Recent Gene Duplication and Subfunctionalization Produced a Mitochondrial GrpE, the Nucleotide Exchange Factor of the Hsp70 Complex, Specialized in Thermotolerance to Chronic Heat Stress in Arabidopsis

The duplication and divergence of heat stress (HS) response genes might help plants adapt to varied HS conditions, but little is known on the topic. Here, we examined the evolution and function of Arabidopsis ( Arabidopsis thaliana ) mitochondrial GrpE (Mge) proteins. GrpE acts as a nucleotide-exchange factor in the Hsp70/DnaK chaperone machinery. Genomic data show that AtMge1 and AtMge2 arose from a recent whole-genome duplication event. Phylogenetic analysis indicated that duplication and preservation of Mge s occurred independently in many plant species, which suggests a common tendency in the evolution of the genes. Intron retention contributed to the divergence of the protein structure of Mge paralogs in higher plants. In both Arabidopsis and tomato ( Solanum lycopersicum ), Mge1 is induced by ultraviolet B light and Mge2 is induced by heat, which suggests regulatory divergence of the genes. Consistently, AtMge2 but not AtMge1 is under the control of HsfA1, the master regulator of the HS response. Heterologous expression of AtMge2 but not AtMge1 in the temperature-sensitive Escherichia coli grpE mutant restored its growth at 43 (cid:1) C. Arabidopsis T-DNA knockout lines under different HS regimes revealed that Mge2 is speciﬁcally required for tolerating prolonged exposure to moderately high temperature, as compared with the need of the heat shock protein 101 and the HS-associated 32-kD protein for short-term extreme heat. Therefore, with duplication and subfunctionalization, one copy of the Arabidopsis Mge genes became specialized in a distinct type of HS. We provide direct evidence supporting the connection between gene duplication and adaptation to environmental stress.

Temperature is an environmental factor that affects the growth and distribution of plants. To be successful in the survival game of nature, plants need to cope with adverse temperatures caused by daily or seasonal temperature fluctuations and by climate change. Understanding how plants adapted to temperature stress can help improve tolerance in crops (Katiyar-Agarwal et al., 2003;Shou et al., 2004).
Plants are often challenged by elevated temperature, which causes heat stress (HS). Knowledge of the plant response to HS is accumulating quickly because of the progress of genomic and functional genomic studies (Kotak et al., 2007). Many genetic components form a complex signaling transduction network and chaperone machineries. Some of these components belong to multigene families, most notably heat shock proteins (HSPs; Vierling, 1991) and heat shock transcription factors (HSFs; von Koskull-Dö ring et al., 2007). The HSP and HSF paralogs apparently arose from gene duplications and subsequent diversification. Gene duplication and divergence increased the capacity of organisms to adapt to environmental change (Conant and Wolfe, 2008). In addition to responding to elevated temperature, HS genes are responsive to other environmental changes. Thus, the duplication and divergence of HS genes might have allowed plants to adapt to different environmental stresses or varied HS conditions.
However, linking the adaptation and duplication of HS genes is difficult because a loss of function of one of the duplicates may not lead to an alteration in phenotype such as thermotolerance, which was often attributed to gene redundancy. Mutant phenotypes may not be revealed because the assays are not performed under the right conditions, as was shown with several Arabidopsis (Arabidopsis thaliana) HS genes (Charng et al., 2006(Charng et al., , 2007Meiri and Breiman, 2009;Meiri et al., 2010). Thus, determining an effective phenotyping method is critical for revealing the functions of HS genes per se and for shedding light on their evolution.
To determine the biological functions of HS response genes, we have used reverse genetic approaches by phenotyping Arabidopsis T-DNA knockout (KO) mutants under varied HS conditions (Charng et al., 2007;Liu et al., 2011). Here, we focused on the function and evolution of the mitochondrial GrpE genes. GrpE is a highly conserved and ubiquitous protein in prokaryotes and eukaryotes. It functions as a cochaperone in the DnaK/Hsp70 complex, consisting of DnaK (Hsp70), DnaJ (Hsp40), and GrpE. The complex is a chaperone machine involved in the folding of the nascent protein peptides, organellar protein translocation, and disaggregation and refolding of the denatured proteins (Mayer and Bukau, 2005). DnaK and Hsp70 homologs consist of an N-terminal ATPase domain and a C-terminal substrate domain. Current models suggest that the machine operates by propelling DnaK/Hsp70 in cycles of alternating low-affinity (ATP-bound) and high-affinity (ADP-bound) states for substrates by interaction with cochaperones, in which nucleotide dissociation catalyzed by GrpE is the rate-limiting step for substrate release (Mayer and Bukau, 2005). In Escherichia coli, both GrpE and DnaK are inducible by heat treatment. Mutants of GrpE and DnaK are sensitive to adverse high temperature (Paek and Walker, 1987;Ang and Georgopoulos, 1989).
In eukaryotes, multiple sets of the Hsp70 chaperone machineries can be found in cytosol and other organellar compartments (Sung et al., 2001). However, only those found in mitochondria and chloroplasts closely resemble their prokaryotic kin. Only one mitochondrial GrpE, Mge1, was found in yeast. Mge1 is not induced by heat shock (Ikeda et al., 1994) but is essential for yeast viability (Bolliger et al., 1994). Mge1 acts in concert with mitochondrial Hsp70s in the translocation and maturation of preproteins (Laloraya et al., 1995). Our knowledge of plant GrpEs and their biological role is limited. Two types of GrpE cDNAs were cloned from tobacco (Nicotiana tabacum) BY2 cells and found to encode mitochondrial proteins that interact with mtHsp70 when reintroduced into BY2 protoplasts (Padidam et al., 1999). In Chlamydomonas reinhardtii, the chloroplastic GrpE homolog Cge1 is strongly induced by heat shock and forms ATP-sensitive complexes with Hsp70B, the counterpart of bacterial DnaK (Schroda et al., 2001). In vitro studies showed that the Hsp70B-Cdj2-Cge1 chaperones facilitate both the assembly and disassembly of the vesicle-inducing protein in plastid oligomers (VIPP1) involved in the biogenesis of thylakoid membranes .
In this paper, we investigated the phylogenetic, regulatory, and functional aspects of duplicated copies of Arabidopsis Mges. Subfunctionalization at the expression and functional levels may have allowed Arabidopsis Mges to respond more efficiently to distinct HS regimes. We provide direct evidence to sup-port a connection between gene duplication and adaptation to HS.

Arabidopsis Mitochondrial GrpE Genes Were Duplicated in a Recent Whole-Genome Duplication Event
The Arabidopsis genome contains two homologous genes putatively encoding for mitochondrial GrpEs, which we named Mge1 (At5g55200) and Mge2 (At4g26780). The At4g26780 locus was previously annotated to have unknown function. Corresponding to these cochaperones are two copies of Hsp70 homologs, mtHsc70-1 (At4g37910) and mtHsc70-2 (At5g09590), encoding for mitochondrial DnaK/Hsp70 proteins. The homology between Mge1 and Mge2 suggests that they are derived from gene duplication. The Arabidopsis genome had experienced at least two rounds of whole-genome duplication (WGD) events that led to massive gene duplication (Blanc et al., 2003). To determine whether the Mges arose from either one of these events, we used the online analytic tool Paralogons to examine the locations of the Mges in the duplicated regions of the Arabidopsis genome (Blanc et al., 2003). Mge1 and Mge2 each locate in two corresponding duplicated blocks formed after the most recent WGD event (Fig. 1). However, their putatively interacting partners, mtHsc70s, came from a more ancient WGD event. Unlike the duplication time for the Mge pair, that for the two chloroplastic GrpE genes, Cge1 (At1g36390) and Cge2 (At5g17710), could not be determined, which suggests that they arose from a time before the traceable WGD events. To our surprise, the chloroplastic DnaK/Hsp70 genes, cpHsc70-1 (At5g49910) and cpHsc70-2 (At4g24280), came from the same WGD as the Mges, because they locate on the same duplicated blocks (Fig. 1). To confirm this observation, we used molecular clock theory to estimate the time of divergence of each homologous pair by using Arabidopsis genomic synonymous substitution rates (Ks) in a described formula (Lynch and Conery, 2000;Raes et al., 2003). Table I shows the Ks and time of divergence of Arabidopsis Mges, Cges, mtHsc70s, and cpHsc70s. The data are in good agreement with the chronicle of the WGD events described (Blanc et al., 2003).

Duplication of Mge Genes Occurred Independently after the Divergence of Many Distantly Related Plant Species
With the knowledge that Arabidopsis Mge genes arose from a recent WGD event, we investigated a phylogenetic relationship among plant Mges. With a BLAST search, we retrieved and aligned the amino acid sequences homologous to Arabidopsis Mges from various species (Supplemental Fig. S1). The aligned sequences were used to generate a phylogenetic tree based on the neighbor-joining method (Fig. 2). The tree showed two major clades of GrpE proteins, one from mitochondrion and E. coli and the other from chloroplast and cyanobacterium, which is in good agreement with the endosymbiotic theory. Two Mge homologs exist in most plants investigated in this study except soybean (Glycine max), which has three. The Mges of most plant species, except monocots or close relatives such as tobacco and tomato (Solanum lycopersicum), clustered together by species, which implies that the Mges duplicated after the divergence of the distantly related species and before the divergence of monocots and closely related species. In contrast, their chloroplastic homologs showed a different clustering pattern. Regardless of plant species, except for moss, Cge1 and Cge2 clustered together by protein type (Fig. 2), which suggests that these proteins were derived from duplication before the speciation of various plants.

Retention of an In-Frame Intron Is Common in Higher Plant Mges
Arabidopsis Mge1 contains six exons and Mge2 has five (Fig. 3). From comparisons of the genomic DNA structures and protein sequences, exon 2 of Mge2 likely contains a retained intron that presumably corresponds to intron 2 of Mge1 (Fig. 3). This putatively retained intron in Mge2 apparently can no longer be processed by the spliceosome because of a loss of the essential features for RNA splicing and, therefore, becomes part of the coding sequence. By contrast, intron 2 of Mge1 does not contain an in-frame sequence and would disrupt the open reading frame if retained. Interestingly, we found retention of the in-frame intron 2 common in alternatively spliced (AS) transcripts for one of the two Mge genes in other higher plant species (Table II). As most of the Mge genes listed here were not annotated, we propose that the Mge genes without intron 2 retention be annotated as Mge1 (or Mge1 and Mge3 in soybean) and the Mge genes with intron 2 retention be annotated as Mge2. Notably, the sequences derived from the retained introns are conserved with a motif of four to six consecutive Arg or Lys residues (Supplemental Fig. S1). Therefore, retention of the in-frame intron 2 may be under selection pressure in higher plants. By contrast, neither Mge1 nor Mge2 of the moss Physcomitrella patens contains an in-frame intron 2.

Mge2 Is Localized in Mitochondria
Previously, mtHsc70-1, mtHsc70-2, and Mge1 were identified in the mitochondrial proteome of Arabidopsis (Millar et al., 2001;Heazlewood et al., 2004), so these proteins are localized in mitochondria. However, the localization of Mge2 was not known. The presence of a predicted N-terminal mitochondrial targeting peptide (mTP) in Mge2 (Supplemental Fig. S1) suggests that this protein is also localized in mitochondria. This prediction was further confirmed by transient expression of the Mge2:GFP fusion protein in Arabidopsis protoplasts. We found that fusion proteins localized in the punctate foci overlapped those stained by the mitochondria-specific dye MitoTracker Orange, with the GFP control localized in the cytosol (Fig. 4).

Mges Are Differentially Expressed
Duplicate genes usually undergo subfunctionalization at the regulatory and/or protein level (Zhang, 2003). Subfunctionalization at the regulatory level Cge2 Undefined n/a n/a cpHsc70-1 cpHsc70-2 Recent 0.71 59 often leads to differential expression of homologous genes. A search of the public microarray database revealed the paralogs AtMge1 and AtMge2 to be responsive to UV-B or heat treatment. Mge1 was upregulated by UV-B, and Mge2 was preferentially induced by heat treatment at 38°C (Supplemental Fig.  S2). Analysis of the cis-elements in the promoters showed the Mge1 promoter containing putative UV-B-responsive elements, such as UV boxes and MRE ANAC13 (Safrany et al., 2008), and the Mge2 promoter containing one canonical heat shock element (HSE; Fig. 3). HSE is the conserved cis-element binding with trimerized HSF for transcription activation of HS genes in eukaryotes (Kroeger et al., 1993). We then tested whether the heat induction of Mge2 requires HsfA1s, the master regu-lators of the HS response in Arabidopsis (Liu et al., 2011;Yoshida et al., 2011). Semiquantitative reverse transcription (RT)-PCR showed no induction of Mge2 with heat treatment in the null mutant of HsfA1 genes, which indicates that Mge2 is under the control of HsfA1 (Fig. 5A). However, Mge1 still maintained the basal expression level regardless of HsfA1s. These results agree with the absence or presence of HSE in the promoter regions of the Mges.
To determine whether Mges are also differentially expressed in species other than Arabidopsis, we examined the transcript levels of Mge1 and Mge2 in tomato seedlings with heat and UV-B treatment. The mRNA expression of Mge1 was up-regulated in tomato by UV-B treatment, whereas that of Mge2 was induced by heat, as in Arabidopsis (Fig. 5, B and C). Therefore, the two paralogs had undergone a similar subfunctionalization process at the regulation level after the divergence of Arabidopsis and tomato.

Heterologously Expressed Arabidopsis Mge2 Confers Thermotolerance in the Heat-Sensitive E. coli grpE Mutant at Higher Temperature Than Does Mge1
Eukaryotic GrpE homologs, such as the distantly related Cge, can replace the function of E. coli GrpE (Deloche et al., 1997;Schroda et al., 2001). Because Arabidopsis Mges are differentially expressed under HS, we wondered whether Mge1 and Mge2 could replace the function of bacterial GrpE in different capacities. We performed a functional complementation assay with the E. coli grpE mutant (DA16; Ang and Georgopoulos, 1989). The mutant strain carries a point mutation that substitutes a conserved Gly at position 122 with Asp in GrpE (Harrison et al., 1997;Grimshaw et al., 2005). The mutant line exhibited a temperature-sensitive phenotype with growth abolished at 40°C (Fig. 6). We ligated Arabidopsis Mge1 and Mge2 cDNAs without the coding sequence of the predicted mTPs onto an expression vector for transformation into the mutant line to determine whether the temperature-sensitive phenotype could be rescued. The mTPs predicted as putative targeting peptides differed significantly in length in Mge1 and Mge2, with N-terminal 58 and 38 amino  acids, respectively, by both TargetP (Emanuelsson et al., 2007) and MitoProt (Claros and Vincens, 1996). However, because the 20-amino acid difference is in a relatively conserved region, part of the functional region of the protein might be removed along with the stretch of peptide sequence. Therefore, we created two constructs for each gene, with one set of constructs having 38 amino acids from the N terminus removed from both Mge1 and Mge2 (designated the "long form"; Supplemental Fig. S1) and another set with 54 or 58 amino acids removed from the N terminus of both proteins (designated the "short form").
At control temperature (30°C), all lines, whether Mge-transformed or empty vector-transformed DA16 lines, grew similar to the wild-type strain DA15. When grown at a higher temperature (40°C), DA16 with the empty vector showed the temperature-sensitive phenotype and failed to grow. However, except for the one with the long-form Mge1, the Mge-transformed lines were able to grow as normally as the wild type, which suggests that the Arabidopsis proteins and the prokaryotic homologs could function similarly. Nevertheless, at 43°C, the two Mge-transformed lines showed different behavior. DA16 transformed with Mge2, whether the short or long form, could at least in part complement the mutation, whereas Mge1 could not, and it showed abolished growth, like DA16 transformed with the empty vector (Fig. 6). The difference between Mge1 and Mge2 was not due to the level of the recombinant proteins expressed in E. coli, because the thermotolerance level was not significantly affected with or without the expression inducer anhydrotetracycline (data not shown). Therefore, Arabidopsis Mge2 could function at higher temperatures than Mge1.

Mge2 Specifically Confers Thermotolerance to Chronic HS in Arabidopsis
Because Mge2 was highly induced by elevated temperature, we wondered whether it was required for thermotolerance in Arabidopsis. We characterized two homozygous T-DNA KO lines for Mge2; the T-DNA KO line for Mge1 was not available during this work. The locations of the T-DNA insertion are shown in Figure 3 and were confirmed by PCR analysis of the mutants' genomic DNA (Supplemental Fig. S3A). RT-PCR analysis confirmed that these mutants no longer generated transcripts like the wild type (Supplemental Fig. S3B), so the lines were null mutants. Under  normal conditions, the mutants did not differ in phenotype from the wild type, which suggests that Mge2 is not essential for growth and development. Seedlings of the mutants then underwent treatment with four different HS regimes to measure the capacity for basal thermotolerance (BT), short-term acquired thermotolerance (SAT), long-term acquired thermotolerance (LAT), and thermotolerance to moderately high temperature (TMHT), and their survival rates were measured. The first three assays involved an extremely high temperature, at 44°C, and the condition for TMHT mimicked chronic HS. The assay for SAT differed from those of LAT in recovery time between the acclimation treatment and acute HS challenge (Fig. 7A). Previously, we showed that the HS-associated 32-kD protein (Hsa32) is required for LAT but not SAT, whereas the heat shock protein 101 (Hsp101) is required for both (Charng et al., 2006(Charng et al., , 2007. Interestingly, the two KO mutants for Mge2, mge2-1 and mge2-2, were defective in TMHT but not the other three types of thermotolerance (Fig. 7). By contrast, the hsp101 and hsa32 mutants showed defective phenotypes in at least two of the assay conditions for BT, SAT, and LAT, without substantial alteration in TMHT (Fig. 7).
To further examine the function of Mge2 in TMHT, we assessed mRNA and protein expression patterns in seedling samples collected at different times during heat treatment. The mRNA expression of Mge2 and other HSP genes, including Hsp101 and Hsa32, was upregulated throughout the chronic HS (Fig. 8A). To detect Arabidopsis Mge2 protein, we raised rabbit antiserum against Mge2 recombinant protein. The antibody recognized a protein band of about 38 kD in the wild type but not the mge2-2 mutant (Fig. 8B). The size determined by migration on SDS-PAGE was the same as for the short-form Mge2 expressed in E. coli described earlier, which suggests that the N-terminal 54 amino acids in the preprotein sequence of Mge2 is the mTP (Supplemental Fig. S1). With the TMHT assay, the Mge2 protein level gradually increased and remained high after 9 d (Fig. 8B). By contrast, the other HSP levels were high at day 1 but gradually decreased during treatment and were barely detectable at day 9.

DISCUSSION
In this study, we show that in Arabidopsis, the two mitochondrial GrpE proteins arose from a recent WGD event, estimated to occur about 24 to 75 million years ago (Lynch and Conery, 2000;Simillion et al., 2002;Blanc et al., 2003). This finding is consistent with our estimated divergence time for the Mges, about 57 Figure 6. Differential complementation of the E. coli grpE mutant strain with Arabidopsis Mge1 and Mge2. Heat-sensitive E. coli mutant DA16 grpE was transformed with Arabidopsis Mges encoding long-and short-form Mge proteins with 38 and 58 putative mTPs removed from the N terminus, respectively. DA16 transformed with short Mge1, long Mge1, short Mge2, and long Mge2 are labeled 1S, 1L, 2S, and 2L, respectively. DA16 transformed with empty vector (V) was used as a negative control. The wildtype strain DA15 (WT) with an isogenic background to DA16 was transformed with the same empty vector to confer the same antibiotic resistance as the others and included as a comparison. Cell cultures were serially diluted and spotted onto LB plates and incubated overnight at 30°C, 40°C, 43°C, or 45°C. Figure 5. RT-PCR analyses of the differential expression of Mge genes. A, Mge1 and Mge2 were differentially regulated by HsfA1s in Arabidopsis. The RT-PCR products of Mge1 and Mge2 were derived from the wild type (W) and a quadruple KO mutant of HsfA1s (Q). RNA samples were purified from 5-d-old seedlings without (Con) or with (HS) heat treatment at 37°C for 1 h. B, RT-PCR products of Mge1 and Mge2 derived from Arabidopsis (At) and tomato (Sl) seedlings without (C) or with (H) heat treatment at 37°C for 1 h or 40°C for 2 h, respectively. C, RT-PCR products of Mge1 and Mge2 derived from Arabidopsis (At) and tomato (Sl) seedlings without (C) or with (U) UV-B treatment. Actin expression was used as a loading control. million years ago (Table I). Intriguingly, our phylogenetic analysis revealed similar independent duplication events for the Mge genes in other higher plants (Fig. 2). Therefore, these Mge paralogs also likely emerged from recent WGD events, which were found to occur in many plant species around the Cretaceous-Tertiary boundary about 65 million years ago (Fawcett et al., 2009). However, we could not exclude that segmental duplication instead of WGD is the cause for some species. Nevertheless, this coincidence implies that selection acted to preserve both copies of the Mge genes in these distantly related species.
Preservation of the two Mge genes could have conferred tolerance to a wide range of environmental conditions. However, retaining both Mge copies might cause dosage imbalance, because overexpression of GrpE caused a defect in the DnaK chaperone system in E. coli (Sugimoto et al., 2008). Subfunctionalization at the regulatory level could alleviate the problem of dosage imbalance (Veitia et al., 2008). The RT-PCR and microarray data revealed Arabidopsis Mges to be differentially expressed under UV-B and HS, which is consistent with the presence of associated cis-elements in the promoters of corresponding genes (Fig. 3). We suspected that for certain eudicots, the promoter of the ancestral Mge contained all of the cis-elements required for both types of stress responses, because the tomato Mges were differentially expressed, similar to the Arabidopsis genes (Fig. 5B). As time elapsed from gene duplication, the cis-elements were differentially lost in the two promoters, probably to optimize response to HS or UV-B. This inference agrees well with the degenerative complementation model of evolution in regulatory elements after gene duplication (Papp et al., 2003;Li et al., 2005) and with the rapid expression divergence of duplicate genes in response to external changes in Arabidopsis (Ha et al., 2007). Here, we provide genetic evidence that Mge2 is involved in thermotolerance, specifically TMHT. However, the role of Mge1 in UV-B stress tolerance remains to be elucidated.
Besides expression divergence, the functions of Mge1 and Mge2 proteins were also subfunctionalized. Mge1 but not Mge2 is present in the mitochondrial proteome under a non-HS condition in cultured cells (Heazlewood et al., 2004), which suggests that Mge1 is Figure 7. Thermotolerance assays of Mge2 T-DNA KO lines. A, Phenotypes of Arabidopsis seedlings of the wild type (WT) and various T-DNA KO mutants after 10 d of recovery from the assays under four different HS regimes as schematically shown on the right of each row. Seedlings in the same row were grown on the same plate. B, Survival rates of seedlings after the assay treatments shown in A. The labels for mutants are mge2-1 (m2-1), mge2-2 (m2-2), hsp101 (h101), and hsa32 (h32). The bars present means 6 SD of three replicates for each treatment (n $ 35 each). * P , 0.01 (versus wildtype plants by Student's t test). responsible for crucial housekeeping functions such as protein import (Truscott et al., 2003). We cannot exclude the role of Mge1 in HS tolerance. Mge1 may be required for basal or acquired thermotolerance, in that it could be involved in protein folding during recovery at ambient temperature. Unfortunately, the T-DNA KO line for Mge1 was not available at the time of the study, which might be due to it being essential for viability like its counterparts in yeast and E. coli (Ang and Georgopoulos, 1989;Laloraya et al., 1994). The functional complementation assay of the E. coli grpE mutant showed the two Arabidopsis Mges with different temperature ranges required for their functions, which is consistent with results in planta. However, that Mge1 can complement the heat-sensitive phenotype at 40°C in the E. coli DA16 mutant (Fig. 6) but is not sufficient to sustain growth at 35°C in Arabidopsis Mge2 KO mutants is intriguing. One reason may be that the thermostability or function of the protein is greatly affected by the milieu; that is, Mge1 might be less stable or less efficient at 35°C in mitochondria than in bacteria cells. Alternatively, the interaction between Mge1 and bacterial DnaK might be more efficient than that between Mge1 and mitochondrial Hsp70s at 35°C. In vitro experiments may help examine the latter possibility.
A closer examination of the genomic sequences of the duplicated Mges in Arabidopsis and comparison with their homologs in other plant species strongly suggest that the two Mges in Arabidopsis might have emerged from a single ancestral Mge that could produce transcripts with two AS forms (Table II). However, following duplication, both Arabidopsis Mge genes lost the ability to generate AS transcripts. Sequence alignment (Supplemental Fig. S1) revealed that the most distinctive difference between the Arabidopsis Mge1 and Mge2 protein sequences is a 30-amino acid stretch encoded by the putatively retained intron 2 of Mge2 (Fig. 3). This stretch is located in a highly diverse region near the N-terminal end of the long a-helix (Supplemental Fig. S1) from the crystal structure of E. coli GrpE (Harrison et al., 1997). The N-terminal 33 amino acids of E. coli GrpE corresponding to this region were found to be unstructured but involved in substrate release from the DnaK (Harrison et al., 1997;Brehmer et al., 2004). Given the conservation of these intron-derived peptide sequences, the intron retention may affect the cochaperone function of Mge. Further studies of the effect of the absence or presence of the in-frame intron 2 in Arabidopsis and other plant species with use of the Arabidopsis Mge2 KO lines, TMHT assay, and E. coli complementation assay established in this study should provide better insights. Of note, temperature-dependent alternative splicing generates Cge isoforms in Chlamydomonas . However, the biological function of the AS forms is also not clear. HSPs during day (d) 0, 1, 4, and 9 of TMHT treatment detected by semiquantitative RT-PCR. Actin expression was used as a loading control. B, Total or mitochondria-enriched fraction proteins were extracted from the TMHT seedling samples in A for western-blot analyses of proteins. For Mge2, 65 mg of proteins from the mitochondria-enriched fraction (MEF) was loaded; for all other HSPs, 60 mg of total proteins (TP) was loaded. Voltage-dependent anion channel (VDAC) was used as a loading control for MEF protein. The band intensity of Mge2 was normalized to that of VDAC, with day 0 assigned as 1, and indicated below the corresponding Mge2 bands. The black arrowhead indicates the protein band of Mge2, and the white arrowhead indicates the nonspecific band. The molecular mass of Mge2 was estimated by molecular mass markers (data not shown). Tubulin was used as a loading control for total proteins. WT, Wild type; m2, mge2-2.
The observation that Mge2 is involved in the thermotolerance of chronic HS but not extreme HS is, to our knowledge, the first evidence for HSPs characterized to date. The role of Mge2 in TMHT agrees well with the expression induced by heat treatment at the transcript and protein levels. Mge2 might be required for importing or folding mitochondrial proteins under moderately high temperature. Of note, one recently identified mutant of Arabidopsis, hit2, which encodes a nuclear transport receptor, was found impaired in growth at moderately high temperature (37°C for 4 d) and defective in BT but not SAT or LAT (Wu et al., 2010). However, Hit2 is not responsive to heat treatment. By contrast, Hsp101 is essential for basal and acquired thermotolerance but not TMHT, which is consistent with previous findings (Queitsch et al., 2000). These observations indicate that plants use different components for diverse HS conditions. Interestingly, despite induction at the transcription level for all HSP genes tested under the TMHT condition, several HSPs, including Hsp101, could not be sustained at the protein level, whereas the level of Mge2 substantially increased with time (Fig. 8B). This phenomenon suggests that the HSPs were under translational or posttranslational control during prolonged exposure to moderately high temperature. Thus, inferring with the function of a gene under different HS conditions based on its transcript level is difficult. Investigating the differential regulation of HSPs at the translational or posttranslational level is needed to better understand the HS response.
The differential evolution of mitochondrial and chloroplastic GrpEs and Hsp70s in Arabidopsis is of interest. Mges arose from a recent WGD and mtHsc70s from an ancient duplication event, and vice versa for their chloroplastic counterparts (Table  I). With the increasing number of plant genomes deciphered, comparison of the evolution of the organellar Hsp70 complexes would help determine how common this phenomenon is in plants. Results from this study can provide basic information regarding the evolution of the chaperone machineries. Of note, two copies of Mges were found in rodents and human (Naylor et al., 1998;Oliveira et al., 2006). How and why two Mges evolved in these organisms is unknown.
Recent progress in our knowledge of chloroplast Hsp70s and cochaperones has shown that the Hsp70 chaperone system is involved in protein translocation into chloroplasts (Shi and Theg, 2010;Su and Li, 2010), development and thermotolerance (Su and Li, 2008), and the biogenesis of thylakoid membranes by regulating the assembly state of VIPP1 . Studies of mitochondrial Hsp70 systems are beginning to reveal the importance of the system in plants. We are interested to know whether mtHsc70-1 and mtHsc70-2 also have different roles in thermotolerance as do their cochaperones, Mges.

CONCLUSION
The duplication of Mges occurred independently in many plant species, which suggests a common tendency in the evolution of the genes. Subfunctionalization of Arabidopsis Mges at both the protein and regulatory levels allowed Mge2 to specialize in the tolerance of long-term exposure to moderately high temperature. Our data suggest that the duplication and subfunctionalization of a highly conserved HSP such as Mge constituted a critical evolutionary adjustment in plants to cope with distinct types of HS and to better adapt to a changed environment.

Phylogenetic Analysis
Sequences of Mge and Cge homologs were found by using the Arabidopsis (Arabidopsis thaliana) Mge2 as query sequence in a TBLASTN search of the databases Dana-Farber Cancer Institute Plant Gene Indices (http://compbio. dfci.harvard.edu/tgi/plant.html), National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/), and Phytozome (http://www.phytozome. net/). The sequences were aligned by use of the ClustalW program in MEGA5 (Tamura et al., 2011). The aligned sequences (for the sections of sequences used, see Supplemental Fig. S1) were then used to generate a phylogenetic tree by the neighbor-joining method.
The determination of Arabidopsis paralogous genes and their duplication events involved the online program Paralogons (http://wolfe.gen.tcd.ie/ athal/dup). To estimate the time of divergence of the two duplicate genes, DnaSP was used to first determine the Ks of each gene pair (Librado and Rozas, 2009). Then, the resulting Ks was used in the formula T = Ks/2l, with l = 6.1 3 10 29 as the average whole-genome Ks of Arabidopsis (Lynch and Conery, 2000;Raes et al., 2003).

In Silico Analyses of Gene Expression and cis-Elements of the Promoter
Transcript expression data for Arabidopsis Mge1 and Mge2 under abiotic stress conditions were obtained from the public microarray database maintained in the Arabidopsis eFP Browser (http://www.bar.utoronto.ca/efp/cgi-bin/ efpWeb.cgi; Winter et al., 2007). Details for the data of HS and UV-B can be found in the browser by choosing the data source "abiotic stress." The consensus sequences for UV-B regulatory elements and HSEs were obtained from Safrany et al. (2008) and Nover et al. (2001), respectively. Sequence about 1,100 bp upstream of the ATG site was considered the promoter region. The elements were found by using the search function of Vector NTI 9.1.0 (Invitrogen).

Plant Materials and Growth Condition
The Mge2 T-DNA insertion lines SALK_075614 (mge2-1) and SALK_082197 (mge2-2) were obtained from the Arabidopsis Biological Resource Center. The mutants are derived from ecotype Columbia (Alonso et al., 2003). Homozygous lines of the mutant allele were identified by PCR analysis as described (Charng et al., 2007). The T-DNA KO mutants of Hsp101, Hsa32, and HsfA1s were described previously (Liu et al., 2011). Seeds were sown on 0.53 Murashige and Skoog (MS) medium plates containing 0.1% or 1% Suc. The sown seeds were imbibed for 3 d at 4°C in the dark before being allowed to germinate and grow at 22°C with 16 h of light (120 mmol m 22 s 21 ).
terminus of Mge2. The finished construct, driven by the cauliflower mosaic virus 35S promoter, was transiently expressed in Arabidopsis mesophyll cell protoplasts as described (Wu et al., 2009). The transformed protoplasts were then stained with the mitochondria-specific probe MitoTracker Orange (Invitrogen). The GFP and MitoTracker Orange signals were observed with a Zeiss LSM 510 Meta confocal microscope.

Heat and UV-B Treatments
Five-day-old Arabidopsis seedlings were subjected to heat treatment at 37°C for 1 h as described (Charng et al., 2006). Seven-day-old tomato (Solanum lycopersicum 'Microtom') seedlings sown in Magenta boxes containing 0.53 MS medium and 1% Suc were subjected to heat treatment at 40°C for 2 h.
For UV-B treatment, 7-d-old Arabidopsis or tomato seedlings were exposed to 150 mJ cm 22 UV-B by using an X1000 UV-B cross-linker (Spectronics).

Semiquantitative Analysis of mRNA
Seedling samples were collected and immediately incubated in liquid nitrogen after treatment. Total RNA was harvested by the Trizol method (Invitrogen) followed by chloroform purification. RNA was reverse transcribed into cDNA by the use of oligo(dT) primers and Moloney murine leukemia virus reverse transcriptase (Promega). PCR involved Ampliqon III Taq polymerase mix with specific primers as described (Charng et al., 2007). Primer sequences are shown in Supplemental Table S1.

Functional Complementation of a Heat-Sensitive Escherichia coli Mutant Containing Defective GrpE
Primers were designed to remove the putative mTPs predicted by TargetP version 1.0 (http://www.cbs.dtu.dk/services/TargetP/) and MitoProt (http://ihg.gsf.de/ihg/mitoprot.html) from Arabidopsis Mge1 and Mge2 protein sequences. The coding sequence corresponding to the first 54 to 58 or 38 amino acids from the N terminus was removed from both Mge1 and Mge2 as mitochondrial targeting sequences. The Mge1 and Mge2 coding sequences for both species were amplified by PCR, cloned onto pASK3 plus expression vector (IBA BioTAGnology), and transformed into DA16 cells (Coli Genetic Stock Center, Yale University). The transformed DA16 lines were grown in Luria-Bertani (LB) medium containing ampicillin (100 mg mL 21 ) at 30°C to about an optical density at 600 nm = 0.8. The liquid cultures then were 10-fold serially diluted, and 4 mL of each dilution was dropped onto an LB plate containing ampicillin. The plates were grown in a growth chamber at 30°C or at 40°C, 43°C, or 45°C in a water bath overnight. To confirm the identities of the expressed recombinant proteins, the protein bands induced by anhydrotetracycline underwent liquid chromatography-electrospray ionization-tandem mass spectrometry analysis after trypsin digestion, performed by the Proteomics Core Laboratory of the Institute of Plant and Microbial Biology/Agricultural Biotechnology Research Center, Academia Sinica.

Thermotolerance Assays
SAT, LAT, BT, and TMHT assays were as described (Liu et al., 2011) with slight modification. For the SAT assay, 5-d-old Arabidopsis seedlings were first acclimated with 37°C heat treatment for 1 h. After a 2-h recovery at 22°C, the seedlings then were challenged with acute HS at 44°C for 150 min. For the LAT assay, 5-d-old Arabidopsis seedlings were acclimated to 37°C heat treatment for 1 h and then underwent a 2-d recovery at 22°C. The seedlings then were challenged at 44°C for 50 min. For the BT assay, 4-d-old seedlings were challenged with 44°C for 23 min without prior acclimation. All phenotypes were assessed after a 10-d recovery at 22°C post-HS. For the TMHT assay, Arabidopsis seeds were sown on 0.53 MS medium plates containing 0.1% Suc. Five-day-old seedlings were then transferred to a growth chamber with temperature of 35°C 6 0.3°C during 16 h of light (110 mmol m 22 s 21 ) and 33.5°C 6 0.3°C during 8 h of darkness. The seedlings then recovered at 22°C for 10 d before phenotypes were assessed.

Immunoblotting
The full-length Arabidopsis Mge2 recombinant protein with a C-terminal tag of six His residues was produced and purified from E. coli cells as described (Charng et al., 2006). Immunization and serum collection were performed by a commercial service (LTK Biotechnology). The antibodies against Hsp101, Hsa32, sHsp-CI, and tubulin were as described (Charng et al., 2006;Chi et al., 2009). The total protein of plant samples was extracted as described (Charng et al., 2006). The mitochondria-enriched protein fraction was extracted with 30 mM MOPS (pH 7.5) buffer with 0.35 M mannitol and 2 mM EDTA. The lysate was first centrifuged at 4,500g for 5 min at 4°C to pellet the undesired crude fraction containing chloroplasts and other cell debris, then the supernatant was centrifuged at 16,000g for 15 min at 4°C. The resulting mitochondria-enriched protein fraction pellet was resuspended with Tris-HCl extraction buffer as for the extraction of total protein. The protein amount was measured by use of DC Protein Assay reagents (Bio-Rad) with bovine serum albumin used as a standard. Immunoblot analysis was performed as described (Charng et al., 2006;Chi et al., 2009).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Protein sequence alignment of Mge homologs.
Supplemental Figure S2. In silico analysis of the expression of Arabidopsis Mges.
Supplemental Table S1. List of primers used in this study.