values obtained for the experimental samples.

Figure 2.

Determination of the molecular shape of DJ-1d. Metal-shadowed images showing the molecular shape of the DJ-1d complex. Arrowheads indicate complexes found in the electron microscope field. The inset shows a series of single images (upper two rows) and representative averaged images (lower row). The scale bar in the inset image represents 20 nm.

Characterization of the catalytic residues by mutagenesis

Because the organization of the catalytic residues found in the DJ-1d enzyme is different from that of its homologs in E. coli and in animals, we investigated their roles in enzyme catalysis by mutating each residue. The primary targets of mutagenesis were the three residues Glu, Cys and His, which are presumed to form a catalytic pocket, and the Glu residue that is found at positions 94 and 287 in the catalytic modules and is predicted to be located in the intermolecular interface. Because each gene encodes a repeated protein structure comprising ‘N’ and ‘C’ domains in a single polypeptide, mutated constructs containing double mutations in both domains were tested alongside single mutation constructs (Fig. S4). When the core cysteines were altered in both domains, C120A and C313A, the activity of the enzyme was completely abolished for both MG and GO substrates (Table 4). Mutation of Glu19 and Glu212 to Ala almost completely eliminated the activity of the enzyme (Table 4). The effect of mutating the two histidines (H121A and H314A) was found to be slightly greater than that of Glu double mutations (E94A and E287A) at the interface. Each single mutation made in the N and C domains reduced the activity of the enzyme by approximately 50% (Table 4). Interestingly, mutation of the additional Glu residues E94 and E287 was found to have a significantly greater effect on the activity of the enzyme than that of the mutations in the three catalytic residues. The reason for this difference is not yet clear.

Arabidopsis glyoxalases are functional in bacterial cells

Because the DJ-1d enzyme exhibits similar biochemical characteristics to that of its E. coli homolog, we tested whether the plant enzyme confers resistance to glyoxals in bacterial cells. In a strain of E. coli lacking the endogenous glyoxalases I (gloA) and III (hchA), the DJ-1d enzyme was expressed under the control of the bacterial promoter (see 'Results'), and was tested for its ability to detoxify MG or GO added exogenously. As shown in Fig. 3, the glyoxalase-deficient strain is unable to grow on LB agar medium containing more than 0.5 mm MG, whereas wild-type bacteria grow even on medium containing 0.7 mm MG. When the wild-type DJ-1d enzyme was expressed, the glyoxalase-negative strain became resistant to 0.7 mm MG, which is comparable to the resistance level exhibited by the wild-type E. coli. The double mutations of each catalytic residue (C120/313A, E19/212A and H121/314A) almost completely abolished the MG resistance conferred by the wild-type DJ-1d enzyme (Fig. 3). Glyoxalase I (gloA) and III (hchA) deficiencies in E. coli do not result in a significant increase in GO sensitivity. However, heterologous expression of the DJ-1d enzyme provided the bacterial cells with additional resistance to GO, thereby making them resistant to GO concentrations of more than 8 mm. The mutations in the DJ-1d catalytic residues were found to eliminate the resistance to high concentrations of GO. Small but clear differences were noted between the effects of the C120/313A, E19/212A and H121/314A double mutations.

Figure 3.

Mutational effects of glyoxalase activity observed in E. coli. pET21a and the plasmids carrying cDNAs encoding the wild-type and mutant (C120/313A, E19/212A and H121/314A) DJ-1d proteins were used to transform the MG1655 ∆gloΙhchA strain. The MG1655 strain was used as the wild-type E. coli strain. The DJ-1d-transformed cells exhibited significant increases in MG and GO tolerances compared to the MG1655 ∆gloΙhchA background, but the effects of the empty pET21a vector and the C120/313A and E19/212A mutants on glyoxalase activity were almost negligible. Because the activity of H121/314A mutants were a little bit exhibited (Table 4), the resistances of MG and GO were higher than other catalytic mutants.


We report here that the DJ-1/PfpI superfamily members found in A. thaliana exhibit glyoxalase activity as found for other family members in bacterial [5] and animal [6] species. The total number of family members exceeds more than 1000 in the Pfam database [20]. A previous attempt to determine the evolutionary and functional relationships within the DJ-1 superfamily was made using 311 homologous sequences obtained by PSI-BLAST with human DJ-1 as the seed sequence [23]. The study identified a sub-group of eukaryotic DJ-1 homologs that were most closely related to the bacterial ThiJ (YajL) gene. Although the ThiJ protein of E. coli was suggested to be a close relative of human DJ-1 on the basis of its 3D structure [24], the function of ThiJ has remained obscure [25]. The six A. thaliana homologs studied here were not specifically clustered in the previous analysis [23], and do not belong to the DJ-1 or ThiJ sub-groups. A further attempt to classify the eight DJ-1 family proteins with known 3D structures used the predicted active-site residues as a basis for comparison [26]. The results indicated the existence of three sub-classes, the first class consisting of DJ-1 and YajL, the second comprising the probable proteases PH1704 of Pyrococcus horikoshii and YhbO of E. coli, and the third group containing the YDR533C protein of Saccharomyces cerevisiae and the Hsp31 (HchA) protein of E. coli. Our previous characterizations of Hsp31 [5] and DJ-1 [6] as glyoxalases revealed the existence of the catalytically essential residues Cys and Glu, together with a His residue that assists catalysis. However, the organization of the catalytic residues in DJ-1 and Hsp31 is different in terms of histidine positioning. The A. thaliana enzyme DJ-1d resembles Hsp31 in that the Cys and His residues are contiguous (Fig. 1). In contrast, the N domain containing predicted catalytic residues in DJ-1a and DJ-1b resembles that of DJ-1β of D. melanogaster [27], in which tyrosine is present instead of histidine. However, the C domain of DJ-1a resembles that of human DJ-1. Despite the similarities in the catalytic core, the local environment of the active site and the oligomeric organization are quite variable between the members of this protein family (Fig. 1 and Table 1).

The similarities between the DJ-1d enzyme and the PH1704 protein of P. horikoshii include their sequence similarity and their predicted catalytic residues, and they are also evident in the results of homology-based structural modeling, gel filtration chromatography, and sedimentation equilibrium analysis. The X-ray structure of PH1704 reveals that it forms a hexamer [22]. Our electron microscopy data for DJ-1d indicate that it has a trimeric ring structure (Fig. 2). The predicted catalytic residues face each other on the oligomeric interfaces (Fig. S4). When we mutated the core residues of the catalytic pocket, the enzyme activity with MG was almost completely lost. This loss of activity is apparent when the kcat values of the E19/212A and C120/313A mutants (< 1 min−1) are compared to that of the wild-type enzyme (1700 min−1) (Table 4). Interestingly, the E94/287A mutation reduces the activity of the enzyme by more than 99% compared to the wild-type. A possible explanation is that the Glu residues are involved in mediating the enzymatic activity of the two facing catalytic pockets (Fig. S4), perhaps via residue interactions. Alternatively, the two active sites may be functionally coupled through the E94/287 residues. This is in line with the observation that the E94 and E287A single mutations reduced the activity with glyoxals by considerably more than 50% (Table 4). The observation that a mutation in either E94 or E287 or in both results in no obvious protein disintegration (Table S4) (gel filtration data not shown) excludes protein disassembly or instability as a possible explanation of the observed effect, although we are unable to detect small conformational alterations in the protein.

The tandem repeat structure of the AtDJ-1 enzymes presents the possibility of a kinetic difference between the monomers. This has been observed in the dimeric glyoxalase from Plasmodium falciparum [28]. The reduction of kcat values by approximately 50% that was observed for the C120A and C313A mutants compared to the wild-type with the MG substrate indicate that the kinetic properties of two domains are not considerably different. However, a substrate-specific difference was noted, as the decreases in the kcat values observed for the C120A and C313A mutant enzymes with GO were 60% and 30%, respectively. We also demonstrated that DJ-1d exhibits in vivo enzyme activity when expressed heterologously in E. coli. This was manifested as a compensatory effect on the MG sensitivity that is caused by deletion of the gloA and hchA genes (Fig. 3). The GlyI enzyme, together with GlyII, is primarily responsible for MG detoxification [29, 30]. Furthermore, the specificity of glyoxalase III (GlyIII) is also directed toward MG [5]. However, multi-copy expression of DJ-1d also conferred additional resistance to glyoxal.

In plants, an increased expression of GlyI was reported to be associated with rapid cell growth [31], a high demand for ATP generation, and enhanced glycolysis [32]. There are three GlyII enzymes in A. thaliana, and the mitochondrial and cytoplasmic isozymes may play different roles in the cell [13-17]. Recently, DJ-1a of A. thaliana was shown to be involved in the response to oxidative stress and associated with cytosolic superoxide dismutase 1 activation [18]. DJ-1b and DJ-1c were reported to be localized to plastids [18], and DJ-1c has also been shown to play an essential role in chloroplast development [19], but the function of DJ-1b is unknown at present. Because DJ-1a and DJ-1b exhibit weak glyoxalase activities, they may be involved in detoxification of glyoxals in the cytosol and the plastids, respectively. The lack of glyoxalase activity associated with DJ-1c appears to be due to the absence of an essential cysteine residue in the active site (Fig. 1). DJ-1d is localized in the cytosol [33] and is widely expressed in seeds, roots, leaves and pollen [34]. Thus, it may be a major detoxification enzyme in A. thaliana. Because DJ-1e and DJ-1f do not have catalytic residues in their C domains, they displayed extremely low enzyme activities. Their exact functions are also unknown.

Our results indicate that the cysteine and glutamic acid residues of DJ-1d play a critical role in catalysis. Based on the predicted 3D structure (Fig. S4), we propose a glyoxalase mechanism (Fig. 4) in which the thiol group of the cysteine residue plays a role similar to that of glutathione in the GlyI/II enzyme [35]. It is likely that the catalytic cysteine (C120 or C313) attacks the aldehyde group of the substrate non-enzymatically and thereby forms a hemithioacetal intermediate (Fig. 4B), and this process may be reversible. The backbone amino groups of the H121/314 residue close to the active cysteine may stabilize this intermediate. The E19/212 residue located close to the side chain of the cysteine appears to be involved in acid–base catalysis, in which the residue abstracts a proton, resulting in formation of an enediol intermediate (Fig. 4C), in which the negative charge may be stabilized by the oxyanion hole formed from the two backbone NH groups of G90 and H121. Subsequently, the oxygen of E19/212 transfers a proton from C1 to C2 during the isomerization step, resulting in formation of S-lactoylcysteine (Fig. 4D). H121/314 may serve as a base and remove a proton from a water molecule, thereby facilitating hydrolysis of the thioester substrate. The positive charges of H121/314 may be stabilized by the nearby E94/287 located in the neighboring domain. Finally, the carbon–sulfur linkage between the enzyme and the substrate is broken by hydrolysis, which leads to the release of lactic acid (Fig. 4E).

Figure 4.

Catalytic scheme for the DJ-1d enzyme. A, Initial state; B, Proton abstraction; C, Enediol intermediate; D, Reaction intermediate; E, Product formation

Experimental procedures

cDNA cloning and protein expression

The protein coding regions of the cDNAs were obtained by PCR amplification using total cDNA prepared from 4-week-old leaves of Arabidopsis thaliana (Col-0), and were cloned into pMD-18T (TaKaRa, Dalian, China), pET15b and pET21a (Novagen, Darmstadt, Germany). The nucleotide sequences of the primers used for PCR are listed in Table S1. The pET15b and pET21a vectors were designed to attach the His tag to the N- or C-termini of the protein. The DJ-1b and DJ-1c genes were cloned without their N-terminal signal sequences.

Mutagenesis of catalytic residues

BglII-digested fragments (619 bp) containing partial coding sequences encoding 111–317 amino acids were amplified from pET21a-DJ-1d-His using the mutated primers for C313A, C120/313A and H314A (Table S1), and were cloned into the pMD-18T vector. The mutated BglII fragments were then substituted for the corresponding regions of the wild-type cDNA in pET21a-DJ-1d-His. The mutated sequences were confirmed by DNA sequencing of both strands. To generate the H121/314A mutant, two PCR products amplified using the primer sets shown in Table S1 were mixed and used as a template for a second PCR reaction. The PCR product was cloned into the pMD-18T vector and used as a substitute for the corresponding wild-type BglII-flanked region in pET21a-DJ-1d-His. For mutagenesis of E19A, E287A, E19/212A and E94/287A in the DJ-1d gene, two PCR products amplified with the primer sets shown in Table S1 were mixed and used as a template for a second PCR reaction. The PCR product was cloned into pET21a. The mutated sequences were confirmed by DNA sequencing of both strands.

Modeling of glyoxalases from A. thaliana

The cDNA sequences of candidate glyoxalases were obtained from the Pfam database [20]. The tertiary structures of the translated sequences were predicted using the SWISS-MODEL program [36]. Comparisons of protein sequences were performed using the ESPript2.2 program [37].

Expression and purification of recombinant AtDJ-1 proteins

The recombinant DJ-1d plasmid was used to transform the BL21 (DE3) strain of E. coli. Cells were grown to an attenuance at 600 nm of approximately 0.4 at 37 °C in Luria–Bertani (LB) medium containing 100 μg·mL−1 ampicillin (USB, Cleveland, OH, USA), and AtDJ-1 protein expression was induced by the addition of 0.25 mm isopropyl-β-d-thiogalactoside (Duchefa, Haarlem, The Netherlands) followed by incubation at 30 °C. After 6 h, the cells were harvested and then resuspended in lysis buffer (50 mm NaH2PO4, 300 mm NaCl, 10 mm imidazole, 10 mm β-mercaptoethanol, pH 8.0) before sonication on ice. The His-tagged AtDJ-1 proteins were purified using Ni2+ affinity columns (Qiagen, Hilden, Germany) in accordance with the manufacturer's instructions, and dialyzed against 20 mm sodium phosphate buffer (pH 6.8) containing 1 mm dithiothreitol and 1 mm EDTA. The protein concentration was determined using the Bradford assay [38], and the purity was confirmed by SDS/PAGE.

Enzyme activity assays

Enzyme activities with MG and GO substrates (M0252 and 50649, respectively; Sigma, St. Louis MO, USA) were determined using a modified 2,4-dinitrophenylhydrazine assay and colorimetry [6]. Briefly, 500 μL of a reaction mixture containing various concentrations of substrate, purified DJ-1a, DJ-1b and DJ-1d proteins and 20 mm sodium phosphate (pH 6.8) was incubated at 37 °C (DJ-1a and DJ-1b) or 45 °C (DJ-1d). At the end of the reaction, 30 μL aliquots were mixed with 90 μL of a solution containing 0.1% 2,4-dinitrophenylhydrazine and 2 m HCl and 210 μL of distilled water. The reaction was stopped by addition of 420 μL of 10% NaOH. The amount of remaining MG or GO was determined using a DU 800 spectrophotometer (Beckman Coulter, Inc., Fullerton, CA, USA) at 540 and 570 nm wavelengths, respectively. The Km values of DJ-1a, DJ-1b and DJ-1d were calculated using Sigmaplot 12.0 (Systat Software Inc., San Jose, CA, USA) .

Electron microscopy and single particle analysis

For metal shadowing, the purified proteins were diluted with a 20 mm sodium phosphate solution containing 50% glycerol at pH 6.8 to give a final concentration of 500 nm, and then sprayed onto freshly cleaved mica for rotary shadowing at an angle of 6° as previously described [39]. Grids were examined using a Technai G2 Spirit Twin transmission electron microscope (FEI, Hillsboro, OR, USA) operated at 120 kV, and the images were recorded on a 4K × 4K Ultrascan 895 CCD camera (Gatan, Inc., Pleasanton, CA, USA). Single particle image processing was performed using the SPIDER program (Health Research Inc., Rensselaer, NY, USA). One hundred and eighty-two well-separated particles from metal-shadowed micrographs were used for processing, and class averages were calculated by the reference-free method as previously described [40].

Estimation of the size of the purified enzymes

Size-exclusion chromatography experiments were performed at 4 °C. The purified proteins were loaded onto a Superdex 200 HR column (GE Healthcare Life Sciences, Piscataway, NJ, USA) at a flow rate of 1.0 mL·min−1 in a solution containing 20 mm sodium phosphate and 100 mm NaCl (pH 6.8). The column was calibrated under the same experimental conditions using the following calibration mixture: bovine thyroglobulin (670 kDa), bovine γ-globulin (158 kDa), chicken ovalbumin (44 kDa), horse myoglobin (17 kDa) and vitamin B12 (1.35 kDa). Sedimentation equilibrium experiments on DJ-1d dissolved in 20 mm sodium phosphate buffer (pH 6.8) were performed at 4 °C in a ProteomeLab XL-A analytical ultracentrifuge (Beckman Coulter, Inc., Fullerton, CA, USA) using an An-60 Ti rotor and two-channel epon charcoal-filled centerpieces at Mokpo National University (South Korea). Samples with an initial absorbance at 280 nm of 0.25 were centrifuged at 11 600 g. Radial absorbance scans of triplicate samples were collected in continuous scan mode at a wavelength of 280 nm and with a step size of 0.001 cm. Global fitting of the datasets was performed using the program SEDPHAT [41]. The partial specific volume of DJ-1d was adjusted using the molecular weight-dependent protein density function [42].

Circular dichroism (CD) spectroscopy

CD spectra for A. thaliana proteins prepared in buffer (50 mm sodium phosphate, pH 6.8) were obtained at 25 °C using a Jasco J-710 spectropolarimeter (Japan Spectroscopic, Tokyo, Japan). CD spectra in the far-UV region (200–250 nm) were acquired at a scan rate of 50 nm·min−1 using a cell with a path length of 0.1 cm. For each sample, the scans were repeated five times, and corrections were made subsequently by subtracting the buffer spectrum. The experimentally obtained ellipticity of the far-UV CD spectra was converted to mean residue ellipticity, from which the secondary structure content was estimated using the CDNN program [43].

Spotting assay

To express the DJ-1d enzyme under the control of the E. coli promoter, two oligonucleotides (5′-CTAGAATTATAGGTACCTATTCCGGCCTGTCAAGT-3′ and 5′-ATACTTGACAGGCCGGAATAGGTACCTATAA TT-3′) containing the −10 box and −35 box consensus sequences derived from E. coli promoters were synthesized, hybridized, and used as a substitute for the region of the pET21a-based DJ-1d expression plasmids flanked by XbaI and AccI. The transformed cells were grown to an attenuance at 600 nm of 1.0, and serially diluted (10−1–10−6) before spotting onto LB plates containing MG or GO at the indicated concentrations, and this was followed by incubation at 37 °C overnight.


This work was supported by grants from the National Research Foundation of Korea (grant number 2012041712) to C.P., from the National Research Foundation of Korea (grant number 2011-0007189) to K.B., and from the Korea Basic Science Institute (grant number T33415A) to J.K.H. We thank Hee-Yoon Lee, Junsang Ko and Seong Oak Park for technical assistance.