Introduction

Glyoxalase I (GlxI) catalyzes conversion of methylglyoxal (MG), a cytotoxic α-oxoaldehyde, to S-D-lactoylglutathione (SLG). The product thus formed, is then converted to D-lactate by the second enzyme of the pathway namely, glyoxalase II (GlxII). This pathway is ubiquitously present in all forms of life, from ancient archaea to bacteria and higher eukaryotes and is the major pathway for detoxification of MG which is generated spontaneously during the course of glycolysis. However, enzyme catalysed synthesis of MG by methylglyoxal synthase, cytochrome P450, and amine oxidase has also been observed in bacteria1,2 but no such activity has yet been reported in plants. In humans, increased levels of MG have been linked with the development of vascular complications of diabetes, such as nephropathy, retinopathy, neuropathy and cardiovascular disease3,4 and glyoxalase system is thought to protect against pathogenesis. In plants this pathway is thought to have an important role in conferring tolerance to multiple abiotic stresses5,6,7,8.

The GlxI enzyme has been characterized from several sources, including plants and animals. It is a metalloenzyme, requiring either Ni(II) or Zn(II) for its catalytic activity9. In plants, GlxI is present in both dicots and monocots, including Brassica juncea5, Glycine max10, Oryza sativa11, Triticum aestivum12 and Thlaspi caerulescens13. Glyoxalase enzymes have also been extensively studied in microbial systems9,14,15. Based on the metal ion specificity, GlxI proteins have been categorized into two groups, one requires Ni(II) and/or Co(II) for activity and was earlier thought to be mainly of prokaryotic origin. The second group enzymes are dependent on Zn(II) for activation and are thought to be of eukaryotic origin. However, the discovery of newer glyoxalase genes suggests the presence of both the metal dependent forms in prokaryotes as well as in eukaryotes, thereby emphasizing the need for revision in classification. For example; Pseudomonas aeruginosa16 and rice code for enzymes from both groups (our unpublished work).

The GlxI enzymes are usually dimeric in nature but monomeric forms of these enzymes have been reported in few species, such as yeast, Plasmodium and wheat. In these organisms the enzymes have two-GlxI domains17,18,19. Moreover, a genome wide analysis of glyoxalase genes in rice and Arabidopsis20 has suggested their presence as a multi-gene family containing 11 GlxI genes.

The present study was taken up with the aim of tracing the origin of metal ion requirement of GlxI and evolutionary pathway leading to expansion of glyoxalase genes as multi-gene family in plants. Our in silico studies revealed that the Ni(II) dependent form of GlxI is the primitive form and subsequent evolution has led to the origin of Zn-dependent GlxI. We also investigated the origin of co-existence of both the metal-dependent forms in an organism. Moreover, an expansion of Ni(II) dependent form was observed in plants, whereas in yeast and animals there was a loss of Ni-GlxI with few a exceptions such as Bombus impatiens. Our results thereby suggest multiple pathways for evolution of different forms of GlxI.

Results

Taxonomic distribution of GlxI and origin of different metal ion specificity

The distribution of GlxI in different subdivisions of the “Tree of Life” is not very well explored and has been extensively studied in only few species16,18,21,22. Our analysis revealed a mosaic distribution of GlxI when mapped on the species tree (data not shown). In bacteria, we could identify GlxI only in proteobacteria and cyanobacteria. However, a Ni activated GlxI has been recently characterized from Clostridium acetobutylicum, a gram positive bacterium15. In order to understand the reason for not picking up the gene from gram positive organisms, we examined the corresponding protein sequence (AAK80149) for the type of Pfam domain unique to GlxI. The C. acetobutylicum GlxI was found to possess Pfam domain ‘PF13669.1’ instead of ubiquitous domain for GlxI (PF00903.20) and exhibited only 22% identity with E. coli Ni-GlxI protein. Therefore it appears that this gene has undergone high level of divergence in gram positive bacteria. Moreover, GlxI was found to be missing in many eukaryotes, such as Euglenoids, Red Algae, Glaucophytes and Haptophytes. Thus, the gene tree of GlxI was patchy.

Next, the individual distribution of Zn- and Ni-GlxI in different organisms was investigated. The results showed that the Zn dependent form is absent in archaea and cyanobacteria indicating that Zn-GlxI may have originated in proteobacteria. Since deltaproteobacteria is the ancient subdivision of proteobacteria23, their genomes were analyzed by BLAST algorithm using sequences of Ni- and Zn-GlxI domains from E. coli and P. aeruginosa (GloA3) respectively. The results revealed the presence of Ni-GlxI and Zn-GlxI in Myxococcales and Bdellovibrionales, respectively. Interestingly, both these species reside in the same ecological niche. We however, failed to identify the instances of simultaneous presence of both forms of GlxI in the above mentioned subgroups. Further, we compared the protein sequence of GlxI domains from Myxococcus xanthus DK 1622 (Ni binding) and Bdellovibrio bacteriovorus HD100 (Zn binding). An insert of two conserved modules (A and B) of 16 aa (DVPTENEANARYAFGR) and 9 aa (YHNGNTEPR) separated by a 18 aa spacer was observed in Zn binding GlxI from B. bacteriovorus in comparison to the Ni-GlxI (Fig. 1). These extra sequences were not unique to Zn-GlxI of B. bacteriovorus but were also present in many other Zn-GlxI proteins with variations in sequence and size. In some organisms B module is further split into two suggesting that these changes may have a bearing on evolution of metal ion requirement and functional adaptation (Fig. 2). We were not able to find the origin of these inserts.

Figure 1: Pairwise alignment of respective Ni and Zn-GlxI domains from delta proteobacterial species, Myxococcus xanthus DK 1622 (Mx108756890) and Bdellovibrio bacteriovorus HD100 (Bb42524256).
figure 1

The bold letters indicate loop regions (A and B) that exist only in the Zn-GlxI.

Full size image
Figure 2: Multiple alignment of representative sequences of Ni and Zn-GlxI domains.
figure 2

Zn-GlxI from Bdellovibrio (Bb42524256) and Glycine max (Gm255630246) cluster together as shown in blue box whereas Ni-GlxI from Myxococcus (Mx108756890) and Glycine max (Gm356531939) form another cluster enclosed in red box. The two domains of Glycine max (Gm356531939) have been taken as separate sequences. Arrows indicate the regions specific to Zn-GlxI.

Full size image

In yeast and animals, reports for the presence of only Zn-GlxI exist18,21, but our analysis showed that even though there is preponderance of Zn-GlxI in animal kingdom, there are organisms which encode both Ni-GlxI and Zn-GlxI (Amblyomma maculatum and Amphimedon queenslandica) or only Ni-GlxI (Bombus impatiens).

Origin of simultaneous existence of both metal ion binding forms of GlxI

Pseudomonas aeruginosa, a gamma proteobacteria, encodes both Ni and Zn forms of the enzyme; GloA1, GloA2 (both Ni binding), and GloA3 (Zn binding)16. This motivated us to track the evolutionary origin of both forms of the enzyme together in one organism. This was done by using BLAST algorithm and domain sequences of Ni- and Zn-GlxI forms respectively from E. coli and Pseudomonas aeruginosa. The sequences of genes, representing ortholog sequences from various subdivisions of three major domains of life, were extracted and cross checked to make sure that these are from respective groups. The two betaproteobacterial species, Methylibium petroleiphilum PM1 and Limnobacter sp. MED105 were found to encode both Ni- and Zn-GlxI genes. Both the organisms belong to the order Burkholderiales, suggesting that either the Zn or the Ni binding GlxI has been horizontally transferred to the ancestor of Burkhoderiales.

Evolution of two-domain GlxI and eventual loss of a two-domain Zn-GlxI

The GlxI originated as a single domain protein irrespective of their metal ion specificity, as both Ni- and Zn-GlxI are single domain proteins in bacterial species (domain length of 120 aa and 142 aa respectively). However during the evolution of eukaryotes from prokaryotes, gene duplication and gene fusion events might have given rise to two-domain GlxI (Fig. 3). The alignment of the representative sequences of Ni- and Zn-GlxI is depicted in Figure 2 whereas the multiple alignment of all the orthologous sequences under study is shown in Supplementary Fig. S1. Ni-GlxI is present as a two-domain protein in all eukaryotes and among the early branching eukaryotes, algae appears to be the first to encode this gene. However in contrast, only a few organisms have two-domain Zn-GlxI (domain length of 144–146). Ectocarpus (Brown Algae) is the only organism that was found to harbor both the single- and the two-domain forms of Zn-GlxI. Other organisms that have two-domain forms of GlxI are diatom Phaeodactylum tricornutum and apicomplexan Toxoplasma gondii. To our knowledge other eukaryotes do not encode the two-domain form of Zn-GlxI. To address as to why the two-domain Zn-GlxI has not been retained further in the eukaryotes, we analyzed the protein sequences of two-domain Zn-GlxI. The sequences of two-domain Zn-GlxI acquired more number of substitutions as compared to the single domain Zn-GlxI and these substitutions are more significant in Ectocarpus. In higher eukaryotes, these substitutions might be unfavorable for gene functionality and thus the two-domain Zn-GlxI has been lost in other eukaryotic genomes with the retention of only the single domain Zn-GlxI proteins.

Figure 3: Pictorial depiction of observed domain architecture in GlxI proteins.
figure 3

The Ni- and Zn-GlxI exist as both single- and two-domain proteins. The GlxI proteins from E. coli and P. putida have been used for representation of single domain GlxI and G. max and E. siliculosus for two-domain GlxI. Metal ion specificity has been indicated at right side for the respective proteins. The domain start and end positions are also mentioned.

Full size image

Sequence motifs and putative sub-cellular localization studies trace the evolutionary mechanism

In order to decipher the mechanism of evolution of metal binding characteristics, we examined the substitution pattern in the metal binding domains of GlxI. For this, we aligned Zn- and Ni-GlxI domain sequences separately. Further, the individual alignments were reordered according to the species tree and each site was examined independently. Proteobacterial Zn-GlxI sequences were placed in a sequential order of evolution as delta, epsilon, alpha, beta and gamma proteobacteria except for Sphingomonas and Candidatus pelagibacter whose sequence motifs were found to be closer to the eukaryotic proteins than that of bacterial counterparts. In bacterial species the 6 amino acid long motif ‘L/FNHTML’ named as start motif (see Supplementary Fig. S2) was found to be conserved with one exception, Sphingomonas where third position of the motif is substituted to ‘Q’. We suggest that Zn-GlxI from Sphingomonas was acquired by the ancestor of the brown algae before separation from plants. Many motifs present in plant and other eukaryotic Zn-GlxI domains were found to be similar to Sphingomonas (Supplementary Fig. S2), suggesting that the Sphingomonas GlxI may be closer to Brown Algae than to other bacterial proteins. This has been further supported by generating Zn-GlxI domain sequence similarity values by pairwise comparison of each of the eukaryotic Zn-GlxI protein under study with Sphingomonas and C. pelagibacter (Supplementary Table S1). The start motif of Zn-GlxI derived from Saccharomyces cerevisiae was however, found to be similar to C. pelagibacter and M. petroleiphilum. The pairwise comparisons indicate the origin of Zn-GlxI in yeast from C. pelagibacter protein (Supplementary Table S1).

In case of two-domain Zn-GlxI, pairwise alignment of both the domains with those of Sphingomonas and C. pelagibacter showed that the two domains are likely to have different evolutionary origin deduced from sequence closeness to either Sphingomonas GlxI domain or to C. pelagibacter respectively. But Toxoplasma gondii is an exception where both the domains have higher similarity to C. pelagibacter than Sphingomonas (Supplementary Table S1). Moreover, single domain Zn-GlxI of plant, green and brown algae are more similar to Sphingomonas GlxI in contrast to the corresponding sequences from human, yeast and Amphimedon queenslandica. The later were found to be closer to C. pelagibacter. The only exception to this pattern is Amblyomma maculatum, an arthropoda, whose GlxI shares higher degree of sequence similarity with GlxI protein from Sphingomonas (a typical of plant GlxI). Taken together we believe there have been two rounds of acquisition of Zn-GlxI. First, an acquisition of a single domain Zn-GlxI from C. pelagibacter has taken place in the ancestor of eukaryotes, and subsequently Zn-GlxI from Sphingomonas got transferred to the ancestor of plant lineage (see Figure 4 for schematic depiction of evolution of Zn-GlxI). Further, the single domain Zn-GlxI from Ectocarpus was found to have higher degree of similarity with the first domain (based on location) of two-domain Zn-GlxI as compared to the second domain. It is likely that in the ancestor of brown algae and diatom for two-domain Zn-GlxI, a duplication of Sphingomonas originated Zn-GlxI might have occurred. Later one copy of the duplicate might have got fused with the pre-existing C. peligabacter Zn-GlxI and hence gave rise to two-domain Zn-GlxI whereas the other copy remained single domain protein (Fig. 4). T. gondii is an exception to this. In Ectocarpus genome, the single copy Zn-GlxI resides along with two-domain Zn-GlxI. But, in P. tricornutum and T. gondii the single domain Zn-GlxI has got deleted.

Figure 4: Proposed mechanism for evolution of Zn-GlxI.
figure 4

The arrows represent point of acquisition of Zn-GlxI from Candidatus pelagibacter (Cp) [black] and Sphingomonas (Sp) [green] in ancestors of eukaryotes and plants respectively. Loss of single domain Zn-GlxI has been marked as red asterisks whereas loss of two-domain Zn-GlxI has been shown by yellow astericks. The box depicts the scheme of evolution of single and two-domain Zn-GlxI in the ancestor of Brown algae and Diatoms (in blue color). Broken arrow represents absence of GlxI in the corresponding sub-division. Modified from Ref. 50. Reprinted with permission from AAAS.

Full size image

Bacterial Ni-GlxI can also be distinguished from the eukaryotic Ni-GlxI proteins based on the start motif ‘LHTML’ and ‘M/F/LLHA/VVY’ respectively (Supplementary Fig. S3). The exception to this rule is the animal GlxI proteins from Amphimedon queenslandica and Bombus impatiens, which are more similar to bacterial GlxI protein. Ni-GlxI from Amblyomma maculatum is placed closer to plant proteins, as seen for Zn-GlxI. We were able to group the proteins from green plants on the basis of first amino acid ‘M/F/LLHAVY’. Among these, the proteins with ‘M’ at the first position of start motif seems to be primitive as this is conserved in P. patens (mosses), the base node in the plant phylogeny and Ectocarpus (Brown algae). In P. patens all the domains have conserved motif ‘MLLHVVY’ whereas in higher plants in some domains M is substituted to ‘F’ or ‘L’. The second domain of Ni-GlxI does not show any kind of grouping based on the sequence motif. Sequence analysis shows more or less vertical descent of Ni-GlxI from prokaryotes to eukaryotes with only one exception (Amblyomma maculatum).

In order to correlate evolutionary information with biological properties an attempt was made to predict intracellular localization of GlxI proteins using experimental data and web based localization prediction tools (see materials and methods). The bacterial Ni- and Zn-GlxI proteins were found to be in the cytosol (Table 1) as also confirmed by in vivo localization studies24,25. However in eukaryotes, Ni and Zn dependent forms of GlxI were differentially localized. The plant Zn-GlxI and Ni-GlxI proteins were predicted to be in the nucleus and chloroplast/cytosol respectively (Table 1). This suggests different mode of acquisition of Zn and Ni dependent forms of the enzymes in plants. In animals both Zn- and Ni-GlxI proteins were predicted to be in the cytosol.

Table 1 Table summarizing the details of GlxI proteins used in the study along with the sub cellular localization site prediction. Abbreviations: Cyto-cytosol; Chlo-chloroplast; Cysk-cytoskeleton; Nucl-nucleus; Mito-mitochondria; Extr-extracellular
Full size table

Phylogenetic analysis

The Maximum Likelihood Tree was constructed from the domain sequences of GlxI using the tool PHYML26. The results clearly showed two major clusters, representing Zn- and Ni-GlxI domains respectively (Fig. 5). The bacterial Ni-GlxI proteins and corresponding protein from Candidatus Nitrosopumilus salaria, an archaea formed a separate group. This was placed closer to the group formed by Zn-GlxI and not with the other Ni-GlxI sequences. Bacterial GlxI domains formed a monophyletic group for both Zn- and Ni-GlxI clusters. However, Zn-GlxI from S. cerevisiae was found to be part of the bacterial Zn-GlxI cluster with C. Nitrosopumilus salaria as the root of bacterial cluster. Moreover, Zn-GlxI of Sphingomonas was found to be part of the eukaryotic Zn-GlxI cluster close to Amblyomma maculatum, a tick and with three other green algae. The single domain Zn-GlxI present in green algae and plants followed the species tree pattern which thereby suggests a vertical descent of the gene. The two-domain Zn-GlxI was observed only in Brown Algae, Diatoms and Apicomplexa. Each domain of the two-domain Zn-GlxI from the above mentioned phyla separated into a group suggesting the independent evolution of two-domain Zn-GlxI.

Figure 5: Phylogenetic analysis of GlxI family.
figure 5

Reconstructed phylogenetic tree obtained from domain sequences of 81 putative GlxI sequences showed two major clusters, representing Zn- and Ni-GlxI domains. The Zn-GlxI from C. pelagibacter and Sphingomonas that have been acquired by horizontal gene transfer events in ancestors of eukaryotes and plants respectively have been marked by asterisk. Bootstrap values above 50% for 1000 bootstrap is depicted on the branches.

Full size image

However, Ni-GlxI is a two-domain GlxI in eukaryotes unlike Zn-GlxI which is essentially a single domain enzyme. Analysis of animal Ni-GlxI proteins suggested that these have followed a different evolutionary trajectory. The two-domain Ni-GlxI proteins and the species placement pattern for the individual domains was found to be different which suggests that the evolution of the two domains is independent of each other. Moreover, each of the two domains of Ni-GlxI does not follow the species evolution pattern. The animal Ni-GlxI from A. queenslandica and B. impatiens are placed in the bacterial Ni-GlxI cluster, but here also A. maculatum is an exception with its placement in plant cluster.

The tree was constructed by Neighbour Joining method as well. It is similar to one constructed using Maximum Likelihood method except for positioning of some GlxI proteins, for example Thalassiosira pseudonana (Supplementary Fig. S4).

Structural comparison of Ni-GlxI and Zn-GlxI

The GlxI belongs to beta-alpha-beta-beta super family of proteins. Each GlxI domain consists of two beta-alpha-beta-alpha subunits, therefore single domain Zn-GlxI has two subunits whereas two-domain Ni-GlxI has four subunits (Fig. 6). The structural topology of the Ni- and Zn-GlxI domains was found to be similar except for the presence of three turn alpha helix in Zn-GlxI. This three turn alpha helix was earlier shown to block the catalytic pocket21. The single domain Zn-GlxI is a homodimer27 whereas two-domain Ni-GlxI remains a monomer. The active sites are present in all the four subunits.

Figure 6: Three dimensional predicted structure of the two-domains of Ni-GlxI (Os115475151) from O. sativa.
figure 6

The N and C termini are not shown. Each domain consists of two beta-alpha-beta-beta subunits.

Full size image

Evolution of role of glyoxalases in stress physiology

The role of glyoxalases in stress physiology is well established where several over-expression studies in plants have determined their ability to confer multiple stress tolerance. Their ubiquitous presence further highlights an important role for these enzymes in biological systems. However, we observed an expansion of glyoxalases as a multi-gene family in plants in contrast to their presence as single genes in most prokaryotes. Specifically, an increase in number of Ni-GlxI was observed. As an approach to understand the need for multiple glyoxalases in plants, a detailed qRT-PCR based expression study was undertaken for the four rice GlxI genes included in the phylogeny analysis to identify the response of each member of the GlxI family to the applied stress treatments (Fig. 7). Two week old IR64 rice seedlings were subjected to different stress treatments for 6 h and 24 h. Expression was analyzed in response to heat, cold, dehydration, wounding, MG, salt and oxidative stress.

Figure 7: Transcript profiling of rice GlxI genes in response to various abiotic stress conditions.
figure 7

(A) Histogram depicting logarithmic expression change of rice GlxI genes based on qRT-PCR analysis. Real time PCR analysis was done with cDNA template generated from shoot tissue of 14 day old stressed or control seedlings. As OsGLYI7.2 and OsGLYI11.1 could not be amplified they are not included in the real-time PCR analysis. (B) Heat map and hierarchical cluster display of expression profile for GlxI genes showing different levels of expression in response to stress at 6 h and 24 h. Colour bar at the bottom represent expression values in terms of logarithmic fold, green (lowest), black (medium) and red (highest) expression levels.

Full size image

OsGLYI11, encoding the cytosol localized Ni-GlxI was found to be the most stress responsive gene amongst all the members of the glyoxalase family. OsGLYI11 possesses three spliced forms, OsGLYI11.1, OsGLYI11.2 and OsGLYI11.3 but OsGLYI11.1 could not be amplified. However, transcript levels of the other two spliced forms varied in response to different stress conditions. OsGLYI11.2 was induced in response to all applied stress conditions except under wounding stress. Highest up regulation was observed under drought stress at 6 h where it was found to be 4.5-fold upregulated. OsGLYI11.3 also showed higher expression levels under multiple stress treatments. Other two GlxI genes, OsGLYI2 and OsGLYI7, encoding chloroplast localized Ni-GlxI, were induced 1.7- and 1.3-fold under MG and salinity stress respectively. OsGLYI8 which is the only Zn-GlxI encoding gene present in the rice glyoxalase family showed a ~2-fold increase in expression in response to MG and oxidative stress at 6 h.

Similarly, multiple stress inducibility of GlxI from Arabidopsis has also been documented by Mustafiz et al. (2011). They have reported a high stress inducibility of Ni-GlxI encoding AtGLYI3 and AtGLYI6 in response to salt, drought, wounding, cold and heat treatments in both roots and shoots whereas Zn-GlxI encoding AtGLYI2 was found to be heat inducible. Moreover, microarray data analyzed with respect to different vegetative and reproductive developmental stages revealed expression of both rice and Arabidopsis GlxI at all stages of development. Thus, the transcript profiling study revealed an important role of GlxI, particularly Ni(II) activated forms of the enzyme, in stress physiology.

Catalytic efficiency of Ni-GlxI and Zn-GlxI characterized from various species

The kinetic properties of characterized GlxI proteins from various species were obtained to compare the catalytic efficiency of enzymes belonging to different metal activated forms (Supplementary Table S2). We found that the Km value for Ni-GlxI were always lower than respective values for Zn-GlxI in both prokaryotes and eukaryotes except for human GlxI protein which had relatively lower Km comparable to Ni-GlxI enzymes. Since, Km value indicates the affinity of the enzyme for the substrate; we found that Ni-GlxI in general possessed higher affinity for their substrate than Zn-GlxI. But Kcat values indicated higher turnover number of Zn-GlxI indicating greater efficiency of Zn-GlxI in converting substrate to product once the enzyme-substrate complex is formed. However, inspection of overall catalytic efficiency represented by Kcat/Km value signified similar efficiency for both the forms.

Discussion

Rules that govern genome evolution are becoming clearer with the use of comparative genomics as more and more genomes are getting sequenced. However, it has become apparent that all parts of genomes do not behave in the same way and different genes/regions follow their own evolutionary trajectory depending on the environmental and ecological history. Multi-gene families are good markers for studying genome evolution as different members of the family may follow their own path. The classical example being hemoglobin genes and evolution of fetal and adult forms28. Here we have studied GlxI family, a multi-gene family whose members can be distinguished from each other based on divalent metal ion dependency. We have analyzed both prokaryotes and eukaryotes with special emphasis on the expansion of the gene family in plants.

Our analysis revealed that the Ni dependent form of GlxI has evolved before the Zn form. This conclusion is based on the observation that archaea and cyanobacteria encode only Ni-GlxI. We suggest that Bdellovibrionales (deltaproteobacteria) are the first organisms to possess Zn-GlxI. They are known to be the obligate parasites of other gram negative bacteria29. However, facultative bacteria such as Myxococcales, another sub-division of deltaproteobacteria, encode Ni-GlxI. Absence of Ni-dependent GlxI in obligate parasites is probably due to non-existence of Ni transporters in these organisms30. In the absence of Ni(II) homeostasis in these organisms, Zn(II) which possesses similar characteristic properties as of Ni(II) may have led to evolution of Zn specificity for GlxI. This may have been achieved by insertion of short sequences in the enzyme31. One of the insert has been shown to form a three-turned α-helix that blocks one side of the catalytic pocket21 being located next to the active site and thus plays a significant role in dictating the metal ion specificity.

Primitive orders of protoebacteria encode either Ni- or Zn-GlxI gene. However, in many organisms both forms of the enzymes are present. One of the first reported examples is Pseudomonas aeruginosa, a gammaproteobacteria16. However, our analysis suggested that both forms of GlxI evolved even before gammaproteobacteria, most likely in betaproteobacterial species, such as M. petroleiphilum PM1 and Limnobacter sp MED105. We believe that the horizontal gene transfer may have helped mobilization of one of the enzymes leading to the presence of both forms32. However, their evolutionary history of acquisition may differ from organism to organism. For example, Pseudomonas (gammaproteobacteria) may have got the gene from one of the deltaproteobacteria, Bdellovibrio bacteriovorus which happens to be a predator of Pseudomonas33. There are many instances where genes of endosymbiont or predator get incorporated in the host genome. For example, Drosophila is known to accommodate fragments of Wolbachia genome34. Episodes of horizontal gene transfer occurred not only in prokaryotes but also in eukaryotes during the evolution of GlxI. Our results suggest that different organisms did not follow the same evolutionary pathway. For example, in the ancestor of algae and plants there was a horizontal gene transfer event from Sphingomonas and this underwent duplication and fusion leading to the two-domain Zn-GlxI encoding gene. However, the gene is retained only in Brown Algae, Diatoms and Apicomplexa, and deleted thereafter in the ancestor of green algae and plants. We believe that the domain duplication and fusion events confer advantage in enzyme functionality and hence have been preserved during evolution. For example, tandem duplication of a helix-loop-helix polypeptide resulted in a more thermally stable protein35.

Predicted sub cellular distribution of GlxI partly supported results of sequence analysis. In plants, while Ni-GlxI is expected to be in chloroplast and cytosol, Zn-GlxI is likely to be found in nucleus. Since chloroplasts are believed to be derived from cyanobacterial ancestors, the localization of Ni-GlxI in chloroplast suggests a cyanobacterial origin of Ni-GlxI.

GlxI proteins have been characterized from various plants and their over-expression confers multiple stress tolerance in plants5,6,7,8,12. In order to find a possible link between their copy number and physiological function, we examined the expression levels of the entire gene family under various stresses. Multiple stress-inducible nature of glyoxalase family was observed. OsGLYI11, a Ni-GlxI encoding gene was found to be highly stress inducible amongst the rice glyoxalase family. Similarly, Arabidopsis Ni-GlxI encoding genes were found to be more responsive to various stress treatments compared to Zn-GlxI20. Additionally, comparison of kinetic parameters of Ni- and Zn-GlxI revealed Ni-GlxI to have lower Km values compared to that of Zn-GlxI, indicating higher affinity of the Ni-dependent forms for the substrate MG. This probably justifies the existence of multiple Ni-GlxI in plants which due to their high stress-responsiveness and greater affinities towards MG may help in enhanced scavenging of this potent cytotoxin, thereby reducing the deleterious effects. Moreover, since MG can disrupt photosynthetic reactions in chloroplasts by acting as an intrinsic mediator catalyzing the photo reduction of O2 at photosystem I leading to O2·− production36, the presence of multiple GlxI can confer evolutionary advantage. For instance, MG/oxidative stress-induced and chloroplast localized Ni-dependent forms, namely, OsGLYI2 and OsGLYI7, can prevent damage to photosynthetic machinery caused by MG and subsequently generated oxidative stress in chloroplasts.

Taken together, the detailed inspection of evolutionary trajectory of GlxI reveals several interesting facts about these enzymes. We propose that the evolution of multiple metal dependent forms and single/two domains are the result of the adaptation of these genes under different circumstances. The fact that these enzymes are needed for removing a toxin, accumulated under stress conditions may have been a deciding factor in evolution.

Methods

Homologous gene extraction

The orthologs for E. coli Ni-GlxI and P. aeruginosa Zn-GlxI (GloA3) were initially searched using a less stringent e-value cutoff of 1E-3. One hundred and nineteen proteins were identified by the reciprocal BLAST best hit analysis from all the sub-domains of Tree of Life. Glyoxalase domain (PF00903) was identified from the PFAM database37 in all the homologs of GlxI. Due to high variation in the N and C termini sequences of the GlxI proteins, only the domain sequences were used for multiple alignment analysis by MUSCLE v3.738. Second domain of each of the two-domain GlxI proteins has been distinguished using the suffix ‘D’ after the respective accession IDs. We discarded the poorly aligned sequences and finally 81 protein sequences were selected for further analysis, all of which showed >40% sequence identity. In order to retain the two-domain Zn GlxI in our study, the sequence identity criterion was relaxed for the second domain of Ectocarpus GlxI protein (Es298711504) possessing 38% identity. Each orthologous sequence represents a domain in a Tree of Life. Orthologs were then used to extract the paralogous sequences in their respective genomes. The paralogs with greater than 40% sequence identity were kept. JALVIEW was used to edit and view the alignment39. The columns which had more than 20% gaps have been removed from the alignment.

Gene tree construction

To construct a robust phylogeny, we used both Maximum Likelihood and Neighbour Joining methods. Both the methods were used in batch mode using 1000 sets of data. The 1000 dataset were generated by “seqboot” of PHYLIP40. The Maximum Likelihood tree was generated using PHYML26 with default LG evolution model and substitution rate category 4. The Neighbor Joining Method was used to construct the tree with evolution model JTT using PHYLIP. The tree has been re-rooted at C. nitrosopumilus salaria. The bootstrap values on all the branches above 50% are shown. The phylogenetic tree was visualized using the program iTOL v2.241.

Assessment of metal ion specificity of GlxI

A total of 81 GlxI proteins identified were used in the study covering all the subdomains of Tree of Life. The metal ion specificity of GlxI from several prokaryotic species, human and yeast was already known from the available literature. Due to lack of information regarding plant GlxI, the metal ion dependency of the GlxI proteins from plants was experimentally verified using rice as a model plant. A rice GlxI protein from each of the two major clusters observed in the analysis and assumingly representing the two different forms of GlxI was checked for the metal ion activation. Glyoxalase activity was measured in the presence of different divalent metal ions. Predicted Ni- and Zn-GlxI enzymes showed maximal activation on addition of respective metal ions (our unpublished work). Based on the above interpretation, metal ion specificity was assigned to the uncharacterized GlxI according to their position in the phylogenetic tree.

Putative sub cellular localization prediction

Analysis using WoLF PSORT42 was performed in combination with the available experimental data to identify the putative intracellular localization site of GlxI. The prediction of localization site for prokaryotic GlxI proteins was performed using PSORTb v3.0.243.

Structure prediction and comparison of glyoxalase proteins

Modeller44 has been used to construct the model for OsGLYI11 (Os115475151), a Ni(II) activated GlxI enzyme of O. sativa. 1F9z was used as a template to build the 15 model of the gene. Energy minimization has been done in Modeller and structure with minimum energy was obtained. All the models were checked with PROCHECK45 to test whether they fall in the allowed region in the Ramachandran plot. The domain topology of Ni(II)/Co(II) and Zn(II) activated enzymes have been compared by first identifying the templates of OsGLYI11 (Os115475151) and OsGLYI8 (Os218196491) in PDB46 and their comparison in PDBsum47.

Plant material and stress treatment for qRT-PCR analysis

Seedlings of IR64 rice cultivar were grown under control conditions in growth chamber at 28 ± 2°C and 16 h photoperiod. The seeds were surface sterilized with 1% Bavistin for 20 min and allowed to germinate in a hydroponic system. Germinated seeds were supplied with modified Yoshida medium48. Fourteen-day-old seedlings were subjected to various stress treatments. For cold or heat treatment, seedlings kept in Yoshida media were transferred to a cold chamber maintained at 4 ± 1°C or an incubator at 42 ± 1°C, respectively. For dehydration stress, seedlings were removed from media and kept on a 3-mm blotting paper. For mechanical wounding, each leaf of seedlings was given a small cut with a blade. For salt, MG and oxidative stress, seedlings were kept in Yoshida medium supplied with 200 mM NaCl, 5 mM MG, or 5 mM hydrogen peroxide, respectively. Tissue samples were harvested after 6 h and 24 h of stress treatment and frozen in liquid nitrogen. Untreated seedlings were used as control.

Real-time PCR

Total RNA was isolated from the shoot tissue of control and stressed plant samples using RaFlex™ solution I and solution II (GeNei, India) as per the manufacturer's protocol. First-strand cDNA synthesis was done using RevertAid™ RNAse H minus cDNA synthesis kit as per manufacturer's instructions (Fermentas Life Sciences, USA). Same primer sequences were used for real-time PCR as described by Mustafiz and co-workers20. The nomenclature used for the GlxI genes has also been adopted from Mustafiz et al.20. The PCR mixture contained 5 μl of cDNA (50 times diluted), 12.5 μl of 2× SYBR Green PCR Master Mix (Applied Biosystems, USA) and 200 nM of each gene-specific primer in a final volume of 25 μl. The qRT-PCR was performed using ABI Prism 7500 Sequence Detection System and software (PE Applied Biosystems). All the PCRs were performed in 48-well optical reaction plates (Applied Biosystems). The specificity of the amplification was tested by dissociation curve analysis and agarose gel electrophoresis. eEF-1α gene was used as internal control. Three technical replicates were analyzed for each sample. The relative expression ratio of each glyoxalase gene was calculated using delta Ct or comparative Ct value method49.